Adsorption of Metals by Geomedia II: Variables, Mechanisms, and Model Applications (Developments in Earth and Environmental Sciences 7)

Elsevier Radarweg 29, PO Box 21 1, 1000 AE Amsterdam, The Netherlands Linacre House, Jordan Hill, Oxford OX2 8DP, UK F...

Author: Mark O. Barnett | Douglas B. Kent

113 downloads 1187 Views 8MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

Elsevier Radarweg 29, PO Box 21 1, 1000 AE Amsterdam, The Netherlands Linacre House, Jordan Hill, Oxford OX2 8DP, UK

F irst editio n 2008 Copyright

© 2008

Elsevier B.V. A ll rights reserved

No part o f this publication may be reproduced , s tored in a retrieval system o r tra nsmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without the prior written permission of the publisher Permissio ns may be sought directly from Elsevier's Science & Technology Rights Department in Oxford, U K : pho ne (+ 44) (0) 1865 84 3830; fax (+ 44) (0) 1865 853333; email: [email protected]. Alte rnatively you can submit your request online by visiting the Elsevier web site at http:/fwww.elsevier.com/ locatefpermissions, a nd selecting

Obtaining permission to use Elsevier material Notice No responsibility is assumed by the publisher for a ny injury a nd/ or d amage to persons o r property as a matter o f products liabili ty, negligence or otherwise, o r from a ny use o r o peration o f any methods, products, instructio ns o r ideas contained in the material herei n. Because of rapid advances in the medical sciences, in particula r, independent verification of diagnoses and drug dosages should be made Library of Congress Cataloging-in-Publication Data A ca talog record for this book is available from the Library of Congress British Library Cataloguing in Publication Data A ca talogue record for this book is available from the British Library

ISBN: 978-0-444-532 12- 1 ISSN: 157 1-9 197 For information o n all Elsevier publications visit our website at books.elsevier.com

Printed and bound in H ungary 08 09 10 I I 12

10 9 8 7 6 54 3 2

Working together to grow libraries in developing countries www.els.evier.com

I

www.bookaid.org

I

\vww.sabre.org

CONTENTS

Foreword

XI

CHAPTER I. Surface Structure a nd Reactivity of Iron Oxide- Water Interfaces Chose, S. K. , S. C. Petitto, K. S. Tanwar, C. S. Lo, P. J. Eng, A. M. Chaka and T P. Trainor 1.1. Introduction 2 1.2. Surface X-ray D iffraction Method: C rystal Truncation Rod (CTR) Technique 4 1.3. Examples of Structural Models of Different Iron Oxide I nterfaces 9 1.4. Perspecti ves and Applications to Surface Reactivity 21 !.5. Summa ry 23 References 24

C HAPTER 2. Anion Sorption T opology on Hematite: Compa rison of Arsena te and Silicate Waychunas, G. A., Y-S. Jun, P. J. Eng, S. K. Chose and T P. Trainor 2. 1. Introduction 2.2. Arsenate Crystal C hemistry in Minerals a nd on Surfaces 2.3. Silicate Crystal Chemistry as a Mo nomer and Small Polymer in Structures and on Surfaces 2.4. Structure of the H ema tite Surface 2.5. Results 2.6. Discussion 2.7. Prospects for F urther Studies 2.8. Conclusions References

31

32 33 38 44 46 56 59 61 62

vi

Contents

CHAPTER 3. Molecular Structure of Lead(II) Coprecipitated with Iron(Ill) Oxyhydroxide Kelly, S. D. , P. Lu, T. Bolin, S. Chattopadhyay, M. G. Newville, T. Shibata and C. Zhu 3. 1. 3.2. 3.3. 3.4. 3.5.

Introduction Experimental Experimental Results X-ray Absorption Modeling Discussion and Conclusions References

C HAPTER 4. Tracking the Interaction of Transition Metal Ions with Environmental Interfaces using Second Harmonic Generation Konek, C. T., M. J. Musorrafiti, A. B. Voges and F M. Geiger 4.1. 4.2. 4.3. 4.4. 4.5. 4.6.

Introduction Experimental Surface Characterization R esults Ion Binding Environmental Implicatio ns and Summary References

C HAPTER 5. Prions, Metals, a nd Soils Charlet, L. , Y. Chapron, G. R oman-R oss, C. Hureau, D. P. Hawkins and K. V. Ragnarsdottir 5. 1. 5.2. 5.3. 5.4. 5.5. 5.6.

Introduction Geochemistry of H otspots The Double Nature of the Prion Protein Prion Sorption and Transformation on C lays H orizontal Infectivity Conclusio ns References

C HAPTER 6. Associations between Iron Oxyhydroxide Nanoparticle Growth and Metal Adsorption/Structural Incorporation Kim, C. S., C. J. Lentini and G. A. Waychunas 6. 1. Introduction 6.2. Experimenta l

67

68 70 72 75

84 91

95

95 97 99 105 Ill 11 8 120 125

126 130 136 139 145 146 146

153

154 156

Contents

6.3. Results 6.4. Discussion 6.5. Conclusio ns References

CHAPTER 7. Temperature and Aging Effects on the Surface Speciation of Cd(ll) a t the Goethite- Water Interface Grafe, M. , G. Mustafa, B. Singh and R. S. Kookana 7.1. 7.2. 7.3. 7.4. 7.5.

In traduction Experimenta l Results Discussion Conclusions References

C HAPTER 8. Cadmium a nd Lead Desorption from K aolinite Srivastava, P. , M . Grafe, B. Singh and M. Balasubramanian 8. 1. 8.2. 8.3. 8.4. 8.5.

Introduction Experimental Results Discussion Conclusio ns References

C HAPTER 9. Mechanism of M olybdenum Adsorption on So ils a nd Soil Minerals Evalua ted Using Vibra tional Spectroscopy a nd Surface Complexation Modeling Goldberg, S., C. T. Johnston, D. L. Suarez and S. M. Lesch 9. 1. 9.2. 9.3. 9.4.

Introduction Experimenta l Results and Discussio n Conclusions References

vii

161 175 179 18 1

187

188 192 194 197 200 20 1

205

205 208 2 13 225 228 229

235

236 239 244 26 1 262

viii

Contents

CHAPTER 10. Blind Prediction and Parameter Uncertainty - A Sorption Test Case Richter, A. and V. Brendler 10.1. 10.2. 10.3. 10.4. 10.5.

Introduction Methodology Modeling Results and Discussion Summary and Conclusions References

CHAPTER II. Biogeochemical Uranium Redox Transform ations: Potential Oxida nts of Uraninite Ginder- Vogel, M. and S. Fendatf ll.l. 11.2. 11 .3. 11.4. 11 .5. 11 .6.

Introduction Uranium Oxidation- Reductio n Reactions Experimental Results Discussion Implications for Biogeochemical Uranium Cycling References

CHAPTER 12. Phosphate Interactions with Iron (Hydr)oxides: Mineralization Pathways and Phosphorus Retention upon Bioreduction Barch, T and S. Fendatf 12. 1. 12.2. 12.3. 12.4. 12.5.

Introduction Experimental Results Discussion Conclusion and Implications References

CHAPTER 13. Influence of Phosphate on Adsorption a nd Surface Precipitation of Lead on I ron O xide Surfaces Xie, L. and D. E. Giammar 13. 1. 13.2. 13.3. 13.4.

Introduction Experimental Results and Discussion Summary References

267

267 27 1 274 282 287 288

293

294 294 298 303 307 310 3 13

32 1

322 325 328 336 341 342

349

350 352 36 1 370 37 1

Contents

C HAPTER 14. Ura nium(VI) Release fro m Contamina ted Vadose Zone Sediments: Estimatio n o f Po tentia l Contributio ns fro m Dissolutio n a nd Desorptio n Bond, D. L. , J. A. Davis and J. M. Zachara 14.1. 14.2. 14.3. 14.4. 14.5.

Introductio n Experimental Results Discussio n Concluding R ema rks References

C HAPTER 15. Arsenic Specia tion in Solid Phases of G eotherm al Fields Alsina, M . A., I. Saratovsky, J. -F Gaillard and P. A. Pasten 15. 1. In troductio n 15.2. Arsenic in Geotherma l System s 15.3. Qua lita ti ve a nd Q ua ntitative C ha racterizatio n o f H o t Spring Deposits 15.4. X AS Ana lysis of Arsenic Solid-Phase Specia tio n in Ho t Springs 15.5. Concluding R ema rks References CHAPTER 16. Reactive Tra nspo rt a nd R esidence Times in U nsatura ted F ractured R ocks fro m F ield-Scale Ex periments Pili, E., S. Bureau, F. Perrier, D. Patriarche, L. Charlet, P. M. Adler and P. R ichon 16. 1. 16.2. 16.3. 16.4. 16.5. 16.6. 16.7.

Introd uctio n Setting Experimenta l Set up, Tracer Introduction and Recovery Reacti ve Tra nspo rt Tracer Tra nsport F racture Networks a nd F low Simula tio ns Conclusio ns References

Subject Index

ix

375

376 379 388 409 41 2 41 3 41 7

41 8 41 9 423 427 434 435

44 1

442 443 446 455 46 1 464 465 467

469

Foreword In 1996, international researchers gathered for several days at the Spring meeting of the American Chemical Society (ACS) to discuss the latest advances in the field of metal, metalloid, and radionuclide adsorption to soils, sediments, and rocks (geomedia). This symposium resulted in the publication of an outstanding peer-reviewed volume ‘‘Adsorption of Metals by Geomedia: Variables, Mechanisms, and Model Applications’’ (Academic Press, 1998, ISBN 0-12-384245-X). Everett Jenne’s introductory chapter of that volume, ‘‘Adsorption of Metals by Geomedia: Data Analysis, Modeling, Controlling Factors, and Related Issues,’’ remains an excellent summary introduction to the field. Our understanding of mineral–water interfacial reactions on geomedia has continued at a rapid pace since 1996, spurred by innovative research conducted over a wide range of spatial scales, from molecular-scale spectroscopic and microscopic techniques to new approaches to modeling mineral–water interfacial reactions at the field scale. In 2006, researchers from five continents convened at a follow-up to the original symposium, ‘‘Adsorption of Metals by Geomedia II,’’ held at the Spring ACS meeting, March 27–29, in Atlanta, Georgia, USA. The purpose of this 10-year anniversary symposium was to highlight the advances since 1996 by way of presentations and discussions among the global community of researchers addressing this vital area and to produce a peer-reviewed volume that reflects these advances. Over 60 contributions to the symposium spanned a wide range of topics related to understanding the adsorption of metals, metalloids, and radionuclides to geomedia, from pure laboratory studies to large-scale field studies. The objective of this book is to highlight some of the advances through a selection of peer-reviewed research papers drawn from the presentations at the symposium. The book begins with two chapters in which molecular-scale computational chemistry and surface-sensitive X-ray scattering and absorption spectroscopic techniques are applied to characterizing the types of surface functional groups and surface structures responsible for adsorption. In Chapter 1, Ghose et al. characterized the types of surface functional groups available at important crystallographic faces on goethite and hematite. These surface functional groups are favorable to both cation and oxyanion

xii

Foreword

adsorption. The configuration of the surface is influenced by sample preparation and, under hydrated conditions, is different from that determined under ultra-high vacuum owing to the presence of structured hydrated layers, cation vacancies, and relaxation effects. In Chapter 2, Waychunas et al. show that, consistent with configurations identified in iron or aluminum arsenate and iron or aluminum silicate minerals, silicate tends to form surface polymers whereas arsenate adsorbs as individual ions in various bidentate configurations available on different crystal faces of hematite. Further investigations of polymerization of silicate at iron oxide surfaces should shed light on the potentially important impact of adsorbed silicate on iron oxide adsorption properties. In Chapter 3, Kelly et al. employed X-ray diffraction (XRD), highresolution electron microscopy, and X-ray spectroscopy to show that coprecipitation of lead(II) and iron(III) at near-neutral pH results in formation of nanoparticulate, lepidocrocite-like sheets with lead substituted for iron. Approximately 20% of the iron or lead ions are at or near the surface and the structural environment of lead is distinct from that of lead adsorbed onto other iron(III) oxides and oxyhydroxides. In Chapter 4, Konek et al. describe how second harmonic generation and other non-linear optical techniques probe changes in surface charge density in optically transparent systems. Rapid and reversible adsorption of chromium(VI) on alumina was demonstrated by the achievement of stable and reproducible signals within minutes of changing adsorbate concentration. In Chapter 5, Charlet et al. synthesize work conducted at the molecular and field scale on potential relationships between physical and chemical properties of soils and the binding of prions implicated in the spread of scrapie, a fatal neurodegenerative disease in sheep. The authors put forward provocative hypotheses regarding the role of manganese(III), copper(II), and clay minerals in the pathways of infection. In Chapter 6, Kim et al. used laboratory experimental studies and synchrotron XRD and X-ray absorption data to show that adsorption of the strongly binding cation copper(II) or strongly binding oxyanion arsenic(V) impedes growth of nanoparticulate goethite. Growth of goethite particles during aging is accompanied by increases in pH, which drive increased adsorption of strongly adsorbing cationic metals like zinc and mercury(II). Chapter 7, by Gra¨fe et al., illustrates how refinements in X-ray absorption measurement and analysis tools have significantly increased the information that can be obtained from samples with low concentrations of target metal ions. The types of surface complexes formed by cadmium on goethite changed with aging time, but there was no evidence for formation of cadmium clusters. In contrast, Srivastava et al. (Chapter 8) found that aging

Foreword

xiii

of lead adsorbed on kaolinite resulted in increased formation of polynuclear lead–OH complexes at the surface. Concomitant decreases in the amount of lead that could be rapidly desorbed with increasing aging suggest that these polymerization reactions could lead to irreversible lead uptake. In Chapter 9, Goldberg et al. used infra-red and Raman spectroscopy to examine solution speciation of molybdenum(VI) and surface speciation of molybdenum(VI) on iron and aluminum oxyhydroxides. Good correspondence between spectroscopic data for aqueous molybdenum and molybdenum adsorbed onto aluminum oxyhydroxide at high pH suggests the dominance of outer-sphere surface complexes. In contrast, significant differences in band intensities between aqueous and adsorbed molybdenum at low pH, suggesting the importance of inner-sphere surface complexes. Empirical isotherm approaches (e.g., KD, Langmuir, and Freundlich isotherms) are not consistent with current scientific understanding of metal ion and metalloid adsorption to geomedia. In Chapter 10, Ritcher and Brendler point out that these approaches are being replaced in contaminant transport problems by those based on application of surface complexation models. They present methods for obtaining estimates of the extent of adsorption in systems where appropriate surface complexation model parameters are not available. Chapters 11 and 12 examine interfacial reactions under anoxic conditions. In Chapter 11, Ginder-Vogel and Fendorf showed that uraninite (UO2) generated by microbial reduction of uranium(VI) can be oxidized by ferrihydrite (poorly ordered iron(III) oxyhydroxide) at neutral pH, with the extent of oxidation increasing with ferrihydrite concentration or surface area. Thermodynamic considerations suggest that uranium(VI)–uranium(IV) and iron(III)–iron(II) redox couples are close in energy and, therefore, whether the oxidation can occur will depend on the concentration of iron(II) and other reactants and products. In Chapter 12, Borch and Fendorf present the results of microbial batch incubations and column experiments conducted with phosphate-loaded, ferrihydrite-coated quartz sand. During microbial reduction of iron(III) to iron(II), at low phosphate loadings, most of the phosphate was retained on the solids as a result of phosphate adsorption on magnetite or green rust but at high phosphate loadings, phosphate retention resulted from precipitation of vivianite. In Chapter 13, Xie and Giammar present the results of an experimental study of lead and phosphate adsorption on goethite-coated silica. Phosphate adsorption enhances lead adsorption and surface complexation model calculations indicate that the observed synergistic effects are consistent with the impact of phosphate adsorption on decreasing the unfavorable coulombic contribution to lead adsorption on goethite.

xiv

Foreword

The final three chapters present findings from case studies with complex natural geomedia. In Chapter 14, Bond et al. examined the release of uranium(VI) from contaminated sediments in which uranium(VI) occurs as surface complexes, the uranium(VI)–mineral metatorbernite, and uranium(VI) co-precipitated with iron oxides and oxyhydroxides. Short-term release is likely dominated by desorption but only accounts for a fraction of the total amount of uranium contamination in the sediments. A surface complexation model with relatively few parameters that were calibrated using laboratory experimental data successfully describes uranium(VI) adsorption–desorption over the range of chemical conditions applicable to the uranium-contaminated field site. In Chapter 15, Alsina et al. present Xray spectroscopic data on arsenic in silica-rich sinters and iron-rich material likely to be of microbial origin from the El Tatio geothermal field in Chile. The predominant forms of arsenic are arsenic(V) surface complexes on iron oxides, similar in structure to arsenic(V) surface complexes on synthetic and natural Fe oxides determined in laboratory studies. In Chapter 16, Pili et al. present a field study of reactive transport in a thick unsaturated zone developed in fractured rock. Time scales of transport through the complex flow paths are determined. The study documents the importance of cation exchange reactions during transport in cases involving the introduction of water whose composition differs from that of pore fluid in the fractured rock and overlying soil. The results of a wide variety of spectroscopic and microscopic studies have demonstrated that adsorption of metal and metalloid ions is the result of chemical reactions between adsorbing solutes and specific sites at mineral surfaces. These reactions include formation of surface complexes, formation of oligomers and polymers, formation of surface precipitates, epitaxial growth of minerals (especially clay minerals), and nucleation and growth of precipitates. The nature and extent of these reactions depend on solution conditions, especially pH and the concentration of the adsorbing solute. Surface complexation models (and extensions to include polymerization and surface precipitation) can describe quantitatively solute uptake in a manner compatible with these underlying mechanisms. Integrating surface complexation models with the entire network of solution speciation, precipitation–dissolution, and oxidation–reduction reactions influencing solute partitioning allows one to achieve a quantitative description of the influence of variable chemical conditions on solute uptake by geomedia. Coupling these comprehensive chemical reaction models with models describing fluid flow and solute transport has provided quantitative descriptions of the fate and transport of adsorbing solutes over applicable ranges of chemical conditions in field applications.

Foreword

xv

ACKNOWLEDGMENTS We would like to express our appreciation to those who made this book possible, particularly to Everett Jenne for organizing the original symposium. The editors and authors were well served by the thoughtful reviews provided by colleagues too numerous to mention individually. We also thank the presenters and session chairs, who made the symposium a scientific success, as well as Yoko Furukawa (Program Chair of the Geochemical Division of the American Chemical Society), Gautham Jeppu, and Vijay Loganathan, whose support contributed significantly to the success of the symposium. Finally, we gratefully acknowledge the help of Linda VersteegBuschman and Tirza van Daalen at Elsevier. Mark O. Barnett Auburn, AL, USA Douglas B. Kent Menlo Park, CA, USA

Developments in Earth & Environmental Sciences, 7 Mark O. Barnett and Douglas B. Kent (Editors) r 2008 Elsevier B.V. All rights reserved DOI 10.1016/S1571-9197(07)07001-2

Chapter 1

Surface Structure and Reactivity of Iron Oxide–Water Interfaces Sanjit K. Ghose1,, Sarah C. Petitto2, Kunaljeet S. Tanwar2, Cynthia S. Lo2,3, Peter J. Eng1, Anne M. Chaka3 and Thomas P. Trainor2 1

Consortium for Advanced Radiation Sources, The University of Chicago, Chicago, IL 60637, USA 2 Department of Chemistry and Biochemistry, University of Alaska Fairbanks, Fairbanks, AK 99775, USA 3 Physical and Chemical Properties Division, National Institute of Standards and Technology, Gaithersburg, MD 20899, USA

ABSTRACT The surface structure and composition of the three distinct iron-(hydr)oxide sys¯ tems, goethite (1 0 0), hematite (1102), and magnetite (1 1 1) were determined under hydrated conditions at room temperature using crystal truncation rod (CTR) analysis. The prediction of surface protonation states and the overall chemical plausibility of the experimental surface models are performed using a bond-valence (BV) analysis. Further analysis of the surface energetics is carried out using ab initio density functional theory (DFT). The analysis of three common iron-(hydr)oxide surface systems reveals the differences in interface structure and distribution of hydroxyl groups at substrate–water interfaces. The goethite (1 0 0) interface structure is determined to have a relaxed double hydroxyl termination with the presence of two semi-ordered water layers that expose a surface with A-type (Fe-OH2) and B-type (Fe2-OH) hydroxyl groups. The ¯ hydrated hematite (1102) interface structure has vacancies in the near surface metal sites, resulting in three types of surface functional groups: A type, B type, and C type (Fe3-O). The interface structure of magnetite (1 1 1) shows two chemically nonequivalent oxygen surface terminations in the surface ratio of 70 O4-Feoh-O4-Fetd1ohtd2:30 O4-Fetd1ohtd2-O4-Feoh suggesting that the octahedral irons are the principal irons involved at the environmental interfaces. In the above three systems, there also is evidence for multiple domains with fractional Corresponding author. Tel.: +1-630-252-0433; Fax:+1-630-252-0436;

E-mail: [email protected] (S.K. Ghose).

2

S. K. Ghose et al.

ordered unit cell steps determined by atomic force microscopy (AFM). Results obtained for the structure of the iron-(hydr)oxide–water interfaces from the CTR and DFT analyses are different from stoichiometric termination of the bulk structure or hydroxylation of the ultra high vacuum (UHV) determined surface structures.

1.1. Introduction Iron-(hydr)oxide phases occur extensively in natural aquatic systems, often comprising major component of soils, sediments, and suspended solids in natural waters (Cornell and Schwertmann, 2003). Heterogeneous reactions at the iron-(hydr)oxide–water interface play a key role in controlling the dissolved concentration of trace elements through adsorption or co-precipitation reactions (Sposito, 1984; Davis and Kent, 1990; Stumm, 1997; Cornell and Schwertmann, 2003) and are important for the heterogeneous transformation of environmental contaminants (Stumm and Morgan, 1996; Elsner et al., 2004; Hofstetter et al., 2006). The biological availability and geochemical cycling of iron also are controlled by heterogeneous processes including biotic and abiotic redox reactions involving soluble Fe(II) and insoluble Fe(III) (Sulzberger et al., 1989; Lovley et al., 1991; Stumm and Sulzberger, 1992; Zhu et al., 1997; Hansel et al., 2004; Kerisit and Rosso, 2006). The incorporation of heterogeneous processes in conceptual and quantitative models of geochemical processes requires a detailed understanding of the rates, extents, and product identities resulting from reactions at the iron(hydr)oxide–water interface, and how these factors depend on environmental and mineralogical variables. For example, the phase identity, variations in particle size, morphology, crystallographic orientation, and surface treatment have all been shown to strongly affect the observed reactivity of iron(hydr)oxide substrates (Bargar et al., 1996, 2004; Templeton et al., 2001; Chambers and Yi, 1999; Trainor et al., 2002c; Gaboriaud and Ehrhardt, 2003; Hansel et al., 2004; Waychunas et al., 2005). The variations in reactivity most likely result from differences in the numbers and types of hydroxyl moieties present at the solid–solution interface (Cornell and Schwertmann, 2003) that can interact with solution species through Brønsted and Lewis acid–base reactions or through ligand exchange (Stumm and Morgan, 1996). The reactivity of oxide surface hydroxyl groups are often related to their cation coordination (Russel et al., 1974; Sposito, 1984; Cornell and Schwertmann, 2003): monodentate (FeOH) (A-type or hydroxo), bidentate

Surface Structure and Reactivity of Iron Oxide–Water Interfaces

3

(Fe2OH) (B-type or m-hydroxo), and tridentate (Fe3OH) hydroxyls (C-type or m3-hydroxo). From a crystal chemistry perspective, the most stable protonation state of these hydroxyl groups depends upon the degree of saturation of the oxygen anion. In addition to the influence of metal coordination number, variations in the metal–oxygen bond length and the degree of hydrogen bonding also influence oxygen saturation, and the combined effect of these variables leads to expected variations of pKa values from roughly 0 to 13 (Hiemstra et al., 1989; Hiemstra and Van Riemsdijk, 1996; Rustad et al., 1996; Stumm and Morgan, 1996). Similarly, the coordination environment of surface hydroxyl groups is expected to control metal binding affinities, and the overall reactivity of a surface will depend upon the numbers and types of exposed hydroxyl groups, as well as their topographic arrangement at the interface (Hiemstra et al., 1989; Hiemstra and Van Riemsdijk, 1996, 1999, 2006; Rustad et al., 1996; Smith and Ferris, 2001; Rahnemaie et al., 2006). Therefore, predicting interfacial reactivity requires a comprehensive understanding of the substrate surface structure including the crystallographic orientation of the terminating surface and any subsequent modifications to the surface after exposure to water or other potentially surface modifying solutes (Stumm, 1997; Hiemstra and Van Riemsdijk, 2006; Kerisit et al., 2006). A systematic investigation of mineral surface structures, in particular iron-(hydr)oxide surfaces, is needed to further understand the factors influencing interface structure and reactivity to improve upon mechanistic models of environmental interfacial processes (Sposito, 1989; Brown et al., 1999). A number of approaches have been utilized to obtain detailed structural and chemical information from specific surfaces of iron-(hydr)oxides. For example, ultra high vacuum (UHV) studies have investigated the surface symmetry, composition, and reactivity using low energy electron diffraction (LEED) and photoemission studies (Liu et al., 1998; Brown et al., 1999; Rakovan et al., 1999; Ritter and Weiss, 1999; Kendelewicz et al., 2000; Henderson, 2002). Single crystal surfaces also have been studied on in-situ systems (under bulk fluid conditions or ambient atmospheric conditions) using scanning probe microscopies (SPM) (Lennie et al., 1996; Rakovan et al., 1999; Eggleston et al., 2003) that have provided detailed information regarding surface topography of iron-(hydr)oxides. However, both UHV and SPM techniques have limited sensitivity to the atomic scale threedimensional (3-D) structure of the interface. In this chapter, we discuss recent surface X-ray scattering studies for investigating the solid–solution interface structure of three different iron(hydr)oxide systems: goethite (1 0 0), hematite (1 1¯ 0 2), and magnetite (1 1 1). The use of hard X-ray scattering and spectroscopic techniques have found

4

S. K. Ghose et al.

widespread application in the study of mineral–fluid interface phenomena (Brown and Sturchio, 2002), in large part due to the penetrating nature of X-rays and their sensitivity to molecular-scale structure. Crystal truncation rod (CTR) diffraction, a technique sensitive to the 3-D arrangement of atoms at an interface, has emerged as a powerful tool for determining the structure of mineral–fluid interfaces (Eng et al., 2000; Fenter, 2002; Trainor et al., 2002a, 2004; Fenter and Sturchio, 2004). The iron-(hydr)oxide mineral phases presented here are common environmental substrates spanning a diverse range of structure and composition. Goethite (a-FeOOH) is the most common crystalline iron oxide phase found in soils (Cornell and Schwertmann, 2003; van der Zee et al., 2003). Hematite (Fe2O3) and magnetite (Fe3O4) are also naturally abundant iron oxides (Hochella et al., 1999; Cornell and Schwertmann, 2003; Frankel and Bazylinski, 2003) that have been particularly well studied due to both their environmental and technological importance (Weiss and Schlogl, 2000; Reiss and Huetten, 2005).

1.2. Surface X-ray Diffraction Method: Crystal Truncation Rod (CTR) Technique 1.2.1. Background CTRs are diffuse streaks of X-ray scattering intensity that line up between bulk Bragg peaks in a direction perpendicular to the surface and result from the abrupt termination of a semi-infinite crystal lattice (Fig. 1.1). The intensity along a CTR is extremely sensitive to the ordered arrangement of atoms at the surface (Robinson, 1986, 1991; Robinson and Tweet, 1992). The analysis of the CTR intensity provides information regarding the identity of the prevalent surface termination (Fenter and Park, 2004; Trainor et al., 2004), surface roughness and relaxations (Robinson, 1986; Vlieg et al., 1989), solid–liquid interfacial structure (Eng et al., 2000; Fenter and Sturchio, 2004), and the structure of adsorbed species on crystalline terminations (Fenter and Sturchio, 2004; Catalano et al., 2005). CTR measurements can be performed using a variety of sample environments, allowing the analysis of surface structures under controlled conditions, such as an aqueous solution and under UHV. A drawback of CTR diffraction is the limited sensitivity to low Z elements due to their weak scattering power (e.g., hydrogen) and the insensitivity to poorly ordered components of the interface such as weakly bound/disordered sorbates and physisorbed water. These limitations are addressed by comparing the structural models determined from the


5

Figure 1.1: A Schematic Picture of the X-Ray Scattering Geometry from a Truncated Bulk Crystal. The Momentum Transfer Q, is Related to Incident (Ki) and Reflected (Kr) X-Rays Wave Vectors by Q ¼ Kr Ki. A Reciprocal Space Schematic Showing Every Bragg Peaks (Black Dots) Intersected by a CTR (Dashed Line) in a Direction Perpendicular to the Surface. A Typical CTR Profile is Shown with Structure Factor Variation in Perpendicular Reciprocal Space Vector (L) (See Text for Details). experimental analysis to the results of theoretical calculations of interfacial structure and energetics (e.g., ab initio and molecular dynamics methods). The scattering intensity measured at a position on a CTR is proportional to the square of the structure factor jF T j2 ; which describes how the incident beam is scattered by the atoms in a unit cell of the crystalline lattice. The magnitude of the CTR structure factor is calculated for a particular reciprocal lattice setting (HKL) (Robinson, 1986; Vlieg et al., 1989; Vlieg, 2000) from, F T ¼ SrðF bc F CTR þ F sc Þ

(1.1)

where Fbc is the structure factor of the bulk unit cell (Warren, 1969), F CTR ¼ 1=½1 expði2pLÞ the CTR complex form factor, S an overall scale factor, and r a roughness factor (Robinson, 1986). Fsc is the structure factor of the surface unit cell " 2 # n X jQj F sc ¼ (1.2) yj f j expðiQ rj Þ exp Bj 4p j¼1 with the sum taken over all n atoms of the surface unit cell having an atomic scattering factor fj, site occupancy yj, fractional coordinate rj, isotropic Debye-Waller factor Bj, and the scattering vector Q. The surface under study is indexed using the reciprocal vector indices (HKL), where H and K

6

S. K. Ghose et al.

correspond to the in-plane momentum transfer, and L is the perpendicular momentum transfer (Fig. 1.1). 1.2.2. Measurement Methods and Instrumentation The data were collected at The University of Chicago, Center for Advanced Radiation Sources (CARS) at Advanced Photon Source (APS) Sector 13 undulator beamline. CTR measurements were performed at room temperature under a thin layer of water (1–20 mm) or a water saturated He atmosphere (relative humidity>90%, pH2O>20 Torr) to ensure that the surfaces remain fully hydrated during the course of measurements (Liu et al., 1998). X-rays with energies ranging from 10 to 12 keV were delivered to the sample using beamline optics consisting of a LN2 cooled double crystal Si(1 1 1) monochromator and Rh coated vertical and horizontal focusing (and harmonic rejection) mirrors. Sample orientation and scanning were performed using a 2+2+kappa-geometry diffractometer equipped with a sample cell with X-ray transparent windows (Trainor et al., 2006). Nonspecular CTR intensities were collected by performing a continuous (trajectory) rocking scan of the diffractometer f-axis at a particular reciprocal lattice setting using a fixed incident angle of 2–41, and the specular rods were collected using the a-axis scans. Individual structure factors were determined by taking the square root of the background subtracted intensity of the rocking curves and correcting for active area, polarization, scan speed, and Lorentz factor (Robinson, 1991; Fenter, 2002). A typical full data set used for structural analyses consists of all CTRs that are accessible within the Ewald sphere defined by the X-ray wavelength, including symmetry equivalent rods. 1.2.3. CTR Analysis and Model Comparison The CTR data analysis procedure is described schematically in Fig. 1.2. The measured structure factors are fit with a calculated CTR profile using Eq. (1.1) and a model consisting of a fixed bulk crystal structure (A) and an adjustable surface crystal structure (B) (Fig. 1.2a). As described in Fig. 1.2b, analysis starts by focusing first on the well-ordered region B of the interface, where the positions and presence or absence (occupancies) of the surface atoms in this region are responsible for the major features on a CTR and dominate the initial fit. A nonlinear least squares fit of region B is performed where atomic displacements, occupancies, and Debye-Waller factors for the atoms within this surface region are allowed to vary, though constrained by


7

symmetry, assumptions regarding maximum/minimum bond lengths, and coordination numbers (Vlieg, 2000; Trainor et al., 2002b). The quality of the fit is characterized using a reduced w2, and the best fit model is selected based on the lowest w2 and the chemical plausibility of individual bond lengths, occupancy, and coordination of the atoms in the top layers. An additional check on the chemical plausibility of the model is performed using a bondvalence (BV) analysis (Pauling, 1960) where the Fe-O bond valence values are calculated using the empirical bond length–bond strength relationship of Brown and Altermatt (1985), while the O-H bond valence contributions are calculated based on the approach provided by Bargar et al. (1997). Under aqueous conditions, interfaces may have a semi-ordered hydration layer(s) that persist for roughly 1–3 molecular layers above the crystalline termination of the bulk crystal shown in Fig. 1.2a (region C) (Fenter and Sturchio, 2004). An indication that the semi-ordered region contributes to the CTR data is the inability to provide an adequate fit based solely on models of the crystalline surface termination (region B) or larger than expected surface relaxations. The effect of hydration layer(s) on CTR data is directly observed in the specular and low-Q off-specular rods resulting from the greater interlayer (perpendicular to the surface) order than intra-layer (parallel to the surface) order exhibited by these semi-ordered species. To test for the presence of semi-ordered hydration layer(s), the surface model is modified to include atoms in region C. The new model is analyzed leaving the

Figure 1.2: (a) A Schematic Picture of the Atomic Layer Model for CTR Analysis: A, Crystalline Bulk Structure; B, Crystalline Surface Structure; and C, Semi-Ordered Physisorbed Water or Sorbates. (b) A Flow Chart Depicting the ROD CTR Analysis Procedure (See Text for Details).

8

S. K. Ghose et al.

Figure 1.2: (Continued ).


9

atoms in region B at their previous refinements and the atoms in region C are allowed to vary to an optimum refinement. Regions B and C are then varied simultaneously and the resultant final-fit model is checked for statistical significance using the Hamilton’s R-ratio test (Hamilton, 1965). As discussed above, the models resulting from the direct analysis of the scattering data are most sensitive to the presence of heavier elements (e.g., Fe and O), and thus provide no direct information regarding the protonation states of the surface hydroxyls. However, the protonation states of the various surface oxygens may be estimated by calculating the number of O-H bonds and/or hydrogen bonds that are required for saturation of the surface oxygens. This approach provides a relatively complete model for the structure of the iron-(hydr)oxide–solution interface, with a degree of uncertainty regarding the positions of hydrogens and disordered interfacial species. The utilization of an independent theoretical means of predicting the interface structure provides a check on the plausibility of the results and gives additional information regarding the driving forces that dictate surface structure. Recent examples of this approach are the combined CTR and DFT studies of the hydrated C-cut hematite (0 0 0 1) surface (Trainor et al., 2004), the adsorption of Zn2+ and Sr2+ on the TiO2(1 1 0)–electrolyte interface (Zhang et al., 2006), and the hydrated R-cut hematite (1 1¯ 0 2) surface described below (Tanwar et al., 2007).

1.3. Examples of Structural Models of Different Iron Oxide Interfaces 1.3.1. Structure and Reactivity of Goethite a-FeOOH(1 0 0)–Water Interface The bulk goethite (a-FeOOH) structure has a space group symmetry of Pnma and can be described as a distorted hexagonally close packing of O and OH groups with Fe3+ occupying half of the octahedral interstices. Each of the six-coordinated iron sites have three short Fe-O bonds (1.96 A˚) and three long Fe-OH bonds (2.10 A˚) and there are four repeat formula units (FeOOH) present within each unit cell (Fig. 1.3a) (Cornell and Schwertmann, 2003). Natural and synthetic goethites have three predominant crystal faces: (1 1 0), (0 2 1), and (1 0 0) (Weidler et al., 1998; Cornell and Schwertmann, 2003). In this study, we characterized the structure of the (1 0 0) cleavage face, which has six possible chemically distinct terminations of the bulk structure shown in Fig. 1.3a: a double hydroxyl layer (layer 1), a single hydroxyl layer (layer 2), an iron over double oxygen layer (layer 3), a double oxygen layer

10

S. K. Ghose et al.


11

(layer 4), a single oxygen layer (layer 5), and an iron over double hydroxyl layer (layer 6). The presence of the screw axis within the unit cell results in chemically equivalent termination pairs (e.g., layers 1 and 7) that have opposite handedness (and are therefore crystallographically distinct), which results in the possibility of half unit cell steps between regions that are chemically and presumably energetically equivalent. Evidence of half-ordered steps has been observed using atomic force microscopy (AFM) step height analyses (not shown) on the prepared samples described below. The orientated a-FeOOH(1 0 0) cleaved single crystal samples, 1 1 0.2 mm (natural crystals from Cornwall, England), were prepared using a standard acid–base cleaning, and CTR measurements performed at room temperature under a water saturated He atmosphere. A subset of the full CTR data set with the calculated CTRs for the best possible termination models are shown in Fig. 1.4. The unrelaxed double hydroxyl ((OH)2-Fe-O2Fe-OH-R, where R represents the repeat of the bulk stoichiometry stacking, Fig. 1.3a) termination resulted in the best overall match to the data when compared to other termination models. Nonlinear least squares fits are initially performed on the (OH)2-Fe-O2-Fe-OH-R model, where the surface and near surface atom lattice positions, Debye-Waller factors, and occupancies were allowed to vary resulting in a fit with a w2 ¼ 2.53. The fit has reasonable bond lengths for the Fe-OH bonds, and a chemically viable coordination environment for the terminal Fe atoms. However, closer examination of the fit model shows poor agreement to the experimental data in the (0 0 L) and (1 0 L) rods (Fig. 1.4). As described in Section ‘‘Surface X-ray diffraction method: crystal truncation rod (CTR) technique,’’ a fit to the Figure 1.3: Layer Stacking Sequence Along cs Axis for a-FeOOH(1 0 0): (a) The Unrelaxed Bulk Structure Showing the Different Possible Stoichiometric Bulk Terminations (Where R Represents the Repeat of the Bulk Stoichiometry), Layer Numbering, and the Stoichiometric Unit Cell (Dashed Box). (b) The Best Fit Model: Double-Hydrated Relaxed Double Hydroxyl ((OH2)2-(OH)2-Fe-O2-Fe-OH-R) Termination with Experimentally Derived Layer Spacings and Percent Layer Relaxations in the z-Direction. The Atoms in Bold Face in the Best Fit Model Formula Represent the Layers Added Above the First Layer (Layer 1) of the Stoichiometric Termination. The Large Spheres Represent Oxygen Atoms, the Small Spheres Represent Iron Atoms, and the Smallest Spheres Represent H Atoms. The Adsorbed Water Molecules are Shown Above Layer 1 with Arbitrary Water Molecule Size and Orientation. The A and B Notation Represents Oxygen Atoms Coordinated to One and Two Iron Atoms, Respectively, and the Arrows on the Terminal OH Show the y-Displacements.

12

S. K. Ghose et al.

Figure 1.4: Experimental (Circles) and Theoretical Structure Factors (FHKL) as a Function of Perpendicular Momentum Transfer (L, in Reciprocal Lattice Units) for the a-FeOOH(1 0 0) Surface. The Dotted Lines Represent Calculated CTRs for the Unrelaxed Double Hydroxyl Termination ((OH)2-Fe-O2-Fe-OH-R), the Dashed Lines are the Calculated CTRs for the Relaxed Double Hydroxyl Termination ((OH)2-Fe-O2-Fe-OH-R), and the Solid Lines Represent the Best Fit Model, the Double-Hydrated Relaxed Double Hydroxyl Termination ((OH2)2-(OH)2-Fe-O2-Fe-OH-R) (Fig. 1.3b). experimental data using a model with only region B surface atoms has apparent missing Fourier components, which implies that the addition of a semi-ordered hydration layer(s) (region C) is justified. Two hydration layers (oxygen layers) were then added to the model resulting in a best fit hydrated model for the FeOOH(1 0 0) surface that is described as a double-hydrated relaxed double hydroxyl ((OH2)2-(OH)2-Fe-O2-Fe-OH-R) termination with a w2 ¼ 2.11 (at confidence level 95%) (Fig. 1.4). The structural model and layer displacements are shown in Fig. 1.3b, and additional details regarding the analysis procedure and atomic coordinates for the best fit model are given in Ghose et al. (2007). The best fit a-FeOOH(1 0 0) CTR model resulted in in-plane (y) relaxations of the top two terminal atoms (layers 1 and 2) in addition to layer


13

contractions for layers 1–2, 3–4, and 8–9 and expansions for layers 2–3, 4–5, 5–6, 6–7, and 7–8 (Fig. 1.3b). The atomic relaxations resulted in Fe-hydroxyl bond length of 2.0670.02 A˚, which is comparable to the bulk Fe-OH bond length of 2.10 A˚ (Cornell and Schwertmann, 2003) with the two water layers at 1.770.1 A˚ and 2.970.2 A˚ above the surface termination layer. As previously stated CTR is insensitive to proton positions and therefore is unable to provide information regarding the protonation states of surface terminated oxo groups under hydrated conditions. Ab initio DFT calculations are used to gain further insight into the stability and structure of the surface by calculating the energetics for the possible terminations and structural models that then can be compared to the best fit CTR model. The DFT model for hydrated iron oxide surfaces are constructed from semi-infinite slabs, which are generally 16–22 atomic layers thick and separated by 10 A˚ of vacuum. The structures are optimized using the DMol3 DFT code that implements the PBE-GGA functional commonly used in condensed matter calculations (Delley, 1990, 2000; Perdew et al., 1996). Thus, a variety of surface stoichiometries, protonation states, and water adlayer structures can be calculated to determine the most stable surface termination (for details refer Lo et al., 2007). The surface free energy of the system, including equilibrium with the gas or liquid phase in contact with the surface, can be calculated from (Reuter and Scheffler, 2002): " # X 1 gðT; p; fN i gÞ ¼ N i mi ðT; pÞ (1.3) G slab ðT; p; fN i gÞ 2A i where Gslab is the Gibbs free energy of the solid substrate in contact with a gas phase reservoir at a given pressure (p) and temperature (T); Ni and mi the number and chemical potential, respectively, of the ith type of atom, and A the surface area of the slab. The chemical potentials of the constituent elements are then related to the chemical potentials of the major species in the system, assuming chemical equilibrium between the solid and vapor phases. For example, in a metal-oxide system, the upper limit of oxygen chemical potential is defined as the point at which gaseous O2 would condense on the solid phase, and the lower limit of the oxygen chemical potential is defined as the point below which the oxide would decompose into solid metal and gaseous O2 (Wang et al., 1998). The surface free energies for a-FeOOH(1 0 0) as a function of the oxygen chemical potential at 0 K are shown in Fig. 1.5. The DFT calculations for different termination models without any hydration layers resulted in the double hydroxyl ((OH)2-Fe-O2-Fe-OH-R) termination as the lowest possible

14

S. K. Ghose et al.

Figure 1.5: The Surface Free Energies of Various Surface Termination Models of a-FeOOH(1 0 0) as a Function of Oxygen Chemical Potential (mo). The Termination Models are Provided in the Legend. surface energy model, which is in agreement with the CTR results. Further detailed ab initio DFT calculations on the hydrated surface and protonation models are currently underway. BV analyses are used to check the chemical plausibility of the best fit CTR model and to estimate the protonation states of the surface oxo groups by calculating the number of O-H bonds and/or hydrogen bonds that are required for complete saturation of the surface oxygen ions. The BV sum for the top Fe layer for the double hydroxyl termination is 3.0970.04, which is comparable to the bulk saturated value of Fe (2.96). The layer-1 terminal hydroxyls are A-type (monodentate) with a bond valence sum of 1.2770.07 considering contributions from Fe and a single proton. This site is considerably undersaturated, even with the addition of 1–2 additional hydrogen bonds (e.g., 0.2–0.4 v.u. additional BV). However, the addition of a second proton (thus making this site an Fe-OH2) would result in saturation of the layer 1 oxygen. The layer-2 terminal hydroxyls are B-type (bidentate) with BV sum 1.7070.04 with contributions from two Fe and a single proton. This


15

site is roughly saturated if a single hydrogen bond is added (Ghose et al., 2007). Previous analysis of the goethite (1 0 0) surface by Rakovan et al. (1999) have shown that the surface is hydroxylated, and they predict that the terminal iron-layer (layer 3) should be in six-fold coordination in the presence of water resulting in a surface with A- and B-type surface functional groups. Our combined CTR and DFT analysis presented above suggest a surface stoichiometry consistent with (OH2)2-(OH2)-(OH)-Fe-O2-Fe-OH-R in agreement with Rakovan et al.’s analysis. Furthermore, the experimentally derived structural model suggests that the layer 1 (A-type) surface hydroxyls are substantially undersaturated and hence are likely to protonate, resulting in terminal aquo groups. These results for the goethite (1 0 0) surface can be used to constrain the analysis of surface proton exchange under variable pH conditions, and the reactivity of the surface functional groups with respect to adsorption of aqueous solutes (Hiemstra et al., 1989, 1996; Rahnemaie et al., 2006). 1.3.2. Structure and Reactivity of Hematite a-Fe2O3(1102)–Water Interface The bulk structure of hematite a-Fe2O3 (space group symmetry of R-3c) has a distorted hexagonal close-packed (HCP) sequence of oxygen anions, where Fe3+ cations occupy two-thirds of the octahedral (oh) sites. Natural a-Fe2O3 has two predominant growth faces, the (1102) and (0 0 0 1) (Cornell and Schwertmann, 2003). For a-Fe2O3(1 1¯ 0 2) surface, the ideal stoichiometric termination (Fig. 1.6a) is reported to be stable under UHV conditions at temperatures below 600 K and at higher temperatures there is evidence for a (2 1) reconstruction potentially due to the ordering of oxygen vacancies (Lad and Henrich, 1988; Gautier-Soyer et al., 1996; Henderson et al., 1998). Water is known to react dissociatively on the stoichiometric termination of a-Fe2O3(1 1¯ 0 2) resulting in bridging and terminal hydroxyls, which also completes the coordination shell of first layer of iron atoms (Fig. 1.6b) (Henderson et al., 1998; Henderson, 2002). A number of previous studies have investigated both the clean (UHV prepared) and water reacted clean surface structure of a-Fe2O3(1 1¯ 0 2). We present here the surface structure of the hydrated a-Fe2O3(1 1¯ 0 2) surface prepared using a wet chemical– mechanical polishing (CMP) procedure under ambient conditions (Tanwar et al., 2007). Natural single crystals of a-Fe2O3(1 1¯ 0 2) 10 10 2 mm (obtained from Bahia, Brazil) were prepared for CTR data collection using a wet CMP procedure followed by acid etching and multiple rinses with ultra-pure water

16

S. K. Ghose et al.

(a)

(b)

Figure 1.6: Layer Stacking Sequence Along cs Axis for a-Fe2O3(1 1¯ 0 2) (a) Ideal Bulk Stoichiometric Termination and (b) Hydroxylated Stoichiometric Termination. The Large Spheres Represent Oxygen and Small Spheres Represent Iron Atoms. The IO and IIIO Represent Oxygen Atoms Coordinated to One and Three Iron Atoms, Respectively, and the VFe and VI Fe Represents Iron Atoms Coordinated to Five and Six Oxygen Atoms, Respectively. (Tanwar et al., 2007). Freshly prepared samples were analyzed by AFM (not shown) prior to CTR measurements that revealed a terrace-step-morphology with half unit cell step heights and surface roughness less than 5 A˚. A subset of the full CTR data set along with the calculated CTR for unrelaxed


17

stoichiometric and hydroxylated stoichiometric a-Fe2O3(1 1¯ 0 2) surface is presented in Fig. 1.7a. The calculated CTRs for both models are in poor agreement to the experimental data. Furthermore, allowing relaxations in these models did not improve the quality of the fit suggesting that the a-Fe2O3(1 1¯ 0 2) stoichiometric bulk termination models can be eliminated and the resultant surface is highly dependent on the surface preparation and experimental conditions. To identify the surface structure for the CMP prepared, hydrated a-Fe2O3(1 1¯ 0 2) surface, numerous chemically distinct Fe-O stoichiometric models were tested (Lo et al., 2007; Tanwar et al., 2007). The model that results in the best fit with the experimental data is a termination where the first layer of Fe cations are absent in comparison to the stoichiometric termination (Fig. 1.7b). However, this surface model has an excess of negative charge on the surface resulting from the missing Fe3+ (two per unit cell). Though under hydrated conditions, the excess negative charge is presumably balanced by the hydroxylation of surface oxo groups by adding back six H+ ions. To gain insight into the protonation states, several slab models of the hydrated surface were optimized using DFT, and their surface free energies are calculated using Eq. (1.3) and shown in Fig. 1.8 (Lo et al., 2007). The (H2O)2-X-(HO)2-Fe2-O2-R surface termination, with zero occupancy of the topmost Fe cations (X) is the most energetically and thermodynamically stable at 0 K, which is consistent with the CTR results shown above. Furthermore, the proposed DFT model is in excellent agreement with the surface relaxations obtained with the experimental CTR data analysis (Fig. 1.7b). Using both the DFT and CTR results, the best fit model has predominantly three types of surface functional groups: Fe-OH2 (A type), Fe2-OH (B type), and Fe3-O (C type) (Fig. 1.7b) (Lo et al., 2007; Tanwar et al., 2007) that were not previously identified on the UHV prepared surfaces (Lad and Henrich, 1988; Gautier-Soyer et al., 1996; Henderson et al., 1998). 1.3.3. Structure and Reactivity of Magnetite Fe3O4(1 1 1)–Water Interface Magnetite (Fe3O4) crystallizes in an inverse spinel cubic crystal lattice with a space group symmetry of Fd3m with a lattice parameter of 8.3963 A˚. Half the octahedral (oh) sites are occupied with both Fe2+ and Fe3+ cations, oneeighth of the tetrahedral (td) sites are filled with Fe3+ cations, and the facecentered cubic (FCC) lattice sites are occupied with O2 anions (Wycoff, 1982; Cornell and Schwertmann, 2003). The most common growth face of Fe3O4 is the (1 1 1) termination, which has six possible unreconstructed

18

S. K. Ghose et al.

surface terminations in the unit cell with the repeat unit of O4-Feoh-O4Fetd1ohtd2 (Cornell and Schwertmann, 2003). In hydrated environments, oxygen terminated surfaces, O4-Feoh-O4-Fetd1ohtd2, and O4-Fetd1ohtd2O4-Feoh, are expected (Fig. 1.9), while under UHV conditions, iron terminated surfaces, Feoh-O4-Fetd1ohtd2-O4, or a combination of the mixed-iron Fetd1ohtd2-O4-Feoh-O4 surface, are predicted. Previous UHV studies observed the mixed-iron terminations with either a quarter monolayer (ML) of

(a)

(b)


19

Fetd2-O4-Feoh-O4 (Weiss et al., 1993; Barbieri et al., 1994; Ritter and Weiss, 1999; Kendelewicz et al., 2000; Zhu et al., 2006) or a half ML of Feohtd2-O4Feoh-O4 (Lennie et al., 1996) on a distorted HCP oxygen layer where both terminations have vacant Fetd1 lattice sites. The orientated Fe3O4(1 1 1) single crystal sample 10 10 2 mm (Commercial Crystal Laboratories, Inc.) was prepared using a standard colloidal silica CMP procedure (Tanwar et al., 2007) followed by a base and acid wash with subsequent CTR data acquired in hydrated conditions at circum-neutral pH. The crystalline surface was of high quality and multiple domains, and minimal surface roughness (r2 A˚ RMS) was confirmed with AFM analyses (not shown). The preparation procedure results in reproducible surfaces with excellent CTR intensity. Hydrated Fe3O4(1 1 1) CTR data was analyzed using a nonlinear least squares fit with fixed bulk and adjustable surface parameters as described in Section ‘‘Surface X-ray diffraction method: crystal truncation rod (CTR) technique’’ of this chapter. The best fit model for the hydrated Fe3O4(1 1 1) surface is a surface contribution of the two chemically nonequivalent oxygen surface terminations in the surface ratio of 70 O4-Feoh-O4-Fetd1ohtd2:30 O4-Fetd1ohtd2-O4-Feoh (Fig. 1.10). However, there is partial occupancy within the mixed-iron termination surface layer where the Fetd1 lattice sites are half-full and the Feoh and Fetd2 iron sites are subsequently more occupied (Petitto et al., 2007). The multiple domain structure is consistent with the AFM analysis of fractional ordered unit cell steps. An ordered water layer could not be constrained within the Figure 1.7: (a) Experimental (Circles) and Theoretical Structure Factors (FHKL) as a Function of Perpendicular Momentum Transfer (L, in Reciprocal Lattice Units) for the a-Fe2O3(1 1¯ 0 2) Surface. The Dashed Lines Represent Calculated CTRs for the Ideal Stoichiometric Termination (O2-Fe2-O2-Fe2-O2-R), the Dotted Lines are the Calculated CTRs for the Bulk Termination with an Added Oxygen Layer (O2-O2-Fe2-O2-Fe2-R), and the Solid Lines Represent the Best Fit Model Termination ((H2O)2-(H2O)2O2-X-O2-Fe2-O2-R). The Atoms in Bold Face in the Best Fit Model Formula Represent the Layers Added Above the First Layer of the Stoichiometric Termination. (b) Layer Stacking Sequence Along cs Axis for the Best Fit Model Along with Experimental and DFT Derived Percent Layer Relaxations in the z-Direction. The Large Spheres Represent Oxygen, and the Small Spheres Represent Iron Atoms. The Adsorbed Water Molecules are Shown Above Layer 1 with Arbitrary Water Molecule Size and Orientation. The IO, II O, and IIIO Represent Oxygen Atoms Coordinated to One, Two, and Three Iron Atoms, Respectively, and the VIFe Represents Iron Atoms Coordinated to Six Oxygen Atoms.

20

S. K. Ghose et al.

Figure 1.8: The Surface Free Energies of Various Water Free Surface Terminations of a-Fe2O3(1 1¯ 0 2) as a Function of Oxygen Chemical Potential (mo). The Stoichiometries of Models are Provided in the Legend and ‘‘X’’ Represents Absence of a Given Layer.

Fe3O4(1 1 1) CTR fit model possibly due to the presence of two distinct terminations having different arrangements of near surface waters, thus making the overall contribution to the CTR particularly weak. This point is currently under further examination. The environmentally relevant hydrated surface structure, which incorporates both oxygen termination layers of Fe3O4(1 1 1) does not correspond with the previously determined mixed-iron terminations found in UHV, though discussion remains on which UHV termination is most stable. Furthermore, the hydrated surface is dominated by the O4-Feoh-O4-Fetd1ohtd2 termination layer; a layer that is not stable in UHV and has both Fe2+ and Fe3+ cations in the uppermost iron layer. This implies that the octahedral irons are the principal irons involved at environmental interfaces and in controlling the surface reactivity of magnetite. Preliminary results also suggest that the Fe3O4(1 1 1) surface structure is dynamic in the presence of water with evolution of the surface structure over time. Continued research


21

Figure 1.9: Layer Stacking Sequence Along the cs Axis for the Oxygen Terminated Fe3O4(1 1 1) (a) O4-Feoh-O4-Fetd1ohtd2 and (b) O4-Fetd1ohtd2O4-Feoh Where the Large Spheres Represent Oxygen Atoms and the Small Spheres Represent Iron Atoms. into the surface structure/reactivity is underway including ab initio DFT calculations to determine the most thermodynamically stable surface structures and the protonation states of the surface oxo groups.

1.4. Perspectives and Applications to Surface Reactivity The overall surface reactivity of natural iron oxides is dictated by a wide range of factors, including bulk structure, surface orientation, presence of defects, and interaction of surface-modifying species. In this chapter, we have provided some insights into the atomic structure of the mineral–fluid structure for several different crystal-face specific iron oxide systems. Such model system studies provide a molecular-scale view of the processes that dictate reactivity, namely the detailed understanding of the coordination environment of surface functional groups. These studies also serve as a basis for the systematic analysis of surface reactivity using detailed experimental and theoretical methods, building off a well-constrained understanding of the

22

S. K. Ghose et al.

Figure 1.10: Experimental (Circles) and Theoretical Structure Factors (FHKL) as a Function of Perpendicular Momentum Transfer (L, in Reciprocal Lattice Units) for the Fe3O4(1 1 1) Surface. The Dashed Line Represents Calculated CTRs for the Ideal Stoichiometric Termination (O4-FeohO4-Fetd1ohtd2), the Dotted Line is the Calculated CTRs for the Ideal Mixed-Iron Stoichiometric Termination (O4-Fetd1ohtd2-O4-Fe), and the Solid Line Represents the Best Fit Model Termination 70 O4-Feoh-O4Fetd1ohtd2:30 O4-Fetd1ohtd2-O4-Feoh. surface structure. For example, these results provide a basis for the application of structure-based models for prediction of surface charge and proton exchange behavior (Hiemstra et al., 1989; Hiemstra and Van Riemsdijk, 1996, 2006; Rustad et al., 1996; Smith and Ferris, 2001; Rahnemaie et al., 2006; Sverjensky and Fukushi, 2006). Furthermore, results such as these coupled with crystal face specific reactivity studies provide a means for interpreting reactivity based on both the coordination chemistry and topology of surface functional groups (Bargar et al., 1996, 2004; Weidler et al., 1998; Templeton et al., 2001; Chambers and Yi, 1999; Trainor et al., 2002c; Gaboriaud and Ehrhardt, 2003; Chun et al., 2006; Kerisit and Rosso, 2006). For instance, differences in surface reactivity on the alumina (0 0 0 1) and


23

(1 1¯ 0 2) and hematite (0 0 0 1) and (1 1¯ 0 2) surfaces (Bargar et al., 1996, 2004; Templeton et al., 2001; Trainor et al., 2002c) are interpreted to result from the predominance of double-coordinated surface hydroxyls on the alumina (0 0 0 1) (Eng et al., 2000), which are predicted to be particularly stable and nonreactive with respect to binding of aqueous Pb(II). The singleand triple-coordinated surface hydroxyl groups found in varying proportions on the other surfaces (Trainor et al., 2002a, 2004; Tanwar et al., 2007) are likely to have much greater proton lability and be much more reactive with respect to the binding of aqueous Pb(II) and other cations. Thus, as an initial step of interpretation, the surface structural differences obtained by the CTR analysis method are consistent with the observation of reactivity differences on different faces of hematite and alumina crystals. Results such as these are awaiting further detailed model calculations in order to provide a complete interpretation in terms of the electronic structure differences of the surfaces, in addition to the differences in surface coordination chemistry.

1.5. Summary The results described above show that CTR diffraction is a powerful tool for determining the 3-D molecular structure of the solid–liquid interface under environmentally relevant conditions of pressure and temperature. The analysis of three common iron-(hydr)oxide surface systems – goethite (1 0 0), hematite (1 1¯ 0 2), and magnetite (1 1 1) – reveals the differences in interface structure and distribution of hydroxyl groups at substrate–water interfaces. Goethite (1 0 0) interface structure is determined as relaxed double hydroxyl termination with the presence of two semi-ordered water layers that exposes a surface with A- and B-type hydroxyl groups. However, the hematite (1 1¯ 0 2) and magnetite (1 1 1) interface structures show vacancies in the near surface metal occupancies and different distributions of surface hydroxyl groups. In the above three systems, there is evidence for multiple domains with fractional ordered unit cell steps determined by AFM. Analysis of the CTR and DFT data also show that a priori prediction of the surface stoichiometry based on the bulk structure or hydroxylation of the UHV determined surface structures may not be correct. Rather, the presence of water and the chemical/physical history of the surface during preparation or reaction are likely to result in modifications of the surface structure and hence affect reactivity. The surface structure information obtained from these studies of the hydrated iron oxide interfaces provides structural constraints for testing

24

S. K. Ghose et al.

structure specific reactivity models. Continued work to investigate the detailed relationships between environmental variables, surface reactions, and the resulting surface structures, as well as the structure/reactivity relationships, are needed to further the development of detailed models of interfacial processes in geochemical modeling applications.

ACKNOWLEDGMENTS We thank Glenn A. Waychunas, Jeffrey G. Catalano, and Lahsen Assoufid for their help with the sample preparation, characterization, and experimentation. We also sincerely thank Gordon E. Brown Jr. for his insightful discussions. This research was supported by NSF grants BES-0404400 and CHE-0431425 and the ACS Petroleum Research Fund. Portions of this work were performed at Arctic Region Supercomputing Center and Advanced Instrumentation Laboratory at the University of Alaska Fairbanks and at GeoSoilEnviroCARS (Sector 13) and MHATT-CAT (Sector 7) of the Advanced Photon Source (APS), Argonne National Laboratory. Use of the APS was supported by DOE Basic Energy Sciences, Office of Energy Research, under Contract No. W-31-109-Eng-3.

REFERENCES Barbieri, A., Weiss, W., Van Hove, M. A., & Somorjai, G. A. (1994). Magnetite Fe3O4(111): Surface structure by LEED crystallography and energetics. Surf. Sci., 302, 259–279. Bargar, J. R., Towle, S. N., Brown, G. E. Jr., & Parks, G. A. (1996). Outer-sphere Pb(II) adsorbed at specific surface sites on single crystal a-alumina. Geochim. Cosmochim. Acta, 60, 3541–3547. Bargar, J. R., Towle, S. N., Brown, G. E. Jr., & Parks, G. A. (1997). XAFS and bond-valence determination of the structures and compositions of surface functional groups and Pb(II) and Co(II) sorption products on single-crystal a-Al2O3. J. Colloid Interface Sci., 185, 473–492. Bargar, J. R., Trainor, T. P., Fitts, J. P., Chambers, S. A., & Brown, G. E. Jr. (2004). In situ grazing-incidence extended X-ray absorption fine structure study of Pb(II) chemisorption on hematite (0001) and (1 1¯ 0 2) surfaces. Langmuir, 20, 1667–1673. Brown, G. E. Jr., Henrich, V. E., Casey, W. H., Clark, D. L., Eggleston, C., Felmy, A., Goodman, D. W., Gratzel, M., Maciel, G., McCarthy, M. I., Nealson, K. H., Sverjensky, D. A., Toney, M. F., & Zachara, J. M. (1999). Metal oxide surfaces and their interactions with aqueous solutions and microbial organisms. Chem. Rev., 99, 77–174.


25

Brown, G. E. Jr., & Sturchio, N. C. (2002). Applications of synchrotron radiation in low-temperature geochemistry and environmental sciences. Rev. Mineral. Geochem., 49, 1–115. Brown, I. D., & Altermatt, D. (1985). Bond-valence parameters obtained from a systematic analysis of the inorganic crystal structure database. Acta Crystallogr. B, 41, 244–247. Catalano, J. G., Trainor, T. P., Eng, P. J., Waychunas, G. A., & Brown, G. E. Jr. (2005). CTR diffraction and grazing-incidence EXAFS study of U(VI) adsorption onto a-Al2O3 and a-Fe2O3 (1 1¯ 0 2) surfaces. Geochim. Cosmochim. Acta, 69, 3555–3572. Chambers, S. A., & Yi, S. I. (1999). Fe termination for a-Fe2O3(0001) as grown by oxygen-plasma-assisted molecular beam epitaxy. Surf. Sci., 439, L785–L791. Chun, L. C., Penn, R. L., & Arnold, W. A. (2006). Kinetic and microscopic studies of reduction of organic contaminants on goethite. Environ. Sci. Technol., 40, 3299–3304. Cornell, R. M., & Schwertmann, U. (2003). Iron Oxides: Structure, Properties, Reactions, Occurrence, and Uses. 2nd Ed., Wiley-VCH, Weinheim. Davis, J. A., & Kent, D. B. (1990). Surface complexation modeling in aqueous geochemistry. In: M. F. Hochella Jr., & A. F. White (Eds). Mineral–Water Interface Geochemistry (Rev. Min.), Vol. 23, pp. 177–260. Delley, B. (1990). An all-electron numerical method for solving the local density functional for polyatomic molecules. J. Chem. Phys., 92, 508–517. Delley, B. (2000). From molecules to solids with the DMol3 approach. J. Chem. Phys., 113, 7756–7764. Eggleston, C. M., Stack, A. G., Rosso, K. M., Higgins, S. R., Bice, A. M., Boese, S. W., Pribyl, R. D., & Nichols, J. J. (2003). The structure of hematite (a-Fe2O3) (001) surfaces in aqueous media: Scanning tunneling microscopy and resonant tunneling calculations of coexisting O and Fe terminations. Geochim. Cosmochim. Acta, 67, 985–1000. Elsner, M., Schwarzenbach, R. P., & Haderlein, S. B. (2004). Reactivity of Fe(II)bearing minerals toward reductive transformation of organic contaminants. Environ. Sci. Technol., 38, 799–807. Eng, P. J., Trainor, T. P., Brown, G. E. Jr., Waychunas, G. A., Newville, M., Sutton, S. R., & Rivers, M. L. (2000). Structure of the hydrated a-Al2O3(0001) surface. Science, 288, 1029–1033. Fenter, P. (2002). Applications of synchrotron radiation in low-temperature geochemistry and environmental sciences. Rev. Mineral. Geochem., 49, 149–216. Fenter, P., & Park, C. (2004). Termination interference along crystal truncation rods of layered crystals. J. Appl. Crystallogr., 37, 977–987. Fenter, P., & Sturchio, N. C. (2004). Mineral-water interfacial structures revealed by synchrotron X-ray scattering. Prog. Surf. Sci., 77, 171–258. Frankel, R. B., & Bazylinski, D. A. (2003). Biologically induced mineralization by bacteria. Rev. Mineral. Geochem., 54, 95–114.

26

S. K. Ghose et al.

Gaboriaud, F., & Ehrhardt, J.-J. (2003). Effects of different crystal faces on the surface charge of colloidal goethite (a-FeOOH) particles: An experimental and modeling study. Geochim. Cosmochim. Acta, 67, 967–983. Gautier-Soyer, M., Pollak, M., Henriot, M., & Guittet, M. J. (1996). The (1 2) reconstruction of the a-FE2O3 (.HIVIN.1012) surface. Surf. Sci., 352–354, 112–116. Ghose, S. K., Waychunas, G. A., Eng, P. J., Lo, C. S., Catalano, J. G., & Trainor, T. P. (2007). Structure of hydrated goethite (100) surface. (to be submitted.) Hamilton, W. C. (1965). Significance tests on crystallographic R factor. Acta Crystallogr., 18, 502–510. Hansel, C. M., Benner, S. G., Nico, P., & Fendorf, S. (2004). Structural constraints of ferric(Hydr)oxides on dissimilatory iron reduction and the fate of Fe(II). Geochim. Cosmochim. Acta, 68, 3217–3229. Henderson, M. A. (2002). Insights into the (1 1)-to-(2 1) phase transition of the a-Fe2O3(012) surface using eels, LEED and water TPD. Surf. Sci., 515, 253–262. Henderson, M. A., Joyce, S. A., & Rustad, J. R. (1998). Interaction of water with the (1 1) and (2 1) surfaces of a-Fe2O3(012). Surf. Sci., 417, 66–81. Hiemstra, T., & Van Riemsdijk, W. H. (1996). A surface structural approach to ion adsorption: The charge distribution (CD) model. J. Colloid Interface Sci., 179, 488–508. Hiemstra, T., & Van Riemsdijk, W. H. (1999). Effect of different crystal faces on experimental interaction form and aggregation of hematite. Langmuir, 15, 8045–8051. Hiemstra, T., & Van Riemsdijk, W. H. (2006). On the relationship between charge distribution, surface hydration, and the structure of the interface of metal hydroxides. J. Colloid Interface Sci., 301, 1–18. Hiemstra, T., Venema, P., & Van Riemsdijk, W. H. (1996). Intrinsic proton affinity of reactive surface groups of metal (hydr)oxides: The bond valence principle. J. Colloid Interface Sci., 184, 680–692. Hiemstra, T., de Wit, J. C. M., & Van Riemsdijk, W. H. (1989). Multisite proton adsorption modeling at the solid/solution interface of (hydr)oxides: A new approach ii. Application to various important (hydr)oxides. J. Colloid Interface Sci., 133, 105–117. Hochella, M. F. Jr., Moore, J. N., Golla, U., & Putnis, A. (1999). A TEM study of samples from acid mine drainage systems: Metal-mineral association with implications for transport. Geochim. Cosmochim. Acta, 63, 3395–3406. Hofstetter, T. B., Neumann, A., & Schwarzenbach, R. P. (2006). Reduction of nitroaromatic compounds by Fe(II) species associated with iron-rich smectites. Environ. Sci. Technol., 40, 235–242. Kendelewicz, T., Liu, P., Doyle, C. S., Brown, G. E. Jr., Nelson, E. J., & Chambers, S. A. (2000). Reaction of water with the (100) and (111) surfaces of Fe3O4. Surf. Sci., 453, 32–46. Kerisit, S., Ilton, E. S., & Parker, S. C. (2006). Molecular dynamics simulations of electrolyte solutions at the (100) goethite surface. J. Phys. Chem. B, 110, 20491–20501.


27

Kerisit, S., & Rosso, K. M. (2006). Computer simulation of electron transfer at hematite surfaces. Geochim. Cosmochim. Acta, 70, 1888–1903. Lad, R. J., & Henrich, V. E. (1988). Structure of a-ferric oxide single crystal surfaces following argon+ ion bombardment and annealing in oxygen. Surf. Sci., 193, 81–93. Lennie, A. R., Condon, N. G., Leibsle, F. M., Murray, P. W., Thorton, G., & Vaughn, D. J. (1996). Structures of Fe3O4(111) surface observed by scanning tunneling microscopy. Phys. Rev. B, 53, 10244–10253. Liu, P., Kendelewicz, T., Brown, G. E. Jr., Nelson, E. J., & Chambers, S. A. (1998). Reaction of water vapor with a-Al2O3(0001) and a-Fe2O3(0001) surfaces: Synchrotron X-ray photoemission studies and thermodynamic calculations. Surf. Sci., 417, 53–65. Lo, C. S., Tanwar, K. S., Chaka, A. M., & Trainor, T. P. (2007). Structure of clean and hydrated a-Fe2O3(1 1¯ 0 2) surfaces, studied via density functional theory. Phys. Rev. B, 75, 075425-0–075425-15. Lovley, D. R., Phillips Ellzabeth, J. P., & Lonergan, D. L. (1991). Enzymatic versus nonenzymatic mechanisms for Fe(III) reduction in aquatic sediments. Environ. Sci. Technol., 25, 1062–1967. Pauling, L. (1960). The Nature of the Chemical Bond. 3rd Ed., Cornell University Press, Ithaca, NY. Perdew, J. P., Burke, K., & Ernzerhof, M. (1996). Generalized gradient approximation made simple. Phys. Rev. Lett., 77, 3865–3868. Petitto, S. C., Tanwar, K. S., Ghose, S. K., Eng, P. J., & Trainor, T. P. (2007). Surface termination structure of hydrated magnetite (111). (to be submitted.) Rahnemaie, R., Hiemstra, T., & van Riemsdijk, W. H. (2006). Inner- and outersphere complexation of ions at the goethite-solution interface. J. Colloid Interface Sci., 297, 379–388. Rakovan, J., Becker, U., & Hochella, M. F. Jr. (1999). Aspects of goethite surface microtopography, structure, chemistry, and reactivity. Am. Miner., 84, 884–894. Reiss, G., & Huetten, A. (2005). Magnetic nanoparticles: Applications beyond data storage. Nat. Mater., 4, 725–726. Reuter, K., & Scheffler, M. (2002). Composition, structure, and stability of RuO2(110) as a function of oxygen pressure. Phys. Rev. B, 65, 035406-1–035406-11. Ritter, M., & Weiss, W. (1999). Fe3O4(111) surface structure determined by LEED crystallography. Surf. Sci., 432, 81–94. Robinson, I. K. (1986). Crystal truncation rods and surface roughness. Phys. Rev. B, 33, 3830–3836. Robinson, I. K. (1991). Surface crystallography. In: G. S. Brown, & D. E. Moncton (Eds). Handbook on Synchrotron Radiation, Vol. 3. North-Holland: Elsevier, Amsterdam, pp. 221–226. Robinson, I. K., & Tweet, D. J. (1992). Surface X-ray diffraction. Rep. Prog. Phys., 55, 599–651.

28

S. K. Ghose et al.

Russel, J. D., Parfitt, R. L., Fraser, A. R., & Farmer, V. C. (1974). Surface structure of gibbsite, goethite and phosphated goethite. Nature, 248, 220–221. Rustad, J. R., Felmy, A. R., & Hay, B. P. (1996). Molecular static calculations of proton binding to goethite surfaces: A new approach to estimation of stability constants for multisite surface complexation models. Geochim. Cosmochim. Acta, 60, 1563–1576. Smith, K. L., & Ferris, F. G. (2001). Proton binding by hydrous ferric oxide and aluminum oxide surfaces interpreted using fully optimized continuous pKa spectra. Environ. Sci. Technol., 35, 4637–4642. Sposito, G. (1984). The Surface Chemistry of Soils. Oxford University Press, New York. Sposito, G. (1989). Surface reactions in natural aqueous colloidal systems. Chimica, 43, 169–176. Stumm, W. (1997). Reactivity at the mineral-water interface dissolution and inhibition. Colloids Surf., 120, 143–166. Stumm, W., & Morgan, J. J. (1996). Aquatic Chemistry: Chemical Equilibria and Rates in Natural Waters. 3rd Ed., Wiley, New York. Stumm, W., & Sulzberger, B. (1992). The cycling of iron in natural environments: Considerations based on laboratory studies of heterogeneous redox processes. Geochim. Cosmochim. Acta, 56, 3233–3257. Sulzberger, B., Suter, D., Siffert, C., Banwart, S., & Stumm, W. (1989). Dissolution of iron(III) (hydr)oxides in natural waters: Laboratory assessment on the kinetics controlled by surface coordination. Mar. Chem., 28, 127–144. Sverjensky, D. A., & Fukushi, K. (2006). A predictive model (ETLM) for As(III) adsorption and surface speciation on oxides consistent with spectroscopic data. Geochim. Cosmochim. Acta, 70, 3778–3802. Tanwar, K. S., Lo, C. S., Eng, P. J., Catalano, J. G., Walko, D. A., Brown, G. E. Jr., Waychunas, G. A., Chaka, A. M., & Trainor, T. P. (2007). Surface diffraction study of the hydrated hematite (1 1¯ 0 2) surface. Surf. Sci., 601, 460–474. Templeton, A. S., Trainor, T. P., Traina, S. J., Spormann, A. M., & Brown, G. E. Jr. (2001). Pb(II) distributions at biofilm-metal oxide interfaces. Proc. Natl. Acad. Sci., 98, 11897–11902. Trainor, T. P., Chaka, A. M., Eng, P. J., Newville, M., Waychunas, G. A., Catalano, J. G., & Brown, G. E. Jr. (2004). Structure and reactivity of the hydrated hematite (0001) surface. Surf. Sci., 573, 204–224. Trainor, T. P., Eng, P. J., Brown, G. E. Jr., Robinson, I. K., & De Santis, M. (2002a). Crystal truncation rod diffraction study of the a-Al2O3 (1102) surface. Surf. Sci., 496, 238–250. Trainor, T. P., Eng, P. J., & Robinson, I. K. (2002b). Calculation of crystal truncation rod structure factors for arbitrary rational surface terminations. J. Appl. Crystallogr., 35, 696–701. Trainor, T. P., Templeton, A. S., Brown, G. E. Jr., & Parks, G. A. (2002c). Application of the long-period X-ray standing wave technique to the analysis of surface reactivity Pb(II) sorption at a-Al2O3/Aqueous solution interfaces in the presence and absence of Se(VI). Langmuir, 18, 5782–5791.


29

Trainor, T. P., Templeton, A. S., & Eng, P. J. (2006). Structure and reactivity of environmental interfaces: Application of grazing angle X-ray spectroscopy and long-period X-ray standing waves. J. Electron Spectrosc. Relat. Phenom., 150, 66–85. Vlieg, E. (2000). ROD: A program for surface X-ray crystallography. J. Appl. Crystallogr., 33, 401–405. Vlieg, E., Van der Veen, J. F., Gurman, S. J., Norris, C., & Macdonald, J. E. (1989). X-ray diffraction from rough, relaxed and reconstructed surfaces. Surf. Sci., 210, 301–321. Wang, X. G., Weiss, W., Shaikhutdinov, Sh. K., Ritter, M., Petersen, M., Wagner, F., Schlogl, R., & Scheffler, M. (1998). The hematite (a-Fe2O3) (0001) surface: Evidence for domains of distinct chemistry. Phys. Rev. Lett., 81, 1038–1041. Warren, B. E. (1969). X-ray Diffraction. Addison-Wesley, New York. Waychunas, G. A., Kim, C. S., & Banfield, J. F. (2005). Nanoparticulate iron oxide minerals in soils and sediments: Unique properties and contaminant scavenging mechanisms. J. Nanopart. Res., 7, 409–433. Weidler, P. G., Hug, S. J., Wetche, T. P., & Hiemstra, T. (1998). Determination of growth rates of (100) and (110) faces of synthetic goethite by scanning force microscopy. Geochim. Cosmochim. Acta, 62, 3407–3412. Weiss, W., Barbieri, A., Van Hove, M. A., & Somorjai, G. A. (1993). Surface structure determination of an oxide film grown on a foreign substrate: Fe3O4 multilayer on Pt(111) identified by low energy electron diffraction. Phys. Rev. Lett., 71, 1848–1851. Weiss, W., & Schlogl, R. (2000). An integrated surface science approach towards metal oxide catalysis. Top. Catal., 13, 75–90. Wycoff, R. W. G. (1982). Crystal Structures. 2nd Ed., Interscience, New York. van der Zee, C., Roberts, D. R., Rancourt, D. G., & Slomp, C. P. (2003). Nanogoethite is the dominant reactive oxyhydroxide phase in lake and marine sediments. Geology, 31, 993–996. Zhang, Z., Fenter, P., Cheng, L., Sturchio, N. C., Bedzyk, M. J., Machesky, M. L., Anovitz, L. M., & Wesolowski, D. J. (2006). Zn2+ and Sr2+ adsorption at the TiO2(110)-electrolyte interface: Influence of ionic strength, coverage, and anions. J. Colloid Interface Sci., 295, 50–64. Zhu, L., Yao, K. L., & Liu, Z. L. (2006). First-principles study of the polar (111) surface of Fe3O4. Phys. Rev. B, 74, 035409-1–035409-10. Zhu, X. R., Prospero, J. M., & Millero, F. J. (1997). Diel variability of soluble Fe(II) and soluble total Fe in north african dust in the trade winds at barbados. J. Geophys. Res., 102, 21297–21306.


Chapter 2

Anion Sorption Topology on Hematite: Comparison of Arsenate and Silicate Glenn A. Waychunas1,, Young-Shin Jun1,2, Peter J. Eng3, Sanjit K. Ghose3 and Thomas P. Trainor4 1

Earth Sciences Division, E. O. Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA 94720, USA 2 Department of Earth and Planetary Sciences, University of California at Berkeley, Berkeley, CA 94720, USA 3 Consortium for Advanced Radiation Sources, University of Chicago, Chicago, IL 60637, USA 4 Department of Chemistry and Biochemistry, University of Alaska Fairbanks, Fairbanks, AK 99775, USA

ABSTRACT Arsenate and silicate are tetrahedral anions that strongly sorb to positive Fe oxide surfaces over the pH range 2–7. Both are important agents for modification of Fe oxide surface reactivity, and notably passivate against other sorption reactions. Arsenate is a significant health hazard as a sorbed pollutant associated with acid mine drainage, while silicate is a common anion in natural solutions. Our aim is to understand the types of sorption complexes that form with these anions on different crystal faces, and whether polymerization occurs with the silicate units. Silicate polymerization could dramatically alter Fe oxide surface reactivity. The structural characterization is conducted using both grazing incidence extended X-ray absorption fine structure (GIEXAFS) at Stanford Synchrotron Radiation Laboratory (SSRL) and the National Synchrotron Light Source (NSLS), and surface diffraction (using crystal truncation rod (CTR) analysis) at the Advanced Photon Source (APS). GIEXAFS yields interatomic distances from arsenic and silicon to their oxygen first neighbor shell and second Fe or other next neighbor shell, and thus allows identification of the local geometry of sorption. Polarized X-ray fine structure spectroscopy further allows determination of the orientation and density of the complexes on the various Fe surface planes. However, this information is incomplete as any response of the surface to sorption is Corresponding author. Tel.:+630-252-7376; Fax:+510-495-7152;

E-mail: [email protected] (G.A. Waychunas).

32

G. A. Waychunas et al.

not revealed, and hydrogen bonding and water molecule arrangement at the surface can be changed due to the sorption process. To access these we use CTR experiments and compare the results with samples without sorbed anions. GIEXAFS results for both hematite (0 0 0 1) and ð1102Þ planes show arsenate sorbed in two ways: bidentate binuclear and bidentate mononuclear. Most of the latter type of sorption geometry appears to be present on surface step edges on the ð1102Þ surface, while there is little or no such attachment to the ð1102Þ surface terraces. These results appear consistent with preliminary structural models obtained by CTR work for the clean and sorbed wet hematite surfaces, and further analysis is in progress. In the case of silicate, strong changes in the CTR are observed indicating considerable surface interactions. CTR measurements of the silicate on the ð1102Þ surface of hematite suggest that a monodentate sorption geometry is more dominant than other possible sorption geometries. However GIEXAFS analysis done on analogous samples at the Si K-edge shows second shell contributions that are similar to amorphous silica, and are largely independent of sorption density. These results suggest that there are at least two kinds of silicate on the hematite surface: a chemically bound fraction present as a silicate–Fe3+ octahedral complex, and an incoherent portion which is not detectable by the CTR measurements but appears to dominate the GIEXAFS results. The complexed silicate on the hematite ð1102Þ surface is linked by a single oxygen to surface Fe, i.e., a monodentate connection, with an interatomic Si-Fe distance close to those observed in the nontronite and acmite structures. This is the first evidence to our knowledge that identifies silicate as a well-defined sorption complex rather than only as an amorphous surface precipitate. More significantly, the monodentate complex is apparently favored over a more strongly bound bidentate complexation geometry, suggesting that polymerization at the interface may originate synchronously with sorption and be mitigated by the position of surface sites consistent with a polymerized network. These findings have important significance for the nature of passivation processes on Fe oxides and other important reactive natural surfaces.

2.1. Introduction Arsenate and silicate anions in the monomeric form are tetrahedral units that strongly sorb to positive Fe (hydr)oxide surfaces over pH ranges from 2 to 7. Both anions can compete with each other for the reactive sorption sites on Fe (hydr)oxide surfaces in the same systems along with phosphate ions (Meng et al., 2002; Luxton et al., 2006). The sorption process can lead to passivation of surfaces and reduction of particle growth, even at quite high pH values (Waychunas et al., 1993, 1996). Arsenate is a widespread toxic component of acid mine drainage and is generally created from dissolved arsenite due to mixing of mine waters with more oxygenated surface waters (Welch et al.,

Anion Sorption Topology on Hematite

33

2000; Nordstrom, 2002). Silicate, as both monomers and larger polymeric units, is ubiquitous in the environment owing to the high solubility of silica at near-circumneutral pH (typically millimolar concentration, (Icopini et al., 2005)). This relatively high silicate concentration may have a large passivation effect on mineral surfaces, especially Fe and Mn (hydr)oxides, rendering them less active chemically than expected. Monomeric silicate can also react with solvated Fe3+ and polymerized Fe3+ species in solution forming more complex clusters (Pokrovski et al., 2003). The transformation of ferrihydrite to goethite is strongly affected by the presence of silicate anions, and the coprecipitation of Fe3+ with silicate often results in amorphous or poorly crystalline Fe oxides (Zeng, 2003). Silicate differs from arsenate inasmuch as polymerization can occur after or along with monomer sorption (Meng et al., 2002). Hence whereas arsenate sorption can be assumed to be monomeric in practically all cases, a wide range of silicate attachment to surfaces might be contemplated. Such varying attachment and sorption geometry may lead to several different pathways for passivation, including local formation of a new silicate phase by chemical reaction with the Fe oxide substrates. In this work, we present recent grazing-incidence X-ray absorption spectroscopy results on arsenate and silicate sorbed onto single crystal hematite surfaces, and initial crystal truncation rod (CTR, surface diffraction) analysis of silicate sorbed to single crystal hematite.

2.2. Arsenate Crystal Chemistry in Minerals and on Surfaces 2.2.1. Bond Valence Calculations Bond valence analysis suggests that arsenate can link to Fe3+ polyhedra at surfaces or within minerals, in several ways. Octahedral Fe3+ bonded to arsenate yields a valence sum on the mutual oxygen of 1.75 (0.5+1.25), which could adjust to near 2.0 by bond contraction or the addition of a strong hydrogen bond. Arsenate oxygens connected with two octahedral Fe3+ would have a bond valence of 2.25 (0.5+0.5+1.25), but this could potentially be reduced by increased bond length. Mitigating against such a topology is the relative strength of the As–O bond, and the inability of an oxygen involved in forming three bonds to move appreciably within a structure. However, this situation could be relaxed on surfaces. Finally, tetrahedral Fe3+ could bind to arsenate to yield a perfect bond valence sum (0.75+1.25). Without consideration of the stability of Fe–O–As bond angles, both edge-sharing and corner-sharing binding appears permissible.

34


2.2.2. Arsenate Linkages in Mineral Structures Arsenate monomers are found in a large number of mineral structures, both as arsenates and as substituents in phosphate and vanadate minerals. In the case of arsenates there at least 25 species that have been named and characterized with mainly Fe3+ or Al3+ polyhedra (a representative set is listed in Table 2.1), but not all have had their crystal structures solved. In those for which structures are known, the arsenate always has a similar type of topology in the structure; namely, one or more of the arsenate apical oxygens are connected to Fe3+ polyhedra (generally always slightly distorted six-coordinated shapes). For example, in wilhelmkleinite (ZnFe3+ 2 (AsO4)2(OH)2) there are 3+ edge-sharing Fe O6 octahedral chains along the b axis that are bridged by arsenate tetrahedra such that each Fe3+O6 octahedron has two attached 3+ arsenates; in pharmacosiderite (KFe3+ O6 4 (AsO4)3(OH)4 7(H2O)) the Fe octahedra form an edge-sharing tetrahedral cluster with four Fe, each Fe3+O6 octahedron connected by corners to three arsenate groups (Fig. 2.1A); in scorodite (Fe3+(AsO4) 2(H2O)) each Fe3+O6 octahedron is attached to four arsenate groups (Fig. 2.1B); in durangite (NaAl(AsO4)F) each Al3+O6 octahedron is connected by the four oxygen apices (the other two are fluorine) to arsenate tetrahedra; and in angelellite (Fe4(AsO4)2O3) there are chains of edge-sharing Fe octahedra that are connected by arsenate groups where the arsenate tetrahedra share two apices with single apex of the Fe3+O6 octahedra, while the other two apices share an oxygen with two octahedra at a shared edge (Fig. 2.1C). In no case do we observe the arsenate attached by sharing a polyhedral edge with an Fe3+ or Al3+ polyhedron. Arsenate can also bond to itself to form a dimer, or As2O7 group, though this is much less common. The few mineralogical examples include petewilliamsite ((Ni,Co,Cu)30(As2O7)15) and a few related structures. However the rarity of these structures and the fact that they are anhydrous suggest that they would be unstable on crystal surfaces in equilibrium with water (Cotton and Wilkinson, 1966), with hydrolysis favored over polymerization. This survey indicates that the most likely types of sorbed complexes on Fe (hydr)oxide surfaces would be arsenate sharing one or more available Fe3+O6 octahedron apices, but without any polyhedral edge-sharing. This allows for bidentate binuclear complexes, the most common observed by extended X-ray absorption fine structure (EXAFS) or deduced from other measurements (Brown, 1990; Brown and Sturchio, 2002), and the possibility of tridentate trinuclear or monodentate complexes. However, we also note that arsenate may act to bind more distant elements together, forming connections across chains of octahedra, or between disconnected octahedra. Hence the possibility exists for surface complexation of a more

Table 2.1: Minerals with Arsenate-(Fe,Al)O6 Polyhedra Connections. Mineral (composition)

Structure type

AsO4 linkages

Pharmacosiderite (KFe3+(AsO4)3(OH)4.7H2O)

Four Fe octahedra form clusters, each connected by arsenates and water

Tetradentate tetranuclear (bidentate binuclear)

Wilhelmkleinite (ZnFe3+ 2 (AsO4)2(OH)2)

Chains of corner-sharing Fe octahedra linked by arsenate tetrahedra and zinc octahedral. Zinc and Fe octahedral share faces

Durangite (NaAl(AsO4)F)

Cluster unit Four-edge sharing FeO6 octahedra

As–O–Fe3+ –O–Fe3+ –O–Fe3+ –O–Fe3+

Buerger et al. (1967)

Tetradentate hexanuclear (monodentate mononuclear; monodentate binuclear)

Complex chains of Fe octahedral and Zn octahedral forming pseudo sheets

As–O–Fe3+ –O–Fe3+ –O–Fe3+,Zn –O–Fe3+,Zn

Adiwidjaja et al. (2000)

Chains of AlO6 octahedra connected by arsenates

Tetradentate tetranuclear (monodentate mononuclear)

Chains of AlO6 connected by corners

As–O–Al –O–Al –O–Al –O–Al

Kokkoros (1938)

Scorodite (Fe(AsO4).2(H2O))

Isolated octahedra linked by arsenates

Tetradentate tetranuclear (mononuclear monodentate)

None; isolated octahedra


Kitahama et al. (1975)

Angelellite (Fe4(AsO4)2O3)

Chains of edge-sharing octahedra bridged by arsenate

Tetradentate hexanuclear (bidentate trinuclear)

Edge-sharing octahedral chains

As–O–2Fe3+ –O–Fe3+ –O–2Fe3+ –O–Fe3+

Wright et al. (2000)

Petewilliamsite ((Ni,Co,Cu)30(As2O7)15)

Sheets of distorted polyhedra sharing edges tied together by diarsenate groups

Tetradentate heptanuclear (monodentate binuclear; monodentate mononuclear)

Sheets of distorted edgesharing polyhedra

As–O–2Ni –O–2Ni –O–2Ni –O–As

Roberts et al. (2004)

Adamite (Zn2(AsO4)(OH))

Chains of edge sharing Zn octahedra joined by arsenate groups

Tetradentate tetranuclear (bidentate binuclear)

Chains of edge-sharing Zn octahedra

As–O–Zn –O–Zn –O–Zn –O–Zn

Hawthorne (1976)

Arseniosiderite (Ca2Fe3(AsO4)3O2 3H2O)

Sheets comprising edgesharing Fe octahedra connected by arsenate tetrahedra; sheets tied vertically by arsenate

Tetradentate heptanuclear (monodentate binuclear; monodentate mononuclear)

Sheets of edge-sharing Fe octahedra and arsenate groups

As–O–2Fe3+ –O–2Fe3+ –O–2Fe3+ –O–Fe3+

Moore and Araki (1977)

35

References


Topology

Structure type

AsO4 linkages

Cluster unit

Parascorodite (FeAsO4 2H2O)

Isolated Fe octahedra connected by isolated arsenate tetrahedra; only three apices of octahedra bridged; other three have OH2


Single Fe octahedra

Sewardite (Ca2Fe2(AsO4)2(OH)2)

Chains of edge and apex sharing Fe octahedra bridged by two types of arsenate

Tetradentate tetranuclear (monodentate mononuclear; bidentate binuclear)

Edge and corner sharing octahedral chains

Topology As–O–Fe –O–Fe –O–Fe –O–Fe


References Perchiazzi et al. (2004)

Roberts et al. (2002)


Mineral (composition)

36

Table 2.1: (Continued ).


37

Figure 2.1: Arsenate Crystal Structures. (A) Tetrahedral Cluster Motif in the Pharmacosiderite Structure. Central Part is Four Edge-Sharing FeO6 Octahedra Forming a Super-Tetrahedron. Arsenate is then BidentateAttached to All Adjacent Pairs of Free Oxygen Apices. Stoichiometry is Fe4As6O28. (B) Scorodite Structure. Each Slightly Distorted Fe Octahedron is Attached Apically to Four Arsenate Groups that Bridge Nearby but Unconnected Octahedra. The Arsenate Groups have All Apices Connected to Fe Octahedra. (C) Angelellite Structure. Fe3+ Octahedra Share Edges and Form Kinked Octahedral Chains Running along the c Axis. Individual Arsenate Groups Bind these Chains Together. different form if there is chemical restructuring of the Fe3+O6 octahedra on the surface. A discussion of surface crystal chemistry must allow for the loss of symmetry at a surface, and the degree of relaxation of bonding angles and distances. It should also allow for the stabilizing effect of easy hydrogen bonding from surface water groups. The structure of pharmacosiderite is particularly interesting in this regard as the main units of the structure

38


(Fig. 2.1A) constitute clusters with multiple bidentate binuclear arsenateFe3+O6 octahedra attachments which are themselves bound by arsenate to one another with water groups surrounding the framework and stabilizing it. On mineral surfaces the water groups can be considered from two opposing views: More hydrogen bonds are available to satisfy surface oxygen bond valence, but the lability of the water groups creates increased disorder in the surface complex binding. Hence we might infer somewhat unusual topologies (e.g., surface tetrahedral Fe3+, or five-coordinated As5+) could be stabilized and also vary with pH, but many of these geometries would be unstable with respect to changes in surface hydration.

2.3. Silicate Crystal Chemistry as a Monomer and Small Polymer in Structures and on Surfaces 2.3.1. Bond Valence Calculations A thick silicate surface layer may present a hydrophobic interface to structural or surface water. The hydrophobicity arises due to total charge compensation of Si–O–Si linkages, and hence there would be no strong water–silicate interaction except at pH values where S–OH bond formation would be favorable. For lower concentrations of silicate on surfaces bond valence sums suggest that besides two octahedral Fe3+ ions sharing an oxygen corner with silicate (bond valence sum 0.5+0.5+1.0 ¼ 2.0), a tetrahedral Fe3+ polyhedron could share an oxygen with silicate and a strong hydrogen bond (bond valence sum 0.75+1.0+0.25 ¼ 2.0). Indeed, solution Fe K-edge EXAFS studies show the presence of tetrahedral Fe3+ at basic pH values with silicate present (Pokrovski et al., 2003). Octahedral Fe3+ sorbed onto silicate surfaces is also observed to have tetrahedral coordination at basic pH values (Waychunas et al., unpublished). A more complicated surface complex would have a tetrahedral Fe3+ sharing an oxygen corner with an octahedral Fe3+ unit and a silicate group. The bond valence sum is 0.75+0.5+1.0 ¼ 2.25, which could be reduced by lengthening of the Fe3+–O bonds. Such a complex would create tetrahedral bound Fe3+ on a surface, and can promote surface dissolution. 2.3.2. Relevant Solution Studies Past studies indicate that the presence of a monodentate-linkage geometry of SiO4 with aqueous Fe3+ is expected at high dilutions (Weber and Stumm,


39

1965). In addition, adsorption modeling of silicate anions on iron oxides often assumes monodentate binding (Swedlund and Webster, 1999; Zeng, 2003). Using Fe K-edge EXAFS, Doelsch et al. (2002) suggested the possibility that Fe–O–Si linkages could hinder further Fe2+ condensation in bulk solution (at Si/Fe2+ ¼ 1). Further, they proposed that monomeric silicate could be bound with Fe2+O6 octahedra as a monodentate structure (single corner linkage) but did not specify the precise complexation geometry. Pokrovski et al. (2003) studied the effect of aqueous silicate anions on the hydrolysis of Fe3+ in solution and suggested that SiO4-tetrahedra bind as binuclear bidentate ( ¼ double corner) to Fe3+O6-octahedra within the Fe3+ polymer. However, they were not able to determine whether monodentate SiO4 linkages (single corner) to Fe3+O6-octahedra exist, possibly due to the high degree of vibrational and static disorder associated with such monodentate geometry in solution. Such linear linkages may produce EXAFS signals that cancel one another out due to the wide range of Si–O–Fe bond angles and resulting Si–Fe distances. A somewhat analogous effect could also occur with CTR measurements (Waychunas et al., 2005), so that we can see only the most ordered surface silicate units. In that case a quantitative evaluation of the measured surface chemical density and our calculated occupancy would indicate whether a significant portion of such units were disordered. Doelsch et al. (2001, 2003) utilized Fourier transform infrared spectroscopy at pH 3, 5, 7, and 10 with solution Si/Fe molar ratios in the range 0.25–4. They found that for Si/Fe3+r1 Si and Fe atoms do not form separate silicate and FeOOH particles, and the presence of Si–O–Fe bonds hindered growth of FeOOH particles. In contrast, when Si/Fe3+ ¼ 2 and 4 at pH ¼ 3 and Si/Fe3+ ¼ 4 at pH ¼ 5, no Si–O–Fe bonds are detected by FTIR, suggesting separation of silica and Fe (hydr)oxide phases. These results are analogous to behaviors observed in surface sorption experiments with a wide range of sorbates; namely that a slight tendency for solution polymerization can be enhanced on a surface.

2.3.3. Silicate Linkages in Mineral Structures Whereas arsenate is rarely polymerized, solutions with high concentrations of silicate are polymerized over much of the typical circumneutral pH range in solution. Hence silicate sorption may in part include various types of polymeric clusters either directly sorbed from solution, or preferentially formed on a favorable surface. As with arsenate, we first consider the

40


structural motifs for Fe3+O6 and Al3+O6 polyhedral linkage with silicate monomers and polymeric units (Table 2.2).

2.3.3.1. Structures with Silicate Monomer (Nesosilicates) The SiO4 tetrahedra units are isolated and bound to each other only by ionic bonds from interstitial cations. Andradite (Ca3Fe3+ 2 (SiO4)3), yoderite andalusite (Mg2(Al,Fe3+)6Si4O18(OH)2), almandine (Fe2+ 3 Al2(SiO4)3), (Al2SiO5), and chloritoid ((Fe2+,Mg)2Al4Si2O10(OH)4) are among the most common nesosilicates, and have (Fe,Al)O6 octahedra sharing only corners with silicate groups.

2.3.3.2. Structures with Silicate Dimers (Sorosilicates) The Si2O7 unit is known in many minerals, sometimes occurring with other types of polymerized silicate units, or with monomeric silicate. The bestknown minerals of this type are probably the closely related epidote (Ca2(Fe,Al)Al2(SiO4)(Si2O7)O(OH)) and zoisite (Ca2Al3(SiO4)(Si2O7)O(OH)) structures where Si2O7 groups bridge corner-sharing (Fe,Al)O6 polyhedra. Other minerals having Si2O7 units include lawsonite (CaAl2Si2O7 (OH)2 (H2O)), and melilite ((Ca,Na)2AlSi2O7). In most cases the individual silicate groups of the Si2O7 unit ‘‘point’’ in opposite directions, or are canted with respect to one another (Fig. 2.2A).

2.3.3.3. Structures with Silicate Chains In pyroxenes (e.g., acmite NaFe3+(Si2O6); Fig. 2.2B) the silicate unit is a single chain kinked in one of three basic variants o, s, or e (Fig. 2.2: D1, D2, and D3, respectively). The o or s configurations allow for connection with metal cation polyhedra with smaller dimensions (15%) than the e configuration. In all cases the silicate units point all in one direction. Somewhat similar geometry is observed in amphiboles having double chains.

2.3.3.4. Structures with Silicate Sheets The most relevant chemistry is observed with annite (KFe2+ 3 AlSi3O10 3+ (OH,F)2), nontronite (Na0.3Fe2 Si3AlO10(OH)2 4(H2O)), and hisingerite

Table 2.2: Minerals with Silicate-(Fe,Al)O6 Polyhedra Connections. Mineral (composition)

Structure type

SiO4 linkages

Cluster unit

Topology

References

Andradite (Ca3Fe3+ 2 (SiO4)3)

Tetrahedra share one edge with a calcium decahedron and four corners are connected with four Fe octahedra


Only corner sharing with Fe octahedra

Si–O–Fe3+ –O–Fe3+ –O–Fe3+ –O–Fe3+

Yoderite (Mg2(Al,Fe3+)6Si4O18(OH)2)

Four corners are connected with four Al or Fe octahedra


Only corner sharing with Al or Fe octahedra

Si–O–Al,Fe3+ –O–Al,Fe3+ –O–Al,Fe3+ –O–Al,Fe3+

Almandine (Fe2+ 3 Al2(SiO4)3)

Isolated silicate tetrahedra are connected with Al octahedra via corner sharing

Identical to Andradite

Only corner sharing with Al octahedra

Si–O–Al3+ –O–Al3+ –O–Al3+ –O–Al3+

Hazen and Finger (1989)

Andalusite (Al2SiO5)

Chains of Al octahedra are coupled to isolated silicate tetrahedra

Tetradentate hexanuclear (monodentate mononuclear and monodentate binuclear)

Edge-sharing AlO6 octahedral chains

Si–O–2Al3+ –O–2Al3+ –O–Al3+ –O–Al3+

Winter and Ghose (1979)

Chloritoid (Fe2+,Mg)2Al4Si2O10(OH)4

Al, Fe2+, Fe3+, and Mg (M) octahedra in a bilayer structure. Isolated silicate tetrahedra are connected to both layers

Tetranuclear; mononuclear (monodentate trinuclear and monodentate binuclear)

Sheets of octahedral sites resembling a clay structure

Si–O–3M –O–2M –O–2M –O–2M

Ivaldi et al. (1988)

Epidote (Ca2(Fe,Al)Al2(SiO4)(Si2O7)O(OH))

Edge-sharing Fe and Al octahedra are connected by isolated silicate tetrahedral and ditetrahedral units

Tetradentate hexanuclear; tetradentate tetranuclear (monodentate mononuclear, binuclear; bidentate binuclear)

Chains of edge-sharing octahedral cations

Si–O–Si –O–Al –O–2(Al,Fe) Si–O–Al –O–Al –OH –OH

Dollase (1971)

Hazen and Finger (1989)

Higgins et al. (1982)

Anion Sorption Topology on Hematite 41

Structure type

SiO4 linkages

Cluster unit

Topology

References

Lawsonite (CaAl2Si2O7(OH)2 (H2O))

Si2O7 ditetrahedral groups bridge chains of edgesharing AlO6 octahedra. All octahedral apices connected to silica units

Tridentate tetranuclear (monodentate mononuclear; monodentate binuclear)

Chains of edge-sharing AlO6 octahedra

Si–O–Si –O–Al –O–Al –O–2Al

Acmite (or Aegirine) (NaFe3+(Si2O6))

Silicate tetrahedral link chains of edge-sharing cation octahedra. Si tetrahedra linked to form chains along the same direction as cation chains

Tetradentate pentanuclear (monodentate mononuclear; monodentate binuclear)

Chains of edge-sharing cations

Si–O–Fe3+ –O–2Fe3+ –O–Si –O–Si

Prewitt and Burnham (1966)

Muscovite (KAl2Si3AlO10(OH)2)

Layer structure. A silicate sheet is connected with an Al octahedral layer in the sequence Si-Al-SiK-Si-Al-Si-K. Al layer is 2/3 occupied. Dioctahedral mica

Tetradentate pentanuclear

Sheets of edge-sharing octahedra

Si–O–Si –O–Si –O–Si –O–2Al

Redhammer et al. (2000)

Nontronite (Na0.3Fe3+ 2 Si3AlO10(OH)2 4(H2O))

Layer structure. A silicate sheet is connected with an Fe octahedral layer



Si–O–Si –O–Si –O–Si –O–2Fe3+

Manceau et al. (1998)

Hisingerite (Fe3+ 2 Si2O5(OH)4 2(H2O))

Layer structure similar to dioctahedral mica.



Si–O–Si –O–Si –O–Si –O–2Fe3+

Ahmad (1984); Manceau et al. (1995)

Biotite (K(Fe,Mg)3Si3AlO10(OH)2)

Layer structure. A silicate sheet is connected with an (Fe, Mg) octahedral layer with sequence Si-Al-Si-K-Si-Al-Si-K. Octahedral layer is nearly fully occupied (trioctahedral mica)



Si–O–Si –O–Si –O–Si –O–2(Fe,Mg)

Rumanova and Skipetrova (1959)

Birle and Tettenhorst (1968)


Mineral (composition)

42



A

B

D1

43

C

D2

D3

Figure 2.2: Silicate Structures. (A) Lawsonite Structure Showing Discrete Si2O7 Units Bridging Edge-Sharing Octahedral (Al,Fe)O6 Chains. Individual Silicate Tetrahedral has Apices Bonded to Three Fe Octahedra and One Silicate Tetrahedron. (B) Pyroxene Structure. Kinked Chains of Edge-Sharing (Fe, Al, Mg)O6 Octahedra are Joined by Chains of Point-Sharing Silicate Tetrahedra. Individual Tetrahedra are Bonded to One Octahedral Apex, One Oxygen Common to a Shared Edge of Two Octahedral, and Two Silicate Tetrahedra. All Silicate Units ‘‘Point’’ in the Same Direction. (C) Nontronite Structure. Alternating Planes of FeO6 Octahedra Sharing Edges, and Silicate Tetrahedral Sharing Three Apices with One Another. (D) Types of Silicate Chain–Octahedral Chain Linkages in Silicates. D1 and D2 are Non-Superimposable, and are Found in Cases with Smaller Cation Octahedra (Al–O, Mg–O Bonds), While D3 is Found in Cases with Larger Octahedra (Fe–O, Ca–O, Na–O Bonds). (Fe3+ 2 Si2O5(OH)4 2(H2O)), but the topology is similar to that in muscovite (KAl2(Si3Al)O10(OH)2), phlogopite ((K,Na,Ba)Mg3(Si3Al)O10(F,OH)2), and more common layer silicates. In annite there are ‘‘e’’ type silicate linkages forming rings within the layers. Their ‘‘points’’ bridge (Al,Fe)O6 polyhedra edges. In nontronite the silicate layers also are formed into rings but adjacent points now bridge metal polyhedra edges or adjacent polyhedra apices (e.g., nontronite, Fig. 2.2C).

44


We note that no edge-sharing of silicate tetrahedra and Fe,Al octahedra are observed in any silicate structures. Summarizing all silicate observations then: monomeric silicate units do not exhibit any edge-sharing geometry, but may share many apices with metal oxide polyhedra; polymerized silicate similarly does not show such edge–edge binding with metal oxide polyhedra; chains and sheets of silicate tend to bind to clusters or layers of metal oxide polyhedra with all of their bonded oxygen apices in the same direction, though the exact topology of the chains and sheets depend on the spacing of the metal oxide polyhedra, i.e., their size relative to the silicate groups; both monodentate and bidentate attachment to chains of cation polyhedra is observed, but layer structures favor monodentate attachment as most Si tetrahedral apices are joined to other silicate groups within the layers.

2.3.3.5. Effect of Water In many silicates containing alkali substituents (e.g., Na), a silicate polymeric layer or ring structure surrounds the large cation site (e.g., in cyclosilicates and tectosilicates). This is the architecture seen in zeolite structures where the cage has oxygens fully charge compensated, and low valence ions can be captured in the tunnels. Water binding via hydrogen bonding may have a similar influence. Hence, it is possible that silicate polymers could form on surfaces with a topology that incorporates surface water and large cations. In other words, a surficial silicate polymer could be partially stabilized by binding to water hydrogen bonds and large alkali ions in solution without leading to further ‘‘large scale’’ polymerization, i.e., thicker layer coverage, on the surface.

2.4. Structure of the Hematite Surface The hematite surfaces we consider here have been investigated in detail over the last few years by our colleagues and us (Trainor et al., 2004; Waychunas et al., 2005; Lo et al., 2007; Tanwar et al., 2007) and have been well characterized by CTR analysis in equilibrium with circumneutral pH water and verified by density functional theory (DFT) calculations. The detailed description of CTR techniques and principles can be found elsewhere (Robinson and Tweet, 1992; Eng et al., 2000; Fenter, 2002; Trainor et al., 2004; Jun et al., 2007). The C plane, (0 0 0 1) surface of hematite is slightly relaxed in comparison to a bulk termination, but the most important feature


45

C

A

bs

c

B as

D cs

as

Figure 2.3: Hematite Surface Models. (A) Domain Model where Topmost FeO6 Octahedra Occur in Dense Regions Equivalent to 50% Occupation of the Fe Layer (Fe Octahedra Sharing Faces with Lower Fe Octahedra are Missing). (B) Representative Picture of Random Occupancy of the Fe Layer at 15% Coverage Indicated by CTR Fitting. (C) CTR Refined Structure of the R Plane, Top View along the (0 0 1) Direction. (D) R Plane Structure Side View along (0 1 0) Direction View.

is the absence of any surface Fe3+O6 polyhedra that share faces with lower (i.e., interior) Fe3+O6 polyhedra. Removal of these types of Fe3+ units on the surface reduce the possible population of the uppermost Fe layer to 50% or less, and the refinements suggest that the actual population of the full layer is 15%. The nature of these Fe3+ polyhedra can be either dispersed or clustered into domains (Fig. 2.3). These two cases present different types of sorption sites to tetrahedral species such as arsenate and silicate: isolated Fe3+ units offer only edges or single apices for binding, while clustered Fe3+ units offer the possibility of additional bidentate and tridentate sites. Scanning tunneling microscopy measurements (Eggleston et al., 2003), as well as our DFT calculations, are consistent with a domain structure, and CTR analysis using a domain model yields slightly improved statistical fits (Trainor et al., 2004). On the R plane, (1102) surface, the structure is also relaxed from the bulk, and the refined termination shows 100% vacancy for the uppermost Fe3+ layer (Tanwar et al., 2007) (Fig. 2.3C and D). The ab initio thermodynamic

46


calculations of Lo et al. (2007) suggest that this vacancy structure is the lowest energy surface stoichiometry at room temperature, and is hypothesized to be stabilized by reducing the number of cation–cation interactions within the surface layer (e.g., removing face-sharing polyhedral interactions). Both the C and R planes have stepped terminations representing symmetry equivalent planes. Hence it is important to consider the topology of the step edges in these surface structures. 2.4.1. Sorption Sites On the C plane surface (Fig. 2.4) we can identify many types of potential sorption sites. It is important to consider that this is the case of a nearly perfect single crystal surface, so that on irregular surfaces many more types of sorption sites might be available. For our purposes here we assume the ordered domain surface model for the C surface, so that the available sites for sorption are associated with a tight arrangement of surface Fe3+O6 octahedra. In this case we can readily identify five types of surface sites: all corner-sharing tridentate, edge-sharing bidentate, edge and corner sharing tridentate, binuclear bidentate, and monodentate. If we include binding at steps, we also have another geometry for bidentate binuclear, a different orientation of edge-sharing mononuclear bidentate, and several different types of mononuclear complexation sites. Though this set of possibilities seems quite large and complex, it is amenable to analysis as each type of complex has specific topological signatures that can be resolved by polarized spectroscopy. For the R plane, the symmetry is reduced, but because of the nature of the CTR-determined termination there are only a few types of attachment sites. Besides monodentate attachment at two nonequivalent octahedral apices, there are bidentate binuclear sites. No bidentate mononuclear sites appear except at step edges, and tridentate complexation does not appear possible.

2.5. Results 2.5.1. Arsenate Sorption on Hematite GIEXAFS Results Arsenate sorption on hematite C and R planes was studied by grazing incidence extended X-ray absorption fine structure (GIEXAFS) measurements on beamline 11-2 at SSRL. Crystals of 25 mm were used allowing a large fraction (20–30%) of the full incident beam to be placed on the sample


A

47

C

B

D

[0001]

[01-10]

Figure 2.4: Potential Sorption Sites on the Hematite C Plane Surface Assuming the Dense Domain Surface Topology. (A) C Plane Map View. (B) C Plane Side View. Types of Sites Shown Include: Tridentate Trinuclear, Tridentate Binuclear, Monodentate Mononuclear, Bidentate Mononuclear, and Bidentate Binuclear. (C) R Plane Map View. (D) R Plane Side View. Sites Shown are Bidentate Binuclear and Bidentate Mononuclear. R Plane Directions Relative to the Trigonal Hematite Cell.

surface at the critical angle (0.21). Other experimental conditions and chemical considerations are described in Waychunas et al. (2005). For each plane, measurements were made with the synchrotron radiation electric vector both along the surface (in-plane), and normal to the surface. For the C plane the azimuthal orientation of the X-ray electric vector inplane should not be important due to the three-fold symmetry of the surface, although this can be compromised if there is a significant miscut (misorientation) so that surface steps are aligned in a particular direction. In the case of the R plane, we oriented the crystals so that the electric vector was along the trace of the mirror plane containing the c axis. The extracted GIEXAFS and pair correlation functions (PCF) resulting from this analysis are shown in Figs. 2.5 and 2.6. Both surfaces show As-Fe

48


6

k3 weighted EXAFS

4 2 0 -2 -4 -6 -8

4

8

6

10

12

k (Å-1)

3.5 3 Transform Magnitude

(0001) surface Dots--surface normal e-vector Solid--surface plane e-vector

2.5 2 1.5 1 0.5

bidentate/tridentate arsenate

0 0

1

2

3

4

5

showing underlying Fe

radial distance (Å)

Figure 2.5: Arsenate GIEXAFS Results and Structure Model for C Plane. Top: Fitted EXAFS for Second Shell As-Fe Region Normal e-Vector. Left: EXAFS Structure Functions for Two Orthogonal Polarizations of e-Vector. Right: Cartoon Showing Arrangement of Sorbed Arsenate and Second/ Fourth Neighbor Fe Atoms. distances (peaks in the PCF) characteristic of two types of arsenate sites: bidentate binuclear (i.e., corner or apical oxygen sharing) and bidentate mononuclear (i.e., edge polyhedral sharing) with the ratios between these changing with electric vector orientation. For the C plane there is less variation between the two orientations, with more edge-sharing sorbates observed in the normal electric vector measurement. However with both data sets the bidentate binuclear sorption sites are the majority. In the R plane analysis, there is almost a complete change in the observation from the

3

1.5

2

1

k3 weighted EXAFS

k3 weighted EXAFS


1 0 -1 -2 -3

49

0.5 0 -0.5 -1 -1.5

4

6

8

10 k

12

4

6

8 -1 k (Å )

(Å-1)

10

12

3.5

Transform Magnitude

3

(1102) surface

2.5 2 1.5 1 0.5 0

0

1

2 3 Radial Distance (Å)

4

5

Figure 2.6: Arsenate GIEXAFS Results for R Plane. Top Left: EXAFS for Second Shell As-Fe Shows Beat Pattern Due to Two Contributions with e-Vector along the Sample Surface. Top Right: EXAFS for Second Shell As-Fe Shows Mainly Only One Contribution for e-Vector Normal to Sample Surface. Bottom: EXAFS Structure Function Showing Second Shell Contributions with Different e-Vector Polarizations. normal electric vector data showing mainly binuclear bidentate sorption, similar to the sorption geometry determined by Catalano et al. (2007), the inplane data showing mainly edge-sharing sorption. We explain these results by using a model for the R plane that includes step edges (Fig. 2.6). Attachment to the sides of these steps must be mainly via the edge-sharing geometry, while attachment to the flat terraces must be via the bidentate binuclear geometry. If this interpretation is correct, it is possible to see a different ratio of edge- to apical-attached arsenate with differently prepared surfaces having altered step/terrace statistics. In the case of the C plane, due to the abundance of separate Fe3+O6 units on the surface, we expect a different type of bidentate binuclear complex, with the arsenate group bridging between two separate ferric polyhedra, rather than attached to two

50


edge-sharing polyhedra (Fig. 2.5). However many types of sorption geometries are possible on this surface (assuming the CTR-derived termination) and many of these would yield very similar distances via EXAFS analysis. 2.5.2. First Si K-Edge GIEXAFS Measurements: Silicate Sorption on the R-Plane (1102) Surface of Hematite Recent work at NSLS beamline X15B has enabled us to collect the first soft X-ray Si K-edge GIEXAFS data from any mineral surface. Silicate sorption on hematite R planes was studied using 25–35 mm crystals and 11 beam incidence angle. Due to interference with a small amount of phosphate and monochromator limitations, we were not able to collect data past the P K-edge, leading to a maximum k value of 9 A˚1. However the spectra are representative of possible data quality to higher k values, and future spectra will be collected to at least a k value of 12 A˚1. Model spectra for comparison were collected for nontronite and acmite ( ¼ aegirine) powders, with the nontronite collected after monochromator repair to permit the greater k range (k ¼ 13 A˚1). In the nontronite structure (Fig. 2.2C) we note that each silicate tetrahedron is connected to one Fe3+O6 polyhedron and to three other silicate tetrahedra, and thus EXAFS analysis should show a second peak with multiple components Si–Fe and Si–Si. There are also Si–O distances due to Si correlation with other oxygens on the Fe3+O6 and silicate polyhedra, although these will make a weaker contribution to the second shell (i.e., next nearest neighbor) PCF peak. Comparison of the EXAFS is shown in Fig. 2.7, and nontronite data show acceptable agreement with the FEFF-calculated pattern derived using the published crystal structure (Manceau et al., 1998). In the case of acmite this study is the first Si K-edge EXAFS reported. Comparison with the ideal structure using a FEFF calculation is also good, although we note that a full analysis of these data, including use of a more accurate sample chemistry in the FEFF calculation, is still in progress. PCFs calculated from the GIEXAFS spectra for silicate on hematite are consistent with silicate contributions in the second shell, but do not show features that occur in the simulated PCFs for bidentate and monodentate silicate surface attachment to the ordered R-plane surface (Fig. 2.8). We note that the short k range makes it difficult to sort out additional Si–Fe contributions in the PCF, so that improved longer k-range data are required for confirmation. However, the GIEXAFS for all samples and both horizontal and vertical polarization show similar features, suggesting that the main contribution is essentially similar and isotropic. A further complication is the


51

A 4

nontronite FEFF calculation nontronite measurement

Chi x k

2

2 0 -2 -4 2

4

6 k (inverse anstroms)

B

8

10

acmite measurement nontronite measurement 1 mM silicate on r-plane hematite (H)

4

Chi x k

2

2 0 -2 -4 2

3

4

5 6 k (inverse anstroms)

7

8

Figure 2.7: Model and Calculated EXAFS and GIEXAFS Silicate k-Edge Spectra. (A) Nontronite Powder Spectrum Compared to FEFF Calculation. (B) Comparison of Acmite, Nontronite, and 1 mM Silicate Sorption Sample. EXAFS is Dominated by First Shell Si–O, which is the Same in All Samples. presence of silicate even in carefully cleaned samples, presumably due to the colloidal silica used in high quality polishing procedures. We have characterized this material prior to CTR data collection, and found it to be present on surface defects and steps, but not on flat terraces – the by-far most dominant part of the polished sample surface. Sorption of monomeric silicate on the hematite surfaces resulted in a doubling of the EXAFS signal in our samples. Hence we would expect to detect changes due to sorption and the development of attendant Fe–Si correlations, but these were not observed. Hence it is likely that the silicate treatment adds mainly to the remnant colloidal silica present, as well as complexing to terrace sites, but the colloidal silica type contributes most of the signal. When compared to a quartz cluster (central silicate surrounded by four silicate tetrahedra and full

52


Figure 2.8: Comparison of EXAFS Structure Functions (PCFs) from FEFF Calculations and Silicate-Sorbed Hematite. Thin and Thick Curves for Silicate-Sorption Samples are for 1.0 and 0.5 mM Silicate Monomer Sorption Solutions, Respectively. Bidentate and Monodentate Models are for Small Clusters of R-Plane Hematite Surface (Eight Fe atoms) with Attached Silicate Group. Quartz FEFF Cluster is for Five Silicate Tetrahedra. oxygen coordinations), the sample GIEXAFS PCFs have a smaller second shell Si–Si amplitude, indicating fewer Si–O–Si linkages. 2.5.3. Arsenate and Silicate CTR Observations ¯ For the (1102) surface of hematite silicate sorption creates significant changes in the 0 0 Ls, 0 2 Ls, 1 1 Ls, and 1 0 Ls rods (Fig. 2.9). The main changes in the CTR profiles are due to increased structure factor amplitude compared to those from clean hematite, e.g., in the 0 0 Ls rod around 1.5 and 3.0 reciprocal lattice units (rlu), and in the 1 0 Ls rod notably distinct new features occurred at 1.5 and 1.5 rlu. Such changes can be attributed to a


2 8 6 4

F(HsKsLs)

F(HsKsLs)

8 6 4

(0 0 Ls)

1000

2

100

Clean r-cut hematite Si on r-cut hematite As on r-cut hematite

8 6 4

(0 2 Ls)

2

1000 8 6 4 2

100

2 1

2

3

4

5

-4

-2

Ls(r.l.u.)

1048

(1 0 Ls)

2

0

2

4

2

4

Ls(r.l.u.)

4

(1 1 Ls)

6 4

1000

8 6 4

F(HsKsLs)

F(HsKsLs)

53

2

2

1038 6 4

100

8 6 4

2 2

10 -2

0

Ls(r.l.u.)

2

-4

-2

0

Ls(r.l.u.)

Figure 2.9: Arsenate and Silicate CTR Observations for Hematite R Plane.

number of factors, including changes in the roughness of the surface, sorption of silicate, precipitation of a ferric silicate surface phase which has chemical binding to the surface atoms, and changes in the surface water structure and thickness, with each factor making distinctive contributions to the scattering data. As the overall rod intensity away from the Bragg peaks increases with sorption relative to the wet hematite surface, these changes are not due to a change in surface roughness. For roughened surfaces, intensity at the positions between Bragg peaks is reduced. The features are thus due to sorption or surface precipitate formation with well-defined geometry, and/or changes in the ordered termination of the hematite surface, including order water adlayers. Disordered surface sorption, or disorganized water would not give rise to any features in the CTRs. Arsenate sorption was studied with two treatments of the sample. The first treatment producing well defined changes in the 0 0 Ls, 0 2 Ls, 1 0 Ls, and 1 2 Ls rods, which were increased after a second sorption exposure. Some overall loss of rod intensity after the second treatment was noted, suggesting disordering of the surface due to roughening and precipitation reactions.

54


Analysis of the CTR data proceeds by starting with the wet hematite surface models previously defined, and using the arsenate GIEXAFS data as a guideline. The arsenate CTR analysis is complex, owing to the several types of complexes observed with the GIEXAFS analysis, and is still in progress. A similar procedure is followed for the case of silicate sorption on the R-plane surface. Significant intensity changes on sorption are evident in the (0 0 Ls), (0 2 Ls), (1 0 Ls), and (1 1 Ls) rods when compared to the clean a-Fe2O3 surface scattering. Therefore, these four rods were investigated in the analysis. Several complexation or substitutional geometries were considered as follows: (1) silicate substitution to the missing Fe position on the R-plane hematite surface; (2) monodentate binding to FeOH– apical sites; (3) bidentate mononuclear binding to adjacent FeOH– or 2Fe–O sites (edge sharing complex); (4) bidentate binuclear binding to two FeOH– sites; and (5) tridentate mononuclear binding to three FeOH– or 2Fe–O sites. These possible arrangements are shown in Figs. 2.4C and D. The analysis assumed that the silicate anions are present only as monomers in solutions before the silicates are reacted with hematite surfaces. This is a reasonable assumption because the silicate sorption solutions were prepared at low concentrations (0.5 and 1 mM) where monomeric silicates are dominant species during our experimental time frame (Qi et al., 1993; Powell et al., 1999). Results show that case number 1 is not physically plausible because the fit results in unrealistic occupancy and very large mismatches in all four rods. Cases #3 and 5 can be excluded for the flat perfect surface due to steric hindrances, as face-sharing or other unlikely topology would be necessary to place the silicate anions in such positions. The exception would be if we allowed bidentate mononuclear binding to step edges on a defective surface (see below). For case #4, the measured structure factors for the (0 0 Ls) rod show large discrepancies although the occupancies and the surface structures are chemically plausible (Fig. 2.10). The best fit for the four rods resulted from case #2, monodentate complexation of silicate ¯ tetrahedra on the ð1102Þ surface of hematite (Fig. 2.11). We note that the occupancy of the silicates and oxygens from silicate tetrahedra in the monodentate fit are 0.55 and 0.58, respectively, and thus a further consideration is whether the silicate is clustered or isolated on the surface. If isolated, we would need to explain why the other free apices of the silicate tetrahedra do not rotate into the surface plane and bind to nearby free Fe polyhedra apices, thus forming bidentate complexes. One answer is that the silicate has polymerized, forming chain-like units on the surface, by attachment of adjacent tetrahedra (Figs. 2.2B and C). Indeed, with more than 50% silicate occupancy, even randomly placed silicate units would have a high proportion of adjacency. Using the CTR fitted occupation, the


A

55

C (02( L s)

(00 L s)

FHKL

FHKL

datam fitting

bs 0

2

Ls

6 -5

4

-3

Ls

1

3

5

3

5

(11 L s)

(10 L s) as FHKL

cs

as

FHKL

B

-1

-4

-2

0 Ls

2

4

-5

-3

-1

1 Ls

Figure 2.10: Silicate R Plane CTR Fit. Bidentate Binuclear Sorption Geometry. Fit is Poor on (0 0 L) Rod Near 1.0 rlu and in Several Places for the Other Rods. Cartoon Shows the Two Types of Silicate Position Possible on the Surface. probability of a silicate having at least one silicate neighbor is 80%, and of having two neighbors is 30%. As we continue to refine these results, other possible models may produce better fits, especially via adjustments to the overlying water or oxygen layers. The structural motifs for Fe3+O6 and Al3+O6 polyhedral linkages with silicate monomers and polymeric units in natural silicate minerals appear consistent with the CTR results as only corner sharing rather than edge-sharing geometry is seen, and a related geometry is found in the chain and layer silicates. However, these silicates all have polymerized silicate chains or layers, reinforcing the idea that the silicate on the hematite surface must indeed be polymerized. To our knowledge, the CTR and GIEXAFS work presented here are the first detailed molecular analysis of silicate adsorption geometry on iron oxide surfaces, and perhaps on any mineral surface. As noted earlier, the CTR data from silicate and arsenate adsorption to the R-plane surface of hematite are significantly different (Fig. 2.9) despite the fact that both anions are tetrahedral structures in solutions. Hence we may infer very significant differences in binding, as indeed are suggested by the arsenate GIEXAFS results and the silicate CTR and GIEXAFS results. The CTR fitting results allow testing of

56


A

C (02 ( L s)

(00(L s )

FHKL

FHKL

data fitting

bs 0

2

Ls

4

6

-5

-3

-1

Ls

1

3

5

3

5

(11 L s )

(10 L s ) as FHKL

FHKL

B cs -4

-2

0 Ls

2

4

-5

-3

-1

1 Ls

as

Figure 2.11: Silicate R Plane CTR Fit. Monodentate Mononuclear Sorption Geometry. Cartoon Shows Monodentate Tetrahedral Connected Together to Form Chains Analogous to the Structure in Pyroxenes. the different possible adsorption geometries and are powerfully complementary to the GIEXAFS analysis. As we discuss below, the detailed comparison of results from these methods allow us to make estimations on the nature of disorder of the surface sorbate, and make additional inferences on the sorption–precipitation process. This work constitutes the first evidence that silicate is sorbed to the surface at well-defined sites relative to the surface crystal structure, rather than precipitating only as an (incoherent) amorphous surface layer. Further, the well-ordered chemical binding geometry of silicate to the hematite surface differs from other possible models of surface passivation, and thus will help guide our understanding of passivation reactions for both natural Fe oxide and other natural surfaces.

2.6. Discussion 2.6.1. Agreement of GIEXAFS and CTR Results and Complementarity of Approaches The key difference between CTR and surface EXAFS measurements is the difference in sensitivity connected with the degree and nature of disorder.


57

Assuming that a crystalline surface is suitably smooth, i.e., roughness on the order of the X-ray wavelength, then CTR measurements will quantitatively reproduce the atomic structure of the surface subject to a Debye-Waller type disorder factor (Robinson, 1986). This means that any atom with a high vibrational amplitude will make less of a contribution to the scattering. Analogously, any atom with a large diversity of positions separate from vibrational disorder (so-called ‘‘static disorder’’) will also contribute much less scattering intensity to the CTR. This contrasts with EXAFS analysis, where the main sensitivity to disorder occurs with vibrations or variations in the absorber–backscatterer interatomic distance. Hence if one has arsenate attached to the edge of an Fe3+O6 unit and the distance is relatively fixed by this binding topology, it matters little what the overall orientation of the interatomic vector is (except that due to polarization effects). Thus arsenate disordered on a mineral surface but bonded locally in the same way may contribute differently to the CTR and surface EXAFS. Contrary to being a problem, this in effect makes the two methods complementary in a very significant way. By making a set of surface EXAFS measurements at different angles to the surface normal, the resulting analyses have sufficient information content to resolve several types of surface topologies as we have shown (i.e., much more than in a powder experiment with all directions averaged together). This reveals the total types of arsenate (or silicate) on the surface. The CTR measurement then reveals that part of the arsenate (or silicate) that is ordered into the same types of sites on the surface. By analyzing and contrasting both sets of data we can then determine, in principle, the populations that are likely to have different bonding arrangements and energies. Why should arsenate groups possess site disorder within a topological attachment type? This should not happen inside of a structure, but at the surface the arsenate group is also connected to water molecules and is subject to varying degrees of protonation/hydrolysis. Each local termination of the surface, for example at an edge step, can lead to a slightly different positioning of the average waters about the arsenate group, and thus a slightly different positioning of the arsenate. This is seen if one visualizes the large number of possible edge step terminations that are possible. As we know that water is often locally ordered to two or three interatomic distances above a surface (Ostroverkhov et al., 2004; Jun et al., 2007), it is easy to suppose that variations in lateral step structure can affect the water structure locally, and thus attending sorbates. This argument requires simulation calculations to test, and is beyond the scope of the present paper. However it is well to consider that mineral–water interfaces are likely to be much more complex than one might expect from basic double layer theory and surface complexation models.

58


2.6.2. Determining Clustering or Polymerization of Surface Silicate Given the results presented herein, the polymerization of surface sorbed silicate monomers on hematite R-planes appears very likely. We have evidence from the CTR analysis that a significant fraction of the surface reactive sites are bound with silicate in a coherent fashion, and that the proximity of these sites is likely to produce surface polymers. What we have not yet been able to determine is if the increase in colloidal or poorly ordered silica observed with GIEXAFS is due to growth of a layer of this material on the already bound surface complexes, or if it is primarily associated with remnant silica remaining after polishing. A most important case is that where a large quantity of silicate may interact with a reactive surface, as may happen in natural systems, with the silicate having a wide range of polymerization. In this case the existing geometry of the polymers may restrict the degree of coherence that can develop between the polymer and the surface. In contrast, if polymerization develops along with sorption, the resultant polymer may be entirely coherent with the surface structure. These two cases are relevant, as they will confer quite different degrees of passivation to the surface, as well as be subject to much different later geochemical changes. For example, consider the case of large silicate polymers in solution that sorb incoherently on an Fe oxide surface. Such sorption isolates the surface from the reactive solution, minimizing further reactions, and produces an unstable silicate–oxide interface region that will reconfigure with time. This case may well be the most relevant to natural systems, and its overall reaction pathways are not yet understood. In contrast, sorption of silicate monomers (and small polymeric units) allows establishment of coherent bonding with the surface before continued sorption and polymerization leads to amorphous structure development, thus allowing a well-defined interfacial structure to form. Such a process may result in the initial formation of silicate–metal oxide phases at the sub nanometer scale, a key aspect of nucleation processes, and in principle traceable by the methods detailed in this work.

2.6.3. Silicate and Low Z GIEXAFS Analysis To obtain complementary GIEXAFS data for silicate or other low atomic number sorbate species, soft X-ray spectroscopy is necessary, and this creates a large set of experimental difficulties. The chief problem is that the


59

experiment must be done in vacuum or in a dry He atmosphere, which would remove most of the surface water. However, strongly bonded water would not be removed, as the samples would not have to be subjected to heating in the vacuum system. This is an important consideration as the effects due to sample drying (polymerization of both silicate and Fe hydroxide units) need to be minimized. A second requirement for such experiments is a well collimated and brilliant soft X-ray source, such as is available on SSRL beamline 11-2 for the hard X-ray arsenate GIEXAFS work. Few beamlines in the world have this capability: many are bending magnet lines with low flux, especially in the necessary energy range (1500–2500 eV); others do not have sufficiently brilliant sources (so-called small emittances); and still others do not have appropriate monochromators or focusing mirrors to work in this energy range with necessary collimation (omilliradian). Our zeal for soft X-ray GIEXAFS is fueled by several advantages (Waychunas, 2002). Smaller sample sizes can be used for measurements because of the larger critical angles for total external X-ray reflection mode EXAFS compared to heavier elements. For example, on hematite surfaces at 1840 eV, the critical angle would be 1.341 while for 11,650 eV (As edge) it is 0.21. This means that samples can be about seven times shorter in the beam direction to intercept the full beam spread and thus maximize signal. This is a special advantage for natural samples where available crystal sizes are restricted (e.g., goethite crystals larger than 1–2 mm do not occur). Another advantage is reduced sensitivity to surface roughness when using a larger incidence angle, though polishing of samples must still be much better than so-called ‘‘optical’’ quality. Finally, there is the advantage that Bragg reflections that can overwhelm a GIEXAFS signal at higher energies will not be possible for lower soft X-ray energies. For example, at 1500 eV the d-spacing of a crystal that would diffract at 901 into a fluorescence detector is 6 A˚. Hence for most materials Bragg diffraction would occur at larger angles or not at all.

2.7. Prospects for Further Studies A number of highly significant sorbates and their interactions with common substrates can be approached by the methods we have described. Chief among these is phosphate, an important agricultural nutrient, as well as a principal agent involved in human metabolic processes. Phosphate is ubiquitous in natural systems, and hence is a competitor with other sorbants during complexation reactions. Other accessible species include Na, Mg, Al,

60


S, Cl, and K, with sulfate surface chemistry being of considerable geochemical and biogeochemical importance. The key aspect of CTR measurements for these species is the ability to sense only coherent ordered sorption geometry, and hence determine interface binding geometry even if significant additional incoherent sorption/precipitation occurs. Comparison with GIEXAFS analysis then allows the relative fractions of coherent and incoherent precipitate to be determined, and differences in their local coordination geometry, thus affording an improved description of the interface reaction process. Complementary tools that aid in the analysis of interfacial reactions include atomic force microscopy imaging, which can be used with minimal disruption of the interface, and electron imaging methods such as high resolution-transmission electron microscopy and scanning transmission electron microscopy, although the latter group can disrupt interfacial structure due to dehydration reaction in vacuum. Another important aspect of sorbate interface chemistry is the relationship to crystal growth processes. In essence, the sorption processes studied at low surface density are closely akin to nucleation. We are familiar with the classical depiction of nucleation phenomena due to the tradeoffs of surface and internal energy of a growing new embryo, but studies of the type that we describe here allow dissection of this process on a molecular basis. Fundamental questions densely populate this process: the size of critical nuclei? The form and identity of nuclei? The density of nuclei prior to the successful germination of a new interfacial phase? The topology and growth processes of interfacial layers? Entrance into the nucleation arena presupposes that the time regime must also be considered. What is the relative speed of sorption with respect to polymerization? How does the nucleation time for a new phase relate to the growth and first appearance of that phase? What are the kinetics relative to natural phenomena that we want to control? Fortunately, synchrotron radiation-based techniques are well suited to analyses in the time domain, down to the 10–100 ps regime. For such experiments a ‘‘pump-probe’’ process is used: the reaction is initiated by a short duration laser pulse (e.g., heating, photon excitation, or other stimulation) and probed at some delay relative to the excitation by a pulse of synchrotron X-rays. These types of approaches with existing synchrotrons allow measurement of processes whose characteristic times fall within the microsecond to picosecond range. In the case of sorption the pump pulse can be used to free a large number of specific ions from inactive solution states so that they may approach the surface for sorption. Subsequent probe pulses can then follow the reaction as interface reactions develop.


61

2.8. Conclusions CTR analysis of silicate sorbed on the R-plane of hematite identifies a significant fraction as being ordered on the surface, and specifically with a monodentate-like geometry. This geometry is similar to that observed in pyroxene and amphibole minerals, and may explain how the beginnings of a silicate surface phase may precipitate. GIEXAFS measurements show that the bulk of silicate appears to precipitate on the surface as a poorly ordered structure with a small Si–O–Si coordination number. This phase is not unlike the colloidal silica used in polishing samples, and can be promoted by the presence of such contaminants on the surface. It is not yet clear how these two types of silicate are related to one another, i.e., if the ordered sorbates bind to new sorbing silicate creating the disordered phase, or if the disordered phase is separately sorbed or collected. Arsenate GIEXAFS shows at least two types of sorption complexes on both surface of hematite examined, with particular specificity of bidentate–mononuclear complexes and for step edges on the R plane. CTR and GIEXAFS measurements are shown to be strongly complementary tools in the investigation of surface sorbate and precipitate structure, particularly as CTR specifically probes the ordered interfacially coherent structure of any sorbate, while GIEXAFS collects a signal from all sorbates. In addition, low energy X-ray GIEXAFS can be of special significance in the investigation of geochemical species such as silicate, phosphate, and sulfate, though the method is experimentally difficult and cannot be done at many synchrotron facilities.

ACKNOWLEDGMENTS This work was done with support from the DOE-BER program supporting the NSF EMSI molecular environmental science institute at Pennsylvania State University (CEKA). Additional support was provided by DOE-BES to GAW and NSF-CBET and NSF-CHE support to TPT for the study of mineral–water interface reactions. Experimental work at the NSLS was greatly facilitated by Paul Northrup on beamline X15B, and at SSRL on beamline 11-2 by Joe Rogers and John Bargar. Suggestions for CTR data analysis from Kunal Tanwar and Sarah Petitto and valuable discussion about silicate aqueous systems with James Davis, Yuji Arai, and Christine Conrad are greatly appreciated.

62


REFERENCES Adiwidjaja, G., Friese, K., Klaska, K.-H., Moore, P. B., & Schluter, J. (2000). The crystal structure of the new mineral wilhelmkleinite: ZnFeIII 2 (OH)(AsO4)2. Z. Kristallog., 215, 96–101. Ahmad, S. (1984). Hisingerite material from a basalt quarry near Geelong, Victoria, Australia. Clays Clay Miner., 32, 272–278. Birle, J. D., & Tettenhorst, R. (1968). Refined muscovite structure. Mineral. Mag. J. Mineral. Soc., 36 (282), 883. Brown, G. E. (1990). Spectroscopic studies of chemisorption-mechanisms at oxidewater interfaces. Rev. Mineral. Geochem., 23, 309–363. Brown, G. E., & Sturchio, N. C. (2002). An overview of synchrotron radiation applications to low temperature geochemistry and environmental science. Rev. Mineral. Geochem., 49, 1–115. Buerger, M. J., Dollase, W. A., & Garaycohea-Wittke, I. (1967). The structure and composition of the mineral pharacosiderite. Z. Kristallog., 125, 92–108. Catalano, J. G., Zhang, Z., Park, C., & Fenter, P. (2007). Bridging arsenate surface complexes on the hematite (012) surface. Geochim. Cosmochim. Acta, 71, 1883–1897. Cotton, F. A., Wilkinson, G. (1966). Advanced Inorganic Chemistry. Wiley-Interscience, New York. Doelsch, E., Masion, A., Rose, J., Stone, W. E. E., Bottero, J. Y., & Bertsch, P. M. (2003). Chemistry and structure of colloids obtained by hydrolysis of Fe(III) in the presence of SiO4 ligands. Colloids Surf., A 217 (1–3), 121–128. Doelsch, E., Rose, J., Masion, A., Bottero, J. Y., Nahon, D., & Bertsch, P. M. (2002). Hydrolysis of iron(II) chloride under anoxic conditions and influence of SiO4 ligands. Langmuir, 18 (11), 4292–4299. Doelsch, E., Stone, W. E. E., Petit, S., Masion, A., Rose, J., Bottero, J. Y., & Nahon, D. (2001). Speciation and crystal chemistry of Fe(III) chloride hydrolyzed in the presence of SiO4 ligands: 2. Characterization of Si-Fe aggregates by FTIR and Si-29 solid-state NMR. Langmuir, 17 (5), 1399–1405. Dollase, W. A. (1971). Refinement of crystal structures of epidote, allanite and hancockite. Am. Mineral., 56 (3-4), 447–458. Eggleston, C. M., Stack, A. G., Rosso, K. M., Higgins, S. R., Bice, A. M., Boese, S. W., Pribyl, R. D., & Nichols, J. J. (2003). The structure of hematite (alphaFe2O3) (001) surfaces in aqueous media: Scanning tunneling microscopy and resonant tunneling calculations of coexisting O and Fe terminations. Geochim. Cosmochim. Acta, 67 (5), 985–1000. Eng, P. J., Trainor, T. P., Brown, G. E., Waychunas, G. A., Newville, M., Sutton, S. R., & Rivers, M. L. (2000). Structure of the hydrated alpha-Al2O3 (0001) surface. Science, 288 (5468), 1029–1033. Fenter, P. A. (2002). X-ray reflectivity as a probe of mineral–fluid interfaces: A user guide. Rev. Mineral. Geochem., 49, 149–220.


63

Hawthorne, F. C. (1976). A refinement of the structure of adamite. Can. Mineral., 14, 143–148. Hazen, R. M., & Finger, L. W. (1989). High-pressure crystal-chemistry of andradite and pyrope – revised procedures for high-pressure diffraction experiments. Am. Mineral., 74 (3–4), 352–359. Higgins, J. B., Ribbe, P. H., & Nakajima, Y. (1982). An ordering model for the commensurate antiphase structure of yoderite. Am. Mineral., 67 (1–2), 76–84. Icopini, G. A., Brantely, S. L., & Heaney, P. J. (2005). Kinetics of silica oligomerization and nanocolloid formation as a function of pH and ionic strength at 25 degrees C. Geochim. Cosmochim. Acta, 69 (2), 293–303. Ivaldi, G., Catti, M., & Ferraris, G. (1988). Crystal-structure at 25-degrees-C and 700-degrees-C of magnesiochloritoid from a high-pressure assemblage (Monte Rosa). Am. Mineral., 73 (3–4), 358–364. Jun, Y. S., Ghose, S. K., Trainor, T. P., Eng, P. J., & Martin, S. T. (2007). Structure of the hydrated (10–14) surface of rhodochrosite (MnCO3). Environ. Sci. Technol., 11, 3918–3925. Kitahama, K., Kiriyama, R., & Yoshihisa, B. (1975). Refinement of the crystal structure of scorodite. Acta Cryst., B31, 322–324. Kokkoros, P. (1938). Uber die struktur des durangit NaAlF(AsO4). Z. Kristallog., 99, 38–49. Lo, C. S., Tanwar, K. S., Chaka, A. M., & Trainor, T. P. (2007). Density functional theory study of the clean and hydrated (1–102) surfaces. Phys. Rev. B, 75, 075425(15). Luxton, T. P., Tadanier, C. J., & Eick, M. J. (2006). Mobilization of arsenite by competitive interaction with silicic acid. Soil Sci. Soc. Am. J., 70 (1), 204–214. Manceau, A., Chateigner, D., & Gates, W. P. (1998). Polarized EXAFS, distancevalence least-squares modeling (DVLS), and quantitative texture analysis approaches to the structural refinement of Garfield nontronite. Phys. Chem. Miner., 25 (5), 347–365. Manceau, A., Ildefonse, P., Hazemann, J. L., Flank, A. M., & Gallup, D. (1995). Crystal-chemistry of hydrous iron silicate scale deposits at the Salton-sea geothermal-field. Clays Clay Miner., 43 (3), 304–317. Meng, X., Korfiatis, G. P., Bang, S., & Bang, K. W. (2002). Combined effects of anions on arsenic removal by iron hydroxides. Toxicol. Lett., 133, 103–111. Moore, P. B., & Araki, T. (1977). Mitridatite, Ca6(H2O)6[FeIII 9 O6(PO4)9].3H2O: A noteworthy octahedral sheet structure. Inorg. Chem., 16, 1096–1106. Nordstrom, D. K. (2002). Public health: Worldwide occurrences of arsenic in ground water. Science, 296 (5576), 2143–2145. Ostroverkhov, V., Waychunas, G. A., & Shen, Y. R. (2004). Vibrational spectra of water at water/alpha-quartz (0001) interface. Chem. Phys. Lett., 386 (1–3), 144–148. Perchiazzi, N., Ondrus, P., & Skala, R. (2004). Ab initio X-ray powder structure determination of parascorodite, Fe(H2O)2AsO4. Eur. J. Mineral., 16, 1003–1007.

64


Pokrovski, G. S., Schott, J., Garges, F., & Hazemann, J. L. (2003). Iron (III)-silica interactions in aqueous solution: Insights from X-ray absorption fine structure spectroscopy. Geochim. Cosmochim. Acta, 67 (19), 3559–3573. Powell, J. J., Van de Water, J., & Gershwin, M. E. (1999). Evidence for the role of environmental agents in the initiation or progression of autoimmune conditions. Environ. Health Perspect., 107 (Suppl. 5), 667–672. Prewitt, C. T., & Burnham, C. W. (1966). Crystal structure of jadeite NaAlSi2O6. Am. Mineral., 51 (7), 956–975. Qi, G. W., Klauber, C., & Warren, L. J. (1993). Mechanism of action of sodiumsilicate in the flotation of apatite from hematite. Int. J. Miner. Process., 39 (3–4), 251–273. Redhammer, G. J., Beran, A., Schneider, J., Amthauer, G., & Lottermoser, W. (2000). Spectroscopic and structural properties of synthetic micas on the annitesiderophyllite binary: Synthesis, crystal structure refinement, Mossbauer, and infrared spectroscopy. Am. Mineralogist, 85 (3–4), 449–465. Roberts, A. C., Burns, P. C., Gault, R. A., Criddle, A. J., & Feinglos, M. N. (2004). Petewilliamsite, (Ni,Co)30(As2O7)15, a new mineoral from Johanngeorgenstadt, Saxony, Germany: Description and crystal structure. Mineral. Mag., 68 (2), 231–240. Roberts, A. C., Cooper, M. A., Hawthorne, F. C., Criddle, A. J., & Stirling, J. A. R. (2002). Sewardite, CaFe3+ 2 (AsO4)2(OH)2, the Ca-analogue of carminite, from Tsumeb, Namibia: Description and crystal structure. Can. Mineral., 40, 1191–1198. Robinson, I. K. (1986). Crystal truncation rods and surface roughness. Phys. Rev. B, 33 (6), 3830–3836. Robinson, I. K., & Tweet, D. J. (1992). Surface X-ray diffraction. Rep. Prog. Phys., 55, 599–651. Rumanova, I. M., & Skipetrova, T. I. (1959). The crystalline structure of lawsonite. Dokl. Akad. Nauk SSSR, 124 (2), 324–327. Swedlund, P. J., & Webster, J. G. (1999). Adsorption and polymerisation of silicic acid on ferrihydrite, and its effect on arsenic adsorption. Water Res., 33 (16), 3413–3422. Tanwar, K. S., Lo, C. S., Eng, P. J., Catalano, J. G., Walko, D. A., Brown, G. E., Waychunas, G. A., Chaka, A. M., & Trainor, T. P. (2007). Surface diffraction study of the hydrated hematite (1(1)over-bar-02) surface. Surf. Sci., 601 (2), 460–474. Trainor, T. P., Chaka, A. M., Eng, P. J., Newville, M., Waychunas, G. A., Catalano, J. G., & Brown, G. E. (2004). Structure and reactivity of the hydrated hematite (0001) surface. Surf. Sci., 573 (2), 204–224. Waychunas, G. A. (2002). Grazing-incidence X-ray absorption and emission spectroscopy. Rev. Mineral. Geochem., 49, 267–315. Waychunas, G. A., Fuller, C. C., Rea, B. A., & Davis, J. A. (1996). Wide angle X-ray scattering (WAXS) study of ‘‘two-line’’ ferrihydrite structure and the effect


65

of arsenate sorption: Comparison with EXAFS results. Geochim. Cosmochim. Acta, 60, 1765–1781. Waychunas, G. A., Rea, B. A., Fuller, C. C., & Davis, J. A. (1993). Surface chemistry of ferrihydrite: Part 1. EXAFS studies of the geometry of coprecipitated and adsorbed arsenate. Geochim. Cosmochim. Acta, 57, 2251–2269. Waychunas, G. A., Trainor, T. P., Eng, P. J., Catalano, J. G., Brown, G. E., Davis, J. A., Rogers, J., & Bargar, J. (2005). Surface complexation studied by combined grazing-incidence EXAFS and surface diffraction: Arsenate on hematite (0001) and (1102). Anal. Bioanal. Chem., 383, 12–27. Weber, W. J., & Stumm, W. (1965). Formation of a silicato-iron(3) complex in dilute aqueous solution. J. Inorg. Nucl. Chem., 27 (1), 237–239. Welch, A. H., Westjohn, D. B., Helsel, D. R., & Wanty, R. B. (2000). Arsenic in ground water of the United States: Occurrence and geochemistry. Ground Water, 38 (4), 589–604. Winter, J. K., & Ghose, S. (1979). Thermal-expansion and high-temperature crystalchemistry of the Al2SiO5 polymorphs. Am. Mineral., 64 (5–6), 573–586. Wright, J. P., McLaughlin, A. C., & Attfield, J. P. (2000). Partial frustration of magnetic order in synthetic angelellite, Fe4As2O11. J. Chem. Soc. Dalton, doi:10.1039/b005350n. Zeng, L. (2003). A method for preparing silica-containing iron(III) oxide adsorbents for arsenic removal. Water Res., 37 (18), 4351–4358.


Chapter 3

Molecular Structure of Lead(II) Coprecipitated with Iron(III) Oxyhydroxide Shelly D. Kelly1,, Peng Lu2, Trudy Bolin3, Soma Chattopadhyay4, Matthew G. Newville5, Tomohiro Shibata4 and Chen Zhu2 1

Argonne National Laboratory, Biosciences Division, Argonne, IL 60439-4843, USA Department of Geological Sciences, Indiana University, Bloomington, IN 47405-1405, USA 3 Argonne National Laboratory, Advanced Photon Source, Sector 9 BM, Argonne, IL 60439, USA 4 CSRRI-IIT, MR-CAT, APS, Argonne National Laboratory, Argonne, IL 60439, USA 5 University of Chicago, Chicago, IL 60637, USA 2

ABSTRACT Detailed knowledge of the uptake mechanisms of trace metals in geomedia is fundamental to environmental geochemical processes, particularly to the prediction of contaminant transport and mobility in geological media. We investigated the molecule structure by using macroscopic and microscopic structural probes including X-ray diffraction (XRD), high-resolution transmission and analytical electron microscopy (HRTEM-AEM), and extended X-ray absorption fine structure (EXAFS) spectroscopy. The long-range crystalline properties of the PbFe coprecipitate, as measured by XRD, were consistent with poorly crystalline lepidocrocite and two-line ferrihydrite (2LFh). The particle size and shape of the PbFe coprecipitate, as measured by HRTEM, showed a mixture of spheres (2–6 nm in diameter) and needles (20–80 nm 200–300 nm) composed of aggregated crystallites 2–3 nm in diameter. The local atomic structure about Pb and Fe in the PbFe coprecipitate was further elucidated through a series of EXAFS modeling efforts, including molecular moiety modeling, linear combination fitting, and co-refinement based on a FeO6 sheet structure for lepidocrocite. The larger Corresponding author. Tel.: +1-630-252-7376; Fax: +1-630-252-2959;

E-mail: [email protected] (S.D. Kelly).

68

S. D. Kelly et al.

atomic size of Pb as compared to Fe was accounted for in the model of the FeO6 sheet structure by displacing the PbO6 unit perpendicular to the sheet by 0.3070.02 A˚ from the FeO6 unit position. The co-refinement of the Pb LIII-edge and the Fe K-edge EXAFS spectra with the same local atomic environment and the additional PbO6 displacement suggested that Pb formed a solid solution in the PbFe coprecipitate. The coprecipitation of Pb by Fe oxyhydroxides may decrease the mobility and bioavailability of Pb2+ by incorporating Pb into the Fe oxyhydroxide structure.

3.1. Introduction Iron oxyhydroxides occur widely in surficial and subsurface geological environments and are efficient scavengers of trace metals (including Pb2+) and radionuclides (Jambor and Dutrizac, 1998). The uptake of Pb2+ onto Fe oxyhydroxides through the adsorption (ADS) contact method has been studied extensively (Dzombak and Morel, 1990 for review; see Dyer et al., 2003). In these systems the Fe oxyhydroxides were formed first, and then the trace element was added. The coprecipitation (CPT) contact method for Pb2+ sorption by Fe oxyhydroxides is defined as a base titration with both the trace element (Pb2+) and the host element (Fe3+) in solution together, while the pH is increased to form a solid precipitate. Even though many geological, environmental, and industrial systems resemble CPT, they have received little attention. For example, when acid mine drainage or landfill leachate loaded with Fe3+ and trace metals (e.g., Pb, As, Cu.) is mixed with neutral surface water and groundwater or allowed to react with calcite in soils, sediments, and aquifers, the Fe3+ and trace metals (such as Pb2+) are coprecipitated simultaneously. During the pretreatment of wastewater and in analytical chemistry methods, Fe3+ is also coprecipitated with trace metals. Laboratory studies have shown that for some systems macroscopic sorption behaviors (including pH-dependent sorption edges and extent of uptake) can differ if the contact method is CPT versus ADS (Charlet and Manceau, 1992; Crawford et al., 1993; Waychunas et al., 1993; Karthikeyan et al., 1997, 1999); however, other systems show essentially no difference due to the contact method (Crawford et al., 1993; Karthikeyan et al., 1997). The differences or similarities are often assumed to be caused by the atomic environment of the trace metal upon sorption to the host material. Commonly, ferrihydrite is the first phase to be precipitated from an aqueous Fe3+ solution. Ferrihydrite is poorly ordered and metastable. Over time, ferrihydrite is transformed to more ordered phases (e.g., goethite and hematite)

Molecular Structure of Lead(II) Coprecipitated with Iron(III) Oxyhydroxide

69

(Cornell and Schwertmann, 1996). This transformation can lead to structural incorporation of the trace metals, resulting in irreversible sorption also called desorption hysteresis. Direct measurements of the atomic structure of Pb2+ and the Fe oxyhydroxide host material are fundamental to understanding the differences in macroscopic sorption properties of the solids as the contact method, pH, ionic strength, and trace metal concentrations vary. Extended X-ray absorption fine structure (EXAFS) spectroscopy has been used to determine the average local atomic structure of Pb sorption to Fe oxyhydroxides through the ADS contact method (Manceau et al., 1992), with various adsorption rates (Scheinost et al., 2001) and heat treatments (Sorensen et al., 2000). An EXAFS study of Pb2+ uptake by goethite and hematite found Pb adsorbed to the edges of FeO6 as mononuclear bidentate complexes (Bargar et al., 1997). Studies of Pb adsorption onto two-line ferrihydrite (2LFh) as a function of pH, ionic strength, and adsorbate concentration found Pb coordinated by 1.9–2.4 Fe neighbors at distances of 3.32–3.36 A˚ (Trivedi et al., 2003). EXAFS spectroscopy is one of the few tools that can give information about the adsorption and/or incorporation of trace metals in poorly ordered nanoparticulate materials such as ferrihydrite. A systematic study based on changes in the atomic configuration of Pb2+ and geochemical modeling is needed to explore differences in the uptake mechanism of Pb2+ within the Fe oxyhydroxide system by the CTP and ADS contact methods under a variety of conditions. Though a general picture of Pb2+ surface adsorption to Fe oxyhydroxides has been established for the ADS contact method under several different conditions (Manceau et al., 1992; Sorensen et al., 2000; Scheinost et al., 2001; Trivedi et al., 2003), no spectroscopic measurements have investigated the structure of Pb2+ after CPT with Fe3+. The aim of the current study is to define a detailed procedure for determining a structurally incorporated trace metal within a poorly ordered nanoparticulate Fe oxyhydroxide. In this study we investigated the physical properties of one particular PbFe coprecipitate formed after 24 h by the coprecipitation of Pb2+ (0.74 mM) with Fe3+ (6.4 mM) at pH 5.25. This system was generally modeled after acid mine drainage loaded with Fe3+ and Pb2+, forming a precipitate as the acid mine drainage is neutralized by groundwater. Fe and Pb can coexist with a wide range of concentrations in acid mine drainage. The Cassiterite tailings site of North Queensland, Australia, contains Fe at up to 8.8% and Pb at up to 621 ppm (3 mM) (Lottermoser and Ashley, 2006). Our study focuses on the structure of the Fe oxyhydroxides and the local atomic structure of Pb, as determined by X-ray diffraction (XRD), high-resolution transmission and analytical electron microscopy

70

S. D. Kelly et al.

(HRTEM-AEM), and EXAFS spectroscopy. Details of EXAFS spectroscopy are presented, and a general procedure for the determination of structurally incorporated trace metals is developed for systems of poorly ordered nanoparticulate ferrihydrite. Our EXAFS modeling approach used information from increasingly complex models to determine the structural coordination of Pb in the PbFe coprecipitate. The EXAFS analysis was first based on molecular moiety modeling of the Pb and Fe EXAFS spectra to show the similarities and differences between the dominant EXAFS signal components. Linear combination fitting (LCF) of the Fe EXAFS spectrum was used to illustrate that the EXAFS spectrum was consistent with a mixture of disordered lepidocrocite and 2LFh. Next, an EXAFS model was built for the sheet-like FeO6 structure that is the basic unit of lepidocrocite. This model was applied to both the Pb and Fe EXAFS spectra by allowing the PbO6 unit to be displaced from the sheet structure to account for the longer Pb–Fe distances and shorter Fe–Fe distances found through molecular moiety modeling. The co-refinement of both the Pb and Fe EXAFS spectra with a single sheet-like FeO6 structure, but with Pb2+ displaced from the usual Fe3+ position by 0.3070.02 A˚, suggested that the predominant Pb species is a solid solution in the coprecipitate.

3.2. Experimental The batch experiments were conducted at room temperature (221C). All reagents were of analytical grade from Fisher Scientific, Inc., unless otherwise indicated. An Fe3+–Pb2+ solution containing 6.4 mM Fe(NO3)3 (ferric nitrate) and 0.74 mM Pb(NO3)2 (lead nitrate) was made from Fe(NO3)3 9H2O (purity 99.1%) and Pb(NO3)2 in a 10 mM KNO3 background electrolyte solution. The solution pH was monitored by using a Beckman F34 pH meter calibrated with three buffers (pH 4, 7, and 10). Although CO2 is usually present in acid mine drainage or wastewater treatment systems it was excluded to simplify the system as in previous adsorption studies (Scheinost et al., 2001; Trivedi et al., 2003). To minimize the influence of atmospheric CO2, N2 was bubbled through a Teflon capillary during coprecipitation. The solution pH was increased by injecting a 1 M KOH solution at a rate of 0.01 mL min1 with a 0.1 mL Gastights #1700 syringe and Teflon capillary tubing attached to the syringe needle. The PbFe coprecipitate was equilibrated for 24 h, after which time a final pH of 5.25 was measured. Subsequent hourly pH measurements of a duplicate sample, up to 100 h, showed that solution pH had stabilized after 24 h. The solids after 24 h of


71

equilibration were separated by centrifugation at a rate of 3910 RCF (relative centrifugal force or g-force) for 45 min, and the liquid supernatant was decanted. The PbFe precipitates were washed three times in deionized water to remove electrolytes. After the washing procedure, the precipitate was separated by re-centrifugation and immediately transferred to capped vials in ethyl alcohol to prevent further reaction before XRD, HRTEMAEM, and EXAFS measurements, which were performed within 1–3 weeks. The PbFe coprecipitate was prepared for HRTEM-AEM by ultrasonication in ethyl alcohol, while the coprecipitate was cooled to prevent transformation. The resuspended particles were then pipetted onto a holey carbon film supported by a copper mesh (TEM grid) and air dried. The HRTEMAEM measurements included both HRTEM and AEM measurements. The AEM measurements (also called STEM-EDX or scanning transmission electron microscopy-electron dispersive X-ray analysis) were made on the PbFe coprecipitate with a Philips FEI Tecnai F20 TEM/STEM with a point-to-point resolution of 0.24 nm, equipped with a Gatan imaging filter and energy-dispersive X-ray detector for energy-filtered imaging and microanalysis. The HRTEM measurements were made with a JEOL 4000EX HRTEM with a point-to-point resolution of 0.17 nm and a Gatan imaging filter to obtain lattice imaging resolution. The PbFe coprecipitate and 2LFh (synthesized following the procedure of Cornell and Schwertmann, 1996) was prepared for XRD analysis by dispersing the precipitate in acetone to form a thick suspension. This suspension was deposited on a zero-background quartz plate and air dried at room temperature. The XRD measurements of the PbFe coprecipitate were made by using a PANalytical X’Pert PRO Theta-Theta multipurpose diffractometer, equipped with a Cu anode at 45 kV and 40 mA, a divergent beam monochromator, and an X’Celerator detector. The XRD spectra were collected over a 2y scan range from 10.0101 to 99.9681 with a step size of 0.0331. EXAFS standards were prepared by spreading fine powder sample on tape. Iron oxide standard materials include lepidocrocite (Bayferox 943) and goethite (Bayferox 910) purchased from Lanexs. Hematite was purchased from Aldrich, and ferrihydrite was produced by standard procedures (Cornell and Schwertmann, 1996). Lead oxide standard material was also produced by standard procedures (Thoral et al., 2005). The Fe K-edge and Pb LIII-edge EXAFS spectra were collected at room temperature from three different beamlines at the Advanced Photon Source (APS) at Argonne National Laboratory. The Fe K-edge spectra for the PbFe coprecipitate were collected in transmission mode at the CMC-CAT bending magnet beamline. The transmission sample was prepared by spreading the PbFe coprecipitate paste between two layers of Kapton tape. The

72

S. D. Kelly et al.

beamline-specific parameters for CMC-CAT (9-BM) include a Si(1 1 1) double-crystal monochromator and a Rh harmonic rejection mirror. The incident and transmission X-ray intensities were monitored by using ionization chambers filled with He gas. The X-ray profile at the sample was 2 mm (vertical) 0.5 mm (horizontal). The Fe K-edge spectra for lepidocrocite, 2LFh, goethite, and hematite standards (powders on tape) were collected in transmission mode at the MR-CAT insertion device beamline following standard procedures (Segre et al., 2000). The Pb LIII-edge spectra for the PbFe coprecipitate were collected in fluorescence mode at the GSECARSCAT bending magnet beamline following standard procedures (Newville et al., 1999). The fluorescence signal was found to be of higher quality than the transmission signal, with no self-absorption, because the large amount of ethyl alcohol in the paste of the PbFe coprecipitate diluted the sample. The fluorescence sample was prepared by filling a thin Plexiglas sample holder with PbFe coprecipitate paste, sealed with Kapton film windows. The EXAFS spectra were processed by using the procedures outlined in the UWXAFS analysis package (Stern et al., 1995). The EXAFS energy scans were aligned and averaged in energy. Then the background was removed by using Athena (Ravel and Newville, 2005), which implements the IFEFFIT methods (Newville, 2001) with an Rbkg value of 1.0 A˚ (Newville et al., 1993). The theoretical models were built by using FEFF7.02 (Zabinsky et al., 1995) and were based on the crystal structure of lepidocrocite (Zhukhlistov, 2001).

3.3. Experimental Results 3.3.1. X-Ray Diffraction Spectra Figure 3.1 shows the XRD patterns for the PbFe coprecipitate and 2LFh with the diffraction lines of lepidocrocite (ICDD: 44-1415) and goethite (ICDD: 28-0713). The 2LFh spectrum contains two broad peaks at d-spacings of 0.154 and 0.27 nm (Cornell and Schwertmann, 1996). The XRD pattern for the PbFe coprecipitate (Fig. 3.1) is consistent with a mixture of poorly crystalline lepidocrocite and 2LFh.

3.3.2. High-Resolution Transmission and Analytical Electron Microscopy The HRTEM images (Fig. 3.2a) show a representative sample from the PbFe coprecipitate with significant particle overlap, indicating that the


73

1400

Relative intensity

2-line Ferrihydrite 900

2-line Ferrihydrite

400

-100 10

20

30

40

50

60

70

2-theta (degree)

Figure 3.1: XRD Spectrum for PbFe Coprecipitate (Top: Black Spectrum). The XRD Spectrum for 2LFh (Bottom: Gray Spectrum) has been Displayed Beneath by Rescaling the Intensity (by 400). The Relative Line Intensities and Positions for Lepidocrocite (ICDD: 44-1415) and Goethite (ICDD: 28-0713) are also Shown as the Lines Terminated with Triangles and Crosses, Respectively. resuspension process used to isolate individual grains was not completely successful. The difficulty in preparing isolated particles adds to the uncertainty in the interpretation of these images. Although additional higherquality images are currently being made from more dispersed particles, some significant observations can be made from the present images. Figure 3.2 shows that the PbFe coprecipitate is composed of spherical and needle-like particles. The spherical particles have diameters ranging from 2 to 6 nm; most of the spheres are 5 nm in diameter (Fig. 3.2a). The needle-like particles are 8–20 nm across and 200–300 nm long. High-resolution images show that the needles are composed of aggregated crystallites 2–3 nm in diameter (not shown). Fast Fourier transformation of the images (not shown) of the spherical particles that are typical for 2LFh showed several d-spacings, whereas 2LFh particles usually show two broad, intense rings (Janney et al., 2000). The

74

S. D. Kelly et al.

(a) 50

Fe

Counts

40 30

20

O Cu

Fe Cu

10

Pb

Fe

Cu 0 5 (b)

Pb Cu

Pb 10

15

20

Energy (keV)

Figure 3.2: (a) TEM Image showing Spherical and Needle-Like PbFe Coprecipitate Particles. Scale Bar Represents 20 nm. (b) STEM-EDX Spectrum showing the Particle Compositions of Fe, O, and Pb, with a Few Cu Lines from the Cu TEM-Grid. spectrum suggested a more crystalline structure for the spherical particles than is typical of 2LFh. Detailed interpretations of these spectra are the topic of continuing investigations. The STEM-EDX measurements showed that the PbFe coprecipitate contains Fe, O, and Pb (Fig. 3.2b). Quantitative analysis yielded a ratio of Pb to Fe in the PbFe coprecipitate that was close to the mass balance for the solution chemistry of 10% Pb and 90% Fe, suggesting close to 100% uptake


2

lepidocrocite

30

20

1 2LFh

10

Pb-Fe and Model A

0 Pb-Fe and Model B

x (k).k3 (Å-3)

x(k).k3 (Å-3)

75

0

-1

-10 -2 0

(a)

2

4

6

8

10

12

14

k (Å-1)

0

(b)

2

4

6

8

10

12

k (Å-1)

Figure 3.3: EXAFS w(k) k3 Spectra and Models. (a) Fe K-Edge Spectra for (from Top) Lepidocrocite; 2LFh; PbFe Coprecipitate (Open Symbols) with Model A (Solid Line: 60:40 of Disordered Lepidocrocite:2LFh); and PbFe Coprecipitate (Open Symbols) with Linear Combination Model B (Solid Line: 100% Disordered Lepidocrocite). (b) Pb LIII-Edge Spectrum for PbFe Coprecipitate (Open Symbols), Overlaid with EXAFS Fitting Model (Solid Line). of Pb by the PbFe coprecipitate. More than 50 EDX spectra showed Pb homogeneously distributed in the sampled regions of both spherical and needle-like particles.

3.3.3. EXAFS Spectra Figure 3.3a shows the Fe K-edge w(k) spectra for lepidocrocite, 2LFh, and the PbFe coprecipitate, while Fig. 3.3b shows Pb LIII-edge w(k) spectrum for the PbFe coprecipitate. The goethite and hematite EXAFS spectra were found not to be representative of the PbFe coprecipitate sample spectrum (not shown).

3.4. X-ray Absorption Modeling Our modeling approach for determining the local atomic environment about Pb and Fe in the PbFe coprecipitate was as follows: 1. The Pb and Fe EXAFS spectra for the PbFe coprecipitate were modeled by using a molecular moiety approach. The results were compared to determine the similarities and differences in the strongest signals from neighboring Pb and Fe atoms in the coprecipitate. The distribution of Fe

76

S. D. Kelly et al.

neighbors about Fe atoms in the PbFe coprecipitate was also compared to the crystallographic distributions for 2LFh, lepidocrocite, and goethite. 2. The Fe EXAFS w(k) spectrum from the PbFe coprecipitate was shown to be reproduced with a linear combination of disordered lepidocrocite and 2LFh, as predicted by XRD of the bulk sample. In addition, the Fe EXAFS w(k) spectrum from the PbFe coprecipitate was modeled with an additionally disordered lepidocrocite standard alone, with a scaling factor much less than one. On the basis of these results, the molecular structure about the Fe in the PbFe coprecipitate was tested for a structure similar to the FeO6 sheet-like structure of lepidocrocite. 3. A crystallographic model based on a lepidocrocite FeO6 sheet structure was built to describe the lepidocrocite standard EXAFS spectrum. This FeO6 sheet model was relaxed and applied to the Fe and Pb spectra from the PbFe coprecipitate. 4. A model was proposed for substitution of the PbO6 octahedral structure into the sheet-like FeO6 structure. With this model, both the Pb and Fe EXAFS spectra for the PbFe coprecipitate were modeled simultaneously. This model is used to test the spectra for consistency within the local atomic environment of Pb and Fe. 3.4.1. Modeling Pb and Fe EXAFS Spectra on the Basis of Molecular Moieties Both the Pb and Fe EXAFS spectra were modeled with a molecular moiety based on two O and two Fe signals. The model included 11 parameters (NO1, NO2, NFe1, NFe2, DRO1, DRO2, DRFe1, DRFe2, s2O ; s2Fe ; and DE0). The Pb and Fe data (3.0–9.0 A˚1) and fit (1.0–3.8 A˚) ranges resulted in 12 independent points in each measured spectrum (Stern, 1993). The values for S 20 ; based on EXAFS modeling of lead oxide and lepidocrocite, were 0.9670.10 and 0.8970.05 for the Pb and Fe spectra, respectively. The model was optimized to each spectrum processed with k-weight values of 1, 2, and 3 in the Fourier transform. The Fourier transform of the EXAFS spectrum and the model for Pb and Fe are shown in Fig. 3.4. The best-fit EXAFS parameters are listed in Table 3.1. The atomic distribution about Fe is compared to other Fe-oxides in Table 3.2. 3.4.2. Linear Combination Fitting of Fe EXAFS Spectrum The Fe K-edge EXAFS w(k) spectrum for the PbFe coprecipitate was compared to the standard spectra for lepidocrocite, 2LFh, goethite, and


77

0.4 0.3 0.2

Re(FT(χ(k)k2))(Å-3)

0.0

|FT(χ(k)k2)|(Å-3)

0.0 O1

-0.2 O2

-0.3

O1

-0.6 O2

-0.9 Fe1

-0.4

Fe1

-1.2

Fe2

Fe2

-0.6 -1.5 (a)

(b)

4

4


0

0

2

-3

|FT(χ(k)k )|(Å )

2

-2

O1

-4

-4

O1

-8 O2

O2

-12 Fe1

-6

Fe1 Fe2

Fe2

-16 0 (c)

2

4 R(Å)

6

0 (d)

2

4

6

R(Å)

Figure 3.4: The Magnitude (a and c) and Real Part (b and d) of the Fourier Transform of the Fe (a and b) and Pb (c and d) EXAFS Spectra (Symbols) and Molecular Moiety Model Fit (Line) for the PbFe Coprecipitate, Processed with k-Weight ¼ 2 in the Fourier Transform. The Contribution to the Model from Each Shell is Shown Individually Beneath the Measured Spectra.

78

S. D. Kelly et al.

Table 3.1: EXAFS Model Results for Pb LIII-Edge and Fe K-Edge Spectra for the PbFe Coprecipitate, Modeled with Molecular Moiety Structuresa. Neighboring atom label

Degeneracy

R (A˚)

s2 (103 A˚2)

Pb in PbFe coprecipitate Pb–O1 Pb–O2 Pb–Fe1 Pb–Fe2

3.171.1 1.270.4 3.672.8 1.270.8

2.3270.03 2.5770.09 3.3970.02 3.8270.09

1175

Fe in PbFe coprecipitate Fe–O1 Fe–O2 Fe–Fe1 Fe–Fe2

2.870.8 2.170.2 3.870.9 0.770.5

1.9770.01 2.1570.01 3.0770.01 4.0470.06

0.571.0

21710

1373

a

The energy shift values were determined to be 4.571.2 and 0.570.9 eV for the Pb and Fe spectra, respectively.

Table 3.2: Distribution of Fe Neighbors about Fe in the PbFe Coprecipitate and for Several Fe Oxyhydroxides. Number of Fe–Fe neighbors

Fe–Fe distance (A˚)

Fe in PbFe coprecipitate 3.870.9 0.770.5

3.0770.01 4.0470.06

Ferrihydritea 2.0–3.1

3.01–3.15

Lepidocrocite 6 2

3.0 3.9

Goethite 2 2 4 2

3.0 3.3 3.5 4.6

a

The ferrihydrite structure is from an EXAFS study of 2LFh (Zhao et al., 1994).

hematite. The goethite and hematite spectra (not shown) were not similar to the PbFe coprecipitate spectrum. The PbFe coprecipitate spectrum was found to be similar in its broad features to both the lepidocrocite and 2LFh standard spectra (Fig. 3.3a). All three spectra show maxima at 1.8, 3.7,


79

6.2, and 8.5 A˚1. The EXAFS spectrum for the lepidocrocite standard shows larger amplitudes and more distinct features than the 2LFh standard or the PbFe coprecipitate spectra. The additional amplitude and distinctiveness are due to the well-ordered structure of the large-particle-size crystalline lepidocrocite standard, as compared to the 2LFh standard or the PbFe coprecipitate. The PbFe coprecipitate spectrum was modeled with two different linear combinations. Model A contains a mixture of 3873% 2LFh and 3772% lepidocrocite while Model B contains 50% lepidocrocite (Fig. 3.3a). 3.4.3. EXAFS Model for the FeO6 Sheet Structure The lepidocrocite standard EXAFS spectrum was used to develop an EXAFS model for the FeO6 sheet structure. This model became the basis for the fully consistent EXAFS model for both the Pb and Fe EXAFS spectra for the PbFe coprecipitate sample (developed in the next section). A fully consistent EXAFS model was needed to test whether the Pb coordination environment was consistent with the Fe coordination environment needed for a solid solution of Pb within the Fe oxyhydroxide host material. The FeO6 sheet structure, based on lepidocrocite, is shown in Fig. 3.5a. The model represents all atoms within a 4.1 A˚ radius of a central Fe atom, including two Fe shells and five O shells (Table 3.3). The EXAFS model uses seven parameters (one expansion value, four s2 values, one energy shift value, and one S 20 value). The expansion term is a constant value (a) that is multiplied by the reference distance, Reff, as determined from the crystal structure of lepidocrocite. The data range (3.0–14.0 A˚1) and the fit range (1.0–3.8 A˚) result in 20 independent points in the spectrum (Stern, 1993). The model was optimized to the spectrum processed with k-weight values of 1, 2, and 3 in the Fourier transform. The magnitude and real part of the Fourier transform of the Fe K-edge EXAFS spectrum for lepidocrocite and the model are shown in Fig. 3.6. The ability of the model to reproduce the measured spectrum illustrates that this model describes the lepidocrocite structure well. The EXAFS results are all in the expected ranges. The linear expansion/compression term a specifies a 0.470.1% expansion from the crystalline structure. The s2o1 ; s2o2 ; s2Fe1 ; and s2Fe2 values are 0.00770.001, 0.00870.002, 0.006470.0005, and 0.01170.003 A˚2, respectively, while the energy shift and S 20 values are 0.870.5 eV and 0.9870.06, respectively. The FeO6 sheet-like model was applied to the local structure of Fe in the PbFe coprecipitate. As illustrated through visual comparison of the EXAFS

80

S. D. Kelly et al.

3

6

6

2 6

1

1 2

4 1

2

1

1 4 1

1

6

6

4

6

5

3

3

4

6

5

2 1b

2

6

6b 6b

5

1a

4 1b

1a1a

5

(a)

1a

4

2

1

5

1a 1a

6a 6a

4

3

5

6a 6a 4

6b 6b

5

5

(b)

Figure 3.5: Schematic of (a) FeO6 Sheet Structure and (b) PbO6 Substitution for FeO6 in the Sheet Structure. The Red (Small Dark), Blue (Small Light), and Brown (Large) Spheres Represent O, Fe, and Pb Atoms, Respectively. The Atom Labels Match the EXAFS Model Description in Table 3.5. The Fe Atom Labeled 0 Represents the Center Fe Atom.

Table 3.3: EXAFS Model for Fe K-Edge Lepidocrocite Spectrum based on FeO6 Sheet Structure. Neighboring atom label Fe0–O1a Fe0–O1b Fe0–O2 Fe0–Fe1a Fe0–Fe1b Fe0–O3 Fe0–O4 Fe0–Fe2 Fe0–Fe2–O1b Fe0–O5

Degeneracy

Reff (A˚)

DR (A˚)

s2 (A˚2)

2 2 2 4 2 2 4 2 4 4

1.981 2.013 2.071 3.062 3.072 3.655 3.663 3.873 3.917 4.003

aReff aReff aReff aReff aReff aReff aReff aReff aReff aReff

s2o1 s2o1 s2o1 s2fe1 s2fe1 s2o2 s2o2 s2fe2 s2fe2 s2o2


81

2.0

|FT(χ(k)k2)|(Å-3)

1.5

1.0

0.5

0.0 (a)


1

0

-1

0 (b)

2

4

6

R(Å)

Figure 3.6: The Magnitude (a) and Real Part (b) of the Fourier Transform of the Fe K-Edge EXAFS Lepidocrocite Spectrum (Symbols) and Model (Line) Based on the Lepidocrocite Structure. w(k) spectrum (Fig. 3.3a) for the PbFe coprecipitate with the lepidocrocite spectrum, molecular moiety modeling results (Table 3.1), and LCF results, the amplitude of the lepidocrocite-based model needs to be reduced to describe the PbFe coprecipitate spectrum accurately. The amplitude of the spectrum can be reduced by decreasing the coordination number for each signal in the model and/or by increasing the disorder for the atom pairs of the signal (s2 values). Preliminary tests performed by increasing the s2 values

82

S. D. Kelly et al.

without changing coordination numbers proved unsatisfactory. As shown in Table 3.1, the molecular moiety model of the Fe spectrum for the PbFe coprecipitate indicates an Fe coordination number of 3.870.9 at a distance of 3.0770.01 A˚ for the PbFe coprecipitate, as compared to a value of 6 for large particles of lepidocrocite. In addition, HRTEM measurements indicate that the needle-like particles are 2–3 nm in domain size. A 2 nm lepidocrocite particle has 4.28 Fe atoms at 3.0 A˚, which is consistent with both HRTEM and molecular moiety modeling results for the EXAFS spectrum. Therefore, the coordination numbers were held at the values shown in Table 3.4 for a 2 nm lepidocrocite particle. The previous use of a single expansion term, a, to describe the distances to the neighboring atoms was too restrictive for the poorly crystalline PbFe particles, and therefore several DR values (5 for the O shells and 3 for the Fe shells) were independently determined in the model fit. The final model includes 12 parameters: 8 DR values, 3 s2 values, and 1 energy shift value. This final model is the basis for the fully consistent EXAFS model used below to describe both the Pb and the Fe spectra. 3.4.4. Building a Fully Consistent EXAFS Model for Pb and Fe A fully consistent EXAFS model for both the Pb and Fe coordination was constructed to rigorously test the measured EXAFS spectra for consistent FeO6 and PbO6 coordination. The success or failure of this model is presented to confirm or discount Pb incorporation into the FeO6 sheet structure during the coprecipitation process. To test the final structural model, the Fe and Pb spectra were simultaneously refined with an EXAFS model based on the sheet-like FeO6 structure, with the distance to each group of neighboring atoms about Fe and Pb constrained to be the same for both the Fe and Pb EXAFS spectra, except for an additional Pb displacement from the sheet structure. To accommodate the larger Pb atomic size, the Pb atom was allowed to be displaced from the FeO6 sheet as depicted in Fig. 3.5b; the displacement was determined in the fit to the measured spectra. The Pb displacement parameter is shown in Table 3.4 in terms of the function F(DRpby) parameterized by a single value DRpby, which is the displacement of the Pb atom from the Fe site along the direction perpendicular to the sheet, as depicted in Fig. 3.5b. The initial distances for Pb listed in Table 3.4 were calculated with DRpby ¼ 0.5 A˚. The functional form of F(DRpby) depends on the position of the neighboring atoms in the sheet and is not explicitly given. The Pb displacement splits the distance to several groups of atoms at essentially the same distance from Fe into two very different sets of values denoted by subscripts a and b (see Table 3.4 and


83

Table 3.4: EXAFS Model for Fe K-Edge and Pb LIII-Edge Spectra from the PbFe Coprecipitate. Reff (A˚)

DR (A˚)

s2 (A˚2)

Parameters for Fe K-edge spectrum Fe0–O1b 2:2 Fe0–O1a 2:2 Fe0–O2 2:2 Fe0–Fe1a 4:3.5 Fe0–Fe1b 2:1.8 Fe0–O3 2:1.7 4:3.5 Fe0–O4 2:1.7 Fe0–Fe2 Fe0–Fe2–O1b 4:3.4 4:3.5 Fe0–O5

1.98 2.01 2.07 3.06 3.07 3.66 3.66 3.87 3.92 4.00

DR_feo1 DR_feo1 DR_feo2 DR_xfe1a DR_xfe1b DR_xo3 DR_xo4 DR_xfe2 DR_xfe2 DR_xo5

r2feo1 s2feo1 r2feo2 r2fefe s2fefe s2feo2 s2feo2 s2fefe s2fefe s2feo2

Parameters for Pb LIII-edge Pb0–O1b Pb0–O1a Pb0–O2 Pb0–Fe1a Pb0–Fe1b Pb0–O5 Pb0–O3 Pb0–Fe2 Pb0–O4

2.35 2.35 2.45 3.11 3.38 3.62 3.75 3.91 4.10

DR_pbo1 DR_pbo1 DR_pbo2 DR_xfe1a+F(DRpby) DR_xfe1b+F(DRpby) DR_xo5+F(DRpby) DR_xo3+F(DRpby) DR_xfe2+F(DRpby) DR_xo4+F(DRpby)

r2pbo1 s2pbo1 r2pbo2 r2pbfe s2pbfe s2pbo2 s2pbo2 s2pbfe s2pbo2

Neighboring atom label

Degeneracy bulk:nano

spectrum 2:2 2:2 2:2 4:3.5 2:1.8 4:3.5 4:1.7 2:1.7 4:3.5

Bold parameters indicate a parameter determined.

Fig. 3.5b). For example, the first six neighboring Fe atoms are split into two groups. A group of two Fe1b atoms in the horizontal direction (along the sheet structure) has an Fe–Fe1b distance of 3.07 A˚, while another group of four Fe1a atoms in the vertical direction (perpendicular to the sheet structure) has an Fe–Fe1a distance of 3.06 A˚. The Fe–Fe1a and Fe–Fe1b distances are essentially the same, with a difference of 0.01 A˚. This is not the case for the distances between the displaced Pb atom and these same neighboring Fe1a and Fe1b atoms. The Pb–Fe1a distance (3.38 A˚) is much longer than the Pb–Fe1b (3.11 A˚), as calculated with DRpby ¼ 0.5 A˚, because the Fe1a group is more perpendicular to the Pb atom, while the Fe1b group is more horizontal to the Pb atom. As shown in Table 3.4, the first neighboring O atoms of the PbO6 moiety were determined independently from the DRpby displacement, because the PbO6 moiety is expected to be larger than the FeO6 moiety by an expansion that is not simply related to DRpby.

84

S. D. Kelly et al.

The final structural model contains 19 parameters, of which 17 are shown in bold type in Table 3.4. The other two parameters are an energy shift for the Pb spectrum and another for the Fe spectrum. The data ranges are 2.5–9.0 and 3.0–9.5 A˚1 for the Fe and Pb spectra, respectively. The fit range is 1.0–4.0 A˚ for both spectra. The data and fit ranges result in a total of 28 independent points (15 and 13 for the Fe and Pb spectra, respectively) for the co-refined Fe and Pb spectra (Stern, 1993). The model was optimized to both spectra, processed with k-weight values of 1, 2, and 3 in the Fourier transform. The magnitude and real part of the Fourier transform of the Fe and Pb spectra and the FeO6 sheet model are shown in Fig. 3.7. The best-fit values for each shell of atoms in the model are given in Table 3.5.

3.5. Discussion and Conclusions This structural study addressed a PbFe coprecipitate formed by the simultaneous removal of Pb2+ and Fe3+ from a homogeneous aqueous solution upon pH titration. The study used both macroscopic and microscopic structural probes (XRD, TEM, and EXAFS). Our EXAFS modeling approach employed information from increasingly complex models to test specific structural models for Pb atoms in the PbFe coprecipitate. The EXAFS analysis was first based on a molecular moiety model, which indicated that the EXAFS spectra for both Pb and Fe were dominated by two O signals and two Fe signals (Table 3.1 and Fig. 3.4). The numbers of coordinating atoms in the two O and two Fe signals (including the uncertainties of the measurements) were similar for both the Fe and Pb spectra. The Pb–O distances were 0.4 A˚ longer than the Fe–O distances, as expected for the larger Pb atom and the smaller Fe atom. A sixfold-coordinated Pb2+ atom has an ionic radius of 1.18 A˚, while a sixfold-coordinated Fe3+ atom has an ionic radius of 0.55 A˚ (Shannon and Prewitt, 1969). The difference between these radii (0.6 A˚) largely accounts for the longer Pb–O distances and the shorter Fe–O distances. The Pb–Fe distances are 0.3–0.2 A˚ longer than the Fe–Fe distances. If the neighboring Fe atoms about Pb and Fe are of the same structure (to be tested below), then the Pb atoms must be displaced as compared to the Fe atoms in the Fe oxyhydroxide structure. The Fe–Fe distribution in the PbFe coprecipitate is shown in Table 3.2 with the distributions for 2LFh, lepidocrocite, and goethite. Our EXAFS result for the number of Fe neighbors is an average of the coordination environments in the spherical and needle-like particles of the PbFe coprecipitate (Fig. 3.2). The average coordination number determined by EXAFS is affected by mixtures of different structures, as well as by particle sizes and


85

1.2

1.2

0.9 0.6

-3

Re(χ(k)k2))(Å )

-3

|χ(k) k 2)|(Å )

0.9

0.6

0.3 0.0 −0.3

0.3 −0.6 −0.9 0.0 (a)

(b)

0.3 0.3

-3

Re(χ(k)k2))(Å )

-3

|χ(k) k 2)|(Å )

0.2

0.2

0.1

0.1 0.0 -0.1 -0.2 -0.3

0.0 0 (c)

2

4 R (Å)

6

8

0 (d)

2

4 R (Å)

6

8

Figure 3.7: The Magnitude and Real Part of the Fourier Transform of the Fe K-Edge Spectrum (a and b) and Pb LIII-Edge Spectrum (c and d) for the PbFe Coprecipitate (Open Symbols) and the Co-Refined FeO6 Sheet Model (Line).

86

S. D. Kelly et al.

Table 3.5: EXAFS Results for the Fe K-Edge and Pb LIII-Edge PbFe Coprecipitate Spectra, illustrating a Consistent Atomic Environment about Fe and Pb in the PbFe Coprecipitate Samplea. R (A˚)

s2 (103 A˚2)

Fe K-edge PbFe coprecipitate spectrum Fe0–O1b 2 Fe0–O1a 2 Fe0–O2 2 Fe0–Fe1a 3.5 Fe0–Fe1b 1.8 Fe0–O3 1.7 Fe0–O4 3.5 Fe0–Fe2 1.7 Fe0–O5 3.5

1.9770.01 2.0070.01 2.2170.02 3.1370.01 2.9970.02 3.4370.05 3.6970.04 3.8970.02 4.0170.04

4.170.8 4.170.8 773 1072 1072 773 773 1072 773

Pb LIII-edge PbFe coprecipitate spectrum Pb0–O1b 2 Pb0–O1a 2 2 Pb0–O2 Pb0–Fe1a 3.5 Pb0–Fe1b 1.8 Pb0–O5 3.5 Pb0–O3 1.7 Pb0–Fe2 1.7 Pb0–O4 3.5

2.3270.02 2.3270.02 2.6370.03 3.3070.02 3.0170.02 3.7870.04 3.4970.05 3.8170.03 3.9570.06

1271 1271 1476 1973 1973 1476 1476 1973 1476

Neighboring atom label

Degeneracy

a

In addition, energy shift values of 671 and 371 eV were determined for the Pb and Fe spectra, respectively.

domain sizes less than a few nanometers. Crystalline structure of nanometer size results in decreased EXAFS coordination number, because the number of surface Fe atoms in a small particle becomes comparable to the number of interior atoms. For example, if half of the Fe atoms are on the surface of a particle and are coordinated by three Fe atoms at 3.0 A˚, while the other half of the Fe atoms are in the interior of the particles and are coordinated by six Fe atoms at 3.0 A˚, the EXAFS coordination number will be the weighted average of 4.5 Fe atoms at 3.0 A˚. A decrease in coordination number can be expected for the PbFe coprecipitate, on the basis of TEM results indicating a domain size of 2–3 nm for the needle-like particles and 5 nm for the spherical particles. The EXAFS coordination number for Fe neighbors at 3.0 A˚ in a 2–3 nm particle of lepidocrocite is 4.7–6.0, whereas bulk lepidocrocite has six Fe–Fe neighbors


87

at 3.0 A˚, and the spherical 2LFh particles have 2.0–3.1 Fe neighbors at 3.0 A˚. Therefore, our PbFe coprecipitate containing a mixture of 2LFh and lepidocrocite (based on XRD; Fig. 3.1) should have a signal for two to six Fe atoms at a distance of 3.0 A˚. Our TEM images were insufficient for estimating the percentage of needlelike and spherical particles. However, our EXAFS results for the Fe–Fe coordination numbers are consistent with our interpretation, with 3.070.9 Fe atoms at a distance of 3.0770.01 A˚ (Table 3.2). The PbFe coprecipitate also has a second Fe shell containing 0.770.5 Fe atoms at a distance of 4.0470.06 A˚. This signal is more similar to the structure of lepidocrocite (with two Fe–Fe neighbors at 3.9 A˚) than to ferrihydrite (with no second-shell Fe atoms) or to goethite (with Fe neighbors at 3.3, 3.5, and 4.6 A˚). Because the distance of the second shell of Fe atoms in the PbFe coprecipitate does not match exactly with the lepidocrocite structure, we suspect that the high dopant ratio of Pb atoms in the PbFe coprecipitate has expanded the structure. The lower coordination number for the second Fe shell is due to the mixture of phases and their small particle size, as discussed above for the first Fe shell. Hence, the EXAFS coordination numbers and distances for the PbFe coprecipitate are consistent with a mixture of lepidocrocite and 2LFh. The Fe K-edge spectrum from the PbFe coprecipitate was modeled with a linear combination of standard spectra from 2LFh and lepidocrocite. The first model (A) is a linear combination of 3873% 2LFh spectrum and 3772% lepidocrocite spectrum. Model A accurately reproduced the EXAFS spectrum for the PbFe coprecipitate throughout the entire data range, k ¼ 0–11 A˚1 (Fig. 3.3a). The total percent of the two spectra in the LCF does not add to 100, indicating that bulk lepidocrocite and 2LFh are not the complete representation of the Fe coordination in the PbFe coprecipitate. Possible explanations include another disordered Fe species (possibly monomeric Fe) and a decrease in crystallinity or particle size of the standards used in the LCF analysis. The standard lepidocrocite spectrum is based on the EXAFS signal for large particles, though XRD of the PbFe coprecipitate showed that the lepidocrocite is not a well-ordered crystal (Fig. 3.1), a finding supported by the small domain size of 2–3 nm for the needle-like particles (Fig. 3.2). As the lepidocrocite particles become smaller and more disordered, the EXAFS signal, like the XRD signal, becomes weaker. The disordered lepidocrocite spectrum is approximated by reducing the bulk lepidocrocite spectrum by 62% of its original intensity. With this assumption, the fraction of lepidocrocite-like phase can be corrected, so that the linear combination of spectra indicates approximately a 40:60

88

S. D. Kelly et al.

mixture of 2LFh and disordered lepidocrocite. The 60% disordered lepidocrocite spectrum contains 0.62 (37% bulk/0.62 ¼ 60% disordered) of the bulk lepidocrocite spectral amplitude of 37%. Another possible LCF model is based on 100% of a disordered lepidocrocite spectrum (50% of the bulk lepidocrocite spectrum). Model B, shown in Fig. 3.3a, reasonably reproduces data at k ¼ 0–11 A˚1. The success of this model indicated that the structure about Fe in the PbFe coprecipitate (spherical and needle-like particles) can be largely accounted for by damping of the lepidocrocite spectrum (i.e., by multiplying the lepidocrocite spectrum by 0.5). Several of the spectroscopic features of the PbFe coprecipitate spectrum are overly pronounced in Model B; examples are the shoulder at 5 A˚1 and the minimum at 6.8 A˚1. These features can be smoothed by allowing the disorder in the signals responsible for the features to be increased. To build a model that allows for additional disorder, the lepidocrocite spectrum needs to be understood in terms of its component signals. An EXAFS model was built for the sheet-like FeO6 structure (Fig. 3.5a) that is the basic unit of lepidocrocite. This model was tested by using the lepidocrocite standard spectrum (Fig. 3.6). The model was applied to both the Pb and Fe EXAFS spectra from the PbFe coprecipitates by allowing the PbO6 unit to be displaced from the sheet structure to account for the longer Pb–Fe distances and shorter Fe–Fe distances found through molecular moiety modeling (Table 3.1 and Fig. 3.4). The Pb displacement from the Fe octahedral sheet was found to be 0.3070.02 A˚. As shown in Table 3.5, the smaller s2 values for Fe neighbors than for Pb neighbors indicate additional static disorder imposed by the large PbO6 unit in the FeO6 structure. The longer distances for the atomic neighbors about Pb are due to displacement of the Pb atom away from the FeO6 sheet, so that the Pb and Fe atoms are modeled with the identical local atomic environment. The excellent agreement of the highly constrained model with both the Fe K-edge spectrum and the Pb LIII-edge EXAFS spectrum indicates that Pb coordination is similar to Fe coordination in the PbFe coprecipitate. Studies investigating the adsorption of Pb2+ have generally concluded that Pb2+ forms surface complexes on Fe oxyhydroxides (Manceau et al., 1992; Bargar et al., 1997; Scheinost et al., 2001; Trivedi et al., 2003). A schematic of a surface-adsorbed Pb atom on a sheet-like FeO6 particle is shown in Fig. 3.8, with two Fe neighbors at a distance of 3.3 A˚ (Trivedi et al., 2003). Figure 3.8 shows many incorporated Pb atoms, to illustrate the difference between an incorporated Pb atom and an adsorbed Pb atom. The incorporated Pb atoms are uniformly distributed throughout the FeO6 particle. The PbFe coprecipitate particle depicted in Fig. 3.8 contains a ratio of Pb to Fe atoms that is similar to our result of 1 Pb atom for every 10 Fe

atoms. This structure has sufficient sites for Pb atoms so that the Pb–Pb distance is greater than 7 A˚; hence, no Pb–Pb signal was detected in the Pb EXAFS spectrum. A Fe–Pb signal was also considered for the model. Each Fe atom is expected to have 0.35 Pb atoms within the first Fe shell, because the Pb to Fe ratio is 1:10, and the first Fe shell contains 3.5 Fe atoms (Table 3.1). Our uncertainties for the Fe K-edge spectra range from 0.5 to 0.9 atoms (Table 3.1), indicating that 0.35 Pb atoms are difficult to detect in our EXAFS measurement. A model with the added Fe–Pb signal did not improve the model, as expected on the basis of the small number (0.35) of Pb atoms in the first Fe shell. The charge imbalance imposed by displacement of a Pb2+ atom for an Fe3+ atom in the sheet-like FeO6 structure could be

90

S. D. Kelly et al.

compensated by the addition of another H+ atom between the FeO6 sheets, as was previously determined for H+ in the lepidocrocite [FeO(OH)] structure (Zhukhlistov, 2001). The EXAFS spectrum of any element in a sample is dominated by the species that is most abundant and most ordered. Trace species (o10%) are difficult to detect, and they become even harder to detect when they are less well ordered (e.g., surface-adsorbed species versus incorporated species). These limitations suggest that some surface-adsorbed Pb species in the PbFe coprecipitate could exist as a trace species, while the majority of Pb atoms are incorporated into the sheet-like FeO6 structure of the coprecipitate. The uptake mechanism of Pb2+ in the Fe oxyhydroxide system is complex. The Fe oxyhydroxides that form and the uptake mechanism of Pb2+ can depend on the contact method (ADS or CPT) and details of the solution chemistry, including pH, ion concentrations, and ionic strength. This study shows compelling evidence for a solid solution of Pb2+ within poorly ordered nanoparticluate Fe oxyhydroxides formed during the CPT of 0.74 mM Pb and 6.4 mM Fe, at pH 5.25 under limited CO2. The difference in the uptake mechanism of an incorporated Pb species under our experimental conditions under CPT conditions rather than the surface-bound Pb species identified under many different ADS conditions (Manceau et al., 1992; Sorensen et al., 2000; Scheinost et al., 2001; Trivedi et al., 2003) may help explain the difference in macroscopic uptake properties of Pb. Additional studies are needed for a direct comparison of changes in the bulk chemical properties of uptake and release of Pb2+ from the Fe oxyhydroxide host material with changes in Pb speciation. Incorporated trace metals such as Pb within Fe oxyhydroxides formed during the coprecipitation process need to be considered for accurate modeling of the geochemical cycling of trace metals in systems such as acid mine drainage, landfill leachate, and pretreatment wastewater processes where trace metals are coprecipitated with Fe oxyhydroxides.

ACKNOWLEDGMENTS This research extended over a long period of time, and we owe gratitude to many for various kinds of help. Chen Zhu acknowledges the support of the National Science Foundation (EAR-0003816, EAR0423971, and EAR0509775) and the US Department of Energy (DOE) (DE-FG26-04NT42125, DE-FG2603NT41806, and DE-FG0204ER63740). We thank Bruce Ravel (Argonne National Laboratory) and Edward O’Loughlin (Argonne) for sample collection


91

and preparation of Fe oxyhydroxide standards. General user support to Shelly Kelly for EXAFS analysis was provided by DOE’s Environmental Remediation Sciences Division (DE-AC02-06CH11357). Portions of this work were performed at CMC (Sector 9), MR-CAT (Sector 10), and GeoSoilEnviroCARS (Sector 13) of the Advanced Photon Source (APS) at Argonne National Laboratory. CMC is supported in part by the DOE Office of Basic Energy Sciences and by the National Science Foundation Division of Materials Research. MR-CAT operations are supported by DOE and the MR-CAT member institutions. GeoSoilEnviroCARS is supported by the National Science Foundation-Earth Sciences (EAR-0217473), DOE-Geosciences (DEFG02-94ER14466), and the State of Illinois. Use of the APS is supported by the DOE Office of Basic Energy Sciences (DE-AC02-06CH11357). Comments from Douglas B. Kent and three anonymous reviewers greatly improved the quality and presentation of the manuscript.

REFERENCES Bargar, J. R., Brown, G. E., & Parks, G. A. (1997). Surface complexation of Pb(II) at oxide–water interfaces. 1. XAFS and bond-valence determination of mononuclear and polynuclear Pb(II) sorption products on aluminum oxides. Geochim. Cosmochim. Acta, 61, 2617–2637. Charlet, L., & Manceau, A. A. (1992). X-ray absorption spectroscopic study of the sorption of Cr(III) at the oxide–water interface. 2. Adsorption, coprecipitation, and surface precipitation on hydrous ferric-oxide. J. Colloid Interface Sci., 148, 443–458. Cornell, R. M., & Schwertmann, U. (1996). The Iron Oxides: Structures, Properties, Reactions, Occurrence and Uses. VCH, Weinheim, Germany. Crawford, R. J., Harding, I. H., & Mainwaring, D. E. (1993). Adsorption and coprecipitation of single heavy-metal ions onto the hydrated oxides of iron and chromium. Langmuir, 9, 3050–3056. Dyer, J. A., Trivedi, P., Scrivner, N. C., & Sparks, D. L. (2003). Lead sorption onto ferrihydrite. 2. Surface complexation modeling. Environ. Sci. Technol., 37, 915–922. Dzombak, D. A., & Morel, F. M. M. (1990). Surface Complexation Modeling: Hydrous Ferric Oxide. Wiley-Interscience, New York. Jambor, J. L., & Dutrizac, J. E. (1998). Occurrence and constitution of natural and synthetic ferrihydrite, a widespread iron oxyhydroxide. Chem. Rev., 98, 2549–2585. Janney, D. E., Cowley, J. M., & Buseck, P. R. (2000). Structure of synthetic 2-line ferrihydrite by electron nanodiffraction. Am. Mineral., 85, 1180–1187.

92

S. D. Kelly et al.

Karthikeyan, K. G., Elliott, H. A., & Cannon, F. S. (1997). Adsorption and coprecipitation of copper with the hydrous oxides of iron and aluminum. Environ. Sci. Technol., 31, 2721–2725. Karthikeyan, K. G., Elliott, H. A., & Chorover, J. (1999). Role of surface precipitation in copper sorption by the hydrous oxides of iron and aluminum. J. Colloid Interface Sci., 209, 72–78. Lottermoser, B. G., & Ashley, P. M. (2006). Physical dispersion of radioactive mine waste at the rehabilitated Radium Hill uranium mine site, South Australia. Aust. J. Earth Sci., 53, 485–499. Manceau, A., Charlet, L., Boisset, M. C., Didier, B., & Spadini, L. (1992). Sorption and speciation of heavy metals on hydrous Fe and Mn oxides. From microscopic to macroscopic. Appl. Clay Sci., 7, 201–223. Newville, M. (2001). IFEFFIT: Interactive EXAFS analysis and FEFF fitting. J. Synchrotron Radiat., 8, 322–324. Newville, M., Livinsˇ , P., Yacoby, Y., Rehr, J. J., & Stern, E. A. (1993). Near-edge X-ray absorption fine structure of Pb: A comparison of theory and experiment. Phys. Rev. B, 47, 14126–14131. Newville, M., Sutton, S., Rivers, M., & Eng, P. (1999). Micro-beam X-ray absorption and fluorescence spectroscopies at GSECARS: APS beamline 131D. J. Synchrotron Radiat., 6, 353–355. Ravel, B., & Newville, M. (2005). ATHENA, ARTEMIS, HEPHAESTUS: Data analysis for X-ray absorption spectroscopy using IFEFFIT. J. Synchrotron Radiat., 12, 537–541. Scheinost, A. C., Abend, S., Pandya, K. I., & Sparks, D. L. (2001). Kinetic controls on Cu and Pb sorption by ferrihydrite. Environ. Sci. Technol., 35, 1090–1096. Segre, C. U., Leyarovska, N. E., Chapman, L. D., Lavender, W. M., Plag, P. W., King, A. S., Kropf, A. J., Bunker, B. A., Kemner, K. M., Dutta, P., Druan, R. S., & Kaduk, J. (2000). The MRCAT insertion device beamline at the advanced photon source. Synchrotron Radiat. Inst. CP, 521, 419–422. Shannon, R. D., & Prewitt, C. T. (1969). Effective ionic radii in oxides and fluorides. Acta Crystallogr. B, 25, 925. Sorensen, M. A., Stackpoole, M. M., Frenkel, A. I., Bordia, R. K., Korshin, G. V., & Christensen, T. H. (2000). Aging of iron (hydr)oxides by heat treatment and effects on heavy metal binding. Environ. Sci. Technol., 34, 3991–4000. Stern, E. A. (1993). Number of relevant independent points in X-ray-absorption fine-structure spectra. Phys. Rev. B, 48, 9825–9827. Stern, E. A., Newville, M., Ravel, B., Yacoby, Y., & Haskel, D. (1995). The UWXAFS analysis package: Philosophy and details. Physica B, 208–209, 117–120. Thoral, S., Rose, J., Garnier, J. M., Van Geen, A., Refait, P., Traverse, A., Fonda, E., Nahon, D., & Bottero, J. Y. (2005). XAS Study of iron and arsenic speciation during Fe(II) oxidation in the presence of As(III). Environ. Sci. Technol., 39, 9478–9485.


93

Trivedi, P., Dyer, J. A., & Sparks, D. L. (2003). Lead sorption onto ferrihydrite. 1. A macroscopic and spectroscopic assessment. Environ. Sci. Technol., 37, 908–914. Waychunas, G. A., Rea, B. A., Fuller, C. C., & Davis, J. A. (1993). Surfacechemistry of ferrihydrite. 1. EXAFS Studies of the geometry of coprecipitated and adsorbed arsenate. Geochim. Cosmochim. Acta, 57, 2251–2269. Zabinsky, S. I., Rehr, J. J., Ankudinov, A., Albers, R. C., & Eller, M. J. (1995). Multiple-scattering calculations of X-ray-absorption spectra. Phys. Rev. B, 52, 2995–3009. Zhao, J. M., Huggins, F. E., Feng, Z., & Huffman, G. P. (1994). Ferrihydrite – surface-structure and its effects on phase-transformation. Clays Clay Min., 42, 737–746. Zhukhlistov, A. P. (2001). Crystal structure of lepidocrocite FeO(OH) from the electron-diffractometry data. Crystallogr. Rep., 46, 730–733.


Chapter 4

Tracking the Interaction of Transition Metal Ions with Environmental Interfaces using Second Harmonic Generation Christopher T. Konek, Michael J. Musorrafiti, Andrea B. Voges and Franz M. Geiger Department of Chemistry and Institute for Environmental Catalysis, Northwestern University, 2145 Sheridan Road, Evanston, IL 60208, USA

ABSTRACT Surfaces and interfaces are ubiquitous in the environment. Nonlinear optical (NLO) techniques, such as second harmonic and sum frequency generation (SHG and SFG, respectively), allow for highly sensitive surface measurements under environmentally relevant temperature and solute concentration conditions. In this work, we present three experimental studies that demonstrate how NLO methods can be used to investigate the interaction of manganese and chromate ions with environmental interfaces. The environmental implications of this work are discussed in the context of understanding heterogeneous processes in geomedia.

4.1. Introduction Surfaces and interfaces are ubiquitous in the environment (Stumm and Morgan, 1996; Finlayson-Pitts and Pitts, 2000), and range from liquid/ solid interfaces in soil to solid/gas interfaces associated with atmospheric particulate matter. Many environmental processes are dominated by events occurring uniquely at interfaces, such as processes that control the bioavailability of metals, transport and cycling of environmentally important species, and the fate of environmental pollutants. It is therefore critical Corresponding author. Tel. 847-467-6553; Fax: 847-491-7713;

E-mail: [email protected] (F.M. Geiger).

96

C. T. Konek et al.

to understand, on a molecular level, what controls the interaction of molecules and ions with environmental interfaces. Recent advances in surface spectroscopy have led to the development of a number of very useful techniques for studying interfaces (Chung, 2001). These techniques often use X-rays (EXAFS, XANES) (Martinez et al., 1996; Brown and Parks, 2001; Sparks, 2001; Carlson et al., 2003; Bargar et al., 2004) or electrons (LEED, HREELS) (Shaikhutdinov et al., 2000; Nishiyama et al., 2006) to identify structural and binding information, or photons (ellipsometry, interferometry, ATR-FTIR) (Dubowski et al., 2004; Gershevitz and Sukenik, 2004; Usher et al., 2004) to spectroscopically probe the interface. Additionally, imaging techniques such as AFM and TEM (Chung, 2001; Dubowski et al., 2004; Grassian, 2005) have been used to characterize interfaces and obtain kinetic information. A thorough review of surface science techniques is beyond the scope of this chapter and we refer the reader to the appropriate textbooks (Somorjai, 1994; Masel, 1996; Chung, 2001). It is important to note that currently applied spectroscopic techniques are significantly curtailed by their inability to probe aqueous/solid interfaces in real time while they are exposed to environmentally representative solute concentrations. Furthermore, limitations may also include the need for expensive substrates or extensive chemical and physical models for data interpretation. However, continuing progress is being made toward bridging the ‘‘environmental gap’’ – the gap between the capabilities of standard surface analysis techniques and environmental conditions (Brown et al., 1999). Nonlinear optical (NLO) techniques (Shen, 1984; Heinz, 1991; Boyd, 1992; Voges et al., 2005), such as second harmonic generation (SHG) and sum frequency generation (SFG), present unique advantages for studying environmental systems including: high sensitivity for studying environmentally relevant concentrations, surface specificity to enable data collection at monolayer and submonolayer surface coverages without the necessity of bulk signal subtraction, and the ability to study buried interfaces between a wide variety of materials, including insulators. NLO techniques are based on frequency upconversion of visible or infrared light that occurs in noncentrosymmetric media, such as a surface or interface (Shen, 1984; Heinz, 1991; Boyd, 1992), provided that the applied optical fields are intense enough. These optical methods require that at least one of the bulk phases be optically transparent. Most NLO experiments are carried out at the interface of two centrosymmetric media, such as silica and water or single crystalline aluminum oxide and water (Higgins et al., 1998). The reader is referred to a number of excellent recent reviews regarding the application of NLO methods to environmental systems (Allen et al., 2000; Richmond, 2002; Shultz et al., 2002; Eisenthal, 2006; Gopalakrishnan et al., 2006; Shen and Ostroverkhov, 2006). In addition,

Interaction of Transition Metal Ions with Environmental Interfaces

97

we recently reviewed the field with a specific focus on applications of NLO methods for studying heterogeneous systems relevant to the natural environment (Voges et al., 2005). In general, SHG is applicable to real environment and soil samples provided one uses either large, flat, optically reflective substrates or particles suspended in solutions (necessitating a hyper-Raleigh correction, as demonstrated by the Eisenthal group in the mid-1990s (Wang et al., 1998)). The limitations of NLO techniques for studying environmental interfaces are discussed in our 2005 review (Voges et al., 2005). In this work, we present three studies that demonstrate how NLO methods can be used to study the interaction of metal ions with environmental interfaces, with direct implications for heterogeneous processes in geomedia. We begin by reviewing our off-resonance SHG work on silica/water interfaces functionalized with carboxylic acid groups (Konek et al., 2004), which are environmentally important organic moieties in many soil environments containing humic and fulvic acids (Evangelou, 1998; Sutton and Sposito, 2005). We then discuss how the interaction of a cation, namely manganese(II), with carboxylic acid-functionalized silica/water interfaces can be tracked using off-resonance SHG measurements. Understanding manganese mobility in the environment, and how this mobility is controlled by heterogeneous geomedia, is important for assessing the bioavailability of metals and manganese cycling in the environment (Tebo et al., 2004), including acid mine drainage (Scott et al., 2002). Next, we present how ligand-to-metal charge transfer resonantly enhanced SHG measurements can be used to track an anion, namely the EPA priority pollutant chromate (Mifflin et al., 2003a,b) at an aluminum oxide/water interface. We conclude by discussing the environmental implications of this work.

4.2. Experimental 4.2.1. Laser System and Signal Detection SHG experiments are performed using a 120 fs, 1 kHz, regeneratively amplified Ti:sapphire laser system (Hurricane, Spectra Physics) that pumps optical parametric amplifiers (OPA-800CF, Spectra Physics) producing continuously tunable laser light (Mifflin et al., 2003b; Voges et al., 2004). Nonresonant w(3) studies of the carboxylic acid-functionalized silica/water interfaces (with no transition metal ions adsorbed) were carried out using 580 nm probe light. Nonresonant w(3) measurements for tracking the interaction of Mn2+ with the carboxylic acid-functionalized silica/water

98

C. T. Konek et al.

interfaces were carried out using fundamental probe light of 620 nm. Finally, the interaction of chromate with single crystal aluminum oxide was studied via ligand-to-metal charge transfer resonantly enhanced SHG measurements carried out using fundamental probe light of 580 nm.The appropriate power dependencies and signal bandwidth for SHG were verified regularly, and sample damage is avoided by attenuating the laser power of the incident light fields using variable neutral density filters to ensure the input powers are below the optical breakdown threshold. In the SHG experiments presented in this work, the fundamental probe light approaches the silica lens at an angle of 601 and at an angle of 451 for the aluminum oxide lens. These incident angles are near the critical angle of total internal reflection for the silica/water interface and the aluminum oxide/ water interface, respectively. A half-wave plate rotates the fundamental probe light field polarization such that it is plane-polarized perpendicular to the interface (p-in). After the second harmonic (SH) signal is produced at the interface, it is separated by the appropriate optical filters from the reflected fundamental probe light field and guided through a monochromator into a low-noise Hamamatsu photon multiplier tube (0.5 dark counts/s). After signal amplification, the number of SHG photons produced at the interface is counted with a Stanford Research Systems gated photon counter. Typical signal intensities range from 2 to 150 counts/s. 4.2.2. Sample Cell and Flow System For the SHG experiments at the alumina/water interface, the flat side of an aluminum oxide (ISP Optics) hemispherical lens ([1 0 0] surface) is used as a substrate. After cleaning the lens by sonication in methanol (6 min), the lens is dried in an oven and subsequently placed in a plasma cleaner (Harrick PDC-32G) for 30 s at 1,000 mTorr air. For the SHG experiments on a carboxylic acid terminated adlayer, the flat side of a silica (ISP Optics) hemispherical lens is used as a substrate for the deposition of the organic adlayer. After cleaning by sonication in methanol (6 min), and drying in a stream of dry nitrogen, the lens is placed in a plasma cleaner (Harrick PDC-32G) for 30 s at 1,000 mTorr air. Upon removal from the plasma cleaner, the lens is promptly placed in a Teflon reaction vessel and covered until silane deposition. Surface functionalization of the silica substrate occurs in an inert-atmosphere glovebox by exposing the flat side of the hemispherical lens to a solution of 11-(trichlorosilyl)-undecanoic acid methyl ester in dried toluene (1%, v/v) for 1 h. After washing the lens with copious amounts of dried toluene and subsequent removal of the lens from the


99

Figure 4.1: Experimental Setup. The Graphic Shows the Experimental Setup for the Second Harmonic Generation Flow Experiments. See Text for Details. glovebox, it is annealed in an oven at 1101C for 1 h to promote cross-linking of the silane groups. To hydrolyze the methyl ester terminated end of the adlayer and create a carboxylic acid-functionalized surface, the lens is placed in a 2.4 M HCl solution for 2 h at 851C. The experimental SHG setup used in the flow experiments is shown in Fig. 4.1, and is similar to the one previously reported (Mifflin et al., 2003b). A dual-pump approach is used to quickly switch between the flow of an aqueous species of interest and water. The aqueous solution is pumped into a custom-built Teflon flow cell capped with the hemispherical lens (either aluminum oxide or functionalized silicon oxide) held tightly against a Viton O-ring. A UV–vis flow cell positioned downstream from the sample cell allows for in-line detection of the bulk concentration of the species of interest. The aqueous solution then flows into a waste container, where the tube connection can be removed to take aliquots of the solution for off-line concentration measurements or to determine the conductivity of the solution.

4.3. Surface Characterization A variety of surface analysis techniques were used to characterize the organic adlayers and to verify a high degree of surface functionalization and a high

100

C. T. Konek et al.

degree of conversion from the methyl ester to carboxylic acid. Direct characterization of the hemispherical lenses was performed when possible, and on substrates with similar surface chemistry (glass slides or silicon wafers with a native oxide layer) in situations where the hemispherical lens was either too bulky for the instrument or the technique required different optical properties (Al-Abadleh et al., 2004). 4.3.1. Contact Angle Measurements As a qualitative measure of hydrophobicity or hydrophilicity, contact angle measurements were used to characterize the surface of the silica lens after preparation of the adlayer. Contact angle measurements (Tantec CamMicro) sample only the exposed surface. Since the contact angle varies with hydrophobicity, it is an appropriate method for distinguishing between the surfaces discussed in this work. Twelve measurements for each surface were averaged resulting in static contact angles of 32(4)1 for clean silica lenses, 67(6)1 for methyl ester siloxanes, and 46(4)1 for carboxylic acid siloxanes (the values shown in parenthesis indicate one standard deviation). These measurements are consistent with an increase in hydrophobicity from the silica lenses to the methyl ester adlayers, and with a subsequent increase in hydrophilicity upon conversion to the carboxylic acid adlayer. The contact angles agree well with literature data on ester- and acid-terminated siloxanes (Wasserman et al., 1989). 4.3.2. Atomic Force Microscopy In the AFM studies, the ester-silane deposition was carried out on a silicon wafer with a native oxide layer. The images were taken on a Nanoscope IIIa (Digital Instruments, Santa Barbara, CA). Phase mode AFM images indicate a high degree of surface functionalization. Figure 4.2A shows that the degree of surface functionalization approaches 95%. A line scan (Fig. 4.2B) across a hole in the adlayer indicates an adlayer height of 10 A˚, which is consistent with results from ellipsometry measurements presented in the following section. 4.3.3. Ellipsometry To further investigate the adlayer thickness, we carried out ellipsometry measurements for two different ester-functionalized silicon wafers with a


101

Figure 4.2: Atomic Force Microscopy Measurements. (A) AFM Image of a Methyl Ester Terminated Surface. The Sample Area is 100 nm 100 nm. (B) A Line Scan showing the Adlayer has a Height of 10 A˚.

native oxide layer rather than perform the measurements on silica to ensure a significant difference in refractive index between the substrate and the organic adlayer. We averaged seven data points taken at different locations on each sample with two different probe wavelengths (544 and 632 nm) and used a refractive index of 1.55 for the organic adlayer, which is a typical bulk value for many organic compounds (Lide, 1996). Ellipsometry showed the thickness of the organic adlayer to be 10–13 A˚, which is consistent with the

102

C. T. Konek et al.

presence of a slightly tilted 10-carbon atom long chain at the interface (Wasserman et al., 1989).

4.3.4. Time of Flight-Secondary Ion Mass Spectrometry (ToF-SIMS) A ToF-SIMS (PHI TRIFT III, Physical Electronics) study was performed to investigate adlayer coverage and extent of conversion on a glass slide using Ga atoms at 25 keV. To track the ester group, a trichlorosilane molecule was synthesized that contained fluorine atoms instead of hydrogen atoms on the terminal methyl ester group (Voges et al., 2004). The ester coverage is tracked by imaging fluorine when the trifluoro methyl ester is deposited. After the hydrolysis of the ester, a high degree of conversion implies that very few fluorine atoms would appear on an image of the surface. These data are shown in Fig. 4.3 and are consistent with both a high degree of surface functionalization and an efficient conversion of the ester group to a carboxylic acid group.

4.3.5. Vibrational Sum Frequency Generation To further investigate the extent of ester conversion to the acid during hydrolysis, vibrational SFG studies were carried out using broadband infrared radiation (140 cm1 bandwidth at 3 mm) with signal upconversion at the sample surface using a picosecond visible (800 nm) pump beam in external reflection (Voges et al., 2004). The SFG experiments are carried out on functionalized glass microscope slides and probe infrared modes containing a component perpendicular to the interface by using the appropriate light field polarizations. Details of our SFG setup are published elsewhere (Voges et al., 2004, 2005). Vibrational SFG measurements, shown in Fig. 4.4 (Voges et al., 2004), demonstrate that the conversion from the methyl ester to the acid is high. The SFG spectrum of the methyl ester siloxane shows the methylene symmetric (2,850 cm1) and asymmetric stretches (2,930 cm1) of the carbon chain in the methyl ester as well as the CH3 symmetric stretch of the methoxy group (R–O–CH3) at 2,975 cm1. After hydrolysis of the ester, the band at 2,975 cm1 is absent in the vibrational SFG spectrum, and the disappearance of this methoxy CH stretch indicates that the hydrolysis of the methyl ester is highly efficient. This is consistent with the ToF-SIMS experiments presented in the previous section.


103

Figure 4.3: Secondary Ion Mass Spectrometry Mapping Results of a Fluorinated Ester Tracking the Presence of Fluorine. Lighter Areas Represent High Intensity Signals from Fluorine; Darker Areas are Associated with Low Fluorine Signal Intensity. Image (A) shows the SIMS Map of the Fluoride Ester Deposited on the Surface; Image (B) shows the SIMS Map of the Surface after the Hydrolysis of the Ester Group into a Carboxylic Acid. The Scale Bar represents 100 mm.

104

C. T. Konek et al.

2975 cm-1 CH3 symmetric stretch of methoxy group

ISFG [a.u.]

Methyl Ester

3100

Carboxylic Acid

3000

2900

2800

2700

Infrared Wavelength [cm-1]

Figure 4.4: Vibrational Sum Frequency Generation Spectra of a Methyl Ester Functionalized Glass Slide Before (Top) and After (Bottom) Hydrolysis. Note the CH3 Symmetric Stretch at 2,975 cm1 that is Present in the Spectrum for the Methyl Ester Terminated Surface but Absent after the Hydrolysis Reaction.


105

4.4. Results 4.4.1. v(3) Measurements Before studying the interaction of the manganese and chromate ions with the mineral/water interfaces under investigation in this work, we employed SHG measurements that are based on surface titration and charge screening experiments to study the organic adlayers at the silica/water interface. We follow the w(3) technique previously described by Shen and coworkers (Xiao et al., 1989) and Eisenthal and coworkers (Xiao et al., 1991, 1997; Ong et al., 1992). Two kinds of experiments were performed – charge screening experiments in which the bulk ion concentration is varied at constant pH and surface titration studies in which the pH of the bulk water flowing across the interface is changed while maintaining constant ionic strength. The first set of experiments yields quantitative information on the interfacial charge densities, and the second set of experiments yields information on the interfacial acid–base equilibria. In the presence of a static interfacial electric field, F, originating from charged interfacial species, NLO signals at 2o are produced from the interaction of the two incident probe fields at frequency o with the static electric field, F, via the third-order nonlinear susceptibility w(3). This ‘‘w(3) technique,’’ as it has been termed by the pioneering team of Eisenthal and coworkers (Ong et al., 1992; Zhao et al., 1993; Wang et al., 2000), yields the interfacial potential directly according to ~2o / ~ ~o E ~o ~ ~o E ~o F0 ¼ A BF0 E wð2Þ E wð3Þ E

(4.1)

Here, w(2) and w(3) are the second-order and third-order nonlinear susceptibilities of the interface in the presence of the applied fundamental E-field, Eo, and A and B are treated as constants. Surface potentials of metal surfaces (Bradley et al., 1993; Hewitt et al., 1993), mineral oxide/water interfaces (Ong et al., 1992; Lantz et al., 1993; Fitts et al., 2002; Higgins et al., 2002) including colloids (Eisenthal, 2006), and charged molecular species at aqueous surfaces have been determined in this fashion (Zhao et al., 1990; Eisenthal, 1996).

4.4.1.1. Charge Screening At constant bulk solution pH, the interfacial potential can be expressed using the Gouy–Chapman model (Adamson, 1990). Here, the surface charge

106

C. T. Konek et al.

density is related to two parameters: the bulk ion concentration, C, and the static potential, F, at the interface. This relation is shown in the following equation: rffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2kT p F¼ arcsin h s (4.2) ze 2kTC where k is the Boltzmann constant, T the temperature, z the valence of the counter ion used for charge screening, e the elemental charge, s the surface charge density, and e the permittivity of bulk water. By recording the SH E-field signal intensity, which depends directly on F (see Eq. (4.1)), with increasing bulk electrolyte concentration at constant bulk solution pH, Eq. (4.2) can be fit to the data using s as a fitting parameter. Before beginning the charge screening experiments, the entire flow system was purged with Millipore water for 2 h to ensure a low initial ion concentration and a clean flow system. The ion concentration was adjusted in the Erlenmeyer flask by gradually adding previously prepared NaCl solutions of increasing concentration. The ion concentration was monitored downstream from the Teflon flow cell using a calibrated conductivity meter (VWR Expanded Range Conductivity Meter). Figure 4.5 shows that as the bulk ion concentration increases the SHG E-field produced by the negatively charged carboxylic acid groups decreases due to charge screening. The data were fit using Eq. (4.1) and Gouy-Chapman model to obtain a charge density of (2.8 104)7(6 105) C/m2 at pH 6.4 and a charge density of (4.2 102)7(2 103) C/m2 at pH 11.2 (data not shown). These surface charge density experiments indicate that at neutral bulk solution pH, neutral (protonated) carboxylic acid groups exceed the number of ionized (deprotonated) carboxylate groups by two orders of magnitude.

4.4.1.2. Surface Acid–Base Titration With the knowledge of the surface charge densities at pH 6.4 and 11.2, we were now able to study the acid-base chemistry of the carboxylic acid-functionalized silica/water interface. During these surface titration experiments, the bulk solution pH is adjusted in an Erlenmeyer flask with stock solutions made from HCl (Fisher) and NaOH (Fisher). All solutions used for the surface titration experiment contain 0.5 M NaCl (Fisher and VWR) to avoid a significant variation in total ion concentration. The pH is monitored by a pH meter (Thermo Orion 3 Star). Alkaline conditions are known to gradually degrade siloxane adlayers on a timescale of 80 min (Wasserman et al.,


107

10

ESHG [a. u.]

8

6

4

0.0

0.2

0.4 0.6 Ion Concentration [M]

0.8

1.0x10-3

Figure 4.5: Second Harmonic Signal as a Function of Bulk Ion Concentration in Solution Flowing across the Carboxylic Acid-Functionalized Silica/ Water Interface at pH 6.5. The Solid Line is a Fit to the Data based on the Gouy-Chapman Model. The Error Bar is Representative of the Uncertainty in all Data Points and is taken from the Standard Deviation of 300 Data Points. 1989). A control experiment (data not shown) demonstrated that after 100 min of exposure to an aqueous solution held at pH 11, the SH signal intensity dropped and then began to fluctuate in an irregular fashion, which may be caused by a degradation in the sample. Therefore, during each run, exposure times exceeding 80 min were avoided for pH values above 7.0. For the w(3) analysis (Xiao et al., 1989; Ong et al., 1992; Konek et al., 2004), the SHG intensity was recorded for 5 min at a given bulk solution pH. The SHG E-field will track the degree of interfacial deprotonation as the pH of the bulk solution flowing across the interface is changed. Figure 4.6 (Konek et al., 2004) shows the results of these measurements. The data show two inflection points in the surface titration curve, which is consistent with the presence of two acid-base equilibria for the monoprotic carboxylic acid

108

C. T. Konek et al.

1.2

1.0

ESHG [a. u.]

0.8

0.6

0.4

0.2

0.0 0

2

4

6

8

10

12

14

Bulk pH

Figure 4.6: Second Harmonic E-Field from Carboxylic Acid Functionalized Fused Quartz as a Function of the Bulk Solution pH Flowing across the Interface. The Fit to the Data is shown in Black Lines. Data Shown are Averages of Three Separate Runs, and the Error Bars show One Standard Deviation in the Raw Data, Propagated through the Data Analysis. See Text for Details. tethered to the surface. In order to interpret this data, and to obtain absolute pKa values for the two acid-base equilibria, charge screening experiments had to be carried out at constant bulk solution pH. Following Eisenthal and coworkers (Ong et al., 1992), we used the interfacial charge densities determined in the previous section in the following equation: 0 1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 2 B ðf smax p=2CkT Þ þ 1 þ ðf smax p=2CkT Þ C B C pH ¼ pK a þ logB f C @ A 1f (4.3a)


f ¼

Low E 2o E pH 2o High Low E pH E pH 2o 2o

109

! (4.3b)

Here, smax is the maximum surface charge density obtained by a given pH transition, C the solution bulk ion concentration, e the bulk water dielectric High Low constant, and E pH and E pH the lowest and highest electric field values 2o 2o for a given pH transition, respectively. For z ¼ 1, the interfacial charge density is given by s ¼ f smax, where f is related to the fraction of deprotonated carboxylic acid groups at the interface for each given acid-base equilibrium. Fitting Eq. (4.3a) to the data shown in Fig. 4.6 results in pKa values of 5.6(2) and 9(1). These results are consistent with recent ATR-FTIR studies (Gershevitz and Sukenik, 2004). The two acid-base equilibria can be interpreted as follows: carboxylic acid groups at defect sites such as grain boundaries or adlayer holes, where the ionizable carboxyl groups are solvated by water similarly to what they would encounter in bulk water, are associated with the lower pKa value (5.6). This lower value is close to the pKa values of monoprotic carboxylic acid groups in dilute aqueous solutions (pKa 4–5 (Ege, 1999)). The higher pKa value (9) is consistent with the presence of lateral hydrogen-bonding networks within the interfacial plane. Deprotonation of these carboxyl groups is energetically unfavorable (Konek et al., 2004) and the carboxylic acid groups remain protonated, and thus neutral, until the aqueous phase is sufficiently alkaline. The idea of different conditions responsible for protonated and unprotonated carboxylic acid groups is reminiscent of the observation that glacial acetic acid, which is highly concentrated, does not act as an efficient conductor unless water is present to disrupt the strong hydrogen-bonding interactions among the carboxylic acid groups (Jacobsen et al., 2001). By combining the maximum charge density, smax, and the fraction of carboxylic acid groups deprotonated at each pH, which was necessary to fit the data shown in Fig. 4.6, we obtained the surface charge density for every bulk solution pH value. From the surface charge densities and assuming one negative charge for each carboxylate group, we calculated the absolute number of deprotonated carboxylic acid groups for every bulk solution pH value (Fig. 4.7). Figure 4.7 shows that there are at least 3 1013 carboxylate groups per cm2 at the interface at pH 12. We have shown in our earlier work that this negative charge density is associated with a fairly high interfacial free energy density of 100 nJ/cm2 (Konek et al., 2004). Clearly, at neutral bulk solution pH conditions, the interface is mainly neutral, i.e., most of the carboxylic acid groups are present in the molecular form. The sensitivity limit of the w(3) method is demonstrated by the scatter in the y-axis data points that are taken below pH 5.

110

C. T. Konek et al.

1013

NCOO-[cm-2]

1012

1011

1010

109 0

2

4

6

8

10

12

14

Bulk pH

Figure 4.7: Absolute Number of Deprotonated Carboxylic Acid Groups as a Function of Bulk Solution pH. Note the Logarithmic Scale on the Y-Axis. See Text for Details. 4.4.1.3. Uncertainty in the pKa Values To assess the uncertainties in the two pKa values, we carried out a sensitivity analysis with respect to the uncertainties in the ion concentration. We also paid attention to uncertainties in the interfacial charge density and photon shot noise. The ion concentrations may vary by up to 10% due to the pH adjustments. The sensitivity analysis was carried out by generating a model data set of 10 evenly spaced E-field values and then calculating the corresponding pH of the solution from Eq. (4.2). All the parameters used to calculate this model data set were based on the experimental data and laboratory conditions. While varying the ion concentration used in the fits by 710% to simulate that uncertainty, Eqs. (4.3a) and (4.3b) was fit to the model data set with surface charge density inputs that varied by up to 715% to simulate possible uncertainties in the surface charge densities. While fitting Eq. (4.2) to the model data sets, the surface charge densities were allowed to


111

vary by 720%. This is analogous to the one standard deviation used in fitting our experimental data. To simulate photon shot noise, a number of E-field vs. pH model data sets were calculated with data points that were shifted through the use of a random number generator with a normally distributed output. The experimental SHG data discussed in the previous section exhibit a larger relative error for lower pH conditions than for higher pH conditions. This experimental observation was mirrored in the shifted model data sets by modulating the standard deviation of the random number generator (number period in excess of 1018 (WaveMetrics, Inc., 2003)) for two regions within a given pKa transition. For each pKa value, this standard deviation was set to 5.5% of the base electric field value for the lower pH region of the model data set and to 4% for the higher pH region. These percent values in the standard deviations were empirically determined to result in model data sets that allowed for straightforward fit convergence. Fitting Eq. (4.2) to the model data sets allowed us to analyze how the experimental uncertainties in the pKa values depended on the uncertainties in the ion concentration in the bulk solution, the interfacial charge density, and photon shot noise. To compare uncertainty from experimental parameters with uncertainty inherent in the experimental setup (laser shot noise), simulations were run, and are summarized in Fig. 4.8. In Fig. 4.8, the dashed lines indicate the most extreme uncertainties in the pKa values when the photon shot noise was turned on, while the filled symbols are the sensitivity analysis results in the absence of photon shot noise. Clearly, Fig. 4.8 shows that the largest contributor to the uncertainty in the lower pKa value (left three panels) is the photon shot noise, and not the uncertainties in the ion concentration or the interfacial charge density. The same is true for the higher pKa value (right three panels).

4.5. Ion Binding 4.5.1. Manganese Binding Now that it is clear that the charge state of the carboxylic acid-functionalized silica/water interfaces under investigation is mainly neutral near pH 7, we proceeded to expose this interface to divalent manganese ions from solution. All manganese experiments were performed with MnSO4 dissolved in Millipore water, and subsequently adjusted to the appropriate pH. Dilute basic solutions were used to adjust the bulk solution pH to ensure that local

112

C. T. Konek et al.

pKa=9 3

2

2

1

1

0

0

-1

-1

-2

-2

-3 3

-3 3

2

2

1

1

0

0

-1

-1

-2

-2

-3 3

-3 3

2

2

1

1

0

0

-1

-1

-2

-2

-3

Error in pKa Value [%]

Error in pKa Value [%]

pKa=5.6 3

-3 -10

-5

0

5

10

-10

-5

0

5

10

Error in Ion Concentration [%]

Figure 4.8: Error Analysis Regarding the Uncertainties of the Two pKa Values as a Function of Error in the Ion Concentration and for Various Noise Patterns in the Single Photon Counting Experiments. Percent Difference is Calculated from the Difference in Data Generation Value and Fit Value Divided by the Data Generation Value Multiplied by 100.

regions of high pH did not cause the formation of manganese oxide. Additionally, UV-vis spectra show no manganese carbonate solid formation over the timescale of the experiment. If manganese carbonate did form, it is likely that it would lead to SHG resonance enhancement, i.e., signal increases, as opposed to the observed w(3) effect that leads to a signal


113

decrease. The manganese concentration was determined using an ICP-AES with a Varian ICP Spectrometer. The experiments performed probed the association of the aqueous dissolved manganous ion with the carboxylic acid adlayer. Figure 4.9 shows a decrease in the SHG E-field as manganese ions at a bulk solution concentration of 1.2 mM are passed across the interface at pH 7. This E-field decrease is consistent with a w(3) effect resulting from manganese cations binding to the carboxylic acid groups at the interface, which should lead to a decrease in the (negative) interfacial charge density and thus a smaller SHG E-field. Figure 4.9 shows that when the manganese flow is stopped and the water flow turned back on, the SHG E-field increases again to the original level. This is consistent with the notion that the interaction between the Mn(II) ion and the surface is fully reversible and suggests that hydrogen bonding is controlling the interaction. Thus, it is

Mn2+

H2O

H2O

Mn2+

H2O

1

ESHG [a. u.]

0

-1

-2

-3 0

200

400

600 800 1000 Time [seconds]

1200

1400

Figure 4.9: SHG E-Field as a Function of Time for a Carboxylic AcidFunctionalized Fused Quartz/Water Interface Exposed to Manganese(II) Solution at pH 7. The Horizontal Bars show the Time Periods for Manganese and Water Flow. See Text for Details.

114

C. T. Konek et al.

likely that the Mn(II) ion binding to the surface is driven by physisorption mediated by the solvation sphere of the Mn(II) ion. To quantify the thermodynamics of the interaction of manganese with the interfacial carboxylic acid groups, the dependence of the SH E-field on manganese concentration was mapped out (Fig. 4.10) by gradually increasing the concentration of manganese flowing across the interface while maintaining constant electrolyte concentration (NaCl) in the bulk solution. Since the SHG E-field depends on the interfacial charge density, i.e., the amount of manganese adsorbed to the interface, an isotherm can be constructed by tracking the SHG E-field decrease as a function of bulk manganese concentration. This experiment is analogous to the w(3) experiments by Salafsky and Eisenthal (2000) probing cytochrome c, which has a +9 charge, adsorbing to silica at neutral pH. Fitting the data shown in Fig. 4.10 with the Gouy-Chapman model that now contains a surface charge density

6.5

ESHG [a.u.]

6.0

5.5

5.0

4.5 0.0

0.2

0.4 0.6 0.8 Manganese Concentration [M]

1.0

1.2x10-4

Figure 4.10: SHG E-Field Obtained from a Carboxylic Acid-Functionalized Fused Quartz/Water Interface as a Function of Manganese Concentration in Bulk Solution at pH 7. The Solid Line is a Fit of Data Based on the Langmuir-Modified Gouy-Chapman Model. See Text for Details.


115

modulated by the Langmuir adsorption isotherm y¼

KC 1 þ KC

(4.4)

where y is the relative surface coverage of the transition metal at the interface, K the equilibrium binding constant, and C the transition metal ion concentration in bulk solution (Atkins, 1998), we obtain a free energy of adsorption, DG, of 27.8(3) kJ/mol at 300 K from DG ¼ RT ln(KMref). Here, R is the universal gas constant, T the experimental temperature, and Mref the molarity of water (55.5 M) chosen as the reference state (Adamson, 1990). In the fitting, we used an initial interfacial charge density of 2.8 104 C/m2 to describe the number density of interfacial carboxylate groups at pH 6.5 (Konek et al., 2004). The free energy of adsorption is consistent with one to two hydrogen bonds, and surface saturation occurs at 40 mM. If a direct Coulombic interaction was controlling the interaction between the Mn(II) and the carboxylic acid group, a much larger free energy of adsorption would be expected. These results also support the idea that manganese binding is mediated by the solvation sphere. Assuming the manganese cations are physisorbed as Mn(II) cations, the fit of the data shown in Fig. 4.10 to Eqs. (4.1), (4.2), and (4.4) results in an interfacial charge density of 2.0(2) 102 C/m2. With two positive charges on each manganese cation, this charge density corresponds to 6 1012 adsorbed cations/cm2. This relatively low number density of adsorbed manganese ions can be reconciled by Coulomb-repulsion and the size of the hydrated cations limiting monolayer coverage. As discussed in the section ‘‘Charge Screening,’’ it is possible to screen the interfacial potential set up by the adsorbed manganese cations with bulk electrolyte solution. This would also allow us to test whether the SHG E-field decrease observed in Fig. 4.10 is indeed due to a sign switch of the interfacial charges. This is indeed observed in experiments in which the manganese concentration in the aqueous solution was kept constant at 1.0 mM, and the NaCl concentration was increased over the course of the experiment. Figure 4.11 shows that the SHG E-field gradually increases with increasing bulk ion concentration, which is consistent with screening a positively charged interface. Using the interfacial charge density of 2 102 C/m2 in the Gouy-Chapman model yields a reasonable fit to the data.

4.5.2. Chromate Interaction with Aluminum Oxide To demonstrate the capabilities of SHG to probe electronic transitions of charged adsorbates through resonance enhancement, we carried out SHG

116

C. T. Konek et al.

4

ESHG[a.u.]

3

2

1

0 0.00

0.02

0.04

0.06 0.08 NaCl [M]

0.10

0.12

0.14

Figure 4.11: SHG E-Field from a Carboxylic Acid-Functionalized Fused Quartz/Water Interface as Exposed to 1.0 mM Manganese(II) Solution as a Function of Increasing NaCl Concentration. The Solid Line is a Fit of the Gouy-Chapman Model to the Data. See Text for Details. experiments of chromate interacting with aluminum oxide/water interfaces at pH 7. These experiments make use of ligand-to-metal charge transfer resonantly enhanced SHG signals, and are analogous to our previous work on chromate interacting with silica/water interfaces (Mifflin et al., 2003a,b). Figure 4.12 contains SHG vs. time traces for chromate adsorbing from the aqueous phase at pH 7 to an aluminum oxide/water interface followed by subsequent desorption. At time 0, the water flow was stopped and the chromate flow was simultaneously started. The low signal intensities that are apparent in the SHG E-field vs. time traces are most likely due to the increased dielectric contrast for the aluminum oxide/water interface as compared to the silica/water interface. The increased dielectric contrast leads to a higher nonresonant SHG signal, which leads to a decreased signal to noise ratio for an analyte, such as Cr(VI), where the resonantly enhanced SHG component is not very strong.


117

2.0 H2O

H2O

Chromate

B

1.8

ESHG [a.u.]

1.6

1.4 A 1.2

1.0

0.8 -4

0

4

8

12

16

Time [min]

Figure 4.12: SHG E-Field vs. Time for Chromate Interaction with Aluminum Oxide/Water Interfaces at a Bulk Chromate Concentration of (A) 23 mM and (B) 73 mM and pH 7.

As indicated by preliminary isotherm measurements (data not shown), the two concentrations used in the experiments ([CrO2 4 ] ¼ 23 mM and 2 [CrO4 ] ¼ 73 mM) correspond to submonolayer and monolayer surface coverage, respectively. The resultant SHG E-field increase is attributed to the adsorption of chromate to the aluminum oxide/water interface, consistent with the following equation (Boyd, 1992): E 2o ð2Þ / wð2Þ NR þ Nha i ¼ EoEo * + X mu au ð2Þ wNR þ N þ ðou o þ iGu Þðou 2o þ iGu Þ u

ð4:5Þ

118

C. T. Konek et al.

(2) Here, wð2Þ tensor, N NR is the nonresonant portion component of the w (2) the number of adsorbates, /a S the orientational average of the hyperpolarizability tensor, ou the resonance frequency, Gu the line width for the transition at u, mu the transition dipole moment, and au the two photon absorption tensor. The time delay between the SHG E-field increase and the start of the chromate flow (1–2 min) is a convolution of the time required to flow chromate from the chromate on/off valve to the interface, a ramping of chromate concentration (the change in bulk concentration is a sigmoid rather than a step function from 0 to the final value) (Mifflin et al., 2003a,b), and the kinetics of the adsorption and desorption processes (Mifflin et al., 2003a,b). At 10.5 min the chromate flow was stopped and replaced by water flowing at approximately the same flow rate (around 1 mL/s). The resulting SHG E-field decrease is indicative of complete chromate desorption from the interface. This observation is consistent with the notion that chromate adsorption is fully reversible, similar to what we recently reported on the interaction of chromate with silica/water interfaces in the presence and absence of organic adlayers (Mifflin et al., 2003b; Al-Abadleh et al., 2005). Fully reversible binding of chromate to the Al2O3/water interface is consistent with the general notion that chromate is highly mobile in the environment.

4.6. Environmental Implications and Summary In this work, we have shown how nonlinear optics, specifically SHG, can be used to identify important interfacial parameters for geochemical applications. These include interfacial charge densities, the number of ionized species at aqueous/solid interfaces, interfacial pKa values, as well as thermodynamic parameters for solute binding, such as equilibrium binding constants and the related free energy of binding. Our experimental results indicate that the properties of a molecule at a surface can be drastically different from the properties of a molecule in bulk solution; therefore, when attempting to understand or model processes controlled by surface chemistry, it is important to use parameters determined by analytical techniques which directly probe the surface. For instance, Fig. 4.7 indicates that while the bulk solution pH value of water in a given scenario may be neutral, carboxylic acid groups within surface-bound natural organic matter can remain neutral, i.e., fully molecular, if they are located in an environment dominated by intra- and intermolecular hydrogen bonding. Water molecules may more fully solvate carboxylic acid groups in interfacial regions of high disorder. These carboxylic acid groups may then deprotonate in pH regimes similar to bulk, solvated carboxylic acid groups. This result suggests the


119

possibility for local and temporal mapping of pKa values and charge densities for suspended or sedimented mineral particles coated with oxidized organic matter, and is directly related to the observation of pH-dependent structural changes in humic substances (Myneni et al., 1999). The results from this work can be incorporated into the next generation of environmental transport models. We have already shown how the widely used Kd model, which is powerful because of its simplicity, can be modified using interface-specific thermodynamic and kinetic measurements (Al-Abadleh et al., 2004, 2005). Our quantitative, interface specific data could also be incorporated into the MUSIC model (Hiemstra et al., 1989) and its modifications (Piasecki et al., 2001). The methods presented in this work are general for tracking metal ions at buried interfaces in real time. Using environmentally relevant manganese and chromium concentrations, Figs. 4.9 and 4.12 show the reversibility of metal ion binding with carboxylic acid-functionalized silica/water interfaces as well as aluminum oxide/water interfaces. The hydrogen-bonding interactions governing the metal ion/surface interactions are consistent with the observed high mobility of manganese and chromium in most soil environments, provided that redox processes are inactive. Clearly, the possibility for studying redox chemistry, including proton-coupled electron transfer processes, occurring at these interfaces represents a next major step in NLO work on geochemical systems. Together with other interface-specific spectroscopies, NLO thus offers the opportunity to quantitatively determine, on the molecular scale, how heterogeneous processes control the fate of metals in geomedia.

ACKNOWLEDGMENTS The authors would like to thank Catherine Schmidt for machining the Teflon flow cell and various sample reaction chambers, Dr. Amanda Mifflin for the ToF-SIMS images, Prof. Terri Odom for the use of laboratory equipment, Prof. Chad Mirkin and Dr. Khaled Salaita for the use of lab equipment and the AFM images, Prof. SonBinh Nguyen and Dr. Paul Bertin for their synthetic expertise and the synthesis of the trichlorosilane molecules, Prof. James Ibers and Dr. Jerry Carsello for help with Laue measurements of the Al2O3 substrates, and Dr. Saman Shafaie from the Northwestern University Analytical Services Laboratory. A portion of this work was completed at the Northwestern Unversity Analytical Services Laboratory. A description of the facility and a full funding disclosure can be found at http:// pyrite.chem.northwestern.edu/analyticalserviceslab/asl.htm. We are also

120

C. T. Konek et al.

thankful for ellipsometry measurements carried out in the laboratory of Prof. Garth Simpson at Purdue University. We gratefully acknowledge support from the Chemical Sciences, Geosciences, and Biosciences Division, Office of Basic Energy Sciences, Office of Science, United States Department of Energy, the National Science Foundation CAREER program in experimental physical chemistry, the NSF/DOE Institute for Environmental Catalysis at Northwestern University, the NSF Nanoscale Science and Engineering Center (NSEC) at Northwestern University, a Wender student fellowship (MJM), a NASA Graduate Student Fellowship in Earth Systems Sciences (ABV), and a Dow Chemical Company professorship (FMG).

REFERENCES Adamson, A. W. (1990). Physical Chemistry of Surfaces. Wiley, New York, NY. Al-Abadleh, H. A., Mifflin, A. L., Bertin, P. A., Nguyen, S. T., & Geiger, F. M. (2005). Control of carboxylic acid and ester groups on chromium(vi) binding to functionalized silica/water interfaces studied by second harmonic generation. J. Phys. Chem. B, 109, 9691–9702. Al-Abadleh, H. A., Voges, A. B., Bertin, P. A., Nguyen, S. T., & Geiger, F. M. (2004). Chromium(VI) binding to functionalized silica/water interfaces studied by nonlinear optical spectroscopy. J. Am. Chem. Soc., 126, 11126–11127. Allen, H. C., Raymond, E. A., & Richmond, G. L. (2000). Non-Linear vibrational sum frequency spectroscopy of atmospherically relevant molecules at aqueous solution surfaces. Curr. Opin. Colloid Interface Sci., 5, 74–80. Atkins, P. W. (1998). Physical Chemistry. Oxford University Press, New York. Bargar, J. R., Trainor, T. P., Fitts, J. P., Chambers, S. A., & Brown, G. E. (2004). In situ grazing-incidence extended X-ray absorption fine structure study of Pb(II) chemisorption on hematite (0001) and (1-102) surfaces. Langmuir, 20, 1667–1673. Boyd, R. W. (1992). Nonlinear Optics. Academic Press, New York. Bradley, R. A., Georgiadis, R., Kevan, S. D., & Richmond, G. L. (1993). Nonlinearoptical spectroscopy of the Ag(111) surface in an electrolyte and in vacuum. J. Chem. Phys., 99, 5535–5546. Brown, G. E., Henrich, V. E., Casey, W. H., Clark, D. L., Eggleston, C., Felmy, A., Goodman, D. W., Graetzel, M., Maciel, G., McCarthy, M. I., Nealson, K. H., Sverjensky, D. A., Toney, M. F., & Zachara, J. M. (1999). Metal oxide surface and their interactions with aqueous solutions and microbial organisms. Chem. Rev., 99, 77–174. Brown, G. E., & Parks, G. A. (2001). Sorption of trace elements on mineral surfaces: Modern perspectives from spectroscopic studies, and comments on sorption in the marine environment. Int. Geol. Rev., 43, 963–1073.


121

Carlson, D. L., McGuire, M. M., Roberts, A. L., & Fairbrother, D. H. (2003). Influence of surface composition on the kinetics of alachlor reduction by iron pyrite. Environ. Sci. Technol., 37, 2394–2399. Chung, Y. W. (2001). Practical Guide to Surface Science and Spectroscopy. Academic Press, San Diego. Dubowski, Y., Vieceli, J., Tobias, D. J., Gomez, A., Lin, A., Nizkorodov, S. A., McIntire, T. M., & Finlayson-Pitts, B. J. (2004). Interaction of gas-phase ozone at 296K with unsaturated self-assembled monolayers: A new look at an old system. J. Phys. Chem. A, 108, 10473–10485. Ege, S. N. (1999). Organic Chemistry: Structure and Reactivity. 4th Ed., Houghton Mifflin Company, Boston, MA. Eisenthal, K. B. (1996). Liquid interfaces probed by second-harmonic and sumfrequency spectroscopy. Chem. Rev., 96, 1343–1360. Eisenthal, K. B. (2006). Second harmonic spectroscopy of aqueous nano- and microparticle interfaces. Chem. Rev., 106, 1462–1477. Evangelou, V. P. (1998). Environmental Soil and Water Chemistry. Wiley, New York. Finlayson-Pitts, B. J., & Pitts, J. N. Jr. (2000). Chemistry of the Upper and Lower Atmosphere. Academic Press, New York. Fitts, J. P., Dadap, J., Stiopkin, I., Zang, Z. H., Flynn, G. W., Heinz, T. F., & Eisenthal, K. B. (2002). Second harmonic studies of the hematite/aqueous interface. Abstr. Pap. Am. Chem. Soc., 223 , C53. Gershevitz, O., & Sukenik, C. N. (2004). In situ ftir-atr analysis and titration of carboxylic acid-terminated SAMs. J. Am. Chem. Soc., 126, 482–483. Gopalakrishnan, S., Liu, D. F., Allen, H. C., Kuo, M., & Shultz, M. J. (2006). Vibrational spectroscopic studies of aqueous interfaces: Salts, acids, bases, and nanodrops. Chem. Rev., 106, 1155–1175. Grassian, V. H. (2005). Environmental Catalysis. CRC Press, Boca Raton, FL. Heinz, T. F. (1991). Second-order nonlinear optical effects at surfaces and interfaces. In: H.-E. Ponath, & G. I. Stegeman (Eds). Nonlinear Surface Electromagnetic Phenomena. Elsevier Publishers, Amsterdam. Hewitt, T. D., Gao, R., & Roy, D. (1993). Effects of surface charging on the 2nd-harmonic generation from a Cu electrode. Surf. Sci., 291, 233–241. Hiemstra, T., Vanriemsdijk, W. H., & Bolt, G. H. (1989). Multisite proton adsorption modeling at the solid-solution interface of (hydr)oxides – A new approach: 1. Model description and evaluation of intrinsic reaction constants. J. Colloid Interface Sci., 133, 91–104. Higgins, S. R., Stack, A., & Eggleston, C. M. (1998). Proton and ligand adsorption at silica– and alumina–water interfaces studied by optical second harmonic generation (SHG). Mineral. Mag., 62A, 616–617. Higgins, S. R., Stack, A. G., Knauss, K. G., Eggleston, C., & Jordan, G. (2002). Probing molecular-scale adsorption and dissolution-growth processes using nonlinear optical and scanning probe methods suitable for hydrothermal applications. In: R. E. Hellmann (Ed). Water–Rock Interactions, Ore Deposits, and

122

C. T. Konek et al.

Environmental Geochemistry: A tribute to David A. Crerar. The Geochemical Society. Jacobsen, J. J., Johnson, K., Moore, J. W., & Trammell, G. (2001). Chemistry comes alive! volume 5: Abstract of special issue 29, a CD-ROM for organic and biochemistry. J. Chem. Educ., 78, 423–424. Konek, C. T., Musorrafiti, M. J., Al-Abadleh, H. A., Bertin, P. A., Nguyen, S. T., & Geiger, F. M. (2004). Interfacial acidities, charge densities, potentials, and energies of carboxylic acid-functionalized silica/water interfaces determined by second harmonic generation. J. Am. Chem. Soc., 126, 11754–11755. Lantz, J. M., Baba, R., & Corn, R. M. (1993). Optical 2nd-harmonic generation as a probe of electrostatic fields and flat-band potential at single-crystal TiO2 electrodes. J. Phys. Chem., 97, 7392–7395. Lide, D. R. (1996). CRC Handbook of Chemistry and Physics. CRC Press, Boca Raton, FL. Martinez, I., Zhang, J. Z., & Reeder, R. J. (1996). In situ X-ray diffraction of aragonite and dolomite at high pressure and high temperature: Evidence for dolomite breakdown to aragonite and magnesite. Am. Mineral., 81, 611–624. Masel, R. I. (1996). Principles of Adsorption and Reaction on Solid Surfaces. Wiley, New York. Mifflin, A. L., Gerth, K. A., & Geiger, F. M. (2003a). Kinetics of chromate adsorption and desorption at fused quartz/water interfaces studied by second harmonic generation. J. Phys. Chem. A, 107, 9620–9627. Mifflin, A. L., Gerth, K. A., Weiss, B. M., & Geiger, F. M. (2003b). Surface studies of chromate binding to fused quartz/water interfaces. J. Phys. Chem. A, 107, 6212–6217. Myneni, S. C. B., Brown, J. T., Martinez, G. A., & Meyer-Ilse, W. (1999). Imaging of humic substance macromolecular structures in water and soils. Science, 286, 1335–1337. Nishiyama, K., Tanaka, Y., Harada, H., Yamada, T., Niwa, D., Inoue, T., Homma, T., Osaka, T., & Taniguchi, I. (2006). Adsorption of organic molecules by photochemical reaction on Cl:Si(111) and H:Si(111) evaluated by HREELS. Surf. Sci., 600, 1965–1972. Ong, S., Zhao, X., & Eisenthal, K. B. (1992). Polarization of water molecules at a charged interface; Second harmonic studies of the silica/water interface. Chem. Phys. Lett., 191, 327–335. Piasecki, W., Rudzinski, W., & Charmas, R. (2001). 1-pK and 2-pK protonation models in the theoretical description of simple ion adsorption at the oxide/electrolyte interface: A comparative study of the behavior of the surface charge, the individual isotherms of ions, and the accompanying electrokinetic effects. J. Phys. Chem. B, 105, 9755–9771. Richmond, G. L. (2002). Molecular bonding and interactions at aqueous surfaces as probed by vibrational sum frequency spectroscopy. Chem. Rev., 102, 2693–2724. Salafsky, J. S., & Eisenthal, K. B. (2000). Protein adsorption at interfaces detected by second harmonic generation. J. Phys. Chem. B, 104, 7752–7755.


123

Scott, D. T., McKnight, D. M., Voelker, B. M., & Hrncir, D. C. (2002). Redox processes controlling manganese fate and transport in a mountain stream. Environ. Sci. Technol., 36, 453–459. Shaikhutdinov, S., Ritter, M., & Weiss, W. (2000). Hexagonal heterolayers on a square lattice: A combined STM and LEED study of FeO(111) on Pt(100). Phys. Rev. B, 62, 7535–7541. Shen, Y. R. (1984). The Principles of Nonlinear Optics. Wiley, New York. Shen, Y. R., & Ostroverkhov, V. (2006). Sum-frequency vibrational spectroscopy on water interfaces: Polar orientation of water molecules at interfaces. Chem. Rev., 106, 1140–1154. Shultz, M. J., Baldelli, S., Schnitzer, C., & Simonelli, D. (2002). Aqueous solution/ air interfaces probed with sum frequency generation spectroscopy. J. Phys. Chem. B, 106, 5313–5324. Somorjai, G. A. (1994). Introduction to Surface Chemistry and Catalysis. Wiley, New York. Sparks, D. L. (2001). Elucidating the fundamental chemistry of soils: past and recent achievements and future frontiers. Geoderma, 100, 303–319. Stumm, W., & Morgan, J. J. (1996). Aquatic Chemistry. Wiley-Interscience, New York. Sutton, R., & Sposito, G. (2005). Molecular structure in soil humic substances: The new view. Environ. Sci. Technol., 39, 9009–9015. Tebo, B. M., Bargar, J. R., Clement, B. G., Dick, G. J., Murray, K. J., Parker, D., Verity, R., & Webb, S. M. (2004). Biogenic manganese oxides: Properties and mechanisms of formation. Annu. Rev. Earth Planet. Sci., 32, 287–328. Usher, C. R., Cleveland, C. A., Strongin, D. R., & Schoonen, M. A. (2004). Origin of oxygen in sulfate during pyrite oxidation with water and dissolved oxygen: An in situ horizontal attenuated total reflectance infrared spectroscopy isotope study. Environ. Sci. Technol., 38, 5604–5606. Voges, A. B., Al-Abadleh, H. A., & Geiger, F. M. (2005). Applications of nonlinear optical techniques for studying heterogeneous systems relevant in the natural environment. Environ. Catal., 83, 128. Voges, A. B., Al-Abadleh, H. A., Musorrafiti, M. J., Bertin, P. A., Nguyen, S. T., & Geiger, F. M. (2004). Carboxylic acid- and ester-functionalized siloxane scaffolds on glass studied by broadband sum frequency generation. J. Phys. Chem. B, 108, 18675–18682. Wang, H., Borguet, E., Yan, E. C. Y., Zhang, D., Gutow, J., & Eisenthal, K. B. (1998). Molecules at liquid and solid surfaces. Langmuir, 14, 1472–1477. Wang, H., Zhao, X., & Eisenthal, K. B. (2000). Effects of monolayer density and bulk ionic strength on acid–base equilibria at the air/water interface. J. Phys. Chem. B, 104, 8855–8861. Wasserman, S. R., Tao, Y. T., & Whitesides, G. M. (1989). Structure and reactivity of alkylsiloxane monolayers formed by reaction of alkyltrichlorosilanes on silicon substrates. Langmuir, 5, 1074–1087.

124

C. T. Konek et al.

WaveMetrics, Inc. (2003). Igor Pro Help Browser4.09 Carbon ed. WaveMetrics, Inc., Lake Oswego, OR. Xiao, X. D., Vogel, V., & Shen, Y. R. (1989). Probing the proton excess at interfaces by 2nd harmonic-generation. Chem. Phys. Lett., 163, 555. Xiao, X. D., Vogel, V., Shen, Y. R., & Marowsky, G. (1991). Molecular monolayers of charge-transfer complexes – protonation and aggregation studied by 2nd harmonic-generation. J. Chem. Phys., 94, 2315–2323. Xiao, X. D., Xie, Y. L., Jakobsen, C., & Shen, Y. R. (1997). Effects of surface impurities on surface diffusion of CO on Ni(110). Phys. Rev. B, 56, 12529–12538. Zhao, X., Ong, S., Wang, H., & Eisenthal, K. B. (1993). New method for determination of surface pKa using second harmonic generation. Chem. Phys. Lett., 214, 203–207. Zhao, X., Subrahmanyan, S., & Eisenthal, K. B. (1990). Determination of pKa at the air/water interface by second harmonic generation. Chem. Phys. Lett., 171, 558–562.

Developments in Earth & Environmental Sciences, 7 Mark O. Barnett and Douglas B. Kent (Editors) r 2008 Elsevier B.V. All rights reserved DOI 10.1016/S1571-9197(07)07005-X

Chapter 5

Prions, Metals, and Soils L. Charlet1,, Y. Chapron1,2, G. Roman-Ross3, C. Hureau1, D. P. Hawkins4 and K. V. Ragnarsdottir4 1

Environmental Geochemistry Group, LGIT-OSUG, University of Grenoble, BP 53, F-38041 Grenoble, Cedex 9, France 2 Alpine Institute of Environmental Dynamics, Rue du Puy, 38660 La Terrasse, France 3 Department of Chemistry, Faculty of Sciences, University of Girona. Campus de Montilivi, 17071 Girona, Spain 4 Department of Earth Sciences, University of Bristol, Bristol BS8 1RJ, UK

ABSTRACT This study focused on understanding whether trace metals in the natural environment have a role to play in the development of prion diseases. The prion protein (PrP) is the key protein implicated in the development of scrapie, a sheep- and goat-specific transmissible spongiform encephalopathy. The N-terminal tail of the protein includes five copper chelating sites as well as numerous positively charged amino acids, which all may induce a binding of the protein to clay minerals, as shown, e.g., by molecular dynamics (MD) calculations and electron paramagnetic resonance (EPR) spectroscopy. The C-terminal part of the protein has a hydrophobic core that may also interact with low-charge clay surfaces as well as organic matter. The speciation of Cu-PrP chelates is show to change upon adsorption on clay minerals. Cu coordination at a given pH in the adsorbed state in presence of PrP is similar to Cu coordination in a solution of lower pH. This, together with high available Mn2+ concentrations, favors the exchange of Mn2+ for Cu2+, which is shown by MD to occur in three steps. Scrapie-prone farms in France, Iceland, and Italy were shown indeed to have soils with extremely high ‘‘easily reducible Mn’’ and pore waters with low free Cu2+ ion activity. In contrast, in Icelandic young andosols, the correlation between scrapie-free and scrapie-prone areas was not as clear-cut as far as copper and manganese is concerned. Copper was seen to be higher (and can be considered adequate) than in the scrapie areas in France. In all areas manganese is considered to be above toxicity thresholds for sheep.

Corresponding author. Tel.: +33 476828020; Fax: +33 476828101

E-mail address: [email protected] (L. Charlet).

126

L. Charlet et al.

5.1. Introduction 5.1.1. Prion Disease Transmission: The Soil Hypothesis Sheep are prone to the fatal neurodegenerative disease, scrapie, which is of unknown etiology (see Bastian and Fermin, 2005; Collinge, 2005; Purdey, 2005) and has been for over two centuries of relatively minor concern to farmers or governments. The situation changed after the outbreak in the 1980s of bovine spongiform encephalopathy (BSE) in cattle, in the 1990s of new variant Creutzfeld Jacob disease (vCJD) in humans in the UK, and in the past 40 years of chronic wasting disease (CWD) for wild deer (Odoco leus spp.) and waipiti (Cervus elaphus nelsoni) in several parts of North America (Williams et al., 2002). Mostly because these diseases belong to the same group of neurological disorders as scrapie (Hopp et al., 2001), these diseases have strong impact on the local economy (e.g. $1 billion deer hunting economy in Wisconsin; Pedersen J., personal comment). The origin of the BSE epidemic that emerged in the UK in the mid-1980s may never be known with certainty but the hypothesis of a species-crossing infection from scrapie-affected sheep bone meal is thought most plausible (Brown, 2003). This hypothesis is unproven, and will probably remain so (see Phillips, 2000). The infected bone meal is suggested to have brought BSE to countries all over Europe. However, that does not explain the fact that some cattle born after the ban of sheep bone meal have come down with the disease. Other suggested horizontal modes of scrapie infection which include release of prions into the environment via accidental dispersion from storage plants of meat and bone meal, incorporation of meat and bone meal in fertilizers, spreading of effluents of slaughterhouses, rendering plants and gelatin industry, possible natural contamination of pasture soils by grazing heards, and burial of carcasses of contaminated animals (Revault et al., 2005). Horizontal transmission, i.e., transmission of the disease within a community from animal to animal, has been documented for CWD among wild deer, where decomposed carcass or urine excretion is considered sufficient to transmit the disease (Seeger et al., 2005). Wild animals usually ingest 100 g of soil per day either incidentally, by feeding and/or grooming (Abrahams and Steigmajer, 2003), or deliberately (Weeks and Kirkpatrick, 1976) in order to complement grass and other forage plants (see Freer and Dove, 2002) for the 22 essential chemical elements that are necessary to sustain life and to allow reproduction of the species (Underwood and Suttle, 2002; Lindh, 2004). Thus, inadvertent soil ingestion or even dust inhalation could be a pathway for parallel nutrient uptake and scrapie infection. CWD was further reported to jump among species, first to white tail deer and last year to moose. Scrapie

Prions, Metals, and Soils

127

and CWD are therefore unique among transmissible spongiform encephalopathies (TSEs) because they appear to be transmitted between animals, and the disease agents are thought to persist in environments previously inhabited by infected animals. 5.1.2. Prion Disease ‘‘Hotspot’’ Geographical Distribution The link of BSE to environmental factors is difficult to demonstrate due to the fact that animals may be born in one place, raised in another, fattened up in the third, and slaughtered in the fourth. Therefore, this study focuses on sheep, goats, and deer that are more traditionally raised in one location all of their lives. Scrapie prevalence across Europe is not uniform (Fig. 5.1). Outbreaks are most common in Cyprus, Slovenia, Greece, Italy, UK, and France. Rates are far from uniform within one country. For instance, hotspots are clustered in France in Basque Country and SW Massif Central and in Italy in Tuscany, Sardinia, Sicily, and Emilia-Romagna regions. In Iceland they were first reported at the beginning of the twentieth century in the midnorth part of the country. By the middle of the century it had spread to both the NW and NE coasts and an eradication program was set up in 1978. The affected flocks were culled and the farms restocked 2–3 years later with lambs from scrapie-free farms. However, in 33 farms scrapie recurred, in some cases up to 21 years after culling (Georgsson et al., 2006). 5.1.3. The Manganese Hypothesis Recent research that aims to fully understand the biology of prion diseases (including scrapie and BSE) has produced a wealth of research articles devoted to TSEs (e.g., Brown et al., 2002; Benestad et al., 2003; Casalone et al., 2004; Fernaeus et al., 2005; Sharpe et al., 2006). Despite these efforts, the unique TSE biology is still largely unexplained (e.g., Hijazi et al., 2003; Aguzzi, 2005) and has sparked many contradictory theories for their origin, in addition to those outlined above. These include the spontaneous conversion of normal prion proteins (PrPs) to deadly pathogenic forms (e.g., Prusiner et al., 1996), the autoimmune response to soil-borne microbes (Acinetobacter) (Ebringer et al., 1998), the trace metal imbalance in the brain of affected humans and animals (Thackray et al., 2002), and even the ingestion of CJD-contaminated human remains floating down the Ganges River, India, as a form of ‘‘burial’’ which may transmit the disease (Colchester and Colchester, 2005). Despite intense research around the globe for over a decade the etiology of prion diseases is still not understood. It appears, however, likely that trace

128 L. Charlet et al.

Cyprus

3596.0

Slovenia

462.2

Sheep

389.2

Greece Italy

121.2

United Kingdom

93.9 69.8

France Slovakia

34.3

Spain

34.3 32.0

Netherlands Ireland

22.4

Czech Republic

22.3

Belgium

13.5 9.5

Germany Portugal

7.9

Finland

7.5 3.1

Sweden 1

2630.6

Cyprus

76.3

Greece

48.2

Finland

15.3

United Kingdom

3.2

Italy

2.5

Spain

2.1

France 1.0

10

Goats

67.7

Slovenia

10.0

100

100.0

1000

1000.0

10000.0

10000

Figure 5.1: Histogram showing Occurrence of Scrapie (cases/10,000 test) in Goats and Sheep in Europe. Data taken from ‘‘Report on the Monitoring and Testing of Ruminants for the Presence of Transmissible Spongiform Encephalopathy (TSE) in the EU in 2005.’’ Available at http://ec.europa.eu/food/food/biosafety/bse/annual_ report_tse2005_en.pdf.


129

metals have a role to play in the multi-factorial development of these diseases. The community that works with trace metals has produced PrP (precursor to PrPSc) in the presence of manganese in vitro but production of scrapie through injection of these altered prions into laboratory mice has not yet been successful (Brown, D.R., personal comment). Similarly, work with prions performed on yeast cells led to the production of the altered PrP (Treiber et al., 2004). Mice fed on diets deficient in copper have produced spongiform change in the hippocampus, similar to those found in prion disease, but PrPSc has not yet been detected (Grassi Zucconi et al., 2007). The trace metal imbalance hypothesis for PrP misfolding is supported by the reporting of measurements of manganese and copper in brains of CJD diseased victims. PrPs are known to regulate copper activity in cells and particularly in nerve cells. The manganese concentration of these victims was 10 times higher than in normal brains (Wong et al., 2001). The same finding has been documented for sheep afflicted with scrapie (Ru, G., personal comment). In 2000 trace metals in the environment were proposed as a possible cause of prion disease (Purdey, 2000), although this was quickly disregarded (Bush, 2000). Soil and vegetation from regions of the world in which prion diseases occur with higher than average incidence were suggested by Purdey (2000) to be low in copper and high in manganese. The normal metal ligand for the PrP is copper (PrPc) (Brown et al., 1997), and PrP has an antioxidant activity similar to that of superoxide dismutase (SOD) (Wong et al., 2001). Indeed the PrPc function in the body is that of transport of copper from the outer cell to the inner cell space, while PrPc is found to be anchored in the inner cell membrane. There is evidence that when manganese replaces this copper in binding to the PrP the protein structure is altered from its normal form to a misfolded form (PrPsc) (Brown et al., 2000). Prions can have as many as six coppers in their ‘‘tails’’ and it is this copper that is crucial to hold intact in a-helices structure. For these reasons Ragnarsdottir and Charlet (2002) summarized the environmental fate of the trace elements copper and manganese in soils. They concluded that manganese is most mobile under low Eh and pH. These are the conditions where copper has low mobility, e.g., in peats and organic-rich soils. Here we report on the geochemistry of hotspots of scrapie outbreak in Europe, focusing on copper and manganese concentrations in soils from scrapieafflicted and scrapie-free areas. This study reported here was undertaken because a substantial association between scrapie and environmental trace elements has been neither proven (Purdey, 2000) nor dismissed (Chihota et al., 2004). We then discuss the fate of prions in soils, particularly the interaction of soil clay component with copper ions and PrP – because of recent suggestions that such sorption may be a horizontal infection factor.

130

L. Charlet et al.

5.2. Geochemistry of Hotspots 5.2.1. Total Manganese and Copper in Soils Cu and Mn concentrations were measured in soils and plants across Europe and the results of these studies are summarized in Angeloni and Bini (1992). European mean concentrations of soils vary from as low as 310 mg kg1 dry soil for Mn in Austria to 1,815 mg kg1 Mn in Greece. Mean soil values for copper, on the other hand, vary from 17 mg kg1 in soils in Austria to 1,588 mg kg1 Cu in Greece. Where the Cu content of soil falls below 10 mg kg1, deficiency symptoms may develop (Rose et al., 1979) but 100 mg kg1 is considered to be toxic for plants (Angeloni and Bini, 1992). Excessive Mn soil concentration is considered to be 1,500 mg kg1. Chihota et al. (2004) examined total concentrations of trace elements from the UK’s National Soil Inventory and deficiencies reported by farmers. No evidence of manganese or copper as a risk factor for scrapie were found by these authors, although high molybdenum was observed to be linked with scrapie areas in the Shetland Islands. This observation is of note because molybdenum interferes with the metabolism of copper in ruminants (Kendall et al., 2001). Total concentrations of trace elements in soils are usually not a good indicator of element availability to plant in soils. Bioavailability is very much related to the soluble concentration of the metal ions, whether as a free ion 2 3 or as an aqueous complex of inorganic (e.g., OH, Cl, SO2 4 , CO3 , PO4 ) or organic ligands (e.g., humic or fulvic acid or simpler acids such as acetic or oxalic acid). This study therefore focuses on the soluble fractions of copper and manganese, in addition to easily reducible manganese – which is considered to represent a considerable fraction of bioavailable Mn (Ragnarsdottir and Charlet, 2002). 5.2.2. Soil Sampling Soil sampling was conducted according to the procedure published by Salminen et al. (1998) and Salminen (2005). Samples were taken with a cylindrical, or screw auger, sampler. The resulting soil sample was a composite of two to four augers from a depth of 0–25 cm. Living surface vegetation, fresh litter, big roots, and rock fragments were removed. The soil was dried and sieved. The analyses presented here were from the o2 mm fraction. Thirteen scrapieaffected farms from France (1) and Italy (12) were sampled. In Iceland, three scrapie-prone and four scrapie-free farms were sampled.


131

Iceland sampling sites 7

♦

4 6

♦

5

♦ 3

2

1

30

100

La Langlade site

80

60

40

20

Number of affected farms

Number of cases

120

25 20 15 10 5 0

0 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

Italy

1995 1996 1997 1998 1999 2000 2001 2002 2003

Figure 5.2: Location of Sampling and Evolution of Scrapie Cases with Time at France and Italy Sites. In Iceland, Samples 1, 2, 3, and 4 Corre spond to Scapie-Free Farms and Samples 5, 6, and 7 to Scrapie-Prone Farms.

The study is therefore limited to sporadic locations in Europe (Fig. 5.2) due to secretive practices of many European countries regarding prion disease outbreaks. It was, for example, not possible to get geographical distributions of prion diseases in the UK beyond county statistics. In France, Spain, Portugal, and Holland similar difficulties were encountered. Therefore, the only European data made available to us were the country averages for European countries (see, e.g., Fig. 5.1). Such secrecy hampers research on the etiology of these debilitating diseases. The only countries we contacted that were open about the actual geographical locations of outbreaks of scrapie were Iceland, Italy, and Norway. Of interest is to compare the findings presented below with geochemical maps of Europe published by FOREGS (http://www.gtk.fi/publ/foregsatlas/).

132

L. Charlet et al.

5.2.3. Soluble and Free Copper Copper occurs in soil solids and pore water almost exclusively as a divalent cation, Cu2+. However, reduction of Cu2+ (cupric) to Cu+ (cuprous) and Cu0 (metallic copper) is possible under reducing conditions, as in the rhizosphere (Kirpichtchikova et al., 2006) or wherever halide or sulfide ions (‘‘soft’’ bases) are present to stabilize Cu+ (McBride, 1994). In order to evaluate bioavailable Cu, total soluble and free copper was determined in this study. Total soluble copper was assumed to correspond to the amount of copper measured in the 0.01 M CaCl2 extract (McLaren et al., 1983). In the Langlade INRA experimental station, located in SW France, total soluble copper values range from 50 to 150 mg kg1 of soil, whereas in one scrapie-prone farm in the Apennine they range from 40 to 70 mg kg1 and in the many Tuscany scrapie-prone farms sampled, they range from 50 to 100 mg kg1. These concentrations are not correlated with pH values (Roman-Ross et al., in preparation). In the present study we focused on free copper activity (pCu2+) because free copper ion has been shown to be the main biovailable species of copper in aquatic systems (Sauve´ et al., 1997). The free copper(II) activity, i.e., the concentration of free cupric ions as measured with a copper ion-selective electrode (ISE), is reported in pCu2+ units. The pCu2+ is defined as the negative log10 of the molar free Cu2+ ion concentration. In soils, most of the metal is insoluble, and a small fraction is reversibly bound to the soil particles. Generally, only a very small quantity of soluble Cu2+ is actually present as free Cu2+ ions in soil solution. Figure 5.3 shows the pCu2+–pH relationships for the sampled sites. In the Langlade site, France, we found constant values of pCu2+ for variable pH values. In the Apennine farm (Italy) where pCu2+ values are very homogeneous, we observed data scattered in two groups. Topography appears as a control factor of pCu2+ in Apennine soils. Samples from areas with moderate slope variations show higher values of pCu2+ and Cu concentrations in the soil solutions. In Tuscany, pCu2+ values show a slight linear dependency with the pH (data not shown). The very low pCu2+ values obtained for all these samples suggest that a large fraction of Cu2+ is adsorbed by functional groups of soil organics in surface soil horizons. The aqueous Cu2+ ion represents then a very tiny fraction of the total soluble copper concentrations due to the high affinity of Cu2+ for soil colloids and/or complexation with organic acids, two factors that determine the low mobility of copper in these soils. The free Cu activity (pCu2+) of 28 Icelandic soils (Fig. 5.3) showed a similar range (6.99–11.9) to those Sauve´ et al. (1997) found in soils from Canada, Denmark, and the US (6.18–12.2) but is generally higher than pCu2+ measured in the French (generally around 11) and Italian soils (Tuscany 9–20; Apenine


133

La Langlade (France) Apennine farms (Italy) Tus cany farms (Italy) Iceland scrapie-prone farms Iceland scrapie- free farms Solubility limit for Tenorite Solubility limit for Malachite Solubility limit for Azurite

0 2 4 6 8 pCu2+

10 12 14 16 18 20 22 4.0

4.5

5.0

5.5

6.0

6.5

7.0

7.5

8.0

8.5

9.0

9.5 10.0 10.5

pH

Figure 5.3: Free Copper Activity in Soil Solution, at France, Italy, and Iceland Scrapie-Prone Farms (Data from: Roman-Ross et al., in preparation; Hawkins, 2006). Precipitation–Dissolution Equilibria were calculated assuming a PCO2 Constant at 101.5 Bar. 11–14). Mean pCu2+ levels of scrapie-free (9.45) and scrapie-prone (9.67) farms from Iceland were however similar, although scrapie-free farms showed greater variability and tended to have higher free Cu activity (Fig. 5.3; Hawkins, 2006). Precipitation–dissolution equilibrium relationships may also control the activity of free Cu2+ in soil solution. Nevertheless, soil solutions in sampled soils are undersaturated with respect to the least soluble mineral phases of Cu, indicating that precipitation phenomena do not control Cu solubility in these surface soils (Fig. 5.3).

5.2.4. Bioavailable Manganese Actual Mn toxicity as well as Mn bioavailability is associated with manganese forms that are either water soluble exchangeable or easily reducible.

134

L. Charlet et al.

Adams (1984) suggested a maximum reducible Mn range of 50–100 mg kg1 above which Mn toxicity would occur. Exchangeable manganese in sampled soils from scrapie-affected areas in Italy and France shows values between 0.2 and 10 mg Mn kg1 soil (Roman-Ross et al., in preparation). These concentrations can be considered as low and normal values. In Tuscany, soil exchangeable Mn concentrations were undetectable in most of the soils, and only two farms revealed concentrations above the detection limit of the AESICP analytical technique (0.1 mg Mn kg1 soil). Samples from the Appenine farm show very homogeneous pH and exchangeable Mn concentrations. Results reported for scrapie-affected areas in Iceland (Ragnarsdottir and Hawkins, 2005; Hawkins, 2006) show exchangeable Mn values higher than those obtained in France and Italy and a dependence of the Mn concentrations with pH values as observed in the Langlade site. The sampled sites show contrasting soil characteristics. Iceland soils are andosols very poor in layer silicate clay minerals and very rich in imogolites, a tabular zeolite-like mineral, and organic matter. In contrast, soils from the French site contain abundant 2:1 clay minerals. Both sites are drained wetland where Mn is highly available due to changes in redox conditions. The high adsorption capacitiy of imogolites and/or low pH could explain the high exchangeable Mn values obtained by Ragnarsdottir and Hawkins (2005) in Iceland soils with basic pH. In contrast, Italian soils were all organic-matterpoor, clay-rich, well-drained calcareous soils. Easily reducible Mn concentrations for the sites investigated in France, Iceland, and Italy are always very high (Fig. 5.4), clearly above the above-mentioned toxicity limits, excepting some soils from Iceland scrapie-free farms. Easily reducible Mn clearly higher than the above-mentioned toxicity limits was higher in soils from scrapie-prone (100–900 mg g1) than in scrapie-free (below 100 mg g1) farms in Italy and France (Fig. 5.4). All our data are in the range of Eh–pH conditions of environments in contact with the atmosphere, i.e., in optimal conditions of oxygenation. Mineralogical control was verified for rhodocrocite (MnCO3) at two different PCO2 (103.5 and 101.5 atm). All soil solutions are undersaturated with respect to this mineral. Therefore, we can consider that precipitation phenomena do not control Mn2+ solubility in these surface soil horizons. 5.2.5. Discussion of Copper and Manganese in Soils Soil is an important, though inadvertent, part of a sheep’s diet. Animals have been shown to ingest large quantities of soil (over 30% dry-matter intake under certain conditions) while grazing (Abrahams and Steigmajer, 2003)


High

1000 Easily Reducible Mn (mg kg-1)

135

100

Normal 10 La Langlade (France) Apennine farm (Italy) Tuscany farms (Italy) Iceland scrapie-pronefarms Iceland scrapie- free-farms

Low

1 4.0

4.5

5.0

5.5

6.0

6.5

7.0

7.5

8.0

8.5

9.0

pH

Figure 5.4: Bioavailable Mn Concentration in Soils, at France, Iceland, and Italy Scrapie-Prone Farms (Data from: Roman-Ross et al., in preparation; Hawkins, 2006). Lines indicate the Limits between Excess, Normal, and Deficit Concentrations (Loue´, 1986). and can obtain up to 20% of the total dietary intake for some trace elements (notably, Co, Fe, Mn, and I) from soil (Lee et al., 2002). Sheep must also inhale minerals present in dust/soil particles, although this is likely to be a very minor pathway for mineral uptake. In ruminants, the relative gastrointestinal absorption of minerals varies considerably (and is different to nonruminants), with estimates for Cu (1–10%) and Mn (1–20%) much less than Zn (60–70%) (see Whitehead, 2000). It is plausible that sheep from scrapieprone farms in Iceland ingest (or inhale) more Cu, Mn. and Zn per gram of soil (and subsequently absorb different proportions of each mineral) than sheep from scrapie-free farms. There is no way to quantify this, as information concerning the availability/absorption of soil-borne metals by animals is scarce (Abrahams and Steigmajer, 2003). Data reported above show that there is an increased risk of scrapie in cultivated pasture farms with soils of high pH, low organic matter content, high bioavailable Mn (i.e., easily reducible Mn), and low bioavailable Cu (i.e., high pCu2+). This is particularly interesting as farms with high soil-pH and low-OM may have an increased risk of scrapie because of low levels of Mn. There is some evidence that scrapie is more common in ‘‘fine meadows’’

136

L. Charlet et al.

with ‘‘luxuriant vegetation’’ and with animals well-fed on nutritive fodder than on ‘‘poor soils’’ with sparse pasture (McGowan, 1922). Therefore, rich, damp, wet, or marshy ground and herbage could impact the development of the disease. It should be remarked here that these very soil conditions are favorable to high manganese solubility and excessive uptake into forage. McGowan (1922) further pointed out that a number of observers believed rich, damp, wet, or marshy ground and ‘‘unwholesome’’ herbage to cause the disease. The meaning of ‘‘unwholesome’’ was not elaborated upon. Although Sarcocystis parasite infection in sheep can increase as the sheep are moved down from hill pasture to near-building meadow fields, and mass infection produces scrapie-like symptoms, it should be remarked here that these ‘‘unwholesome’’ soil conditions are favorable to high manganese (and molybdenum) solubility and excessive uptake of these trace metals into forage. These soils often rich in organic matter will also have low free copper activity. These observations are therefore supportive of Mark Purdey’s hypothesis of excessive dietary Mn as contributing to scrapie, and possibly other TSEs (Purdey, 2000). However, wetter lowland pastures are also known to be higher in Mo content, and are commonly higher in S as well (Suttle, 1987), both elements which reduce Cu absorption in the ruminant and increase the chance of hypocuprosis. High S content characterizes, for instance, willows eaten by CWD-affected deer (McBride, 2006). Whitelaw et al. (1979) reported a case of lambs confined to improved hill pasture, with forage higher in Mo (2.9 mg/kg) and S (0.43%) than the unimproved control pasture, developing marked Cu deficiency. High S in the forages of heavily industrialized Britain appears to be quite widespread based on the survey of Leech and Thornton (1987). In conclusion, highly available S and Mn and little available Cu, both in soil and in forages, are conditions which appear to favor, in the limited number of farms sampled, the appearance of scrapie and CWD diseases.

5.3. The Double Nature of the Prion Protein The PrP, the key protein implicated in diseases known as TSEs, is composed of two parts: an N-terminal positively charged hydrophillic unstructured tail and a structured (two a-helices) hydrophobic C terminal globular part with negative carboxylate functional groups. The N-terminal tail (residues 23–106) includes 14 positively charged amino acids (6 Lys+, 3 Arg+, and 5 His+) and bears in the 58–91 peptide section the four octapeptide repeats that can bind transition metal ions such as copper or manganese. PrPs are known to regulate Cu2+ activity in cells and the copper binding site present in the N-terminal tail (site numbers 1–4;


137

N-terminal Tail with Octarepeat 59-91 region (sites #1 to #4)

Site #5 92-96

Hydrophobic C-terminal Core

Figure 5.5: Structure of the PrP Protein (Adapted with Permission from Burns et al., 2003). Spirals Depict a-Helices and Arrows b Structures.

Fig. 5.5) is composed of one histidine and three adjacent glycine residues (HGGG) (Milhausen, 2004; Revault et al., 2005). However, when Mn2+ is present, it dictates the physical characteristics of PrPSc, even in the presence of excess Cu (Treiber et al., 2006). The structure of the HGGG site binding either Cu or Mn was obtained from quantum mechanics/molecular mechanics (QM/MM) computations. The structure of the HGGG binding site was first optimized with the GAUSSIAN98-03 program (Frisch et al., 2004), based on the final structure of the molecular dynamic trajectories in aqueous environment. The HGGG final histidine and glycine residues were terminated by hydrogen. Calculations were performed with QM/MM at the HF/3-21G level for the chelator core, and with MM with the UFF force field for the outside part of the complex (Chapron et al., in preparation). Distinct optimizations were carried out with Cu2+, Cu+, Mn2+, and Mn3+ as the central metal ion. The metal ion nearest to water molecule was retained for the calculation. These calculations by Chapron et al. (in preparation) support the hypothesis that copper can be exchanged for manganese in a three-step reaction, starting with the Cu–HGGG complex (Fig. 5.6). First the complexed Cu2+ is reduced to Cu+, by trace amounts of reductants, e.g., ascorbic acid. Upon reduction, the chelating cage volume

138

L. Charlet et al.

Cu2+

Mn2+

Mn3+

Figure 5.6: Complexation of the Octarepeat Peptide with various Transition Metal Ions. Results from Molecular Dynamics and Quantum Chemistry Computations (Adapted with permission from Chapron et al., in preparation).

increases, and, in the presence of excess soluble manganese and of Cu(I) specific chelators, such as bathocuproine disulfonate (BCS) and clioquinol (CQ), the Mn2+ for Cu+ cation exchange is then thermodynamically favored (Treiber et al., 2006). Once Mn2+ is present in the HGGG chelating site, it may spontaneously oxidize to Mn3+, which forms very stable complexes. Such an oxidation, observed in other Mn–organic complexes (Duckworth and Sposito, 2005), may be due to traces of oxygen present in water or to an oxidation by water itself (Gehin et al., 2007 and references therein). The whole Cu2+Cu+-Mn2+-Mn3+ exchange process is quite general, as the same mechanism occurs with multicopper oxidase enzymes in the catalyzed oxidation of Mn, at the surface of Mn-oxidizing bacteria (Tebo et al., 2004).


139

Molecular dynamics (MD) simulations further show that once the manganese ion is oxidized, not only does the diameter of the chelating cage around the central metal ion decrease by up to 0.4 A˚, but a change of coordination is also observed, which involves the nitrogen of the imidazole ring of histidine (Fig. 5.6). This coordination change may then lead to the formation of the misfolded PrPSc prion protein, characterized by an increased b-structure content entirely due to misfolding of the flexible tail starting from the first b-structure of the normal molecule (Brown, 2003). The findings of Chapron et al. (in preparation) and Treiber et al. (2006) therefore support the hypothesis that imbalance between copper and manganese may play a role in the etiology of prion diseases. In contrast, the C-terminal part of the protein (residues 106–230) contains both negatively (four Asp and eight Glu) and positively (five Lys+, eight Arg+, and five His+) charged amino acid residues (Fig. 5.7, top). The 3D globular 55 A˚ 35 A˚ 25 A˚ core acts as a hydrophobic solvent-excluding surface area unit, surrounded by a cage-like structure of H-bond water molecules (Fig. 5.7, bottom). In this part of the molecule the conversion from the a- to the pathogenic b-form occurs. In normal PrPc a-helices comprise 26.4% of the protein, whereas in the misfolded PrPSc conformation a-helices have been replaced by b-sheet-rich units (Revault et al., 2005). The fifth Cu GGTH96 complexing site (site number 5; Fig. 5.5) is located near the junction between the C-terminal part of the protein and the N-terminal tail and may play a critical role in the formation of PrPSc (Brown et al., 2000).

5.4. Prion Sorption and Transformation on Clays 5.4.1. Modes of Sorption Proteins interact with various soil components such as clays and organic matter. The interaction of PrP with clay minerals has been the topic of various studies (Jeffrey et al., 2006; Johnson et al., 2006; Rigou et al., 2006; Hureau and Charlet, in preparation) and may, or may not, induce structural changes of the protein (Norde, 2000). In these various studies, montmorillonite is chosen as model clay mineral, as it is a phyllosilicate clay mineral typical of temperate and semi-arid soils and known to be a strong adsorbent for inorganic cations as well as proteins. PrP may sorb on clays via either electrostatic or hydrophobic interactions. Electrostatic sorption mechanisms on clay mineral surfaces have been studied for decades and one distinguishes: (i) cation exchange occurring in the interlayer space and on basal plane surfaces and (ii) specific pH-dependent

140

L. Charlet et al.

Figure 5.7: Structure of the 150–214 Segment. Top: Hydrophobicity (Maximum in Red and Minimum in Blue). Bottom: Superficial Charge (Positive Charge: Blue; Negative Charge: Red). sorption occurring on clay broken edges (e.g., Sposito, 2004). Cation exchange may occur between single inorganic cations and cationic organic molecules such as alkyl ammonium, nitrobenzene cations, quinones, or dyes. These sorbed entities respond to thermodynamics laws of mass action described by cation exchange equations (Sposito, 1981). Cation exchange is pH-independent. It originates from the presence of a permanent negative structural charge in the alumino-silicate layer created by isomorphic substitutions in the clay lattice. The density of exchange sites can be derived from the structural formula, and is electrically balanced by


141

exchangeable cations or positively charged molecules. The structural surface charge density is for montmorillonite of the order of 1 molc kg1 and limits the maximum amount of adsorbed cationic species, as well as the amount of molecules sorbed by hydrophobic interactions. One negative charge is found on average on the siloxane surface every 0.8 nm, although due to the random distribution of the isomorphically substituted cations in the octahedral layer, the distance between two charges, i.e., the portion of surface area with a hydrophobic character, may be larger or smaller. On low-charge phyllosilicates, such as kaolinite or pyrophyllite, much larger portions of the hydrophobic siloxane surface are, of course, available for hydrophobic interaction. Maximum PrP adsorption onto Na saturated montmorillonite was shown C Sc 1 to be equal to 0:1 g g1 (Jeffrey clay for recombinant PrP or 1 g gclay for PrP et al., 2006; Rigou et al., 2006). This sorption leads to a 1.47–1.22 nm extension of the d001 spacing respectively, i.e., to a larger interlayer space for the full protein than for the truncated toxic protein. IR spectra and ‘‘desorption experiments’’ run at pH 2.5 in the presence of a pH buffer demonstrated that in the absence of co-adsorbed cations, sorption is pH independent. Sorption protects the molecules from transconformational changes, which would otherwise be induced in the absence of the clay mineral, e.g., by a change in solution pH (Revault et al., 2005; Jeffrey et al., 2006). This array of observations confirms that sorption occurs in part via electrostatic interactions on the clay basal plane. Other workers have also shown that proteins, having a high flexibility, are strongly retained on such electronegative surfaces but undergo large conformational rearrangements (Servagent-Noiville et al., 2000). In contrast to inorganic cation exchange processes, which are usually reversible (except for K+ and Cs+), the adsorption of PrP by cation exchange mechanism is irreversible, as shown by desorption experiments run in 1 M NaCl solution (Revault et al., 2005; Jeffrey et al., 2006). In order to understand the strong anchorage of the protein to the clay basal plane, clay–PrP interactions were modeled and simulated by MD using the XPLOR (Brunger, 1992) and NAMD codes (Phillips et al., 2005), with the structure of PrPSc kindly provided by Govaerts et al. (2004). Only the 92–138 peptide part of the molecule was considered since the a helix, N-glycosylation residues, and the loop after 138 are not considered to be responsible for infectivity of PrPsc. Computations were performed with slightly modified CLAYFF (Cygan et al., 2004) and CHARM/XPLOR (Brooks et al., 1983; Brunger, 1992) force fields, for clay mineral and protein, respectively. Water molecules were treated explicitly so that no dielectric was used (e ¼ 1). The anchorage of the protein to the clay surface is found to occur via positively charged histidine and lysine residues (Chapron et al., in preparation; Fig. 5.8).

142

L. Charlet et al.

Figure 5.8: Molecular Dynamics Snapshot of PrPSc Anchorage on a Clay Basal Plane via Positively Charged Histidine and Lysine Residues (Adapted with permission from Chapron et al., in preparation). A pH-dependent ‘‘sorption edge’’ is observed from pH 6 to 8 (Grosclaude, personal comment) that suggests the implication of specific sorption on variable-charge clay-edge sites. Specific sorption on clays is a mechanism related to the acid–base properties of functional groups at the clay edges as opposed to cation exchange which was linked to the permanent charge of the clay structure (e.g., Fletcher and Sposito, 1989; Charlet et al., 1993; Stadler and Schindler, 1993; Zachara and Smith, 1994; Baeyens and Bradbury, 1997; Avena, 2002; Gehin et al., 2007). Clay edges are known to bind specifically a number of organic molecules and, in that way, to catalyze a variety of reactions leading to the transformation of these molecules. However, until recently no protease activity for clay particles had been reported. Johnson et al. (2006) observed that upon desorption, only the C-terminal domain of the protein could be released. They assumed the N-terminal tail to be retained by montmorillonite and therefore concluded that montmorillonite had indeed a protease activity. However, questions arise about the purity of the protein used. If Johnson et al. (2006) had used a mixture of full-length PrP and C-terminal domain, the apparent protease activity could simply be due to an expected stronger retention


143

of the N-terminal domain compared to the C-terminal domain. This could not be excluded based on the methods and data reported. Hydrophobic interactions occur when apolar neutral molecules gain energy through an association with hydrophobic mineral surfaces. The basal plane siloxane surface is hydrophobic, as can be seen for minerals that do not contain isomorphic substitution within the clay structure such as talc or pyrophyllite (Quincampoix and Radcliff, 1992; Stauton and Quinquampoix, 1994). The Cterminal globular part of the protein, which contains the hydrophoboic core of the protein, was modeled with the VMD code (Humphrey et al., 1996) and the hydrophobic part of the molecule is depicted in Fig. 5.7. This part of the protein exhibits a significant, although 10 times weaker, interaction with montmorillonite than the full-length protein (Rigou et al., 2006). Similarly kaolinite, a mineral devoid of negative permanent structural charge, adsorbs 10 times less PrP than montmorillonite, even when the surface coverage is normalized per mineral surface area unit (Johnson et al., 2006). The hydrophobic, entropically driven, interaction is in general responsible for weak interactions between proteins and hydrophobic minerals or organic matter in soil. This interaction can often be reversed by addition of detergent, but in the present case, addition of sodium dodecyl sulfate (SDS) was shown in two concurrent studies to be unable to desorb the C-globular domain (Rigou et al., 2006; Johnson et al., 2006), unless a drastic 10% SDS extraction was used at 1001C or in hot laemmli buffer. Of note is that hot laemmli buffer could only desorb the 18 kDa C-terminal PrP domain (Rigou et al., 2006). Another way to test for hydrophobic interactions is to study the effect of additional molecules that sorb in the same manner. A 500-times excess of fecal calf serum protein compared to recombinant PrPC leads to no PrP desorption. Instead it leads to PrP surface aggregation! Therefore, hydrophobically interacting species do not compete but rather cooperate for adsorption onto clay minerals. In conclusion the Rigou et al. (2006) and Johnson et al. (2006) studies provide strong evidence that soils, and specifically montmorillonite clay particles, can serve as a reservoir for PrPSc and TSE infectivity. They also show that migration of PrP in a soil column may be strongly retarded, and that claybased liners will be very efficient barriers for PrP waste product landfill sites. 5.4.2. Speciation Change upon Adsorption Two types of chelating sites are present in the PrPC protein and can bind transition metal ions: the four octapeptide repeat HGGG complexation sites present in the N-terminal PrP tail and the fifth chelating site (GGGTH), shown to lead to the transconformational change from the normal cellular

144

L. Charlet et al.

PrPC to PrPSc, and present at the junction between the C-terminal part of the protein and the N-terminal tail (Milhausen, 2004). The adsorption of a P5 model protein molecule (GGGTH) on synthetic montmorillonite was investigated in the presence of copper ions (at a 1:1 P5:Cu ratio) in order to elucidate mechanisms of prion retention and transconformation in soils (Hureau and Charlet, in preparation). Montmorillonite was found to have a large and selective adsorption capacity for the P5–Cu2+ complex, as previously reported for the full-length PrP protein. Electron paramagnetic resonance (EPR) was used to characterize the copper ion local environment. EPR spectra indicate that P5–Cu speciation within the clay interlayer space differs from that observed in solution (Table 5.1). While a Cu2+ ion surrounded by four equatorial nitrogen atoms dominates the speciation in solution above pH 8 (N4 coordination; Hureau et al., 2006), this very stable complex is never observed in the clay interlayer, whatever the pH. Instead, the N3O coordination is observed in the presence of clay particles at ‘‘physiological’’ and ‘‘normal soil’’ pH values (6.5opHo8), whereas in solution it occurs only at low pH values (4opHo6.5). These results indicate that the interlayer aqueous medium is characterized by a local acidity, as previously reported in ammonia and pyridine sorption IR studies. These results contrast with a previous FTIR study performed on ovine PrP sorbed on montmorillonite which, in the absence of copper ions, showed no spectral change with pH and thus tended to indicate that PrP–clay interactions were unlikely to induce a change of conformation leading to the pathogenic form of prion (Revault et al., 2005). In the present study, where copper was co-adsorbed with the protein on montmorillonite, changes of the P5–Cu coordination, and, thus complexation upon adsorption, may

Table 5.1: EPR Parameters of Cu2+ Species Obtained with the P5 Ligand in Solution or Adsorbed on Clay (Adapted from Hureau et al., 2006; Hureau and Charlet, 2007). Dominant Cu species

g//

|A//| (104 cm1)

pH

Equatorial binding mode

Solution

Cu(H2O)2+ 6 Cu(P5)2+ Cu(P5H2) Cu(P5H3)

2.41 2.36 2.23 2.20

114 140 180 200

pHo5 4opHo6.5 6.5opHo8 pH48

O4 NO3 N3O N4

Clay

Cu(H2O)2+ 6 Cu(P5)2+ Cu(P5H2)

2.41 2.36 2.22

114 118 178

pHo6 6.5opHo8 pH48

O4 NO3 N3O


145

have profound implications. Since the P5–Cu complex stability decreases from an N4 to an NO3 coordination (Hureau et al., 2006), it will be much easier to exchange copper ions in the clay interlayer than in solution. Morover, in such an oxygenated environment, the Cu(II) ion will become more easily reducible (Hureau et al., 2006). Therefore, as discussed above, the P5– Mn complex formation may induce the formation of PrPSc precursor. PrP pathogenicity could then be strongly increased upon PrP burial in soils rich in easily reducible manganese, which were shown in the first part of this study to often characterize scrapie-prone farms.

5.5. Horizontal Infectivity Horizontal transmission of PrP, i.e., animal infection from PrP protein present in the environment, may occur via the gut following oral absorption of forage and soil particles. Studies on CWD strongly suggest that lush, fresh grass, growing in the immediate vicinity of contaminated feces and buried carcasses, could be an efficient infectivity route (Miller et al., 2004). In the acidic conditions prevailing in the fourth stomach (abomasum), the clay–PrP complex was shown to lead to neither PrP desorption (Rigou et al., 2006) nor b-sheet formation (Revault et al., 2005). Thus, in this infectivity route, the whole PrP–clay complex must cross the gut mucosa and be taken up by epithelial cells, although free PrP molecules can also follow the same route (Jeffrey et al., 2006). The PrP–clay complex would then move via the M cells to Peyer patches, where PrPSc replication would occur in dendritic cells, i.e., along the phospholipidic membrane contact zone in a pH 4.6 local environment. PrPSc would then be extruded via lysosomes to the nerves located behind Peyer’s patches (Grosclaude, personal comment). Although this infectivity pathway needs still to be experimentally confirmed, it is a reasonable hypothesis since CWD infection-specific protease-resistant PrPs (PrPCWD) do accumulate in gut-associated lymphoid tissues of infected mule deer (Miller et al., 2004 and references therein). Although this seems the most probable infectivity pathway, experiments done to date consisted mainly of intracerebral inoculation of PrPSc–clay complexes. In these experiments, not only a large amount of PrPs is injected due to the high sorption capacity of clay minerals, but also prions were shown to remain infectious (Miller et al., 2004; Johnson et al., 2006; Grosclaude, personal comment). The infectivity of the clay–PrP complex was, at first glance, unexpected since sorption of PrP could induce partial unfolding of the PrP protein and thus reduction of its pathogenicity, and since experimental desorption of full-length PrP was unsuccessful (Johnson et al., 2006; Rigou et al., 2006).

146

L. Charlet et al.

5.6. Conclusions In the manganese prion transformation hypothesis, copper ions present in the PrP are exchanged for manganese ions originating from the environment, and this exchange leads to the molecule adopting a pathogenic configuration. Furthermore, recent publications show clay–prion–soil complexes to be a possible mean of prion infectivity. In the present study, the Mn hypothesis was tested by both field and molecular-level investigations. In French, Icelandic, and Italian scrapie-prone farms, soils were found to be very rich in clays (or in imogolite in the case of the Icelandic soils), and to have very low levels of bioavailable copper (i.e., free copper ion concentration) and very high levels of bioavailable manganese (i.e., easily reducible). Thus, the conditions are favorable for the Mn2+ for Cu2+ cation exchange to occur, either in soil or in the animal after excess Mn uptake, but the exchange requires the transient reduction of Cu2+ to Cu+. Prion unfolding subsequently occurs when Mn2+ is oxidized to Mn3+. In soils, EPR spectroscopy demonstrates that the exchange may occur in clay interlayers, where local pH is lower than the macroscopic pH. Therefore, the local environment present in clay aggregates, and the high-Mn/low-Cu prevailing conditions in scrapie-prone farm soils, creates the conditions favorable for the PrP transconformation change to occur, and thus for the persistent PrP buried molecule to be ready, in a concentrated form, for horizontal infectivity. The PrP–clay complexes persist many years in soils and once ingested, they may orally contaminate sheep and deer.

ACKNOWLEDGMENTS This study was funded by the Euroean Commission FP5 Quality of Life programme (FATEPriDE; QLRT-2001-02723) and the Leverhulme Trust (F/00 181/R). We acknowledge fruitful discussions with Dr. Murray McBride, Cornell University, USA; Dr. Francis Eychenne, INRA Langlade, France; Dr. Giuseppe Ru, CEA, Turin, Italy; and Dr. A. Porquet, Univerisity of Geneva. We finally would like to thank farmers and veterinaries from Italy and Iceland for their warm cooperation.

REFERENCES Abrahams, P. W., & Steigmajer, J. (2003). Soil ingestion by sheep grazing the metal enriched floodplain soils of mid-wales. Environ. Geochem. Health, 25, 17–24.


147

Adams, F. (1984). Crop response to lime in the southern United States. In: F. Adams (Ed). Soil Acidity and Liming. Agronomy Monograph, 12, ASA, CSSA and ASA, Madison, WI, pp. 211–265. Aguzzi, A. (2005). Prion toxicity: All sail and no anchor. Science, 308, 1420–1421. Angeloni, M., & Bini, C. (1992). Trace element concentrations in soil and plants in Western Europe. In: D. C. Adriano (Ed). Biogeochemistry of Trace Metals. Lewis Publishers, Ann Arbor, MI, pp. 19–60. Avena, M. (2002). Acid–base behaviour of clay surfaces in aqueous media. Encyclopedia Surf. Colloid Sci., 37–63. Baeyens, B., & Bradbury, M. H. (1997). A mechanistic description of Ni and Zn sorption on Na-montmorillonite. Part I: Titration and sorption measurements. J. Contam. Hydrol., 27, 199–222. Bastian, F. O., & Fermin, C. D. (2005). Slow virus disease: Deciphering conflicting data on the transmissible spongiform encephalopathies (TSE) also called prion disease. Microsc. Res. Tech., 68, 239–246. Benestad, S. L., Sarradin, P., Thu, B., Scho¨nheit, J., Tranulis, M. A., & Bratberg, B. (2003). Cases of scrapie with unusual features in norway and designation of a new type, Nor98. Vet. Rec., 153, 202–208. Brooks, B. R., Bruccoleri, R. E., Olafson, B. D., States, D. J., Swaminathan, S., & Karplus, M. (1983). CHARMM – A program for macromolecular energy, minimization, and dynamics calculations. J. Comput. Chem., 4, 187–217. Brown, D. R. (2003). Prion protein expression modulates neuronal copper content. J. Neurochem., 87, 377–385. Brown, D. R., Hafiz, F., Glassmith, L. L., Wong, B., Jones, I. M., Clive, C., & Haswell, S. J. (2000). Consequences of manganese replacement of copper for prion protein function and proteinase resistance. EMBO J., 19, 1180–1186. Brown, D. R., Nicholas, R. St. J., & Canevari, L. (2002). Lack of prion protein expression results in a neuronal phenotype sensitive to stress. J. Neurosci. Res., 67, 211–224. Brown, D. R., Qin, K., Herms, J. W., Madlung, A., Manson, J., Strome, R., Fraser, P. E., Kruck, T. A., von Bohlen, A., Schulz-Schaeffer, W., Giese, A., Westaway, D., & Kretzschmar, H. A. (1997). The cellular prion protein binds copper in vivo. Nature, 390, 684–687. Brunger, A. T. (1992). XPLOR – Version 3.1. Yale University Press, New Haven, USA. Burns, C. S., Aronoff-Spencer, E., Legname, G., Prusiner, S. B., Antholine, W. E., Gerfen, G. J., Peisach, J., & Millhauser, G. L. (2003). Copper coordination in the full-length, recombinant prion protein. Biochemistry, 42 (22), 6794–6803. Bush, A. I. (2000). Metals and neuroscience. Curr. Opin. Chem. Biol., 4, 184–191. Casalone, C., Zanusso, G., Acutis, P., Ferrari, S., Capucci, L., Tagliavini, F., Monaco, S., & Caramelli, M. (2004). Identification of a second bovine amyloidotic spongiform encephalopathy: Molecular similarities with sporadic Creutzfeldt-Jakob disease. Proc. Natl. Acad. Sci. U.S.A., 101, 3065–3070. Chapron, Y., Charlet, L., & Porquet, A., (in preparation). PrP–Mn and PrP–Clay interactions: A molecular dynamic and quantum chemistry approach.

148

L. Charlet et al.

Charlet, L., Schindler, P. W., Spadini, L., Furrer, G., & Zysset, M. (1993). Cation adsoption on oxides and clays: The aluminium case. Aquat. Sci., 55, 291–303. Chihota, C. M., Gravenor, M. B., & Baylis, M. (2004). Investigation of trace elements in soil as risk factors in the epidemiology of scrapie. Vet. Rec., 154, 809–813. Colchester, A. C. F., & Colchester, N. T. H. (2005). The origin of bovine spongiform encephalopathy, the human prion disease hypothesis. Lancet, 366, 856–861. Collinge, J. (2005). Molecular neurology of prion disease. J. Neurol. Neurosurg. Psychiatry, 76, 906–919. Cygan, R. T., Liang, J.-J., & Kalinichev, A. G. (2004). Molecular models of hydroxide, oxyhydroxide, and clay phases and the development of a general force field. J. Phys. Chem. B, 108, 1255–1266. Duckworth, O. W., & Sposito, G. (2005). Siderophore–manganese(III) interactions. II. Manganite dissolution promoted by desferrioxamine B. Environ. Sci. Technol., 39, 6045–6051. Ebringer, A., Pirt, J., Wilson, C., Thorpe, C., Tiwana, H., Cunningham, P., & Ettelaie, C. (1998). Bovine spongiform encephalopathy: Comparison between the ‘‘prion’’ hypothesis and the autoimmune theory. J. Nutr. Environ. Med., 8, 265–276. Fernaeus, S., Ha¨lldin, J., Bedecs, K., & Land, T. (2005). Changed iron regulation in scrapie-infected neuroblastoma cells. Mol. Brain Res., 133, 266–273. Fletcher, P., & Sposito, G. (1989). The chemical modelling of clay electrolyte interactions for montmorillonite. Clay Miner., 50, 632–645. Freer, M., & Dove, H. (Ed). (2002). Sheep Nutrition. CAB International, Oxon, UK, 383 pp. Frisch, M. J., Trucks, G. W., Schlegel, H. B., Scuseria, G. E., Robb, M. A., Cheeseman, J. R., Montgomery, J. A., Jr., Vreven, T., Kudin, K. N., Burant, J. C., Millam, J. M., Iyengar, S. S., Tomasi, J., Barone, V., Mennucci, B., Cossi, M., Scalmani, G., Rega, N., Petersson, G. A., Nakatsuji, H., Hada, M., Ehara, M., Toyota, K., Fukuda, R., Hasegawa, J., Ishida, M., Nakajima, T., Honda, Y., Kitao, O., Nakai, H., Klene, M., Li, X., Knox, J. E., Hratchian, H. P., Cross, J. B., Bakken, V., Adamo, C., Jaramillo, J., Gomperts, R., Stratmann, R. E., Yazyev, O., Austin, A. J., Cammi, R., Pomelli, C., Ochterski, J. W., Ayala, P. Y., Morokuma, K., Voth, G. A., Salvador, P., Dannenberg, J. J., Zakrzewski, V. G., Dapprich, S., Daniels, A. D., Strain, M. C., Farkas, O., Malick, D. K., Rabuck, A. D., Raghavachari, K., Foresman, J. B., Ortiz, J. V., Cui, Q., Baboul, A. G., Clifford, S., Cioslowski, J., Stefanov, B. B., Liu, G., Liashenko, A., Piskorz, P., Komaromi, I., Martin, R. L., Fox, D. J., Keith, T., Al-Laham, M. A., Peng, C. Y., Nanayakkara, A., Challacombe, M., Gill, P. M. W., Johnson, B., Chen, W., Wong, M. W., Gonzalez, C., & Pople, J. A. (2004). Gaussian 03, revision C.02. Gaussian, Inc., Wallingford, CT. Gehin, A., Greneche, J.-M., Tournassat, C., Brendle, J., Rancourt, D. G., & Charlet, L. (2007). Reversible surface-sorption-induced electron-transfer oxidation of Fe(II) at reactive sites on a synthetic clay mineral. Geochim. Cosmochim. Acta, 71, 863–876.


149

Georgsson, G., Sigurdarson, S., & Brown, P. (2006). Infectious agent of sheep scrapie may persist in the environment for at least 16 years. J. Gen. Virol., 87, 3737–3740. Govaerts, C., Wille, H., Prusiner, S. B., & Cohen, F. E. (2004). Evidence for assembly of prions with left-handed beta-helices into trimers. PNAS, 101, 8343–8347. Grassi Zucconi, G., Cipriani, S., Scattoni, R., Balgkouranidou, I., Hawkins, D. P., & Ragnarsdottir, K. V. (2007). Copper deficiency elicits glial and neuronal response typical of neurodegenerative disorders. Neuropathol. Appl. Neurobiol., 33, 212–225. Hawkins, D. P. (2006). Environmental Factors affecting the Development of Scrapie in Icelandic Sheeps. PhD Thesis, University of Bristol, UK. Hijazi, N., Shaked, Y., Rosenmann, H., Ben-Hur, T., & Gabizon, R. (2003). Copper binding to PrPc may inhibit prion disease progression. Brain Res., 993, 192–200. Hopp, P., Ulvund, M. J., & Jarp, J. (2001). A case control study on scrapie in Norwegian sheep flocks. Prev. Vet. Med., 51, 183–198. Humphrey, W., Dalke, A., & Schulten, K. (1996). VMD – Visual Molecular Dynamics. J. Mol. Graph., 14, 33–38. Hureau, C., & Charlet, L. (in preparation). EPR investigation of the prion and trien (Cu) complex speciation change upon adsorption onto clay minerals. Geochem. T. Hureau, C., Charlet, L., Dorlet, P., Gonnet, F., Spadini, L., Anxolabe´here`-Mallart, E., & Girerd, J.-J. (2006). A spectroscopic and voltametric study of the pHdependent Cu(II) coordination to the peptide GGGTH: Relevance to the fifth Cu(II) site in the prion protein. J. Biol. Inorg. Chem., 11, 735–744. Jeffrey, M., Gonza´lez, L., Espenes, A., Press, C. McL., Martin, S., Chaplin, M., Davis, L., Landsverk, T., MacAldowie, C., Eaton, S., & McGovern, G. (2006). Transportation of prion protein across the intestinal mucosa of scrapiesusceptible and scrapie-resistant sheep. J. Pathol., 209, 4–14. Johnson, C. J., Phillips, K. E., Schramm, P. T., McKenzie, D., Aiken, J. M., Aiken, J. M., & Pedersen, J. A. (2006). Prions adhere to soil minerals and remain infectious. PLoS, 2, 296–302. Kendall, N. R., Mackenzie, A. M., & Telfer, S. B. (2001). Effect of a copper, cobalt and selenium soluble glass bolus given to grazing sheep. Livestock Prod. Sci., 68, 31–39. Kirpichtchikova, T. A., Manceau, A., Spadini, L., Panfili, F., Matthew, A., Marcus, M. A., & Jacquet, T. (2006). Speciation and solubility of heavy metals in contaminated soil using X-ray microfluorescence, EXAFS spectroscopy, chemical extraction, and thermodynamic modelling. Geochim. Cosmochim. Acta, 71, 2163– 2190. Lee, J., Knowles, S. O., & Judson, G. J. (2002). Trace-element and vitamin nutrition of grazing sheep. In: M. Freer, & H. Dove (Eds). Sheep Nutrition. CAB International, Oxon, UK. Leech, A. F., & Thornton, I. (1987). Trace elements in soils and pasture herbage on farms with bovine hypocupremia. J. Agric. Sci. Camb., 108, 591–597.

150

L. Charlet et al.

Lindh, U. (2004). Biological functions of the elements. In: Essentials of Medical Geology: Impacts of the Natural Environment on Public Health. Selinus, Elsevier Academic Press, Burlington, MA, USA, pp. 115–160. Loue´, A. (1986). Oligo-Elements en Agriculture; SCPA, Agri Nathan International, Paris, France, 339 pp. McBride, M. B. (1994). Environmental Chemistry of Soils. Oxford University Press, New York. McBride, M. B. (2006). Neurodegenerative diseases and the environment: The soil– health link. Hammond Lectures on Environment, Energy and Resources, University of Guelp. McGowan, J. P. (1922). Scrapie in sheep. Scott. J. Agric., 5, 365–375. McLaren, R. G., Williams, J. G., & Swift, R. S. (1983). The adsorption of copper by soil samples from Scotland at low equilibrium solution concentrations. Geoderma, 31, 97–106. Milhausen, G. L. (2004). Copper binding in the prion protein. Acc. Chem. Res., 37, 79–85. Miller, M. W., Williams, E. S., Hobbs, N. T., & Wolfe, L. L. (2004). Environmental source of prion transmission in mule deer. Emerg. Infect. Dis., 10, 1003–1006. Norde, W. (2000). Proteins at solid surfaces. In: A. Baszkin, & W. Norde (Eds). Physical Chemistry of Biological Interfaces. Marcel Dekker, New York, pp. 115–136. Phillips, C. J. C. (2000). Further aspects of the use of individual animals as replicates in statistical analysis. Appl. Anim. Behav. Sci., 69, 85–88. Phillips, J. C., Braun, R., Wang, W., Gumbart, J., Tajkhorshid, E., Villa, E., Chipot, C., Skeel, R. D., Kale, L., & Schulten, K. (2005). Scalable molecular dynamics with NAMD. J. Comput. Chem., 26, 1781–1802. Prusiner, S. B., Telling, G., Cohen, F. E., & DeArmond, S. J. (1996). Prion diseases of humans and animals. Semin. Virol., 7, 159–173. Purdey, M. (2000). Ecosystems supporting clusters of sporadic TSEs demonstrate excesses of the radical-generating divalent cation manganese and deficiencies of antioxidant co factors Cu, Se, Fe, Zn. Does a foreign cation substitution at prion protein’s Cu domain initiate TSE? Med. Hypotheses, 54, 278–306. Purdey, M. (2005). Metal microcrystal pollutants: The heat resistant, transmissible nucleating agents that initiate the pathogenesis of TSEs. Med. Hypotheses, 65, 448–477. Quincampoix, H., & Radcliff, R. G. (1992). A 13P NMR study of the adsorption of bovine serum albumin on montmorillonite using phosphate and the paramagnetic cation Mn2+. Modification of conformation with pH. J. Colloid Interface Sci., 148, 343–352. Ragnarsdottir, K. V., & Charlet, L. (2002). Environmental fate of the trace elements copper and manganese in soils. In: D. R. Brown (Ed). Prion Diseases and Copper Metabolism. Horwood, Colchester. Ragnarsdottir, K. V., & Hawkins, D. P. (2005). Bioavailable copper and manganese in soils from Iceland and their relationship with scrapie occurrence in sheep. J. Geochem. Explor., 88, 228–234.


151

Revault, M., Quiquampoix, H., Baron, M. H., & Noinville, S. (2005). Fate of prion in soil: Trapped conformation of full-lengh ovine prion protein induced by adsorption on clays. Biochim. Biophys. Acta, 1724, 367–374. Rigou, P., Rezaei, H., Grosclaude, J., Staunton, S., & Quinquampoix, H. (2006). Fate of prions in soil: Adsorption and extraction by electroelution of recombinant ovine prion protein from montmorillonite and natural soils. Environ. Sci. Technol., 40, 1497–1503. Roman-Ross, G., Charlet, L., Eychenne, F., Perfetti, G., & Ru, G. (in preparation). Exposure to soil high Mn and low Cu concentrations and the development of scrapie. Environ. Geochem. Health. Rose, A. W., Hawkes, H. E., & Webb, J. S. (1979). Geochemistry in Mineral Exploration. 2nd Ed., Academic Press, New York. Salminen, R. (Ed.) (2005). Geochemical Atlas of Europe. Part 1 – Background Information, Methodology and Maps, available online at http://www.gtk.fi/publ/ foregsatlas/ Salminen, R., Tarvainen, T., Demetriades, A., Duris, M., Fordyce, F. M., Gregorauskiene, V., Kahelin, H., Kivisilla, J., Klaver, G., Klein, H., Larson, J. O., Lis, J., Locutura, J., Marsina, K., Mjartanova, H., Mouvet, C., O’Connor, P., Odor, L., Ottonello, G., Paukola, T., Plant, J. A., Reimann, C., Schermann, O., Siewers, U., Steenfelt, A., Van der Sluys, J. B., de Vivo, B., & Williams, L. (1998). FOREGS geochemical mapping. Field Manual. Geologian tutkimuskeskus, Opas – Geological Survey of Finland, Guide 47. Sauve´, S., McBride, M., Norvell, W. A., & Hendershot, W. (1997). Copper solubility and speciation of in situ contaminated soils: Effects of copper level, pH, and organic matter. Water Air Soil Pollut., 100, 133–149. Seeger, H., Heikenwalder, M., Zeller, N., Kranich, J., Schwarz, P., Gaspert, A., Burkhardt, S., Miele, G., & Aguzzi, A. (2005). Coincident scrapie infection and nephritis lead to urinary prion excretion. Science, 310, 324–326. Servagent-Noiville, S., Revault, M., Quinquampoix, H., & Baron, M. H. (2000). Conformational changes of bovine serum albumin by adsorption on different clay surfaces: FTIR analysis. J. Colloid Interface Sci., 221, 273–283. Sharpe, A., McElroy, M., Bassett, H., & Sweeney, T. (2006). Clinical and pathological features of experimental scrapie in Irish blackface mountain sheep. Res. Vet. Sci., 80, 71–78. Sposito, G. (1981). The Thermodynamics of Soil Solution. Clarendon Press, Oxford, NY. Sposito, G. (2004). The Surface Chemistry of Natural Particles. Oxford University Press, New York, NY, 242 pp. Stadler, M., & Schindler, P. W. (1993). Modeling of H+ and Cu2+ adsorption on calcium-montmorillonite. Clays Clay Miner., 41, 288–296. Stauton, S., & Quinquampoix, H. (1994). Adsorption and conformation of bovine serum albumin on montmorillonite: Modification of the balance between hydrophobic and electrostatic interactions by proteins methylation and pH variation. J. Colloid Interface Sci., 166, 89–94.

152

L. Charlet et al.

Suttle, N. F. (1987). Geochemistry and health. In: I. Thornton (Ed). Proceedings of the 2nd International Symposium on Geochemistry and Health, Science Reviews Ltd., Northwood, pp. 69–79. Thackray, A. M., Knight, R., Haswell, S. J., Bujdoso, R., & Brown, D. R. (2002). Metal imbalance and compromised antioxidant function are early changes in prion disease. Biochem. J., 362, 253–258. Tebo, B. M., Bargar, J. R., Clement, B. G., Dick, G. J., Murray, K. J., Parker, D., Verity, R., & Webb, S. M. (2004). Biogenic manganese oxides: Properties and mechanisms of formation. Annu. Rev. Earth Planet. Sci., 32, 287–328. Treiber, C., Simons, A., & Multhaup, G. (2006). Effect of copper and manganese on the de novo generation of protease-resistant prion protein in yeast cells. Biochemistry, 45, 6674–6680. Treiber, C., Simons, A., Strauss, M., Hafner, M., Cappai, R., Bayer, T. A., & Multhaup, G. (2004). Clioquinol mediates copper uptake and counteracts copper efflux activities of the amyloid precursor protein of Alzheimer’s disease. J. Biol. Chem., 279, 51958–51964. Underwood, E. J., & Suttle, N. F. (2002). The Mineral Nutrition of Livestock. 3rd Ed., CAB International, Oxon, UK. Weeks, H. P., & Kirkpatrick, C. M. (1976). Adaptations of white-tailed deer to naturally occurring sodium deficiencies. J. Wildl. Manage., 40, 610–625. Whitehead, D. C. (2000). Nutrient Elements in Grassland: Soil–Plant–Animal Relationships. CAB International, Oxon, UK. Whitelaw, A., Armstrong, R. H., Evans, C. C., & Fawcett, A. R. (1979). Study of the effects of copper deficiency in Scottish blackface lambs on improved hill pasture. Vet. Rec., 104, 455–460. Williams, M. W., Miller, M. W., Kreeger, T. H., Kahn, R. H., & Thorne, E. T. (2002). Chronic wasting disease of deer and elk: A review with recommendations for management. J. Wildl. Manage., 66, 551–563. Wong, B. S., Brown, D. R., Pan, T., Whiteman, M., Liu, T., Bu, X., Li, R., Gambetti, P., Olesik, J., Rubenstein, R., & Sy, M. N. (2001). Oxidative impairment in scrapie-infected mice is associated with brain metal perturbations and altered antioxidant activities. J. Neurochem., 79, 689–698. Zachara, J. M., & Smith, S. C. (1994). Edge complexation reactions of cadmium on specimen and soil derived smectite. Soil Sci. Soc. Am. J., 58, 762–769.


Chapter 6

Associations between Iron Oxyhydroxide Nanoparticle Growth and Metal Adsorption/Structural Incorporation C. S. Kim1,, C. J. Lentini1 and G. A. Waychunas2 1

Department of Physical Sciences, Chapman University, Orange, CA 92866, USA Earth Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA 2

ABSTRACT The interaction of metal ions and oxyanions with nanoscale mineral phases has not been extensively studied despite the increased recognition of their prevalence in natural systems as a significant component of geomedia. A combination of macroscopic uptake studies to investigate the adsorption behavior of As(V), Cu(II), Hg(II), and Zn(II) onto nanoparticulate goethite (a-FeOOH) as a function of aging time at elevated temperature (751C) and synchrotron-based X-ray studies to track changes in both the sorption mode and the rate of nanoparticle growth reveals the effects that uptake has on particle growth. Metal(loid) species which sorb quickly to the iron oxyhydroxide particles (As(V), Cu(II)) appear to passivate the particle surface, impeding the growth of the nanoparticles with progressive aging; in contrast, species that sorb more slowly (Hg(II), Zn(II)) have considerably less impact on particle growth. Progressive changes in the speciation of these particular metals suggest shifts in the mode of metal uptake with time, possibly indicating structural incorporation of the metal(loid) into the nanoparticle; this is supported by the continued increase in uptake concomitant with particle growth, implying that metal species may transform from surface-sorbed species to more structurally incorporated forms. This type of incorporation would have implications for the long-term fate and mobility of metals in contaminated regions, and affect strategies for potential remediation/modeling efforts.

Corresponding author. Tel.: +1-714-628-7363; Fax: +1-714-532-6048;

E-mail: [email protected] (C.S. Kim).

154

C. S. Kim et al.

6.1. Introduction The processes of metal sorption to geomedia represent a critically important area in the field of environmental geochemistry. In particular, the role of iron oxyhydroxides in these processes is significant due to their natural abundance in the environment and the strong affinity they possess for dissolved metal species, thus influencing the fate and transport of such pollutants. The initial rapid adsorption of metals onto iron oxyhydroxides can, with time and/or changing geochemical conditions, lead to slower rate-limiting uptake mechanisms including diffusion into micropores, (co-)precipitation, surface precipitation, and structural incorporation/substitution (Sposito, 1984; Dzomback and Morel, 1990; Scheidegger and Sparks, 1996; Axe and Anderson, 1998; Trivedi and Axe, 1999; Ford et al., 2001), further impacting the mobility, potential bioavailability, and long-term sequestration of metal contaminants. A number of macroscopic uptake studies have been conducted demonstrating the behavior of metal uptake onto iron oxyhydroxide phases (Dzombak and Morel, 1986; Scheidegger et al., 1997; Subramaniam and Yiacoumi, 2001; Trivedi and Axe, 2001; Villalobos et al., 2001; Dyer et al., 2004; Egirani et al., 2005). More recently, investigators have applied X-ray-based spectroscopic methods to study the precise mechanisms of metal uptake onto mineral surfaces at the atomic scale, gaining considerable insight into the different processes by which metals are transferred from the aqueous to the solid phase (Fendorf et al., 1997; Collins et al., 1999b; Morton et al., 2001; Farquhar et al., 2002; Foster et al., 2003; Kim et al., 2004a; Trivedi et al., 2004). Few studies, however, have explored the interaction of metals with nanosized particles, which are postulated to be of particular importance and abundance in natural systems due to their extremely high surface areas, possible enhanced or altered reactivity (e.g., with metal sorbates), and long lifetimes in suspension, leading to long distance transfer in aqueous/ subsurface systems (Vilks et al., 1997; Kersting et al., 1999; Novikov et al., 2006). Furthermore, nanoparticles are likely to aggregate and/or grow in natural settings, although the effects of these processes on either previously sorbed metals or the potential of such multiparticle aggregates for additional uptake have not been characterized. Our initial studies suggest that with increasing particle size at the nanoscale, the mode of uptake may change subtly while the degree of surface coverage may actually increase, indicating that the mode and quantitative amount of uptake can vary with particle size (Waychunas et al., 2005); other macroscopic studies also suggest differences in the pH-based uptake behavior of metals onto nanosized sorbents as a function of particle size and pH (Madden et al., 2006). These recent findings

Associations between Iron Oxyhydroxide Nanoparticle Growth

155

further support the conjecture that nanoparticles may act and react in fundamentally distinct ways from those of their macroscale counterparts, hence motivating the current work. The studies previously referenced establish the basis for combining traditional batch sorption experiments, used to characterize the general uptake behavior of metals onto nanoscale particles, with modern spectroscopic techniques designed to explore differences at the atomic level between the modes of metal sorption onto nanosized particles during growth. The primary goal of this work is to understand the behavior and speciation of metal uptake to nanoscale particles as they are growing, predicting that uptake will both increase and become more associated with structurally incorporated species with continued growth; additionally, surface uptake is anticipated to have a measurable impact on the rate of nanoparticle growth, introducing surface poisoning/passivation effects that can inhibit both aggregation-based and ripening-based particle growth mechanisms. The determination of metal(loid)-specific behavior under these conditions is critical to predicting more accurately the mobility and availability of metals that sorb to nanoparticles and their larger scale aggregates. In this study, both macroscopic and spectroscopic strategies are applied to characterize the sorption and incorporation of four different metal(loid) ions (As(V), Cu(II), Hg(II), and Zn(II)) onto iron oxyhydroxide nanoparticles as a function of nanoparticle aging/growth induced by elevated temperature. Specifically, macroscopic uptake experiments are used to track changes in metal(loid) uptake as a function of time while synchrotron-based X-ray methods including micro-X-ray diffraction (mXRD) and extended X-ray absorption fine structure (EXAFS) spectroscopy are applied to explore both the structural/size evolution of particles as a function of contaminant uptake and the mode of metal uptake as a function of aging time. Together these methods can provide detailed macroscopic and atomic-level information on both the mechanisms of metal uptake onto nanoparticles and the related effects on the dynamic, growing substrate to which the metals are becoming incorporated. Referring to the final chapter (entitled ‘‘Priorities for Future Metal Adsorption Research’’) of the original Adsorption of Metals by Geomedia volume (Jenne, 1998), we have attempted to address some of the proposed goals and considerations for future adsorption research. System characterization was enhanced by addressing issues such as pH changes, Fe(III) release in response to cation adsorption, and aging time (Redden et al., 1998). In an attempt to better characterize the process of metal adsorption to a single adsorbent (nanoscale goethite) a multi-metal data approach was taken as suggested by Kinniburgh et al. (1998). Lastly, time-dependent data

156

C. S. Kim et al.

were analyzed and transitions in modes of uptake were directly modeled using EXAFS spectroscopy.

6.2. Experimental 6.2.1. Nanoparticle Preparation A suspension of goethite (a-FeOOH) nanoparticles was prepared according to Guyodo et al. (2003) by first adding a 0.25 M NaHCO3 solution dropwise to an equivalent volume of a 0.20 M Fe(NO3)3 solution in a LDPE screwcapped bottle, both solutions having been previously filtered through a 0.2 mm filter. The resulting mixture was then placed on an agitator for 30 min with periodic opening of the cap to release CO2 buildup. Once gas release was complete the solution was placed in a conventional kitchen microwave and heated at high power in intervals of 30 s, with periodic mixing between heating intervals, for a period of approximately 3 min or until boiling of the suspension had just begun. This procedure initiates nucleation of nanosized goethite particles (Guyodo et al., 2003). The mixture was then immediately placed into an ice bath to halt the nanoparticle formation process; once cool, the solution was transferred to a 1,000 MWCO dialysis tubing and submerged in a large volume of DI water which was replaced several times a day until reaching a pH of 5.070.1. The resulting nanoparticle suspension was found through previous characterization (Waychunas et al., 2005) to be a monodisperse suspension of 5-nm diameter oblong particles with a solid concentration of 6.74 g/l and a BET surface area of 306 m2/g. The initial unaged suspension was stored in a refrigerator at 41C. As a basis of comparison with the nanoparticulate goethite, a batch of macroscale goethite was prepared by base (NaOH) titration of a ferric nitrate solution followed by equilibration at 601C and dialysis as described by Atkinson et al. (1968), resulting in acicular needles of 200 nm 30 nm dimensions and surface area of 91 m2/g. So-called ‘‘two-line’’ ferrihydrite was also synthesized following the procedure of Schwertmann and Cornell (1991). Eight grams of Fe(NO3)3 9H2O was dissolved in 100 ml of deionized water and the pH adjusted to 7–8 using 1 M KOH and vigorous stirring. Samples were placed into 50 ml centrifuge tubes, centrifuged for 15 min at 3,000 RPM, and either washed with 45 ml of deionized water or dialyzed in a 1,000 MWCO dialysis tubing to remove residual electrolytes before drying. No difference in the mXRD patterns of the ferrihydrite samples was observed as a function of the washing method.


157

6.2.2. Sample Preparation Batch uptake experiments were performed in 1L glass beakers with magnetic stir bars under ambient atmospheric conditions with a background electrolyte concentration of 0.1 M NaNO3. Separate 400 ml portions of the initial nanoparticle suspension were first combined with 60 ml of 5 mM solutions of AsHNa2O4 (374.6 mg/l), Cu(NO3)2 (317.7 mg/l), Hg(NO3)2 (1,002 mg/l) and Zn(NO3)2 (327.0 mg/l). A fifth batch of nanogoethite was exposed to an equivalent volume of deionized water in place of the metal solutions and served as the control/blank for the experiment. Macroscopic uptake of each metal(loid) was induced by adding 0.1 M NaOH in 50 ml aliquots to obtain a pH level of 6.070.1; although the metal cations and oxyanions display opposite pH-dependent uptake behavior (with the cations exhibiting maximum sorption at high pH and the oxyanion at low pH), this pH level was anticipated through preliminary testing to achieve sufficient uptake for our experiments. Appropriate amounts of 0.5 M NaNO3 and deionized water were added to adjust the ionic strength and metal(loid) concentration in the final metal and control solutions to 0.1 and 0.5 mM, respectively, resulting in a final volume of 600 ml with a solid concentration of 4.49 g/l. Following preparation, each metal suspension was separated into labeled 30 ml LDPE bottles with screw-cap lids. Samples were placed in a conventional laboratory oven or hot water bath/shaker at 78721C to induce particle growth in the presence of the metal(loid), with samples collected at varying time intervals up to a total duration of 7 days. Our previous characterization of similar nanoparticle suspensions with mXRD, dynamic light scattering, transmission electron microscopy, and small- and wide-angle X-ray scattering suggests that nanoparticle aggregation plays a significant role in the early growth stages of the particles, although ripening-based growth may become a more significant process for the continued growth of the larger aggregates (Waychunas et al., 2005). The resulting particle growth occurs at a considerably faster rate than at room temperature, where stable nanoclusters have been shown to persist in suspension for 10 weeks with minimal continued aggregation (Gilbert et al., 2007). Upon removal from the elevated temperature conditions, suspensions were immediately centrifuged at 3,000 RPM for 15 min and the supernatant decanted, recorded for pH level, filtered with a 0.2 mm filter, and acidified using 20 ml aliquots of concentrated HNO3 until pHo2 prior to analysis by ICP-OES. The centrifuged solids to be analyzed by mXRD were cleaned by resuspending the particles twice in 30 ml of deionized water followed by additional centrifugation and decanting. The washed pellets were then

158

C. S. Kim et al.

resuspended in a minimal volume of deionized water to maximize recovery from the centrifuge tubes and placed onto watch glasses to be dried in an oven overnight at 601C. Once dry, samples were stored in individual capped glass vials for later mXRD analysis. The centrifuged solids to be analyzed with EXAFS spectroscopy were spread on a filter paper to remove excess liquid, and then loaded as moist pastes in Teflon sample holders and sealed with Kapton tape. All ‘‘0 h’’ (unaged) samples were processed as quickly as possible with no subsequent heating; approximately 30 min elapsed between the preparation of the sample and the separation of pellet and supernatant. 6.2.3. Inductively Coupled Plasma Optical Emission Spectroscopy (ICP-OES) Analysis A TJA IRIS Advantage/1000 Radial ICP Spectrometer was used to determine the metal(loid) concentrations in the filtered and acidified supernatants. Standards of 1, 10, and 100 ppm (mg/l) concentrations of the added metal(loid)s as well as Fe(III) were prepared under the same pH and ionic strength conditions as the samples for proper instrumental calibration prior to analysis. The residual metal(loid) concentrations measured in the supernatants were then used to calculate the percentage of metal(loid) uptake from solution assuming minimal uptake to vessel walls (Kim et al., 2004a). Supernatant Fe(III) concentrations were measured to assess the extent of nanoparticle dissolution or passage through the filters. 6.2.4. Synchrotron X-Ray Microdiffraction Dried samples were powdered with a mortar and pestle, resuspended in 20 ml of ethanol, and placed on a clean low-background quartz (1 1 1) C-cut single crystal to dry prior to mXRD analysis at bend magnet beamline 7.3.3 of the Advanced Light Source (ALS) at Lawrence Berkeley National Laboratory using a four-crystal Si (1 1 1) monochromator tuned to an energy of 6,000 eV. Diffraction images were collected for 1,200 s at a beam size of 7 mm 10 mm onto a MAR CCD detector at a sample-detector distance of 136.617 mm. This analysis allowed data collection over a Q range of 1.3–3.7 A˚1. A diffraction image was collected on a clean spot of the single crystal prior to sample data collection for background subtraction purposes. Integration of the raw mXRD diffraction images was performed using X-ray Microdiffraction Analysis Software Version 5.1 (X-MAS) (Tamura,


159

2007) over a 2y range of 25–751 and a w range of 22 to 2 with a resolution of 0.100. The statistical software analysis program Origin 7.5 (OriginLab Corporation, 2006) was used to quantify the rate of nanoparticle growth through diffraction peak integration. The (0 0 1) diffraction peak was selected for integration and particle size analysis as it corresponds with the primary growth direction along the c-axis of goethite crystals (Rakovan et al., 1999). Each integrated diffraction pattern was processed to minimize background variations and beam flux variability; the latter correction was accomplished by normalizing each data point by the beam intensity (I0) recorded just prior to each data collection step. Data were converted to Q-space and a Gaussian fit performed on the (0 0 1) peak. Following baseline subtraction of the fitted peak, Origin was used to calculate the integrated area of the peak as represented by the best possible Gaussian fit. As a proxy for tracking the growth rates of the nanoparticles along the c-axis, the ratio of the integrated area under the (0 0 1) peak (25–321 2y) to the integrated area of the entire dataset (25–751 2y), or A001/Atotal, was calculated and plotted for each set of aged samples. Application of the Scherrer equation to quantify particle size was found to be unreliable due to variability in baseline determinations and low absolute peak intensity, so this method was not applied. 6.2.5. X-Ray Absorption Spectroscopy Based on results from the macroscopic uptake experiments which showed progressive uptake of Hg(II) and Zn(II) over time (see the section ‘‘Results’’), X-ray absorption spectroscopy (XAS) analysis of selected Hg(II) and Zn(II) sorption samples was performed on wiggler-magnet beamlines 10-2 and 11-2 at Stanford Synchrotron Radiation Laboratory (SSRL). Samples were analyzed as moist pastes on either double Si (1 1 1) (BL 10-2) or Si (2 2 0) (BL 11-2) monochromator crystals detuned 30% down from the maximum beam intensity to reject higher order harmonic signals. Zn K-edge (9,659 eV) and Hg LIII-edge (12,284 eV) XAS spectra were collected on samples at room temperature in fluorescence yield mode using a 13 (BL10-2) or 30 (BL 11-2) element high-throughput germanium detector. This technique is preferred over the use of transmission-collected data or ion-chamber-collected fluorescence for low concentration samples (Stern and Heald, 1979; Waychunas and Brown, 1994). The samples were placed so that their normals were at a 451 angle to the incoming beam and the germanium detector window, in the plane of the synchrotron X-ray electric vector, so that elastic scattering from the sample was minimized. Aluminum

160

C. S. Kim et al.

filters and Soller slits were used to minimize Fe fluorescence and Soller slits were also used to further limit elastic scattering from reaching the detector (Stern and Heald, 1979). Zn metal foil and HgCl2 powder were used as monochromator energy calibration standards. The number of scans for each sample was determined by the amount of uptake measured through the macroscopic uptake experiments, with lower concentrations requiring additional scans; Zn samples required from 7 to 10 scans and Hg samples from 10 to 23 scans. Data were analyzed using the SixPACK software package Version 0.57 (Webb, 2007). Each XAS scan was calibrated in energy by using the first derivative of the relevant calibration standard to detect any changes in assigned monochromator energy, and then correcting each data scan accordingly. To insure correct amplitude measurements of EXAFS spectra due to detector deadtime, sample spectra collected at BL 11-2 were deadtimecorrected from reference data in the SamView interface and then averaged into one file. Background subtraction was automatically performed (k-weighting ¼ 3, R background ¼ 1 A˚) using the automatic background removal algorithm in SixPACK; the resulting spectra were fit using a spline with an average of eight spline knots. After removal of the spline-fit background function the data were converted to k-space with a k3 weighting. Prior to k-space fitting, the spectra were Fourier-transformed to produce EXAFS structure functions. Fitting was performed over a k-range of 2–12 A˚1 for Hg and 3–10.5 A˚1 for Zn while the R-range was expressed from 0 to 4 A˚ for all samples. EXAFS single scattering paths used to fit the background subtracted, k3-weighted data were created in SixPACK using FEFF 6l (Zabinsky et al., 1995). The approach to fitting EXAFS spectra was to model each feature observed in the Fourier transform separately with its own individual path, and having obtained suitable single peak (or ‘‘shell’’ fits), then fitting the spectrum as a whole using parameters generated from the single-shell fitting results. For both Hg and Zn, the feature representing the nearestneighboring atomic shell was first fit to obtain values for oxygen neighbor coordination number (CN), interatomic bond distance (R), and energy shift (E0). The E0 value obtained from fitting the first shell was subsequently used for all further shell fitting. The amplitude reduction factor (S 20 ) for all shells was fixed at 0.9 for Hg and 0.81 for Zn. S 20 for the Hg samples was based on our previous EXAFS fitting in which S 20 was allowed to drift for wellcharacterized crystalline model compounds (Kim et al., 2004a,b). The Zn S 20 was determined by first obtaining seed values for all other parameters and allowing S 20 to float, resulting in a S 20 value within the range of values (0.71–0.9) used in other studies (Trivedi et al., 2001; Waychunas et al., 2002;


161

Roberts et al., 2003; Nachtegaal and Sparks, 2004; Pan et al., 2004; Trivedi et al., 2004; Grafe and Sparks, 2005). The factor s2, which serves as a type of Debye–Waller factor and represents the measure of disorder around the atom of interest, was generally allowed to float when fitting the first shell but was fixed at values consistent with those of sorption complexes for second and third shells (0.01 A˚2) as determined by our own studies and those of other experimenters in which the Debye–Waller factor for these distant features was allowed to float, typically resulting in s2 values of 0.01 A˚2 (Kim et al., 2004a,b). To improve fitting of the Hg(II) data, which suffered from lower quality, s2 was floated for each shell, and then fixed at its optimum value of either 0.005 or 0.007 A˚2 when fitting subsequent shells. Generally, coordination numbers generated from such fitting methods are considered to be accurate within 715–30% of their reported values (Trivedi et al., 2001; Waychunas et al., 2002; Roberts et al., 2003; Grafe et al., 2004; Nachtegaal and Sparks, 2004; Pan et al., 2004); however, fixing the s2 parameter during the fits artificially reduces the errors associated with coordination numbers as the two variables are correlated. In order to obtain a more accurate representation of the uncertainties in the reported coordination numbers, each fit was separately conducted with s2 allowed to float and the resulting errors in the coordination numbers transferred to those of the fitting results where s2 was fixed.

6.3. Results 6.3.1. Macroscopic Uptake Macroscopic uptake results for all metal(loid)s as a function of aging time are shown in Fig. 6.1, with uptake expressed in micromoles of metal(loid) removed per gram of nanogoethite present in suspension. At the earliest sampling point (representing 0 h of uptake at elevated temperature), As(V) and Cu(II) are already completely removed from solution while Hg(II) and Zn(II) exhibit considerably lower initial uptake onto the nanoparticles, with approximately 15 and 40 mmol/g of the metal removed from solution, respectively. As the samples are allowed to age for longer periods of time, progressive uptake of Hg(II) and Zn(II) onto/into the nanoparticles is observed; by the end of 4 days of aging, approximately 50% (60 mmol/g) of the Hg(II) (85% by the end of 7 days; data not shown) and 90% (100 mmol/g) of Zn(II) have been removed. Uptake of As(V) and Cu(II) remains essentially complete over the same aging period. Overall, the extent of uptake of each metal(loid) is ordered as follows: As(V)ECu(II)>Zn(II)>Hg(II).

162

C. S. Kim et al.

Figure 6.1: Macroscopic Batch Metal(loid) Sorption onto Nanoscale Goethite (4.49 g/l) as a Function of Aging Time. [Me]I ¼ 0.5 mM, pH 6, I ¼ 0.1NaNO3, T ¼ 78.01C. Error Bars Accounting for Instrumental Measurement Variability are Contained within the Data Points. Tracking the pH of the reacting suspensions (Fig. 6.2) shows an initial drop in pH within the first 8 h followed by an upward drift in pH with time for the remainder of the experiment. This is thought to influence the observed macroscopic uptake behavior of the different metal(loid)s and is discussed in more detail in the ‘‘Discussion’’ section. Measured concentrations of Fe(III) in all supernatants were uniformly low, corresponding to o0.5% of the total iron initially introduced in the form of nanogoethite. This excludes nanoparticle dissolution or passage of suspended nanoparticles through the filters during the filtration step as significant factors and supports the conclusion that the supernatant metal(loid) concentrations accurately reflect the unsorbed fraction of the specific metal(loid).

6.3.2. Synchrotron X-Ray Microdiffraction Figure 6.3 displays mXRD stack plots of the nanogoethite solids aged in the presence of Hg(II), Zn(II), Cu(II), and As(V) (Fig. 6.3a, c–e) as well as those aged in the absence of any metals (Fig. 6.3b). The latter samples (referred


163

6.4

6.2

6

pH

As 5.8

Blank Cu Hg

5.6

Zn 5.4

5.2 0

20

40

60

80

100

Hours Aged

Figure 6.2: Measured pH Values in Metal(loid)–Nanoparticle Solutions as a Function of Aging. from this point forward as the ‘‘control’’ samples) exhibit a clear trend wherein the diffraction patterns become more resolved and individual peaks more distinct with subsequent aging. These peaks correspond directly to those of goethite as shown by the agreement with the mXRD pattern of the macroscale goethite and the dissimilarity with that of the synthesized ferrihydrite (Fig. 6.4). There is no evidence for shifting of peaks or the appearance of new peaks with aging, implying that no significant compositional change or phase transformation takes place, and hence that the particles remain as goethite throughout the aging process. This is consistent with our previous observations during the aging of goethite nanoparticles from 0 to 25 days (Waychunas et al., 2005), although subtle surface rearrangements may be occurring within the first 24 h that affect the surface structure yet do not appear to have a significant influence on the metal(loid) uptake trends. Our previous XRD and Fe K-edge EXAFS studies (Waychunas et al., 2005) also concluded that this continuous increase in structural ordering as a function of aging was observed as a result of a greater proportion of Fe atoms being associated with the bulk as compared to surface-terminated Fe atoms. Correspondingly, the evolution in the diffraction patterns can be attributed to

C. S. Kim et al.

a

b 100

60

90 50

80

96 hours

72 hours 30

48 hours 20

70

Relative Intensity

Relative Intensity

40

96 hours

24 hours

60

72 hours 50 40

48 hours 30 20

24 hours

10 10

0 hours

0 1.3

0 1.3

1.7

2.1

2.5

2.9

3.3

3.7

0 hours 1.7

2.1

2.5 2.9 Q(Å-1

3.3

3.7

)

Q(Å-1)

c

d 50

60

45 50

40

96 hours 40

96 hours 72 hours

30

48 hours

20

Relative Intensity

35 30

72 hours

25 20

48 hours

15

24 hours

10

24 hours

10

5 0 1.3

0 hours

0 hours 1.7

2.1

2.5

2.9

3.3

0 1.3

3.7

1.7

2.1

Q(Å-1)

2.5

Q(Å-1) e 50

96 hours

45 40

72 hours

35

Relative Intensity

Relative Intensity

164

30

48 hours

25 20 15

24 hours 10 5

0 hours

0 1.3

1.7

2.1

2.5

Q(Å-1)

2.9

3.3

3.7

2.9

3.3

3.7


165

Figure 6.4: Synchrotron X-Ray Diffraction Patterns for Synthesized Goethite and Two-Line Ferrihydrite Compared with the Control at 0 and 96 h Aging.

Figure 6.3: Synchrotron Micro-X-Ray Diffraction (mXRD) Patterns for Nanoparticles Aged in the Presence of: (a) Hg(II); (b) No Metal (Control); (c) Zn(II); (d) Cu(II); and (e) As(V). Patterns are Arranged in Order of Decreasing Aging Time (Most Aged at Top); the Time Interval between Patterns is 8 h.

166

C. S. Kim et al.

increasing particle/aggregate size and structural ordering as a result of aggregation-based nanoparticle growth as documented previously through TEM and SAXS/WAXS analysis (Guyodo et al., 2003; Waychunas et al., 2005). Comparison of the diffraction patterns collected from nanoparticles aged in the presence of metal(loid)s with those of the control samples reveals relatively minimal differences for samples aged in the presence of Hg(II) and Zn(II), indicating that the presence or the uptake behavior of these two metals during the aging process does not significantly impede the growth process of the goethite nanoparticles. In contrast, differences can be clearly observed between the control samples and samples aged in the presence of As(V) and Cu(II). Specifically, the improved peak intensity and definition observed with progressive aging in the control samples is considerably lessened for the As(V)- and Cu(II)-bearing samples, suggesting a retardation of the nanoparticle growth process due to the presence of these metal(loid)s. A similar effect was seen for the uptake of arsenate onto ferrihydrite Waychunas et al. (1993) and was further extensively characterized in Waychunas et al. (1996). Integrated area ratios of the (0 0 1) peak region to the area of the entire diffraction pattern (A001/Atotal) as a function of time are plotted for all samples in Fig. 6.5. This peak ratio, used here as a proxy for particle size, generally increases in a sigmoidal growth pattern (as indicated with best-fit third-order polynomial functions) with inflection points at 48 h for all samples except those containing As(V), where particle size appears to decline after about 32 h of aging. This is supported by re-examining the mXRD patterns generated in the presence of As(V), where the overall intensity and peak resolution decrease among patterns with >32 h aging (Fig. 6.3e). By considering both the mXRD stack plots as well as the peak ratios, the introduced metal(loid)s appear to inhibit the aggregation-based growth of the goethite nanoparticles to different extents, ordered as follows: As(V)>Cu(II)>Zn(II)>Hg(II). This is in agreement with the macroscopic uptake behavior of the metal(loid)s as characterized by ICP-OES analysis. During the later stages of aging there is evidence for a possible acceleration of particle growth in the presence of Hg(II) and Zn(II) relative to the other metal(loid)s and possibly even the control, although data are sparse in this region. This is also consistent with the aforementioned difference in uptake behavior between the metal(loid)s. 6.3.3. X-Ray Absorption Spectroscopy Based on the macroscopic uptake data, samples exposed to As(V) and Cu(II) were not anticipated to demonstrate considerable change in metal(loid)


167

Figure 6.5: Ratios of the Integrated Area of the (0 0 1) Peak (25–321 2y) to Entire Dataset (25–751 2y). Lines Represent Third-Order Polynomial Best Fits Used to Track the Growth Rates of Nanoparticles along the Primary Growth Axis of Goethite (c-Axis). R-Factors of Fits: Blank ¼ 0.7736; As(V) ¼ 0.8194; Cu(II) ¼ 0.9025; Hg(II) ¼ 0.9482; Zn(II) ¼ 0.8802.

speciation with time (Waychunas et al., 1993; Manceau, 1995; O’Reilly et al., 2001). This was corroborated by earlier EXAFS investigations of 0-, 1-, and 5-day aged samples exposed to As(V) and Cu(II) which did not exhibit significant differences between their EXAFS spectra over time (data not shown). As samples exposed to Zn(II) (G ¼ 39.7–99.8 mmol/g) and Hg(II)

168

C. S. Kim et al.

(G ¼ 15.7–84.7 mmol/g) in this study showed measurable changes in macroscopic uptake during the course of the experiment, the potential for changes in speciation with time was considerably greater and so EXAFS analysis was focused on these two systems. Normalized zinc K-edge X-ray absorption near-edge structure (XANES) spectra of selected Zn(II)-sorbed nanogoethite samples are shown in Fig. 6.6. Two distinct edge features can be observed in the XANES spectra: an initial peak at 9,665 eV and a second peak at 9,668 eV. Simulated XANES spectra calculated by Waychunas et al. (2003) using model spinel clusters resulted in two similar peaks, with the second peak attributed to a greater number of second-nearest neighbor Fe atoms around the central absorbing Zn atom. Changes in the coordination of Zn with increased aging are corroborated by the EXAFS spectra and associated Fourier transforms of the same samples (Fig. 6.7), with the latter (representing radial distribution functions that indicate the distance(s) between a central absorbing Zn atom and its nearest-neighboring atoms) showing a clear increase in amplitude with time at distances consistent with second-nearest neighbors (Fig. 6.7b). Quantitative fitting of these features (Table 6.1) reveals both a Zn–Fe scattering interaction at 3.4470.03 A˚ which increases in coordination number after 8 h of aging as well as the appearance of a second Zn–Fe feature at 2.9970.02 A˚ at 8 h of aging which increases slightly in coordination by the end of the aging period. Qualitative changes in the EXAFS spectra appear to track these increases in coordination as well (Fig. 6.7a). The shortest, firstneighbor scattering interaction was fitted for all samples with a Zn–O scattering path at an interatomic distance of 1.9870.02 A˚ and coordination numbers of 4 indicating tetrahedral coordination, consistent with other studies which predict that octahedrally coordinated aqueous Zn(II) is typically converted to a tetrahedrally coordinated species upon sorption to different metal oxide phases (O’Day et al., 1998; Trainor et al., 2000; Trivedi et al., 2001, 2004; Waychunas et al., 2003; Lee and Anderson, 2005). While at sufficiently low sorption densities Zn species have been shown to retain their octahedral hydration shell upon sorption to goethite (Schlegel et al., 1997), the EXAFS spectra of these samples do not appear to indicate octahedral Zn coordination. Using the parameters generated through EXAFS fitting analysis, potential sorption geometries of Zn(II) onto nanogoethite were modeled using the Spartan ’04 molecular modeling program (Kong et al., 2000). These models associate the two Zn–Fe distances of 3.44 and 2.99 A˚ most closely with a binuclear bidentate corner-sharing (Fig. 6.8) and mononuclear bidentate edge-sharing (Fig. 6.9) inner-sphere surface sorption complex, respectively. These findings are generally in agreement with other Zn–Fe oxide studies,


169

Figure 6.6: Normalized XANES Spectra for Zn(II) Sorption onto Nanoscale Goethite, with Arrows and Vertical Lines Pointing Out Relevant Features. The Second Peak (2) Has Been Correlated to an Increase in the Number of Next-Nearest-Neighbor Fe Atoms Based on Previous Spinel Cluster Calculations (Waychunas et al., 2003).

170 C. S. Kim et al.

Figure 6.7: Fits for (a) k3-Weighted Zn(II) K-Edge EXAFS and (b) Fourier Transforms of Goethite Nanoparticle Samples Aged in the Presence of Zn(II) for Varying Lengths of Time. The Solid Line Represents the Experimental Data and the Gray Line Indicates the Best Fits.

Sample reaction time (h)

0 2 4 8 16 24 48 72 96

Zn–O

Zn–Fe

Zn–Fe

CN

R (A˚)

s2 (A˚2)

CN

R (A˚)

s2 (A˚2)

CN

R (A˚)

s2 (A˚2)

3.7(3) 3.6(5) 3.7(3) 3.5(4) 3.7(5) 3.5(4) 3.5(4) 3.7(5) 3.6(5)

1.988(7) 1.97(1) 1.96(1) 1.97(1) 1.98(1) 1.98(1) 1.98(1) 1.98(1) 1.99(1)

0.007(2) 0.007(2) 0.0071(7) 0.006(1) 0.007(2) 0.007(1) 0.007(1) 0.008(2) 0.007(2)

– – – 0.4(4) 0.4(5) 0.4(4) 0.4(4) 0.6(4) 0.8(4)

– – – 2.97(6) 3.00(7) 2.97(6) 2.99(6) 3.00(5) 3.00(4)

– – – 0.01a 0.01a 0.01a 0.01a 0.01a 0.01a

1.2(3) 0.7(5) 1.4(5) 2.5(5) 2.3(6) 2.3(5) 2.2(4) 2.2(5) 2.2(5)

3.43(2) 3.41(5) 3.44(3) 3.44(2) 3.44(2) 3.45(2) 3.44(2) 3.43(2) 3.43(2)

0.01a 0.01a 0.01a 0.01a 0.01a 0.01a 0.01a 0.01a 0.01a

Corresponding EXAFS spectra and Fourier transforms can be seen in Fig. 6.7. CN: coordination number, R: interatomic distance, and s2: Debye–Waller factor. a Set values; S20 ¼ 0:81; E0 values ranged from 0.05 to 2.6 eV.


Table 6.1: Zn K Edge EXAFS Fitting Results for Zn(II) Sorption to Nanoscale Goethite.

171

172

C. S. Kim et al.

Figure 6.8: Simulated Corner-Sharing Bonding Configuration for Zn onto Fe(O,OH)6 Octahedra Representing the Goethite Structure Using Spartan Pro. The Interatomic Distance between Zn and Fe is Represented by the Gray Dotted Line and Equals 3.41 A˚.

Figure 6.9: Simulated Edge-Sharing Bonding Configuration for Zn onto Fe(O,OH)6 Using Spartan Pro. The Interatomic Distance between Zn and Fe is Represented by the Gray Dotted Line and Equals 2.98 A˚.

with the first shell being fit with 4 O atoms at a atomic distance of 1.97 A˚ and the second shell fit with Fe/Zn at a distance of 3.44–3.51 A˚ (representing corner-sharing bidentate complexes) (Trivedi et al., 2001, 2004; Waychunas et al., 2002). The shorter Zn–Fe distance again appears to be consistent with


173

Figure 6.10: Fits for (a) k3-Weighted Hg(II) LIII-Edge EXAFS and (b) Fourier Transforms of Goethite Nanoparticle Samples Aged in the Presence of Hg(II) for Varying Lengths of Time. The Solid Line represents the Experimental Data and the Gray Line Indicates the Best Fits.

either edge-sharing complexes or a surface precipitate, e.g., Zn(OH)2(s), although the latter is somewhat unlikely considering the relatively low surface coverages measured and the absence of evidence for octahedrally coordinated Zn. Mercury LIII-edge EXAFS spectra and associated Fourier transforms of selected Hg(II)-sorbed nanogoethite samples are shown in Fig. 6.10. Due to lower total amounts of uptake on these samples (see Fig. 6.1), their EXAFS spectra are of poorer quality than those featuring Zn(II) uptake. Nevertheless, clear trends in both the EXAFS spectra and Fourier transforms can be observed, specifically an increase in the first Fourier transform feature

174

C. S. Kim et al.

Table 6.2: Hg LIII EXAFS Fitting Results for Hg(II) Sorption to Nanoscale Goethite. Sample reaction time (h)

0 3 4 6 12 24 72 120 168

Hg–O

Hg–Hg

CN

R (A˚)

s2 (A˚2)

CN

R (A˚)

s2 (A˚2)

0.6(2) 1.0(2) 0.7(2) 0.8(2) 1.2(3) 1.3(2) 1.8(3) 1.7(3) 1.7(3)

2.09(2) 2.04(2) 2.08(2) 2.06(2) 2.07(1) 2.060(8) 2.066(5) 2.062(9) 2.072(7)

0.007a 0.007a 0.007a 0.007a 0.007a 0.007a 0.007a 0.007a 0.007a

0.9(2) 0.8(4) 0.5(3) 0.5(3) 0.5(3) 0.4(2) 0.3(3) 0.3(2) 0.3(2)

2.51(2) 2.53(1) 2.55(1) 2.54(1) 2.54(1) 2.56(1) 2.53(2) 2.56(1) 2.52(1)

0.005a 0.005a 0.005a 0.005a 0.005a 0.005a 0.005a 0.005a 0.005a

Corresponding EXAFS spectra and Fourier transforms can be seen in Fig. 6.10. CN: coordination number, R: interatomic distance, and s2: Debye–Waller factor. a Set values; S20 ¼ 0:90; E0 values ranged from 7.7 to 13.0 eV.

(generally corresponding to lower frequency k-space EXAFS oscillations) and a concomitant decrease in the more distant Fourier transform feature (generally corresponding to higher frequency k-space EXAFS oscillations). EXAFS fitting analysis (Table 6.2) identifies these two features as an Hg–O scattering path at 2.0670.03 A˚ and a Hg–Hg scattering path at 2.5370.03 A˚, respectively. Attempts to fit either peak with a Hg–Fe scattering path resulted in a much poorer quality of fit, while inclusion of the same path with the Hg–O and Hg–Hg interactions did not substantially improve the quality of the fit; thus, Hg–Fe scattering paths did not merit inclusion in the final fit results presented. The coordination numbers determined from EXAFS fitting follow the trends observed qualitatively in the Fourier transforms, with the Hg–O coordination increasing from 0.6 to 1.8 and the Hg–Hg coordination decreasing from 0.9 to 0.3 with progressive aging. Coordination numbers less than 1 are consistent with only a minor fraction of Hg in the sample present in that coordination environment. The increase in the Hg–O coordination number indicates a progressive tendency toward inner-sphere adsorption as observed in other studies (Kim et al., 2004a); although the data quality is not sufficient to identify any weaker second-neighbor scattering components such as Fe, the first-shell fitting results likely support the formation of bidentate corner-sharing surface complexes as observed by Collins et al. (1999b) and Kim et al. (2004a) in similar Hg(II)–goethite sorption systems. The Hg–Hg interatomic


175

Figure 6.11: Simulated Bonding Configuration for Hg2(OH)2 Species.

distance determined from EXAFS fitting is consistent with Hg(I)–Hg(I) distances, which average 2.43–2.69 A˚ (Grdenic and Tunell, 1969; Pervukhina et al., 1999) (e.g., Hg2Cl2), and implies some degree of photoreduction resulting from exposure to synchrotron radiation. Our earlier studies (Kim et al., 2004a) identified similar Hg–Hg distances for Hg(II)–g-alumina sorption samples and attributed them to the photoinduced reduction of loosely sorbed or unsorbed free Hg(II) to the mercurous ion Hg(I), which is known to form dimeric complexes such as the Hg2(OH)2 aqueous species (Fig. 6.11) (Kong et al., 2000). Therefore, the presence of Hg–Hg scattering interactions can be taken to represent the fraction of Hg(II) that is weakly associated, perhaps via outer-sphere sorption with the substrate surface, and indicates that poorly sorbed Hg(II) is considerably more susceptible to photoreduction than inner-sphere-sorbed Hg(II).

6.4. Discussion This study verifies our proposed hypotheses correlating the effects of progressive metal(loid) uptake with changes in the speciation of the sorbed metal(loid) and predicting that such uptake would have a measurable impact on nanoparticle growth rates. It also identified clear metal(loid)-dependent differences in uptake behavior that additionally influence speciation mode and effects on nanoparticle growth. Such outcomes demonstrate the utility of combining macroscopic uptake experiments, which track the total extent of metal(loid) sorbed to a given geologically relevant sorbent, with X-ray synchrotron-based methods which allow detailed characterization of both the sorbent and the sorbate as uptake is progressing. A synthesis of complementary techniques is of particular benefit when the sorbent itself, in this case a suspension of nanoscale iron oxyhydroxide particles, is

176

C. S. Kim et al.

undergoing substantial changes (i.e., aggregation-based growth) during the sorption process. Considerable agreement among the results of the different methods indicates that variations in the rate and extent of macroscopic uptake as a function of the specific sorbing metal(loid) correspond with effects on particle growth as well as on the mode(s) of metal complexation. A fundamental difference in sorption behavior as a function of the metal(loid) introduced was apparent in this study, with As(V) and Cu(II) displaying immediate and total uptake while Hg(II) and Zn(II) demonstrating initially low uptake but progressively more uptake with aging. These differences are likely to be in part a result of the metal(loid)-specific uptake dependence on pH to the Fe-oxyhydroxide nanoparticle sorbent. That is, the pH at which the experiments were begun (6.070.1) was sufficient to yield maximum uptake of As(V) and Cu(II) but only partial uptake of Hg(II) and Zn(II). Along the standard macroscopic uptake curves for each metal(loid), this would be akin to the top of the curves for the former species and somewhere along the curve for the latter species (Benjamin and Leckie, 1981; Ponthieu et al., 2006). For those metal cations not yet at their maximum uptake levels (i.e., Hg(II) and Zn(II)), then the continuous pH increase for all supernatants shown in Fig. 6.2 induces progressive metal uptake as shown directly in the case of Hg(II) (Fig. 6.12). This early decrease in pH is attributed to re-equilibration through protonation of the substrate surface following base titration to the initial experimental pH; although metal cation adsorption could be a source of proton release to solution (Stumm, 1992), the pH drop is also observed in the control experiment where no metal was introduced, precluding the likelihood of cation adsorption as the primary factor in the initial decline in pH. Such an initial decline in pH would also be consistent with a transformation of ferrihydrite to goethite, resulting in the loss of protons to solution; a comparable change may be taking place during possible nanoparticle surface structural rearrangements hypothesized to occur early in the aging process as mentioned earlier. The subsequent increase in pH over time, also observed by Subramaniam and Yiacoumi (2001) for the uptake of copper onto ferric oxide, could be attributed either to the slight dissolution of the solid, thus releasing hydroxyl species to solution, or perhaps to particle growth, as the loss of surface area through aggregation could result in a net dehydroxylation effect that increases pH and, correspondingly, the uptake of Hg(II) and Zn(II). Although such a process has yet to be conclusively proven, a schematic reaction involving the bonding of two Fe octahedra and subsequent release of hydroxyls might follow a path similar to the following: FeðO; OHÞ6 þ FeðO; OHÞ6 ! Fe2 ðO; OHÞ10 þ 2OH

(6.1)


177

Figure 6.12: Comparison of Percent Mercury Uptake and pH as a Function of Aging. Error Bars Accounting for Instrumental Measurement Variability are Contained within the Data Points. The differences in uptake rate as controlled by the varying pH-dependent uptake behaviors of the selected metal(loid)s are clearly reflected in the mXRD results, with species exhibiting slower, progressive uptake (Hg(II) and Zn(II)) demonstrating minimal to moderate effects on the particle growth process while those featuring rapid uptake (As(V) and Cu(II)) having an immediate and more significant impact on growth. The extent to which the sorption of the specific metal(loid) impedes growth correlates directly with the extent of uptake as shown when comparing the mXRD peak ratio data and the macroscopic uptake data, with the general ordering as follows: As(V)ZCu(II)>Zn(II)>Hg(II). The trends of particle growth based upon the peak ratio measurements remain roughly the same among the different systems, with a sigmoidal shape suggesting two separate stages of growth or recrystallization. This is not to suggest that only a single mechanism is responsible for growth at any given time; in fact, multiple processes – aggregation-based growth, ripening, and surface restructuring/ recrystallization – are likely taking place in various proportions throughout

178

C. S. Kim et al.

the aging experiments. However, the shape of the peak ratios suggests that there is a transition between two dominant growth mechanisms that occurs in the presence of all metal(loid)s studied in addition to the control. This would agree with our previous studies suggesting that the aggregation-based growth of nanoparticles takes place in the early stages of growth, at which point more traditional ripening-based growth mechanisms dominate the growth process (Waychunas et al., 2005). The effects of rapid metal(loid) uptake on nanoparticle aggregation and growth suggest that surface passivation (site deactivation/poisoning), through either modification of surface charge or alteration of the surface structure/composition, inhibits growth processes during aging, as has been observed and/or predicted with other aqueous species including arsenate, silicate, and other oxyanions (Fuller et al., 1993; Waychunas et al., 1993; Myneni et al., 1997; Stumm, 1997). Classical growth is likely restricted by reduction of active attachment sites for new atoms, e.g., reduction in ‘‘kink’’ or ‘‘edge’’ sites. The effect ought to be related to the strength and hence stability of such complexation reactions. In the case of aggregation, while some degree of aggregation may still occur, the process is reduced due to the presence of metal impurities on aggregation interfaces which would need to be desorbed or incorporated to allow aggregation to proceed. The latter effect could result in aggregated particles with poor overall structural coherence, and hence XRD patterns that do not show improved crystallization (internal order). In systems where metal uptake is initially low and further uptake occurs over time, nanoparticle growth is still impeded but at a considerably lesser rate. This direct apparent relationship between the extent of metal(loid) uptake and the rate of particle growth supports the persistence of nanoscale and/or less structured/amorphous mineral phases in natural systems, where any number of aqueous ligands in addition to metal ions present in natural waters may similarly sorb to and passivate such particle surfaces. The progressive metal adsorption observed in the Hg(II) and Zn(II) systems allows the observation of real-time transitions between distinct modes of uptake using the atomic-scale probe of EXAFS spectroscopy. In surface complexation models (Dzomback and Morel, 1990), sorption is often characterized as a two-step process, with metals first binding to the most high-energy sorption sites first; once these sites are saturated yet conditions for further sorption are still favorable, higher surface loading levels are achieved by continued sorption to lower energy sorption sites. This appears to be taking place in the case of Zn(II), where inner-sphere bidentate corner-sharing surface complexes are initially formed, followed later by inner-sphere bidentate edge-sharing surface complexes. These results


179

are consistent with other studies (Stumm, 1997; Collins et al., 1999a; Kim et al., 2004a; Paktunc et al., 2004; Peacock and Sherman, 2004; Trivedi et al., 2004) that find bidentate corner-sharing surface species as a preferred mode of metal complexation in a number of metal sorption systems. As sites where such species are likely to form tend to be concentrated at the edges, steps, and defects of a mineral surface, which typically feature undercoordinated oxygen atoms in the Fe(O,OH)6 octahedra that form the building blocks of iron oxyhydroxide phases, it follows that these sites will serve as high-energy locations for the initial rapid stages of metal sorption. Once these sites are saturated, a second mode of Zn(II) sorption comprised of edge-sharing surface complexes appears to dominate for the duration of the experiment, representing an additional and increasing proportion of the macroscopic uptake observed from 8 h of aging onwards. While nanosized Zn–Fe precipitates cannot be conclusively ruled out, the relatively low surface coverages exhibited throughout the experiment suggest direct chemical sorption as the primary means of uptake. Although less clear, progressive changes in the mode of Hg(II) uptake appear to be occurring as well, with the initial formation of loosely bound, potentially outer-sphere Hg(II) surface complexes which are easily converted through beam photoreduction into Hg(I) dimers. With continued aging, however, there appears to be an increased preference for inner-sphere surface complexation, evidenced by the relatively short Hg–O distances consistent with direct binding to the surface and a reduced degree of Hg(II) photoreduction resulting in lower Hg–Hg coordination numbers.

6.5. Conclusions The work presented in this study provides substantial evidence indicating that metal(loid) uptake onto iron oxyhydroxide nanoparticles either occurs very rapidly or increases progressively with continued aging/growth of the particles at elevated temperatures. At the same time, such uptake serves to inhibit particle growth, likely through surface passivation effects. In systems where the process of metal uptake occurs simultaneously and continuously with particle growth, as observed from macroscopic uptake data, a proportion of the initially sorbed metal will inevitably be located at particle interfaces that can later join as the particles aggregate. Since the macroscopic data show that nanoparticle growth in the presence of Hg(II) and Zn(II) does not result in a net desorption of metal from the surface, but rather occurs alongside increasing macroscopic uptake, it is likely that some of the metal that is initially sorbed onto nanoparticle surfaces exposed directly to solution

180

C. S. Kim et al.

becomes incorporated into aggregation zones connecting particles and eventually the larger aggregates’ bulk structure with time. The speciation of Zn(II) and Hg(II) at the greatest aging times, as determined in this study by EXAFS spectroscopy, may represent a point along the continuum between surface-sorbed and structurally incorporated metals. A deeper level of sequestration as would be implied by such structural incorporation would presumably restrict metal remobilization into solution and lessen the potential environmental impacts of such metals by retaining them more fixedly in the solid phase. Further evidence for the structural incorporation of metal(loid)s through simultaneous sorption and nanoparticle aggregation could be obtained macroscopically by conducting desorption experiments to gauge the relative extent of metals that can be released at different stages of aggregation and spectroscopically by identifying evidence of metal(loid) (co-)precipitation with or incorporation into the nanoparticle substrate. For example, an increase in Zn–Fe coordination numbers beyond that predicted for the surface sorption complexes proposed would indicate a higher degree of Fe coordination around the average Zn atom (Waychunas et al., 2002), as would be the case for zinc that has become structurally incorporated into the nanoparticle bulk region. Identification of the geochemical and aging conditions that maximize such methods of metal removal from solution would help more permanently entrain metals in the solid phase and hold implications for the effective remediation of metal-contaminated waters.

ACKNOWLEDGMENTS We wish to acknowledge the invaluable staff support at the ALS (Nobumichi Tamura, Bryan Valek) and SSRL (John Bargar, Joe Rogers, Sam Webb) for their assistance with mXRD and EXAFS data collection, respectively. We also thank Guangchou Li at Stanford University for help with ICP-OES measurements. Finally, we would like to thank Kim Environmental Geochemistry group members Lauryn DeGreeff, Brian Reinsch, and Henry Yan for their help with sample preparation and laboratory experimentation. The insightful comments of editor M. Barnett and two anonymous reviewers contributed greatly to the improvement of the manuscript. The Advanced Light Source is supported by the Director, Office of Science, Office of Basic Energy Sciences, the U.S. Department of Energy under Contract No. DE-AC02-05CH11231. Portions of this research were carried out at the Stanford Synchrotron Radiation Laboratory, a national user facility operated by Stanford University on behalf of the U.S. Department of


181

Energy, Office of Basic Energy Sciences. This study was supported by the Wilkinson College of Letters and Sciences, Chapman University, and grants from the American Chemical Society – Petroleum Research Grant, PRF #44721-GB10, and the National Science Foundation, Division of Earth Sciences, Grant #0618217.

REFERENCES Atkinson, R. J., Posner, A. M., & Quirk, J. P. (1968). Crystal nucleation in Fe(III) solutions and hydroxide gels. J. Inorg. Nucl. Chem., 30, 2371–2381. Axe, L., & Anderson, A. J. (1998). Intraparticle diffusion of metal contaminants in amorphous oxide minerals. In: E. A. Jenne (Ed). Adsorption of Metals by Geomedia. Academic Press, San Diego, pp. 193–205. Benjamin, M. M., & Leckie, J. O. (1981). Multiple-site adsorption of Cd, Cu, Zn, and Pb on amorphous iron oxyhydroxide. J. Colloid Interface Sci., 79 (1), 209–221. Collins, C. R., Ragnarsdottir, K. V., & Sherman, D. M. (1999a). Effect of inorganic and organic ligands on the mechanism of cadmium sorption to goethite. Geochim. Cosmochim. Acta, 63 (19–20), 2989–3002. Collins, C. R., Sherman, D. M., & Ragnarsdottir, K. V. (1999b). Surface complexation of Hg2+ on goethite: Mechanism from EXAFS spectroscopy and density functional calculations. J. Colloid Interface Sci., 219 (2), 345–350. Dyer, J. A., Trivedi, P., Scrivner, N. C., & Sparks, D. L. (2004). Surface complexation modeling of zinc sorption onto ferrihydrite. J. Colloid Interface Sci., 270 (1), 56–65. Dzombak, D. A., & Morel, F. M. M. (1986). Sorption of cadmium on hydrous ferric oxide at high sorbate/sorbent ratios: Equilibrium, kinetics, and modeling. J. Colloid Interface Sci., 112 (2), 588–598. Dzomback, D. A., & Morel, F. M. M. (1990). Surface Complexation Modeling: Hydrous Ferric Oxide. Wiley, New York, NY, 393 pp. Egirani, D. E., Baker, A. R., & Andrews, J. E. (2005). Copper and zinc removal from aqueous solution by mixed mineral systems I. Reactivity and removal kinetics. J. Colloid Interface Sci., 291 (2), 319–325. Farquhar, M. L., Charnock, J. M., Livens, F. R., & Vaughan, D. J. (2002). Mechanisms of arsenic uptake from aqueous solution by interaction with goethite, lepidocrocite, mackinawite, and pyrite: An X-ray absorption spectroscopy study. Environ. Sci. Technol., 36 (8), 1757–1762. Fendorf, S., Eick, M. J., Grossl, P., & Sparks, D. L. (1997). Arsenate and chromate retention mechanisms on goethite.1. Surface structure. Environ. Sci. Technol., 31 (2), 315–320. Ford, R. G., Scheinost, A. C., & Sparks, D. L. (2001). Frontiers in metal sorption/ precipitation mechanisms on soil mineral surfaces. In: Advances in Agronomy. Elsevier Inc., San Diego, CA, Vol. 74, pp. 41–62.

182

C. S. Kim et al.

Foster, A. L., Brown, G. E., & Parks, G. A. (2003). X-ray absorption fine structure study of As(V) and Se(IV) sorption complexes on hydrous Mn oxides. Geochim. Cosmochim. Acta, 67 (11), 1937–1953. Fuller, C. C., Davis, J. A., & Waychunas, G. A. (1993). Surface-chemistry of ferrihydrite. 2. Kinetics of arsenate adsorption and coprecipitation. Geochim. Cosmochim. Acta, 57 (10), 2271–2282. Gilbert, B., Lu, G., & Kim, C. S. (2007). Stable cluster formation in aqueous suspensions of iron oxyhydroxide nanoparticles. J. Colloid Interface Sci., 313, 152–159. Grafe, M., Nachtegaal, M., & Sparks, D. L. (2004). Formation of metal-arsenate precipitates at the goethite–water interface. Environ. Sci. Technol., 38 (24), 6561–6570. Grafe, M., & Sparks, D. L. (2005). Kinetics of zinc and arsenate co-sorption at the goethite–water interface. Geochim. Cosmochim. Acta, 69 (19), 4573–4595. Grdenic, D., & Tunell, G. (1969). Mercury: Crystal chemistry. In: K. H. Wedepohl (Ed), Handbook of Geochemistry. Springer-Verlag, New York, NY, pp. 8. Guyodo, Y., Mostrom, A., Penn, R. L., & Banerjee, S. K. (2003). From nanodots to nanorods: Oriented aggregation and magnetic evolution of nanocrystalline goethite. Geophys. Res. Lett., 30 (10), 1512. Jenne, E. A. (Ed). (1998). Adsorption of Metals by Geomedia: Variables, Mechanisms, and Model Applications. Academic Press, San Diego. Kersting, A. B., Efurd, D. W., Finnegan, D. L., Rokop, D. J., Smith, D. K., & Thompson, J. L. (1999). Migration of plutonium in ground water at the nevada test site. Nature, 397 (6714), 56–59. Kim, C. S., Rytuba, J. J., & Brown, G. E. Jr. (2004a). EXAFS study of mercury(II) sorption to Fe- and Al-(hydr)oxides: I. Effects of pH. J. Colloid Interface Sci., 271, 1–15. Kim, C. S., Rytuba, J. J., & Brown, G. E. Jr. (2004b). EXAFS study of mercury(II) sorption to Fe- and Al-(hydr)oxides: II. Effects of chloride and sulfate. J. Colloid Interface Sci., 270, 9–20. Kinniburgh, D. G., Van Riemsdijk, W. H., Koopal, L. K., & Benedetti, M. F. (1998). Ion binding to humic substances. In: E. A. Jenne (Ed). Adsorption of Metals by Geomedia. Academic Press, San Diego, pp. 484–517. Kong, J., White, C. A., Krylov, A. I., Sherrill, D., Adamson, R. D., Furlani, T. R., Lee, M. S., Lee, A. M., Gwaltney, S. R., Adams, T. R., Ochsenfeld, C., Gilbert, A. T. B., Kedziora, G. S., Rassolov, V. A., Maurice, D. R., Nair, N., Shao, Y. H., Besley, N. A., Maslen, P. E., Dombroski, J. P., Daschel, H., Zhang, W. M., Korambath, P. P., Baker, J., Byrd, E. F. C., Van Voorhis, T., Oumi, M., Hirata, S., Hsu, C. P., Ishikawa, N., Florian, J., Warshel, A., Johnson, B. G., Gill, P. M. W., Head-Gordon, M., & Pople, J. A. (2000). Q-chem 2.0: A high-performance Ab initio electronic structure program package. J. Comput. Chem., 21 (16), 1532–1548. Lee, S. W., & Anderson, P. R. (2005). EXAFS study of Zn sorption mechanisms on hydrous ferric oxide over extended reaction time. J. Colloid Interface Sci., 286 (1), 82–89.


183

Madden, A. S., Hochella, M. F., & Luxton, T. P. (2006). Insights for size-dependent reactivity of hematite nanomineral surfaces through Cu2+ sorption. Geochim. Cosmochim. Acta, 70 (16), 4095–4104. Manceau, A. (1995). The mechanism of anion adsorption on iron-oxides – evidence for the bonding of arsenate tetrahedra on free Fe(O,Oh)(6) edges. Geochim. Cosmochim. Acta, 59 (17), 3647–3653. Morton, J. D., Semrau, J. D., & Hayes, K. F. (2001). An X-ray absorption spectroscopy study of the structure and reversibility of copper adsorbed to montmorillonite clay. Geochim. Cosmochim. Acta, 65 (16), 2709–2722. Myneni, S. C. B., Traina, S. J., Logan, T. J., & Waychunas, G. A. (1997). Oxyanion behavior in alkaline environments: sorption and desorption of arsenate in ettringite. Environ. Sci. Technol., 31 (6), 1761–1768. Nachtegaal, M., & Sparks, D. L. (2004). Effect of iron oxide coatings on zinc sorption mechanisms at the clay-mineral/water interface. J. Colloid Interface Sci., 276 (1), 13–23. Novikov, A. P., Kalmykov, S. N., Utsunomiya, S., Ewing, R. C., Horreard, F., Merkulov, A., Clark, S. B., Tkachev, V. V., & Myasoedov, B. F. (2006). Colloid transport of plutonium in the far-field of the mayak production association, Russia. Science, 314 (5799), 638–641. O’Day, P. A., Carroll, S. A., & Waychunas, G. A. (1998). Rock–water interactions controlling zinc, cadmium, and lead concentrations in surface waters and sediments, US tri-state mining district. 1. Molecular identification using X-ray absorption spectroscopy. Environ. Sci. Technol., 32 (7), 943–955. O’Reilly, S. E., Strawn, D. G., & Sparks, D. L. (2001). Residence time effects on arsenate adsorption/desorption mechanisms on goethite. Soil Sci. Soc. Am. J., 65 (1), 67–77. OriginLab Corporation (2006). OriginLab Corporation. Northampton, MA. http:// www.originlab.com/index.aspx?s=1&1m=123 Paktunc, D., Foster, A., Heald, S., & Laflamme, G. (2004). Speciation and characterization of arsenic in gold ores and cyanidation tailings using X-ray absorption spectroscopy. Geochim. Cosmochim. Acta, 68 (5), 969–983. Pan, G., Qin, Y. W., Li, X. L., Hu, T. D., Wu, Z. Y., & Xie, Y. N. (2004). EXAFS studies on adsorption–desorption reversibility at manganese oxides–water interfaces I. Irreversible adsorption of zinc onto manganite (gamma-MnOOH). J. Colloid Interface Sci., 271 (1), 28–34. Peacock, C. L., & Sherman, D. M. (2004). Copper(II) sorption onto goethite, hematite and lepidocrocite: A surface complexation model based on ab initio molecular geometries and EXAFS spectroscopy. Geochim. Cosmochim. Acta, 68 (12), 2623–2637. Pervukhina, N. V., Romanenko, G. V., Borisov, S. V., Magarill, S. A., & Palchik, N. A. (1999). Crystal chemistry of mercury(I) and mercury(I, II) minerals. J. Struct. Chem., 40 (3), 461–476. Ponthieu, M., Juillot, F., Hiemstra, T., van Riemsdijk, W. H., & Benedetti, M. F. (2006). Metal ion binding to iron oxides. Geochim. Cosmochim. Acta, 70 (11), 2679–2698.

184

C. S. Kim et al.

Rakovan, J., Becker, U., & Hochella, M. F. (1999). Aspects of goethite surface microtopography, structure, chemistry, and reactivity. Am. Mineral., 84 (5–6), 884–894. Redden, G. D., Li, J., & Leckie, J. O. (1998). Adsorption of UVI and citric acid on goethite, gibbsite, and kaolinite: Comparing results for binary and ternary systems. In: E. A. Jenne (Ed). Adosorption of Metals by Geomedia. Academic Press, San Diego, pp. 292–313. Roberts, D. R., Ford, R. G., & Sparks, D. L. (2003). Kinetics and mechanisms of Zn complexation on metal oxides using EXAFS spectroscopy. J. Colloid Interface Sci., 263 (2), 364–376. Scheidegger, A. M., Lamble, G. M., & Sparks, D. L. (1997). Spectroscopic evidence for the formation of mixed-cation hydroxide phases upon metal sorption on clays and aluminum oxides. J. Colloid Interface Sci., 186 (1), 118–128. Scheidegger, A. M., & Sparks, D. L. (1996). A critical assessment of sorption– desorption mechanisms at the soil mineral/water interface. Soil Sci., 161 (12), 813–831. Schlegel, M. L., Manceau, A., & Charlet, L. (1997). EXAFS study of Zn and ZnEDTA sorption at the goethite (alpha-FeOOH)/water interface. J. Phys. IV, 7 (C2), 823–824. Schwertmann, U., & Cornell, R. M. (1991). Iron Oxides in the Laboratory: Preparation and Characterization. VCH Publishers, Inc., New York, NY, 137 pp. Sposito, G. (1984). The Surface Chemistry of Soils. 1st Ed., Oxford University Press, New York, NY, 234 pp. Stern, E. A., & Heald, S. M. (1979). X-ray filter assembly for fluorescence measurements of X-ray absorption fine structure. Rev. Sci. Instrum., 50 (12), 1579–1582. Stumm, W. (1992). Chemistry of the Solid–Water Interface: Processes at the Mineral–Water and Particle–Water Interface in Natural Systems. 1st Ed., Wiley, New York, NY, 428 pp. Stumm, W. (1997). Reactivity at the mineral–water interface: Dissolution and inhibition. Colloids Surf. A: Physicochem. Eng. Aspects, 120 (1–3), 143–166. Subramaniam, K., & Yiacoumi, S. (2001). Modeling kinetics of copper uptake by inorganic colloids under high surface coverage conditions. Colloids Surf. A: Physicochem. Eng. Aspects, 191 (1–2), 145–159. Tamura, N. (2007). X-Ray Microdiffraction Analysis Software (XMAS); http:// X-raysweb.lbl.gov/microdif/user_resources.htm Trainor, T. P., Brown, G. E. Jr., & Parks, G. A. (2000). Adsorption and precipitation of aqueous Zn(II) on alumina powders. J. Colloid Interface Sci., 231 (2), 359–372. Trivedi, P., & Axe, L. (1999). A comparison of strontium sorption to hydrous aluminum, iron, and manganese oxides. J. Colloid Interface Sci., 218 (2), 554–563. Trivedi, P., & Axe, L. (2001). Ni and Zn sorption to amorphous versus crystalline iron oxides: Macroscopic studies. J. Colloid Interface Sci., 244 (2), 221–229.


185

Trivedi, P., Axe, L., & Tyson, T. A. (2001). An analysis of zinc sorption to amorphous versus crystalline iron oxides using XAS. J. Colloid Interface Sci., 244 (2), 230–238. Trivedi, P., Dyer, J. A., Sparks, D. L., & Pandya, K. (2004). Mechanistic and thermodynamic interpretations of zinc sorption onto ferrihydrite. J. Colloid Interface Sci., 270 (1), 77–85. Vilks, P., Frost, L. H., & Bachinski, D. B. (1997). Field-scale colloid migration experiments in a granite fracture. J. Contam. Hydrol., 26 (1–4), 203–214. Villalobos, M., Trotz, M. A., & Leckie, J. O. (2001). Surface complexation modeling of carbonate effects on the adsorption of Cr(VI), Pb(II), and U(VI) on goethite. Environ. Sci. Technol., 35 (19), 3849–3856. Waychunas, G. A., & Brown, G. E. Jr. (1994). Fluorescence yield XANES and EXAFS experiments: Application to highly dilute and surface samples. Adv. X-Ray Anal., 37, 607–617. Waychunas, G. A., Fuller, C. C., & Davis, J. A. (2002). Surface complexation and precipitate geometry for aqueous Zn(II) sorption on ferrihydrite I: X-ray absorption extended fine structure spectroscopy analysis. Geochim. Cosmochim. Acta, 66 (7), 1119–1137. Waychunas, G. A., Fuller, C. C., Davis, J. A., & Rehr, J. J. (2003). Surface complexation and precipitate geometry for aqueous Zn(II) sorption on ferrihydrite: II. XANES analysis and simulation. Geochim. Cosmochim. Acta, 67 (5), 1031–1043. Waychunas, G. A., Fuller, C. C., Rea, B. A., & Davis, J. A. (1996). Wide angle X-ray scattering (WAXS) study of ‘‘two-line’’ ferrihydrite structure: Effect of arsenate sorption and counterion variation and comparison with EXAFS results. Geochim. Cosmochim. Acta, 60 (10), 1765–1781. Waychunas, G. A., Kim, C. S., & Banfield, J. F. (2005). Nanoparticulate oxide minerals in soils and sediments: Unique properties and contaminant scavenging mechanisms. J. Nanopart. Res., 7, 409–433. Waychunas, G. A., Rea, B. A., Fuller, C. C., & Davis, J. A. (1993). Surface chemistry of ferrihydrite: 1. EXAFS studies of the geometry of coprecipitated and adsorbed arsenate. Geochim. Cosmochim. Acta, 57 (10), 2251–2269. Webb, S. M. (2007). Sam’s Interface for XAS Package (SixPACK); http:// www-ssrl.slac.stanford.edu/swebb/index.htm Zabinsky, S. I., Rehr, J. J., Ankudinov, A., Albers, R. C., & Eller, M. J. (1995). Multiple-scattering calculations of X-ray absorption spectra. Phys. Rev. B, 52, 2995–3009.


Chapter 7

Temperature and Aging Effects on the Surface Speciation of Cd(II) at the Goethite–Water Interface Markus Gra¨fe1, Ghulam Mustafa1, Balwant Singh1, and Rai S. Kookana2 1

Faculty of Agriculture, Food and Natural Resources, J.R.A. McMillan Building, The University of Sydney, Sydney, NSW 2006, Australia 2 CSIRO Land and Water, Private Mail Bag 2, Glen Osmond, SA 5064, Australia

ABSTRACT Cadmium (Cd) adsorption at the goethite–water interface was studied as a function of aging period (16 h vs. 180 d (d: days)) and equilibrium temperature (201C vs. 401C) using extended X-ray absorption fine structure (EXAFS) spectroscopy. Structural parameters gleaned from non-linear least-squares fitting of EXAFS data revealed that Cd ions form single edge-sharing (1ECdFe) surface complexes on goethite after 16 h regardless of equilibrium temperature. After 180 d, both 20 and 401C EXAFS samples show the formation of double-edge sharing (2ECdFe) surface complexes indicated by circa one Fe atom at an average radial distance (RCdFe) of 3.30 A˚ in addition to circa one Fe atom at 3.13 A˚. Contributions from double-corner sharing (2CCdFe) surface complexes at 3.8070.05 A˚ remained minor, which we ascribed to differing reaction conditions of the samples in this vs. previous studies. With respect to recently published observations regarding the desorption behavior of Cd from goethite as a function of aging period and temperature, we conclude that the formation of 2ECd–Fe surface complexes over time provides increased stability to Cd ions sorbed on the goethite surface, which is in good agreement with recent density functional theory calculations. The effect of temperature on Cd desorption from goethite is likely due to the increased electrostatic attraction of Cd ions to the goethite surface at 401C as a result of a decreasing point of zero charge with temperature.

Corresponding author. Tel.: +61-2-9351-2237; Fax: +61-2-9351-2945;

E-mail: [email protected] (B. Singh).

188

M. Gra¨fe et al.

7.1. Introduction Cadmium (Cd) is a metal contaminant of great environmental concern because of its bioaccumulation and toxicity (Bingham et al., 1986). It is currently ranked eighth in the United States Environmental Protection Agency’s priority list of hazardous substances (ATSDR, 2007) and in humans, causes a variety of adverse effects depending on the dose, route, and duration of exposure. Among several diseases attributed to Cd toxicity are prostate and lung cancer, Itai-itai disease (brittleness of the bones), gastrointestinal and renal dysfunction, and vascular endothelium (Nordberg, 2004; Prozialecka et al., 2006). Cd is not a naturally abundant element in the Earth’s crust, but during the last two centuries various anthropogenic activities, such as electroplating, paint and plastic manufacturing, mining, smelting, paint pigments, and batteries, have released Cd into the environment (Adriano, 1986; Singh, 2001). Cd contamination of agricultural soils commonly occurs from the application of phosphatic fertilizers and Cd-bearing sewage sludge (Loganathan et al., 2003). Cd occurs as an impurity in rock phosphate, where it substitutes for calcium (Ca) in the structure of apatite (Sery et al., 1996) and other related phosphate containing minerals. Its similar ionic radius to Ca2+ (0.95 A˚ vs. 1.00 A˚, respectively) enable Cd2+ to substitute for Ca2+ in other Ca2+-bearing minerals, e.g., calcite (Lorens, 1981; Davis et al., 1987). Cd’s ability to form wurtzite- and sphalerite-like sulfide phases also allows it to substitute for zinc (Zn) in its respective sulfide phases as well as in galena (PbS) (Urabe, 1977; Bortnikov et al., 1995; Levelut et al., 1995). High concentrations of Cd may thus be found in environments where such mineral deposits are naturally abundant. The mobility and bioavailability of heavy metals (including Cd) in soils depend largely on their adsorption to the surfaces of soil components and their precipitation as discrete phases (Sposito, 1989). The adsorption of Cd by soil and model compound surfaces has been studied extensively in the last few decades and is well understood in relation to the effects of solution pH, ionic strength, co- and counter-ion composition, reaction time, and temperature (Benjamin and Leckie, 1982; Johnson, 1990; Gray et al., 1998; Mustafa et al., 2004, 2006). Research on Cd desorption has identified pH, index cations, temperature, and aging as important in controlling the desorption of Cd from soils and minerals (Tiller et al., 1984a,b; Backes et al., 1995; Naidu et al., 1997). Tiller and coauthors pointed out that Cd desorption from variably charged soil components increased with decreasing pH, but was also dependent on the saturation of available surface sites. Similar observations were made later by Backes et al. (1995), Davis and Upadhyaya (1996), and

Temperature and Aging Effects on the Surface Speciation of Cd(II)

189

Gray et al. (1998). Backes et al. (1995) showed that Cd desorption from Fe and Mn oxides decreased with residence or aging time (1 week vs. 15 weeks), which also appeared to be responsible for decreasing the rate of desorption (Wang and Xing, 2004). The effect of aging time has mostly been attributed to intra- and inter-particle diffusion, which may be enhanced by favorable pH conditions in solution (Tiller et al., 1984a,b; Bruemmer et al., 1988). Several surface complexation models have been used to describe Cd adsorption on metal oxides and clay mineral surfaces (Davis et al., 1978; Venema et al., 1996; Boily et al., 2005). Doubts have been raised about the stoichiometries obtained by surface complexation modeling and have increased the emphasis on direct spectroscopic methods such as extended X-ray absorption fine structure (EXAFS) spectroscopy to determine the reaction mechanisms of Cd (and other metals) on mineral surfaces (Spadini et al., 1994; Bochatay et al., 1997; Randall et al., 1999; Manceau et al., 2000; Boily et al., 2005; Ramstedt et al., 2005). In aqueous solutions and at mineral–water interfaces of goethite (a-FeOOH), lepidocrocite (g-FeOOH), and manganite (g-MnOOH), Cd ions are sixfold coordinated with average Cd–oxygen (O) bond distances ranging between 2.24 and 2.31 A˚. Bond valence sum analysis (BVSA) suggests that within this range of Cd–O bond distances, Cd should have octahedral symmetry (Brown and Altermatt, 1985; Mosselmans et al., 1996). Second and third shell coordination environments of Cd on Mn and Fe oxides are dependent on the local surface structure (anionic layer structure of the clay or metal oxide), the surface loading, and thus also on pH, ionic strength, and other conditions that affect surface loading (Spadini et al., 1994; Manceau et al., 2000). Spadini et al. (1994) reported the formation of three different complexes for Cd adsorption on goethite depending on surface loading. At low coverage, Cd was adsorbed mainly on the {0 0 1} planes by sharing edges and corners with surface octahedra, whereas at medium and high surface coverages, adsorption occurred by sharing corners on the dominant {1 1 0} and {1 0 0} planes of goethite. Cd–Fe radial distances for different surface complex geometries (edge-sharing vs. single-corner sharing vs. double-corner sharing vs. face sharing) are well established and in some cases are comparable across different metal oxides surfaces (Table 7.1) (Spadini et al., 1994; Parkman et al., 1999; Randall et al., 1999). Manceau et al. (2000) have argued that the lack of 1,2ECdFe surface complexes in high surface loading EXAFS samples is not equivalent to its complete absence, but rather that at high(er) surface coverage, the number of 2CCdFe surface complexes outweighs contributions from 1,2ECdFe surface complexes to undetectable levels. The number of 1,2 ECdFe surface complexes is small due to the limited number of available

190

Surface complex

Symbol

R70.05 A˚a

1

F E

3.15

2

2.85+3.34

Face sharing Single edge-sharing Double edge-sharing

E

Crystallographic plane (hkl)

Reaction conditions

Method, reference

Not yet observed, but suggested by (Spadini et al., 1994) 0 0 1, 0 1 0 pH 7.5, [Cd]i ¼ 0.7–10.5 mM, EXAFS, Spadini equilibrium periods ¼ 4–12 h et al. (1994) 0 0 1, 0 2 1

3.15+3.40

Single edge-sharing Fe(O,OH)6 with attached aqua Cd complex pH 7.5, [Cd]i ¼ 0.7–10.5 mM, equilibrium periods ¼ 4–12 h

DFTb, Randall et al. (1999) EXAFS, Spadini et al. (1994)

Single cornersharing

1

C

3.52

1 0 0, 1 1 0

Single edge-sharing Fe(O,OH)6 with attached aqua Cd complex

DFT, Randall et al. (1999)

Double cornersharing

2

C

3.80

1 0 0, 1 1 0

pH 5.4–9.3, [Cd]i ¼ 26–445 mM, equilibrium period ¼ 48–72 h pH 6.0–8.21, [Cd]i: not specified, equilibrium period ¼ 24 h pH 5.1–7.7, [Cd]ir3 mM, equilibrium period ¼ 24 h, presence of phthalate

EXAFS, Randall et al. (1999) EXAFS, Boily et al. (2005) EXAFS, Parkman et al. (1999)

a

R: radial distance. DFT: density functional theory.

b

M. Gra¨fe et al.

Table 7.1: Surface Complexation States of Cd on Goethite (Pbnm).


191

surface sites on the {0 2 1/0 0 1} planes of goethite, whereas 2CCdFe surface complexes are believed to form on the more prominent {1 1 0} and {1 0 0} planes. Detection of the 2CCdFe surface complexes, however, only becomes possible after the 1,2ECdFe surface sites have been filled, i.e., high-affinity surface sites bind Cd ions before sorption sites of lesser affinity, and the surface complexation mechanism changes to 2CCdFe surface complexation. This progressive appearance and ‘‘disappearance’’ of surface complexes on goethite with increasing surface coverage is overall consistent with the affinity of surface sites of goethite’s crystallographic planes for cation adsorption (Spadini et al., 1994): {0 0 1}>{0 1 0}>{1 1 0}{1 0 0}. While this trend was established for increasing surface loading, it is unclear whether the surface complexation of Cd on goethite also changes as a function of aging or residence time and temperature. In a recent study by Mustafa et al. (2006), the authors showed significant reductions in Cd desorption from goethite with aging (16 h–180 d (d: days)), pH (5.5–6.0), temperature (20–701C), and initial Cd concentration (180–20 mM Cd, Fig. 7.1). The authors used dissolution studies to partition Cd into surface adsorbed Cd and Cd diffused or entrapped into cracks and goethite particles; however, no direct evidence was provided to show the effects of aging and temperature

Figure 7.1: Effects of Equilibrium Temperature (20 and 401C) and Aging (16 h, 30, 90, and 180 d) on Cd Desorption from Goethite. An Initial Cd Concentration of 180 mM was used to Equilibrate Cd and Goethite at pH 6 for 2 h Prior to Centrifugation and Aging as a Moist Paste. The Surface Density of Cd on Goethite was 0.06 mmol m2 for Samples, which corresponds to 73% Surface Coverage.

192

M. Gra¨fe et al.

on the local bonding environment of Cd adsorbed on goethite. In this paper, we follow up on some of the samples from this study (Mustafa et al., 2006) to understand the effects of aging and temperature on Cd’s local bonding environment on goethite as gleaned from Cd K-edge EXAFS spectroscopy.

7.2. Experimental 7.2.1. Sample Preparation The goethite used in these experiments had a specific surface area of 71 m2 g1 and a point of zero salt effect measured at circa pH 8.25 (Mustafa et al., 2004). One hundred and eighty (180.0) micromolar Cd(NO3)2 was reacted with sufficient goethite (20 g L1 suspension density) to obtain a surface loading of circa 4.29 mmol Cd g1 or 0.06 mmol Cd m2 at pH 6.0 for a 2 h reaction period and represents circa 73% of maximum surface coverage at pH 6.0. Four samples were prepared and allowed to age as moist pastes for either 16 h or 180 d at 20 and 401C. Moist pastes were dried gently at 401C and crushed with a mortar and pestle into powdered samples, which were sealed into the void space of a stainless-steel sample holder by KaptonTM tape for EXAFS analysis. For ease and clarity of communication, the four experimental spectra will be referred in the text as: 16hCdGoe201C, 180d CdGoe201C, 16hCdGoe401C, and 180dCdGoe401C.

7.2.2. Cd K-edge Data Collection X-ray absorption fine structure (XAFS) spectroscopy data at the Cd K-edge were collected to determine changes in Cd’s average coordination environment as a function of aging and temperature. Cd K-edge data were collected at beamline 20B (bending magnet) at PNC-CAT (Advanced Photon Source, Argonne National Laboratory, Argonne, IL, USA). The operating details of beamline 20B at PNC-CAT and the storage ring of the APS have been described in detail elsewhere (Heald et al., 1999). The Si (1 1 1) monochromator was calibrated using a Ag foil (energy resolution (DE/E) of this beamline in the range from 4 to 29 keV is 1.4 104). All experimental spectra, a 10 mM Cd(NO3)2 (aq.), and a Cd-substituted goethite (Huynh, 2001; Huynh et al., 2003) reference standard were collected in fluorescence mode using a 13 element solid state detector, while other reference compounds (Cd(OH)2, CdO, and CdCO3) were collected in transmission mode.


193

The ionization (I0), the transmission (It), and the reference (Ir) columns were filled with Ar gas resulting in circa 10% absorbance in I0 and circa 30% absorbance in It and Ir. Due to the low surface loading (482 mg Cd kg1 goethite) of Cd on goethite, up to 30 scans were collected per sample.

7.2.3. Data Analysis All data reduction was performed using WinXAS 2.1 and later versions (Ressler, 1998) following procedures described elsewhere (Bunker, 2004). Individual spectra were averaged and subsequently corrected for background absorbance and normalization. Spectra were then converted from energy to photo-electron wave vector (k) units (k is the wave vector number with units of A˚1) by assigning the origin, E0, to the first inflection point of the absorption edge (Cd ¼ 26.711 keV). The EXAFS was extracted using a cubic spline function consisting of r7 knots applied over an average range in k-space (1.3–12.0 A˚1). Fourier transformation (FT) of the raw k3w(k) function was performed (2.0–11.50 A˚1) to obtain a radial structure function (RSF) using a Bessel window function and a smoothing parameter (b) of 4 to minimize truncation effects in RSFs. The FEFF 7.02 program (Zabinsky et al., 1995) was used to calculate theoretical phase and amplitude functions of Cd–O, Cd–C, Cd–Fe, and Cd–Cd scattering paths using input files based on the structural refinement of otavite (CdCO3), monteponite (CdO), and Cd-substituted goethite (Graf, 1961; Zhang, 1999; Huynh et al., 2003). Nonlinear least-square shell fits were performed simultaneously on raw k3weighted w(k) spectra and segments of the corresponding RSFs over an average distance of 1.2–3.7 and 4.09 A˚ (R+DR, uncorrected for phase shift). The number of permissible free-floating parameters (Npts) was circa 15 (Stern, 1993). For each fit, the coordination number (N), the radial distance (RA˚), the mean-square disorder of the radial distance (Debye–Waller parameter, s2A˚2), and a single, cross-correlated (linked to the same value) phase shift value (DE0) for all single backscattering paths were allowed to vary. For fits with multiple second shell contributions, s2 was correlated to the same value. The amplitude reduction factor (S20 ) was fixed to 1.0. The estimated accuracies for the fit parameters, N and R, are based on a comparison of the best fit result for CdO with values from refined XRD measurements published in the literature (Graf, 1961; Zhang, 1999): first shell – N713% or 70.8, R70.03 A˚; second shell – N723%, R70.01 A˚ for CdO, but more likely to be R70.05 A˚ for experimental samples due to the lack of a good structural model.

194

M. Gra¨fe et al.

7.3. Results Raw, k3-weighted w(k) spectra (Fig. 7.2a) of Cd sorbed on goethite at varying temperatures and residence times were dominated by a sinusoidal wavefunction, which corresponded to backscattering oxygen (O) atoms by comparison to the Cd(NO3)2 (aq.) reference standard. Upon Fourier transformation, a major peak centered at circa 1.60 A˚ (R+DR) was observed in the respective RSFs (Fig. 7.2b). On average, 6.370.8 O neighbors were observed at an average distance of 2.2670.08 A˚ (Table 7.1); however, in all cases, the first ligand shell environment was modeled better with two distinct O shells at 2.20 and 2.34 A˚. In aqueous solution, Cd(NO3)2 showed up to 7.1 next nearest O neighbors distributed evenly over 2.21 and 2.35 A˚, which was still within the accuracy range of first shell coordination numbers (N713%). In 16hCdGoe201C, NCdO over both distances was equal to 3.0; however, with increasing temperature and aging period, the distribution of O atoms around Cd converged toward four O neighbors at 2.20 A˚ and two O neighbors

Figure 7.2: (a) Raw k3-Weighted w(k) Spectra of Reference and Experimental Samples and (b) Their corresponding Radial Structure Functions (RSFs, Same Order as k3-Weighted w(k) Spectra). RSFs of CdCO3 and Cd(OH)2 appear in Dotted Lines to make the Differentiation between these Spectra and CdO Easier.


195

Table 7.2: Structural Parameters of Cd’s First and Second Shells Derived from Non-Linear Least-Square Fits of RSFs [w(R), A˚3] Within a Range indicated per Reference/Sample.

CdO Cd–O Cd–Cd

Na

Rb (A˚)

s2c (A˚2)

5.2 9.1

2.32 3.33

(1.30–3.60 A˚ (R+DR)) 0.008 2.0 0.008 2.0

5.6

5.1

3.5

DE0d (eV)

Res.e

Cd(NO3)2 (aq.) Cd–O(1) Cd–O(2) Cd–Cd(1) Cd–Cd(2)

3.5 3.6 2.2 1.9

2.21 2.35 3.59 3.83

(1.40–3.95 A˚ (R+DR)) 0.003 7.6 0.003 7.6 0.011 7.6 0.011 7.6

Cd–goethite Cd–O(1) Cd–O(2) Cd–Fe(1) Cd–Fe(2)

4.5 1.7 2.3 1.4

2.21 2.35 3.09 3.38

(1.20–3.20 A˚ (R+DR)) 0.005 5.0 0.005 5.0 0.008 5.0 0.008 5.0

3.0 3.0 1.4 1.5 1.7 1.7

2.19 2.35 3.13 3.79 3.58 3.82

(1.20–4.09 A˚ (R+DR)) 0.001 4.6 0.001 4.6 0.009 4.6 0.009 4.6 0.010 4.6 0.010 4.6

4.1 2.5 0.6 0.5 0.8

2.22 2.37 3.15 3.81 3.59

(1.20–3.60 A˚ (R+DR)) 0.006 4.7 0.006 4.7 0.007 4.7 0.007 4.7 0.009 4.6

4.5 4.1

2.21 2.36 3.14 3.28 3.85 3.64

(1.20–3.72 A˚ (R+DR)) 0.003 5.1 0.003 5.1 0.007 5.1 0.007 5.1 0.012 5.1 0.007 5.1

2.7 2.6

16 h, 201C (16hCdGoe201C) Cd–O(1) Cd–O(2) Cd–Fe(1) Cd–Fe(2) Cd–Cd(1) Cd–Cd(2) 16h

16 h, 401C ( CdGoe401C) Cd–O(1) Cd–O(2) Cd–Fe(1) Cd–Fe(2) Cd–Cd(1) 180 d, 201C (180dCdGoe201C) Cd–O(1) Cd–O(2) Cd–Fe(1) Cd–Fe(2) Cd–Fe(2) Cd–Cd(1)

3.5 2.4 1.0 1.0 0.6 0.8

2.0 3.5

196

M. Gra¨fe et al.


180 d, 401C (180dCdGoe401C) Cd–O(1) Cd–O(2) Cd–Fe(1) Cd–Fe(1) Cd–Fe(2) Cd–Cd(1)

Na

Rb (A˚)

5.1 1.7 1.2 0.6 0.4 0.4

2.23 2.35 3.12 3.32 3.83 3.62

s2c (A˚2)

DE0d (eV)

(1.20–3.72 A˚ (R+DR)) 0.007 4.7 0.007 4.7 0.008 4.7 0.008 4.7 0.008 4.7 0.008 4.7

Res.e

2.8 2.7

a

N: coordination number. R: radial distance. c 2 s : Debye–Waller parameter. d DE0: phase shift. e Res.: fit residual. b

at 2.34 A˚, suggesting an octahedral coordination environment with two elongated O positions. None of the four experimental spectra showed second shell RSF peaks of the same magnitude as those of Cd-substituted goethite, Cd(OH)2, CdO (monteponite), or CdCO3 (otavite) which suggested that precipitates or lattice incorporations into the goethite structure were not major constituents of goethite-sorbed Cd (Fig. 7.2b). RSFs of 16hCdGoe201C and 180dCdGoe201C had second shell peaks at circa 2.5 A˚, which broadened and shifted to circa 2.7 A˚ (R+DR) for samples aged at 401C. Above 3 A˚ (R+DR), the peaks in the RSFs of 16hCdGoe201C and Cd(NO3)2 (aq.) coincided, but shifted to lower R+DR values and lost intensity as the temperature and residence time increased. Non-linear least-square multi-shell fits indicated that Cd was coordinated to the goethite surface by 1.5 Fe atoms at 3.15 A˚ after aging for 16 h at 201C (Table 7.2; Fig. 7.3a and b). The two peaks above 3 A˚ (R+DR) for this sample (16hCdGoe201C) were fitted with 1.7 Cd atoms at 3.59 and 3.82 A˚ and thus fits were similar to results obtained for Cd(NO3)2 (aq.). After 180 d aging at 201C, Cd was coordinated at the goethite surface by 1.0 Fe atom at 3.14 A˚, 1.0 Fe atom at 3.28 A˚, and 0.6 Fe atoms at 3.85 A˚. It should be noted that the peak above 3 A˚ (R+DR), which we modeled with Cd–Fe backscattering atoms at 3.85 A˚, could also be fitted with 0.8 Cd atoms at 3.59 A˚ (Table 7.2). These alternative Cd–Cd fits are shown in italics in Table 7.2. Fit results for Cd sorbed on goethite after 16 h at 401C showed 0.6 Fe atoms at 3.15 A˚ and 0.5 Fe atoms at 3.81 A˚. After 180 d of aging at 401C, Cd was coordinated at the goethite surface by 1.2 Fe atoms at 3.12 A˚, 0.6 Fe atoms at 3.32 A˚, and 0.4 Fe atoms 3.83 A˚.


197

Figure 7.3: Non-Linear Least-Square Fit of 16hCdGoe201C (16 h–201C), 180d CdGoe201C (180 d–201C), 16hCdGoe401C (16 h–401C), and 180dCdGoe401C (180 d–401C): (a) k3w(k) Function (2–11.5 A˚1), (b) RSF from 2 to 6 A˚ (R+DR). 1E: Single Edge-Sharing Complex; 2C: Corner-Sharing Complex. The Right-Hand Side Superscript denotes the Backscattering Pair (e.g., Cd–Fe) and the Subscript the Fitted Radial Distance (e.g., 3.15 A˚).

7.4. Discussion Coordination of Cd ions by Fe octahedra at the goethite surface at various radial distances (3.15, 3.30, and 3.80 A˚) suggested that Cd formed innersphere complexes on the goethite surface with varying adsorption geometries (e.g., edge-sharing, corner-sharing, etc.). The fit results of 16hCdGoe201C suggest that Cd formed a 1ECdFe surface complex on the goethite surface. This spectrum, however, also showed significant contributions from Cd(NO3)2 (aq.), which disappeared with aging and temperature and likely reflected an increased removal of Cd from the entrained solution of the moist pastes as a result of surface adsorption reactions. We could not determine with certainty if the peaks above 3 A˚ (R+DR) in the sorption sample (16hCdGoe201C) stemmed solely from Cd backscattering atoms; however, it was likely that contributions existed from Fe backscattering neighbors as goethite’s surface sites were 73% saturated (0.06 mmol Cd m2). The fit

198

M. Gra¨fe et al.

attempt with Fe as a backscattering neighbor suggested 1.2 Fe atoms at 3.79 A˚ (Table 7.2) indicative of a double-corner sharing (2CCdFe) complex with A-type hydroxyls on the {1 1 0} and/or {1 0 0} planes of goethite. After 180 d of aging at 201C, two additional Cd surface complexes on goethite could be identified. RCdFe at 3.28 A˚ in conjunction with RCdFe at 3.14 A˚ indicated double-edge sharing (2E) surface complexes of Cd ions at terminal Fe-octahedra of double-chains at the {0 0 1/0 2 1} growth planes of goethite (Fig. 7.4) (Spadini et al., 1994; Randall et al., 1999). The disappearance of Cd–Cd scattering above 3 A˚ (R+DR) revealed 0.8 neighboring Fe atoms at 3.85 A˚, which was indicative of 2CCdFe complexes on the {1 1 0} and/or {1 0 0} planes of goethite. The radial distance is 0.05–0.1 A˚ longer than those obtained by Randall et al. (1999) and Boily et al. (2005); however, Spadini et al. (1994) suggested that the 2C complex could vary by as much as

Figure 7.4: Simplified Graphical Display of Cd Sorption Complexes on Goethite. g-Cd(OH)2 is shown to demonstrate the Similarity of Crystal Structure to that of Goethite. The Face-Sharing (FCdCd) Distance is 3.11 A˚, the Edge-Sharing (ECdCd) Distance is 3.42 A˚, and the Corner-Sharing (2C) Distance is 3.85–3.91 A˚.


199

0.1 A˚ depending on the phase shift functions used during the fit routine. In this study, we calculated Cd–Fe scattering paths based on the structural refinement of a Cd-substituted goethite sample (Huynh et al., 2003). The accuracy of the second shell radial distance in our study was 0.01 A˚ based on the refinement of the CdO XAFS spectrum with FEFF 7 calculated phase shift and amplitude functions and compared to the structural refinement of XRD data of monteponite (CdO). This accuracy may however not apply to the sorption samples, because there are no proper structural models for Cd–Fe scattering amplitude and phase shift functions despite crystallographic data from the Cd-substituted goethite sample. This spectrum could only be fitted with a 2ECdFe surface complex at 3.09 and 3.38 A˚, which is rather similar to the local coordination environment of Cd in g-Cd(OH)2 and suggested that incorporated Cd ions in goethite did not mimic the average coordination environment of Fe as suggested by Rietveld refinement (Lecerf et al., 1988; Huynh et al., 2003). It is more likely that the accuracy of RCdFe is circa 70.05 A˚, which is within the accepted range of accuracies for second and third shell radial distances (Manceau et al., 2000). Due to the absence of Cd(NO3)2 (aq.) contributions in 16hCdGOe401C, the aging effect was more noticeable at 401C than at 201C. The main difference between the two samples was the emergence of Fe neighbors at 3.32 A˚, which was indicative of 2ECdFe complexes forming on the growth planes {0 0 1/0 2 1} on goethite over time similar to the behavior of the samples aged at 201C. Values for NCdFe were similar between the two 401C samples, but were significantly smaller than their 201C counterparts despite similar Debye–Waller parameters (s2). The reason for the reduced amplitudes of Cd–Fe backscattering pairs in 401C samples is not entirely clear, but smaller peaks in the RSFs of the 401C samples and a significant increase in the signal-to-noise ratio above 8 A˚1 in all samples suggest that it may be an artifact of limited data quality due to the low surface concentration. The effects of temperature (201C vs. 401C) for 16 h aging periods on the surface speciation of Cd on goethite were minor. Fit results for both 16 h samples suggested that Cd was coordinated to the surface largely as an 1 ECdFe complex. The major difference between the two samples was the absence of contributions from Cd(NO3)2 (aq.) in 16hCdGoe401C permitting the refinement of a Cd–Fe shell at 3.81 A˚ indicative of the 2CCdFe complex. Temperature effects on Cd speciation for 180 d aging periods were equally minor, with the major difference being significantly smaller values of NCdFe, particularly for the 2C complex (NCdFe ¼ 0.4 vs. 0.8). Lower NCdFe values in the second shell of Cd are in agreement with reduced amplitudes of the relevant peaks in the RSFs. Visual inspection of the raw k3-weighted w(k) functions shows little or no amplitude or phase differences. Some differences

200

M. Gra¨fe et al.

are observed above 8 A˚1; however, the signal-to-noise ratio is significantly lower above 8 A˚1 as well, suggesting that a comparison of NCdFe is again limited by data quality. In the context of desorption data recently reported by Mustafa et al. (2006), the increased stability of Cd as function of aging time (Fig. 7.1) can be correlated to the emergence of the 2ECdFe complex at circa 3.30 A˚ in the EXAFS data. The absence of this surface complex after 16 h in the 20 and 401C samples may explain why 70 and 50% of Cd desorbed from the surface, respectively, suggesting that 1ECdFe complexes do not provide much stability. The 20% difference in Cd desorption between the 16 h–201C and 16 h–401C samples is likely due to the formation of 2CCdFe complexes as well as greater electrostatic attraction between Cd ions and the surface as the point of zero charge of metal oxide surfaces is known to decrease with increasing temperature (Mustafa et al., 1998; Rudzinski et al., 2000). This may possibly also explain why we observed greater Cd(NO3)2 (aq.) contributions in the 201C–16 h EXAFS sample than in the 401C–16 h EXAFS sample. Randall et al. (1999) reported that the 2ECdFe surface complex is the most stable surface geometry of Cd on goethite based on DFT calculations, which is in good agreement with the observations of this study and those of Mustafa et al. (2006). Despite the 73% surface coverage of Cd on goethite in this study, the emergence of 2CCdFe complexes remained low and did not mask contributions from 1,2ECdFe surface complexes as suggested previously by Spadini et al. (1994) and Manceau et al. (2000). We believe that the observations of 2CCdFe surface complexes are limited to studies in which either the equilibration pH is higher and/or in which the initial Cd concentration in solution is significantly greater (compare Table 7.1 with experimental conditions in this study). Indeed, Randall et al. (1999) and Spadini et al. (1994) achieved increased surface loadings by increasing the equilibration pH and using initial Cd concentrations ranging between 26 and 1,050 mM, which is up to six times greater than the initial Cd concentration used in this study. With pH above 6.0 (this study), the theoretical surface coverage will be significantly greater for Cd sorption on goethite. This suggests that at pH 6, the majority of actively binding surface sites on goethite occur at the terminal Fe-octahedra on the {0 2 1/0 0 1} growth planes as suggested by the observed Cd surface complexes in this study.

7.5. Conclusions Structural parameters gleaned from Cd K-edge XAFS data have related the local bonding environment of Cd sorbed on goethite for varying aging


201

periods and temperatures and have shown that the increasing stability of Cd ions at the goethite surface is primarily related to the time-dependent formation of 2ECdFe surface complexes. EXAFS data do not suggest that the surface complexes after 180 d of aging at 20 or 401C are significantly different. The stability of Cd ions at the surface at elevated temperatures is likely increased due to electrostatic effects at the surface owing to a lowering of the pzc on goethite as a function of elevated temperature. The increasing stability of Cd ions at the goethite surface (Fig. 7.1) suggests therefore that these surface complexes form over time, and may be accelerated at elevated temperatures; however, to prove this hypothesis, additional EXAFS data would be needed for aging periods intermediate of 16 h and 180 d at 20 and 401C to elucidate the rate of 2ECdFe complex formation.

ACKNOWLEDGMENTS This work was supported by the Australian Synchrotron Research Program (GUP-2574), which is funded by the Commonwealth of Australia under the Major National Research Facilities Program. Use of the Advanced Photon Source was supported by the U.S. Department of Energy, Office of Science, Basic Energy Sciences, under Contract No. W-31-109-Eng-38. The authors are grateful for the support received from Dr. M. Balasubramanian and Dr. S.M. Heald at beamline 20 BM-B (PNC-CAT). Ghulam Mustafa appreciated the funding of his Ph.D. research through the F.H. Loxton Scholarship.

REFERENCES Adriano, D. C. (1986). Trace Elements in the Terrestrial Environment. SpringerVerlag, New York. ATSDR. (2007). Agency for toxic substances and disease registry; http://www. atsdr.cdc.gov/cercla/05list.html, 03/2007 Backes, C. A., McLaren, R. G., Rate, A. W., & Swift, R. S. (1995). Kinetics of cadmium and cobalt desorption from iron and manganese oxides. Soil Sci. Soc. Am. J., 59, 778–785. Benjamin, M. M., & Leckie, J. O. (1982). Effects of complexation by Cl, SO4, S2O3 on adsorption behavior of Cd on oxide surfaces. Environ. Sci. Technol., 76, 162–170. Bingham, F. T., Peryea, F. J., & Jarrell, W. M. (1986). Metal toxicity to agricultural crops. In: H. Siegel (Ed). Metal Ions in Biology, Vol. 20. Marcel Dekker, New York, pp. 119–134.

202

M. Gra¨fe et al.

Bochatay, L., Persson, P., Lovgren, L., & Brown, G. E. (1997). XAFS study of Cu(II) at the water–goethite (alpha-FeOOH) interface. J. Phys. IV, 7 (C2), 819–820. Boily, J. F., Sjoberg, S., & Persson, P. (2005). Structures and stabilities of Cd(II) and Cd(II)–phthalate complexes at the goethite/water interface. Geochim. Cosmochim. Acta, 69 (13), 3219–3235. Bortnikov, N. S., Dobrovolskaya, M. G., Genkin, A. D., Naumov, V. B., & Shapenko, V. V. (1995). Sphalerite-galena geothermometers – distribution of cadmium, manganese, and the fractionation of sulfur isotopes. Econ. Geol. Bull. Soc. Econ. Geol., 90 (1), 155–180. Brown, I. D., & Altermatt, D. (1985). Bond valence parameters obtained from a systematic analysis of the inorganic structural database. Acta Crystallogr., B41, 244–247. Bruemmer, G. W., Gerth, J., & Tiller, K. G. (1988). Reaction kinetics of the adsorption and desorption of nickel, zinc and cadmium by goethite: I. Adsorption and diffusion of metals. J. Soil Sci., 39, 37–52. Bunker, G. (2004). Overview of the standard XAFS data analysis procedure; http://gbxafs.iit.edu/training/tutorials.html, 04/2004 Davis, A. P., & Upadhyaya, M. (1996). Desorption of cadmium from goethite (alpha-FeOOH). Water Res., 30 (8), 1894–1904. Davis, J. A., Fuller, C. C., & Cook, D. C. (1987). A model for trace-metal sorption processes at the calcite surface – adsorption of Cd2+ and subsequent solidsolution formation. Geochim. Cosmochim. Acta, 51 (6), 1477–1490. Davis, J. A., James, R. O., & Leckie, J. O. (1978). Surface ionization and complexation at the oxide/water interface: I. Computation of electrical double layer properties in simple electrolytes. J. Colloid Interface Sci., 63, 480–499. Graf, D. L. (1961). Crystallographic tables for the rhombohedral carbonates. Am. Mineral., 46, 1283–1316. Gray, C. W., McLaren, R. G., Roberts, A. H. C., & Condron, L. M. (1998). Sorption and desorption of cadmium from some New Zealand soils: Effect of pH and contact time. Aust. J. Soil Sci., 36, 199–216. Heald, S. M., Brewe, D. L., Stern, E. A., Kim, K. H., Brown, F. C., Jiang, D. T., Crozier, E. D., & Gordon, R. A. (1999). XAFS and micro-XAFS at the PNC-CAT beamlines. J. Synchrotron. Radiat., 6, 347–349. Huynh, T. (2001). Crystallographic and Chemical Properties of Cu and Cd Substituted Goethites. Masters of Science, The University of Sydney. Huynh, T., Tong, A. R., Singh, B., & Kennedy, B. J. (2003). Cd-substituted goethites – a structural investigation by synchrotron X-ray diffraction. Clays Clay Min., 51 (4), 397–402. Johnson, B. B. (1990). Effect of pH, temperature, and concentration on the adsorption of cadmium on goethite. Environ. Sci. Technol., 24, 112–118. Lecerf, A., Riou, A., Cudennec, Y., & Gerault, Y. (1988). Etude physico-chemique et structurale de l’hydroxide de cadmium g Cd(OH)2. Mater. Res. Bull., 23, 1479–1490.


203

Levelut, C., Sainctavit, P., Ramos, A., & Petiau, J. (1995). Linear-X-ray dichroism of cadmium-sulfide with wurtzite and zincblende structures. J. Phys. Condens. Matter, 7 (11), 2353–2367. Loganathan, P., Hedley, M. J., Grace, N. D., Lee, J., Cronin, S. J., Bolan, N. S., & Zanders, J. M. (2003). Fertiliser contaminants in New Zealand grazed pasture with special reference to cadmium and fluorine: A review. Aust. J. Soil Res., 41 (3), 501–532. Lorens, R. B. (1981). Sr, Cd, Mn and Co distribution coefficients in calcite as a function of calcite-precipitation rate. Geochim. Cosmochim. Acta, 45 (4), 553–561. Manceau, A., Nagy, K. L., Spadini, L., & Ragnarsdottir, K. V. (2000). Influence of anionic layer structure of Fe-oxyhydroxides on the structure of Cd surface complexes. J. Colloid Interface Sci., 228 (2), 306–316. Mosselmans, J. F. W., Schofield, P. F., Charnock, J. M., Garner, C. D., Pattrick, R. A. D., & Vaughan, D. J. (1996). X-ray absorption studies of metal complexes in aqueous solution at elevated temperatures. Chem. Geol., 127, 339–350. Mustafa, G., Kookana, R. S., & Singh, B. (2006). Desorption of cadmium from goethite: Effects of pH, temperature and aging. Chemosphere, 64, 856–865. Mustafa, G., Singh, B., & Kookana, R. S. (2004). Cadmium adsorption and desorption behaviour on goethite at low equilibrium concentrations: Effects of pH and index cations. Chemosphere, 57 (10), 1325–1333. Mustafa, S., Dilara, B., Neelofer, Z., Naeem, A., & Tasleem, S. (1998). Temperature effect on the surface charge properties of g-Al2O3. J. Colloid Interface Sci., 204, 284–293. Naidu, R., Kookana, R. S., Sumner, M. E., Harter, R. D., & Tiller, K. G. (1997). Cadmium sorption and transport in variable charge soils: A review. J. Environ. Qual., 26 (3), 602–617. Nordberg, G. (2004). Cadmium and health in the 21st century – historical remarks and trends for the future. Biometals, 17, 485–489. Parkman, R. H., Charnock, J. M., Bryan, N. D., Livens, F. R., & Vaughan, D. J. (1999). Reactions of copper and cadmium ions in aqueous solution with goethite, lepidocrocite, mackinawite, and pyrite. Am. Mineral., 84 (3), 407–419. Prozialecka, W. C., Edwards, J. A., & Woods, J. M. (2006). The vascular endothelium as a target of cadmium toxicity. Life Sci., 79, 1493–1506. Ramstedt, M., Norgren, C., Shchukarev, A., Sjo¨rberg, S., & Persson, P. (2005). Co-adsorption of cadmium(II) and glyphosate at the water–magnetite (g-MnOOH) interface. J. Colloid Interface Sci., 285, 493–501. Randall, S. R., Sherman, D. M., Ragnarsdottir, K. V., & Collins, C. R. (1999). The mechanism of cadmium surface complexation on iron oxyhydroxide minerals. Geochim. Cosmochim. Acta, 63 (19/20), 2971–2987. Ressler, T. (1998). Winxas: A program for X-ray absorption spectroscopy data analysis under MS-windows. J. Synchrotron. Radiat., 5, 118–122. Rudzinski, W., Piasecki, G., Panas, G., & Charmas, R. (2000). Calorimetric effects and temperature dependence of simple ion adsorption at oxide/electrolyte

204

M. Gra¨fe et al.

interfaces: The systems in which PZC and CIP do not coincide. J. Colloid Interface Sci., 226, 353–363. Sery, A., Manceau, A., & Greaves, G. N. (1996). Chemical state of Cd in apatite phosphate ores as determined by EXAFS spectroscopy. Am. Mineral., 81 (7–8), 864–873. Singh, B. (2001). Heavy metals in soils: Sources, chemical reactions and forms. In: Proceedings of the 2nd Australia and New Zealand Conference on Environmental Geotechnics – Geoenvironment 2001, Australian Geochemical Society, Newcastle, NSW, Australia, pp. 77–93. Spadini, L., Manceau, A., Schindler, P. W., & Charlet, L. (1994). Structure and stability of Cd2+ surface complexes on ferric oxides. 1. Results from EXAFS spectroscopy. J. Colloid Interface Sci., 168, 73–86. Sposito, G. (1989). The Chemistry of Soils. Oxford University Press, New York. Stern, E. A. (1993). Number of relevant independent points in X-ray absorption fine structure spectra. Phys. Rev. B, 48, 9825–9827. Tiller, K. G., Gerth, J., & Brummer, G. (1984a). The relative affinities of Cd, Ni and Zn for different soil clay fractions and goethite. Geoderma, 34 (1), 17–35. Tiller, K. G., Gerth, J., & Brummer, G. (1984b). The sorption of Cd, Zn and Ni by soil clay fractions: Procedures for partition of bound forms and their interpretation. Geoderma, 34, 1–16. Urabe, T. (1977). Partition of cadmium and manganese between coexisting sphalerite and galena from some Japanese epithermal deposits. Miner. Depos., 12 (3), 319–330. Venema, P., Hiemstra, T., & vanRiemsdijk, W. H. (1996). Multisite adsorption of cadmium on goethite. J. Colloid Interface Sci., 183 (2), 515–527. Wang, K. J., & Xing, B. S. (2004). Mutual effects of cadmium and phosphate on their adsorption and desorption by goethite. Environ. Pollut., 127 (1), 13–20. Zabinsky, F., Rehr, J. J., Ankudinov, A., Albers, R. C., & Eller, M. J. (1995). Multiple-scattering calculations of X-ray absorption spectra. Phys. Rev. B Condens. Matter, 52, 2995–3009. Zhang, J. (1999). Room-temperature compressibilities of MnO and CdO: Further examination of the role of cation type in bulk modulus systematics. Phys. Chem. Min., 26 (8), 644–648.


Chapter 8

Cadmium and Lead Desorption from Kaolinite Prashant Srivastava1, Markus Gra¨fe1, Balwant Singh1, and Mahalingam Balasubramanian2 1

Faculty of Agriculture, Food and Natural Resources, The University of Sydney, NSW 2006, Australia 2 Advanced Photon Source, Argonne National Laboratory, Argonne, IL 60439, USA

ABSTRACT Experiments were conducted to evaluate the effects of aging and temperature on surface speciation of cadmium (Cd) and lead (Pb) on kaolinite using X-ray absorption fine structure (XAFS) spectroscopy and column desorption experiments. Cd formed outer-sphere complexes on kaolinite and there were no detectable changes in XAFS fitting results due to aging and temperature. A decrease in Cd desorption after 60 days at 401C was observed along with small changes in the XAFS for this sample between 10 and 11 A˚1, which could not be quantified using non-linear fitting. A significant decrease in Pb desorption from kaolinite was due to the formation of polynuclear Pb–hydroxide complexes, which with increased aging stabilized at the kaolinite surface. Increased stability coincided with greater Pb–Pb edge sharing, suggesting an increase in size of polynuclear complexes with age and temperature. The Pb desorption behavior showed that time was a more effective stabilizing factor than temperature, suggesting that polynuclear Pb complexes may undergo Ostwald ripening.

8.1. Introduction Elevated concentrations of cadmium (Cd) and lead (Pb) commonly occur at contaminated sites. Average Cd concentrations in soils not exposed to obvious sources of pollution lie in the range 0.06–1.1 mg kg1 with a Corresponding author. Tel.: +61 2 9351 2237; Fax: +61 2 9351 2945;

E-mail: [email protected] (B. Singh).

206

P. Srivastava et al.

minimum of 0.01 mg kg1 and maximum of 2.7 mg kg1 (Kabata-Pendias and Pendias, 1992). Pb concentrations in uncontaminated soils range from 10 to 106 mg kg1, whereas at mining sites Pb concentration has been reported to be as high as 30,090 mg kg1 (Davies, 1995). Cd and Pb have contrary behaviors in soils. Pb is relatively less mobile, whereas Cd exhibits a greater tendency to dissociate from insoluble inorganic and organic complexes to form soluble ionic species that remain stable at neutral or slightly alkaline pH (Kabata-Pendias and Pendias, 1992). Cd forms relatively weak complexes with hydroxyl ions, such that the hydrolysis of Cd becomes significant only in concentrated solutions at pH values in excess of 7.0. Pb forms relatively strong complexes with hydroxyl ions. Below pH 8.0, only Pb2+ and PbOH+ contribute significantly to total Pb in soil solution (Lindsay, 1979). Adsorption–desorption and precipitation–dissolution processes govern the bioavailability and toxicity of most heavy metals in soils and sediments. A great deal of research has been focused on the understanding of adsorption reactions in soils; however, there has been little research done on desorption reactions involving heavy metals (Gao et al., 2003; Moreno et al., 2006). Adsorption–desorption reactions are often not completely reversible and this nonsingularity or hysteresis has been observed to increase with increased aging of heavy metals with soil constituents (McLaren et al., 1986, 1998; Ainsworth et al., 1994; Backes et al., 1995; Eick et al., 1999; Glover et al., 2002). Both reversible and hysteretic adsorption reactions of Cd and Pb have been observed in soils and soil minerals. For example, in soils and some oxides, increasing the length of the adsorption period before measuring desorption caused a decrease in the proportions of Cd or Co desorbed (Bruemmer et al., 1988; Barrow et al., 1989; Mustafa et al., 2006). Complete reversibility of adsorbed Cd from illite, kaolinite and an Oxisol (Comans, 1987; Puls et al., 1991; Naidu et al., 1997) and of adsorbed Pb from aluminum oxides (Strawn et al., 1998) has also been observed. Mustafa et al. (2006) observed that for any initial surface loading and equilibrium pH, Cd desorption from goethite decreased with aging and increased in temperature. Various explanations have been proposed for effects of aging on desorption of metallic cations from minerals and soils, for instance, solid state diffusion within oxide particles (Bruemmer et al., 1988), diffusion into micropores and intra-particle spaces (Backes et al., 1995), change in the type of surface complex (McBride, 1994), incorporation into the mineral structure via recrystallization (Ainsworth et al., 1994), surface catalyzed oxidation and incorporation into the crystal matrix (Backes et al., 1995) and surface catalyzed hydrolysis and precipitation (Backes et al., 1995). Our knowledge of heavy metal adsorption–desorption processes has been based mainly on the study of equilibrium conditions using thermodynamic

Cadmium and Lead Desorption from Kaolinite

207

approaches. Such approaches can only predict the final state of a soil system from an initial, non-equilibrium state. Analysis of the kinetics may yield important information concerning the nature of the reaction at a given time. In a desorption kinetics study, Backes et al. (1995) observed that a single-site model adequately accounted for the release of Cd and Ni from a neutral soil, which indicated the presence of uniform energy sites responsible for releasing these metals. Eick et al. (1999) and Glover et al. (2002) examined Pb desorption kinetics on goethite and observed that its slow desorption was modeled best by the parabolic diffusion equation. At all sorption densities, desorption rate coefficients and the quantity of Pb desorbed were greater for the short-term (lesser aging periods) experiments than long-term (longer aging periods) experiments; however, the differences were not statistically significant. Kaolinite is the most abundant phyllosilicate in highly weathered tropical soils and possesses a small permanent negative charge (Singh and Gilkes, 1992). The mobility and availability of heavy metals including Cd and Pb in highly weathered soils in tropics and sub-tropics may depend on the adsorption of these metals onto kaolinite. Constant capacitance surface complexation modeling of Cd and Pb adsorption onto kaolinite suggests two kinds of binding sites on kaolinite (Schindler et al., 1987; Angove et al., 1997, 1998; Ikhsan et al., 1999; Srivastava et al., 2005). The first kind of adsorption sites adsorbs Cd and Pb by outer-sphere complexation, whereas the second kind of adsorption sites involves inner-sphere binding of metals to ampholytic surface hydroxyls (SOH) groups. However, direct spectroscopic methods such as X-ray absorption spectroscopy (XAS) are required to confirm these observations. Relatively fewer spectroscopic studies have examined the effects of aging on local environment of adsorbed metals onto kaolinite (Eick et al., 1999; Thompson et al., 2000). Thompson et al. (2000) observed that Co release from kaolinite decreased with increasing adsorption period, which indicated that the adsorbed phase became increasingly stable with adsorption time. They attributed the decreased Co release to the formation of cobalt hydrotalcite-like precipitates over longer aging periods. Extended X-ray absorption fine structure (EXAFS) spectroscopy has been employed to characterize the adsorption of aqueous Pb complexes on Al- and Fe-oxides (Chisholm-Brause et al., 1990; Bargar et al., 1997a,b; Strawn et al., 1998). The EXAFS spectroscopic results of Pb(II) complexes at the g-Al2O3– water interface suggest that Pb(II) bonds directly to the mineral surface as inner-sphere complexes (Chisholm-Brause et al., 1990). However, Bargar et al. (1997a) observed water molecules along with Pb adions showing possibilities of outer-sphere complexes as well. An EXAFS study of Ni complexation and precipitation on kaolinite by Eick and Fendorf (1998) revealed the presence of multinuclear Ni surface precipitates with a similar, but

208


not identical, local structural environment to pure crystalline Ni(OH)2. Sahai et al. (2000) suggested that Sr primarily adsorbed as a hydrated surface complex to kaolinite at pH>7.5 and no Sr precipitates were formed. Quantitative analyses of the EXAFS data of kaolinite samples from this study showed a single first shell of 9–10 (71) oxygen atoms around Sr at an average distance of 2.61 (70.02) A˚, indicating hydrated surface complexes. Aging for up to 57 days and the presence of dissolved phosphate did not change the coordination environment of Sr. In a recent X-ray absorption fine structure (XAFS) study, Gra¨fe et al. (2007) reported that Cd was sorbed onto kaolinite as an outer-sphere complex and a small fraction of Cd ions formed CdOHCl complexes, whereas Pb was sorbed as a polynuclear complex on kaolinite. With the above background, the objectives of this research were to examine the changes in local coordination environments around Cd and Pb on kaolinite surfaces due to aging (30 and 60 days) and temperature (20 and 401C) using XAFS spectroscopy (XANES and EXAFS) and subsequently evaluate the desorption behaviors of Cd and Pb from kaolinite.

8.2. Experimental Acid washed kaolinite (Ajax Chemicals, Sydney, Australia), without any further treatment, was used for the study. X-ray diffraction confirmed the presence of well crystalline kaolinite with traces of mica. Properties of the mineral relevant to cation sorption were determined as described in an earlier study (Srivastava et al., 2005). The cation exchange capacity and BET specific surface area of kaolinite were 115 mmolc kg1 and 14.5 m2 g1, respectively. The point of zero salt effect as measured by batch method (Zelazny et al., 1996) was 4.8. 8.2.1. Adsorption Experiments The adsorption experiments were conducted in a 500 mL borosilicate reaction vessel at controlled room temperature (22711C) under nitrogen environment. Two grams of kaolinite was placed in the reaction vessel to give a surface area concentration (BET surface area/volume of suspension) of 96.3 m2 L1 in 0.01 M NaNO3 background electrolyte. The suspension was stirred overnight (for 16 h) at its natural pH (5.0) to hydrate the mineral surface. The experiments were conducted at two initial solution concentrations (low ¼ 133.3 mM, i.e., 20 mmol g1 and high ¼ 666.7 mM, i.e., 100 mmol g1). The solution conditions were under-saturated with respect to Pb(OH)2 (log(IAP/Kso) ¼ 0.38) and Cd(OH)2 (log(IAP/Kso) ¼ 5.83).


209

The equilibrium pH was maintained at 6.0 by addition of 0.1 M HCl or 0.1 M NaOH. After equilibration for 1 h, the samples were centrifuged at 3,000 rpm for 20 min, the supernatants filtered through Whatman No. 1 filter paper and collected for analysis of respective metal(s) by atomic absorption spectrometer (Varian SpectrAA 220FS). After 1 h of equilibration, the mean amounts of Cd and Pb adsorbed on kaolinite from low solution concentration were 9.95 and 16.17 mmol g1, respectively, and from high initial solution concentration were 19.33 and 43.44 mmol g1, respectively. Kaolinite samples with the above amounts of adsorbed Cd or Pb were aged for different periods as described below.

8.2.2. Aging The moist kaolinite samples with adsorbed Cd or Pb were kept in Teflon tubes under N2 atmosphere for aging at 20 and 401C in combination treatments of 30 and 60 days, respectively. To maintain moisture content in the samples during the aging periods, the tubes were periodically opened and a few drops of double-deionized water were added, while purging the head space vigorously with N2 gas. After aging for the specified periods, the samples were dried at 401C for 48 h and then ground into fine powder using an agate mortar and pestle. The dried samples were stored in plastic tubes for desorption and EXAFS experiments. The sample collected after 1 h shaking at room temperature (‘‘no aging’’ at room temperature, 1 h–201C) was used as control for comparison.

8.2.3. Column Desorption After aging for 30 or 60 days, the dried solid samples were desorbed for individual heavy metal(s) for 72 h with continuous leaching using 0.01 M NaNO3 (pH 6.0) solutions. Metal adsorbed dried powder of aged kaolinite (0.1 g) was placed between 0.2 mm nitrocellulose Millipores filter papers, which were sandwiched between layers of acid washed quartz sand (4 g) in a vertical desorption column (inner diameter ¼ 1.3 cm and length ¼ 20 cm) (Pharmacia Biotechs XK16). Sodium nitrate solution was percolated through the column from bottom to top at a flow rate of 10 mL h1 using a peristaltic pump (Pharmacia Biotechs Pump P-1). Ten milliliters aliquots of the eluate were collected every hour using a continuous flow fraction collector (Pharmacia Biotechs RediFrac) for up to 72 h. The eluates were analyzed for

210


Cd or Pb using flame- and/or graphite furnace AAS. Desorption curves were plotted as cumulative amount of metal desorbed against time of elution.

8.2.4. XAFS Spectroscopy The XAFS spectroscopic experiments were conducted on kaolinite samples with high surface loadings (G) of Cd and Pb being 19.33 and 43.44 mmol g1 aged for 30 and 60 days at 20 and 401C. The samples were mounted into plastic sample holders and sealed on both sides by Kapton tape. For ease of communication, the Cd and Pb XAFS samples will be referred in the text as 30d–201C, 30d–401C, 60d–201C and 60d–401C samples where appropriate. The preparation of the reference materials, including an unaged sorption sample of Cd and Pb on kaolinite at room temperature, is described in detail elsewhere (Srivastava, 2005; Gra¨fe et al., 2007).

8.2.4.1. Data Collection XAFS spectroscopy data collection has been described in detail elsewhere (Gra¨fe et al., 2007). Briefly, Cd K-edge (26.711 keV) XAFS data for Cd sorption samples on kaolinite were collected on beamline 20BM (bending magnet) at PNC-CAT (Advanced Photon Source, Argonne National Laboratories, Argonne, IL, USA) using a 10 element solid state detector. The operating details of this beamline and the APS storage ring have been described in detail elsewhere (Heald et al., 1999). Pb LIII-edge (13.035 keV) XAFS data for Pb sorption samples were collected at the bending magnet beamline 20B of the Australian National Beamline Facility (ANBF) at the Photon Factory (KEK, Tsukuba, Japan) using a 10 element Ge solid state detector. The operating details of this beamline and the Photon Factory storage ring have been described elsewhere (Garrett et al., 1995; Katoh et al., 1998). The ionization (I0), the transmission (It) and the reference (Ir) columns were filled with mixtures of Ar and N2 gases resulting in circa 10% absorbance in I0 and circa 30% absorbance in It and Ir. Higher harmonics were rejected by detuning the incident beam in I0 by 50% at the ANBF, while higher harmonics were rejected with Rh coated mirrors behind the Si (1 1 1) double crystal monochromator at beamline 20BM (PNC-CAT). Approximately 12 scans were collected to improve the signal-to-noise ratio for each sample and were averaged prior to data refinement.


211

8.2.4.2. Data Analysis All data analysis was performed using WinXAS 2.1 and later versions (Ressler, 1998) following procedures described elsewhere (Bunker, 2003). Individual spectra were averaged and subsequently corrected for background absorbance and normalization. Spectra were then converted from energy to photo-electron wave vector (k) units (k is the wave vector number with units of A˚1) by assigning the origin, E0, to the first inflection point of the Cd K-edge and to the major 2p3/2-6d electron transition of the Pb LIII-edge (Cd ¼ 26.711 keV, Pb ¼ 13.035 keV). The EXAFS signal was extracted using a cubic spline function consisting of r7 knots applied over an average range in k-space (Cd: 1.3–12.0 A˚1, Pb: 0.5–11.0 A˚1). Fourier transformation (FT) of the raw k3w(k) function was performed (Cd: 2.0–11.00 A˚1, Pb: 2.0–10.0 A˚1) to obtain a radial structure function (RSF) using a Bessel window function and a smoothing parameter (b) of 4 to minimize truncation effects (ring tones) in RSFs. Raw k3-weighted w(k) Cd spectra were linearly fitted with Cd reference spectra and the resulting residuals were subsequently Fourier transformed and a portion (Cd: R+DR, 2.9–3.9 A˚; uncorrected for phase shift) of these RSFs was Fourier back-transformed for non-linear least-square fitting. This step was necessary to remove significant (475%) contributions of Cd(NO3)2 (aq.) from the CdKaol spectrum. The first shell of Cd was fitted as a k3-weighted, Fourier-filtered w(k) function encompassing the major first shell peak in the RSFs and its satellite peak at circa 2.45 A˚. The coordination number (N), the radial distance (R, A˚), the mean-square disorder of the radial distance (Debye–Waller parameter, s2, A˚2) and a single, cross-correlated (linked to the same value) phase shift value (DE0, eV) for all single backscattering paths were allowed to vary. For fits with multiple second shell contributions, s2 was correlated to the same value. The amplitude reduction factor (S 20 ) for Cd fits was fixed at 1.0 (Table 8.1). RSFs derived from raw k3-weighted w(k) Pb spectra were fitted between 1.2 and 4.6 A˚ (R+DR, i.e., uncorrected for phase shift) using the non-linear leastsquare fitting method. The fit-range was progressively increased until all major shells visible in the RSF were selected. The coordination number (N), R and s2 were allowed to float (first shell) or in some cases (1 h–201C and PbGib) were fixed at 0.010 A˚2 following examples shown by Bargar et al. (1997a). The phase shift function was split during the fitting into first and second shell neighbors due to the lack of a good reference standard for Pb–Al scattering paths which resulted in strong divergence of phase shifts (Table 8.2). This approach is practically similar to fitting the first and second shells separately, but instead provides a continuous fit function for both the raw k3-weighted w(k) function and its RSF. The amplitude reduction factor for Pb fits was fixed at 1.0.

212


Table 8.1: Structural Parameters of First and Second Shells Derived from Non-Linear Least-Square Fitting of Cd Reference Compounds and CdKaol. Sample

N

R

s2

DE0

Residual

30d–201C Cd–O(1) Cd–O(2) Cd–Cl(1) Cd–Cd Cd–Cd

2.2 3.3 0.3 1.1 1.5

2.21 2.33 2.79 3.47 3.73

0.002a 0.002a 0.002a 0.010 0.010

5.3 5.3 5.3 10.7 10.7

12.1


2.5 3.1 0.2 0.7 1.0

2.22 2.34 2.79 3.45 3.75

0.002a 0.002a 0.002a 0.009 0.009

4.4 4.4 4.4 13.1 13.1

7.0


2.4 2.9 0.2 1.1 1.5

2.17 2.34 2.79 3.46 3.73

0.002a 0.002a 0.002a 0.010 0.010

4.9 4.9 4.9 10.9 10.9


2.9 3.1 0.2 1.1 1.9

2.23 2.36 2.79 3.47 3.74

0.002a 0.002a 0.002a 0.011 0.011

3.9 3.9 3.9 11.8 11.8

References Cd(NO3)2 (aq.) Cd–O(1) Cd–O(2) Cd–Cd(1) Cd–Cd(2)

3.2 4.1 3.5 3.6

2.20 2.34 3.62 3.86

0.002 0.002 0.013 0.013

5.7 5.7 8.8 8.8

3.8 2.6 0.3 1.5 2.1

2.24 2.35 2.79 3.46 3.73

0.002 0.002 0.002 0.012 0.012

4.9 4.9 4.9 10.4 10.4

1 h–201C Cd–O(1) Cd–O(2) Cd–Cl(1) Cd–Cd Cd–Cd

22.6

18.9

8.6 22.0

6.9 19.3

8.8 17.1

5.1 17.9


213

Table 8.1: (Continued ). N

R

s2

DE0

Residual

6.7 0.8 2.0

2.25 3.14 3.36

0.013 0.003 0.013

5.8 9.9 9.9

13.7

Cd(OH)2 (s) Cd–O Cd–Cd

8.2 11.7

2.30 3.51

0.008 0.011

7.4 7.4

15.1

CdO Cd–O Cd–Cd

6.0 10.8

2.32 3.33

0.009 0.009

2.3 2.3

15.7

Sample CdGib Cd–O Cd–Cd Cd–Al

The accuracy of the fit parameters in the first shell was N75% and RCd–O70.04 A˚ and in the second shell N79% and for RCd–Cd70.01 A˚. The accuracy of NCd–Al and RCd–Al is probably greater than 9% and 70.01 A˚ due to the absence of an appropriate structural model from which reliable Cd–Al scattering paths can be derived. N ¼ coordination number; R ¼ radial distance, A˚; s2 ¼ Debye–Waller parameter, A˚2; DE0 ¼ phase shift, eV. a Fixed for cross-comparison to 1 h–201C reference sample.

The FEFF 7.02 program (Zabinsky et al., 1995) was used to calculate theoretical phase and amplitude functions of Cd/Pb–O, Cd/Pb–Cl, Cd/Pb–Al and Cd/Pb–Cd/Pb scattering paths using input files based on the structural refinement of monteponite (CdO), otavite (CdCO3), Al-substituted g-Cd(OH)2 and atacamite (b-Cd2(OH)3Cl) (Graf, 1961; de Wolff, 1966; Sokolova and Egorovtismenko, 1990; Zhang, 1999), and massicot (o-PbO), hydro-cerussite (Pb3[CO3]2(OH)2), cerussite (PbCO3), laurionite (Pb(OH)Cl) and Al-substituted massicot (Sahl, 1974; Venetopoulos and Rentzeperis, 1975; Hill, 1985; Martinetto et al., 2002). The number of permissible free floating parameters (Npts) for all non-linear least-square fits was determined from (2DRDk)p1 to limit Npts (Stern, 1993).

8.3. Results 8.3.1. Cadmium Desorption Cd desorption curves for all kaolinite samples were characterized by an initial rapid step followed by a slower step leaning toward a plateau (Fig. 8.1). An apparent aging effect on Cd desorption was observed at both surface

214


Table 8.2: Structural Parameters of First and Second Shells Derived from Non-Linear Least-Square Fitting of Pb Reference Compounds and PbKaol. Sample

N

R

s2

DE0

Residual

30d–401C (1.2–4.6 A˚, R+DR) Pb–O(1) Pb–O(2) Pb–Cl(1) Pb–Pb(1) Pb–Pb(2) Pb–Pb(3) Pb–Al(1)

1.2 2.4 0.3 1.3 1.7 0.4 0.7

2.17 2.36 2.74 3.50 3.70 4.30 4.46

0.005 0.005 0.005 0.009 0.009 0.009 0.009

8.71 8.71 8.71 10.01 10.01 10.01 10.01

3.12


1.0 1.9 0.2 0.7 1.1 0.4 0.4

2.19 2.38 2.78 3.44 3.63 4.19 4.39

0.001 0.001 0.001 0.003 0.003 0.003 0.003

5.92 5.92 5.90 8.03 8.03 8.03 8.03

3.06


1.0 1.8 0.2 1.0 1.3 0.5 0.5

2.21 2.39 2.76 3.45 3.65 4.22 4.41

0.002 0.002 0.002 0.004 0.004 0.004 0.004

6.06 6.06 6.06 8.49 8.49 8.49 8.49

3.93

References PbO (1.2–3.8 A˚, R+DR) Pb–O(1) Pb–O(2) Pb–Pb(1) Pb–Pb(2)

2.8 0.8 6.1 4.7

2.25 2.54 3.58 3.85

0.011 0.011 0.016 0.016

4.25 4.25 4.25 4.25

21.95

PbGib (1.2–4.6 A˚, R+DR) Pb–O(1) Pb–Al(1) Pb–Al(2)

1.5 0.5 1.3

2.26 3.16 3.42

0.005 0.01a 0.01a

9.58 9.58 9.58


215

Table 8.2: (Continued ). N

R

s2

DE0

Residual

Pb–Pb(1) Pb–Al(3) Pb–Pb(2)

1.4 1.7 1.1

3.52 4.28 4.64

0.01a 0.01a 0.01a

9.58 9.58 9.58

4.87

1 h–201C (PbKaol) (1.0–4.6 A˚, R+DR) Pb–O(1) Pb–O(2) Pb–Al(1) Pb–Pb(1) Pb–Pb(2)

2.7 2.1 0.9 0.4 0.7

2.31 3.11 3.22 3.67 4.56

0.015 0.015 0.01a 0.01a 0.01a

12.26 12.26 12.26 12.26 12.26

4.57

Pb(NO3)2 (aq.) (0.9–2.9 A˚, R+DR) Pb–O(1) Pb–O(2) Pb–O(3)

3.0 3.1 1.1

2.45 2.63 2.81

0.009 0.009 0.009

0.73 0.73 0.73

34.39

Sample

The fits were carried out on RSFs in the range as indicated per reference compound or sample. The accuracy of the fit parameters in the first shell was N743% and RPb–O70.04 A˚ and in the second shell, N728% and RPb–Pb70.01 A˚ based on a comparison to published crystallographic values of massicot (PbO, Pbnm) (Hill, 1985). N ¼ coordination number; R ¼ radial distance, A˚; s2 ¼ Debye–Waller parameter, A˚2; DE0 ¼ phase shift, eV. a Fixed at value indicated.

loadings (9.95 and 19.33 mmol g1). At no aging, nearly all of the adsorbed Cd was desorbed over a period of 72 h from the sample at low surface loading, whereas circa 92% of the adsorbed Cd could be desorbed at high surface loading. A decrease in Cd desorption was observed at low surface loading for the samples aged for 30 and 60 days at 20 and 401C, in comparison to the ‘‘no aging’’ sample. However, at high surface loading, a significant decrease in Cd desorption was observed only for the sample aged for 60 days at 401C, i.e., 40% Cd was desorbed. A temperature effect on Cd desorption was observed for the samples aged for 30 days at low surface loading and for the samples aged for both 30 and 60 days at high surface loading.

8.3.2. Cadmium EXAFS The raw k3-weighted w(k) spectra of the aged Cd sorption samples at 20 and 401C were dominated by a single sinusoidal wave function, which upon FT resulted in a dominant peak at circa 1.9 A˚, a small satellite peak at 2.45 A˚ and two low amplitude, yet very distinct peaks at 3.42 and 3.82 A˚ (R+DR,

216


Figure 8.1: Cadmium Desorption from Kaolinite Samples Aged for Different Periods at Two Temperatures at (a) Low Surface Loading (9.95 mmol g1) and (b) High Surface Loading (19.33 mmol g1).


217

Figure 8.2: (a) k3-Weighted w(k) Cd Spectra with their Corresponding RSFs: (b) First Ligand Shell and (c) Second and Third Nearest Atomic Neighbors. The Spectra 1 h–201C, CdGib and Cd(NO3)2 (aq.) are Taken from Gra¨fe et al. (2007) and are Reproduced here with the Permission of the Authors. uncorrected for phase shift, Fig. 8.2a and b). Non-linear least-square fitting of the major peak at 1.9 A˚ and its satellite at 2.45 A˚ following Fourierfiltering (k3w(k)) revealed two to three and three to four next nearest oxygen (O) neighbors at circa 2.20 and 2.32 A˚ and 0.2 Cl atoms at 2.45 A˚ (fitted radial distances are corrected for phase shift), respectively (Table 8.1). The sum of coordination numbers (NCd–O(1)+NCd–O(2)+NCd–Cl) and radial distances (average RCd–O2.28 A˚, RCd–Cl2.79 A˚) suggested octahedral symmetry of Cd with O and Cl atoms comprising the first ligand shell, which was in good agreement with previous studies (Sokolova and Egorovtismenko, 1990; Spadini et al., 1994; Parkman et al., 1999; Randall et al., 2001). The two low amplitude peaks between 3 and 4 A˚ (R+DR) were strongly suppressed in amplitude due to the large contributions from Cd(NO3)2 (aq.). Linear least-square combination fitting revealed that 93% of the 30d/60d– 201C spectra and 75 and 86% of the 30d/60d–401C spectra were composed of Cd(NO3)2 (aq.) and attempts to fit these two peaks as Fourier-filtered k3w(k)

218


Figure 8.3: (a) Residual k3-Weighted w(k) Cd Spectra Following the Linear Removal of Contributions from Cd(NO3)2 (aq.) and (b) their Corresponding RSFs. (c) Fourier-Filtered k3-Weighted w(k) Functions (J) and the NonLinear Least-Square Fit Function (—). All Spectra are Shown in the Same Order and are Identified in (b). The 1 h–201C Spectra were Taken from Gra¨fe et al. (2007) and are Reproduced here with the Permission of the Authors.


219

functions failed or produced non-meaningful results. Therefore, we attempted to fit the two peaks after linearly subtracting the Cd(NO3)2 (aq.) contributions from the raw k3-weighted w(k) spectra, i.e., from the generated residuals of the linear least-squares fit (Fig. 8.3a and b). The residual wave functions of the experimental Cd sorption samples had contributions from unfitted O and Cl backscattering neighbors below 3 A˚, and shifted the two peaks above 3 A˚ (R+DR; 3.42 and 3.82 A˚) to lower R+DR values: 3.20 and 3.70 A˚ (Fig. 8.3c). Non-linear least-square fitting of the Fourier-filtered residual peaks at 3.20 and 3.70 A˚ (R+DR) revealed circa one Cd atom at 3.47 A˚ and one to two Cd atoms at 3.72 A˚ (Table 8.1). Attempts to fit the peaks with Al and/or Si atoms or a combination of Cd, Al and Si atoms did not converge to meaningful fit results. 8.3.3. Lead Desorption The Pb desorption curves showed a fast desorption reaction initially followed by a slow desorption reaction, which tended to reach a plateau especially at the high surface loading of 43.44 mmol g1 (Fig. 8.4). Pb desorption from kaolinite decreased with aging at both surface loadings, and this effect was more pronounced in the samples with high surface loading. At low surface loading (16.17 mmol g1), only 10% of the adsorbed Pb was desorbed from the ‘‘no aging’’ sample, which decreased to 0.9 and 2.3% for the samples aged for 30 and 60 days at 201C. A negligible amount (less than 0.15%) of the adsorbed Pb was desorbed from the samples aged for 30 and 60 days at 401C. At high surface loading, circa 23% of the adsorbed Pb was desorbed from ‘‘no aging’’ sample, which decreased to circa 17 and 3% for the samples aged for 30 and 60 days at 201C. At 401C, circa 13% of the adsorbed Pb was desorbed from the sample aged for 30 days and very little (circa 0.2%) from the sample aged for 60 days at high surface loading. Many eluates collected from the samples aged at 401C at low surface loading and for sample aged for 60 days at 401C at high surface loading had Pb concentrations below the detection limit of the graphite furnace AAS (1 nM or 0.2 mg L1); hence, the desorbed amounts presented may not be accurate. 8.3.4. Lead LIII -Edge EXAFS Visual inspection of the first derivative spectra of the three aged samples at 20 and 401C showed a broad Gaussian absorption band centered around the main 2p3/2-6d orbital electron transition (Fig. 8.5) (Bargar et al., 1997a).

220


Figure 8.4: Lead Desorption from Kaolinite Samples Aged for Different Periods at Two Temperatures at (a) Low Surface Loading (16.17 mmol g1) and (b) High Surface Loading (43.44 mmol g1).


221

Figure 8.5: First Derivative XANES Spectra of Pb Sorption Samples Aged from 30 or 60 Days at 20 or 401C. The Reference Spectra Including 1 h–201C, PbGib, PbO2, PbCO3, Pb3(OH)2[CO3]2, PbO, Pb(OH)2 and Pb(NO3)2 (aq.) were Taken from Gra¨fe et al. (2007) and are Reproduced here with the Permission of the Authors.

222



223

Specific peaks above and below the 2p3/2 electron binding energy (E0) such as the ones observed in the spectra of Pb sorbed gibbsite (PbGib), Pb(NO3)2 (aq.), Pb(OH)2, PbO and Pb3(CO3)2(OH)2 were missing and instead a broad single peak suggested that the first ligand shell coordination environment was either strongly distorted or Pb existed in multiple coordination states. The aged and temperature treated samples were mostly similar to the 1 h–201C Pb sorption sample, which was shown recently by Gra¨fe et al. (2007) to be similar to the Pb4(OH)4+ sample shown by Bargar et al. (1997a), in which Pb exists 4 as a distorted trigonal pyramid. Visual inspection of raw, k3-weighted w(k) spectra of reference compounds (Fig. 8.6a) and aged Pb sorption samples at 20 and 401C (Fig. 8.7a) showed a significant phase shift between the Pb sorption samples on kaolinite (all of which were in phase with each other) and the PbGib and Pb(NO3)2 reference materials (Figs. 8.6a–c and 8.7a–c). The major peak in the corresponding RSFs centering between 1.8 and 1.9 A˚ (R+DR) corresponded to an uneven distribution of first shell O atoms for Pb sorbed on kaolinite and for Pb(NO3)2 in solution, while PbGib had a nearly perfect bell-shaped first ligand shell peak. Non-linear least-square fit results of the Fourier-filtered first shell contributions showed that Pb was coordinated with 1.0 and 2.0 O atoms at 2.20 and 2.38 A˚, and in addition by circa 0.3 Cl atoms at 2.74 A˚ in the temperature treated and aged samples at the kaolinite surface. This distribution of first shell ligands was different to that of the 1 h–201C PbKaol sorption sample, where the coordination state of Pb was influenced by contributions from Pb(NO3)2 (aq.) and possibly by Pb4(OH)4+ (Bargar et al., 1997a; Gra¨fe et al., 2007). 4 Visual inspection of magnified second shell regions in RSFs (Figs. 8.6c and 8.7c) shows that the moduli and imaginary parts of the Pb sorption samples aged for 30 and 60 days at 20 and 401C were dissimilar to the 1 h–201C PbKaol sorption sample, the PbGib and the Pb(NO3)2 (aq.) samples. Nonlinear least-square fit results show that the peaks above 2.5 A˚ (R+DR) were due to several neighboring Pb atoms at circa 3.47, 3.65 and 4.25 A˚ and 0.5 Al atoms at circa 4.4–4.45 A˚ indicating that Pb formed a polynuclear Pb surface complex on kaolinite (Table 8.2).

Figure 8.6: (a) Raw k3-Weighted w(k) Functions of Pb Reference Compounds and their Corresponding RSFs: (b) First Ligand Shell and (c) Second and Third Nearest Atomic Neighbors. These Spectra were Taken from Gra¨fe et al. (2007) and are Reproduced here with the Permission of the Authors.

224



225

8.4. Discussion 8.4.1. Cadmium Local Coordination and Desorption Behavior Non-linear least-square fit results show that the speciation of Cd at the kaolinite–water interface did not change with time or temperature. The only indication of a change in coordination environment was gleaned from linear least-square fits of the raw k3-weighted w(k) data, which suggested that both samples aged at 401C had circa 10% less contributions from Cd(NO3)2 (aq.). Second and third nearest Cd neighbors modeled between 3 and 4 A˚ (R+DR) and the presence of Cl in the first ligand shell of Cd suggested the possible formation of Cd(OH)Cl clusters, though. At pH 6 and at high initial Cd(NO3)2 solution concentration, the origin of Cl was likely the acid washed kaolinite. Cd(OH)Cl as well as Cd(OH)2 at pH 6 are under-saturated (log(IAP/Kso)5.8) at these solution conditions, which suggested that formation of Cd(OH)Cl clusters would have occurred due to the concentration of Cd ions at the (permanently negatively charged) kaolinite surface (CEC115 mmolc kg1). In atacamite (b-Cd2(OH)3Cl), Cd atoms are either coordinated by two Cl atoms and four O atoms or one Cl atom and five O atoms (Sokolova and Egorovtismenko, 1990). In the latter case, RCd–Cl is 2.77 A˚, which is similar to the Cd–Cl bond distance modeled for the satellite peaks at circa 2.45 A˚ (R+DR) in this study. Inter-atomic Cd–Cd distances (RCd–Cd) in the array of edge-sharing Cd-octahedra in atacamite are 3.45, 3.68 or 3.72 A˚, which agree well with RCd–Cd distances derived from non-linear least-square fits on the Fourier-filtered residual structure (Table 8.1). It was therefore likely that the linear XANES fit indicating 20% contributions from CdO was a proxy for Cd(OH)Cl (or possibly b-Cd2(OH)3Cl) contributions, but due to the absence of this standard from our list of available references, CdO provided enough fit improvement in the XANES spectrum of CdKaol, but not in its w(k) spectrum. The desorption behavior of aged samples at 20 and 401C was consistent with our XAFS results that showed that Cd speciation did not change except for the sample aged for 60 days at 401C. Cd speciation of the adsorption data by surface complexation modeling (Srivastava et al., 2005) showed that at Figure 8.7: (a) Raw k3-Weighted w(k) Functions of Pb Sorption Samples on Kaolinite Aged for 30 or 60 Days at 20 or 401C and their Corresponding RSFs: (b) First Ligand Shell and (c) Second and Third Nearest Atomic Neighbors.

226


pH 6.0 nearly all of the adsorbed Cd (more than 99%) formed outer-sphere complexes at the permanent charge sites (X2Cd), which was in good agreement with linear least-square fit results suggesting 93% contributions from Cd(NO3)2 (aq.). With increasing aging time and temperature, Cd desorption decreased; however, EXAFS fit results did not show any change in local coordination environment of Cd. Visual inspection (Fig. 8.2c) showed that the oscillation at circa 10.5 A˚1 diminished in amplitude with aging and temperature, which causes the saddle feature between 3 and 4 A˚ (R+DR) to diminish as well. This apparent change, however, could not be quantified with non-linear least-square fitting, which suggested that the decreased desorption is a physical rather than a chemical (speciation) effect. The decreased Cd desorption for the sample aged for 60 days at 401C could be due to interor intra-particle diffusion, a phenomenon that cannot be detected by XAFS spectroscopy. Johnson (1990) proposed that at higher temperatures, a decrease in the solvated ionic radii occurs, which would favor diffusion of ions into pore space within or between particles at elevated temperatures. As shown (Fig. 8.1a and b), Cd desorption decreased in the samples aged at 401C for 30 and 60 days, but not in the samples aged at 201C, suggesting that elevated temperatures during the adsorption period are critical to stabilize Cd adions on kaolinite. Since the desorption experiments were carried out at room temperature, desorption would be hindered by the expansion of the ionic radius of occluded Cd ions.

8.4.2. Pb Local Coordination and Desorption Behavior An important difference between the 1 h–201C and the 30d/60d–20/401C PbKaol sorption samples was the appearance of Cl atoms in the first ligand shell of Pb, which suggested that Pb was undergoing Cl–H2O ligand exchange reactions over time. This exchange came possibly at the expense of Pb–Al edge-sharing complexes, which we fitted to the unaged Pb sorption sample, but which could no longer be fitted to the Pb sorption samples that were aged at 20 and 401C for 30 and 60 days. The observation of Cl ions in the first ligand shell of Pb was in good agreement with EXAFS fit results of Cd sorption samples on kaolinite reported in this study, where a similar observation of Cl atoms in the first ligand shell of Cd was made. The number of atoms in the first ligand shell was circa 5 (accounting for circa 60% error), which agrees well with a trigonal pyramidal symmetry state, and in good agreement with the observed similarity of the first derivative XANES spectra of this study and Pb4(OH)4+ in the study by Bargar et al. (1997a). 4


227

Pb–Pb inter-atomic distances between 3.44 and 3.70 A˚ are indicative of Pb–Pb edge-sharing with the distance being controlled by the Pb–O bond distances and the Pb–O–Pb bond angles. In litharge, for example, the Pb–Pb edge-sharing distance is 3.66 A˚ due to a consistent Pb–O bond distance of 2.31 A˚ and a Pb–O–Pb bond angle of 118.61. Conversely, in massicot, two Pb–Pb edge-sharing distances at 3.54 and 3.61 A˚ reflect Pb–O bond distances at 2.20 and 2.48 A˚ and Pb–O–Pb bond angles of 96.6 and 1001, respectively. In laurionite (Pb(OH)Cl), the Pb–Cl bond distance occurs above 3 A˚, and the average Pb–Pb distances are dissimilar to the ones we obtained (Venetopoulos and Rentzeperis, 1975). This suggested that Cl–H2O ligand exchange was limited and that laurionite did not form with temperature and time. Instead, Pb–Pb inter-atomic distances of 3.44 and 3.65 suggested the formation of a network of edge-sharing Pb polyhedra with additional Pb atoms occurring between 4.20 and 4.30 A˚ being indicative of Pb–Pb single-corner sharing complexes as part of the polynuclear Pb complex. It is unclear to what extent Cl atoms may be responsible for bridging the polynuclear complex to the surface or linking Pb polyhedra into a polynuclear complex. The appearance of circa 0.5 Pb atoms at 4.40–4.45 A˚ was probably due to the formation of a single-corner sharing complex to the kaolinite surface. The increased number of Pb–Pb edge-sharing complexes in the samples aged for 30 and 60 days at 20 and 401C compared to the 1 h at 201C sample suggested that temperature and aging promoted the formation of polynuclear Pb complexes. This appears to have occurred at the expense of Pb–Al edge-sharing complexes observed in the 1 h–201C sample. Because the surface loading (43.44 mmol Pb g1) on kaolinite was constant for all Pb kaolinite sorption samples irrespective of aging or temperature, the local coordination environment at each temperature and aging reflected speciation changes of the same Pb atoms. In other words, aging and temperature did not increase the surface loading, which could have caused new Pb surface species to appear and mask EXAFS signal from other surface species. Such effects have been observed for Cd sorption on goethite as a function of surface loading (Spadini et al., 1994). The significant decline of the Debye–Waller parameter (s2) in the 60d–201C and 60d–401C samples suggested that time rather than temperature promoted the internal coherence or order of the complexes, which is consistent with the desorption behavior of Pb from kaolinite (Fig. 8.4b). The general decrease in Pb desorption with time and temperature (Fig. 8.4b) appears to be correlated to the formation of polynuclear Pb complexes as observed by additional Pb–Pb edge-sharing contributions in the EXAFS of the 30d–20/401C sample compared to 1 h–201C sample.

228


The speciation of Pb adsorption data from extended constant capacitance modeling (Srivastava et al., 2005) showed that at pH 6.0, circa 21% of the adsorbed Pb was in the form of outer-sphere complexes (X2Pb) and the rest of the adsorbed Pb formed inner-sphere complexes (SOPbOH) on the variable charge sites. Linear least-square fits of the raw k3-weighted w(k) spectra with reference spectra (data not shown) showed that the spectra were composed of circa 45% Pb(NO3)2 (aq.), representative of outer-sphere complexes at the kaolinite surface, and two times greater than predictions by surface complexation modeling. The discrepancy is likely due to the limitations of the model to predict the formation of polynuclear Pb complexes at the kaolinite surface. As shown by EXAFS, the number of surface bonds decreases over time, while the number of Pb–Pb edge-sharing contributions increased, suggesting that Pb increasingly formed polynuclear complexes. The high positive charge emitted by polynuclear Pb complexes such as Pb4(OH)4+ provides a high degree of electrostatic attraction to the perma4 nently negatively charged kaolinite surface and therefore a significant degree of surface stability even if the number of bonds to the surface are limited. Following the selectivity guidelines of Helfferich (1956) on cation exchange, cations such as Na+ or Ca2+ would be virtually ineffective in competing for negatively charged surface sites on kaolinite (permanent and variably charged sites) with polynuclear complexes of two to four times their charge in addition to the significant steric hindrance posed by the large polynuclear complexes.

8.5. Conclusions Cd was adsorbed on kaolinite as an outer-sphere complex and there is no detectable change in its surface speciation with aging and increased temperature. The decrease in Cd desorption after 60 days at 401C cannot be explained from EXAFS spectroscopy results because any contribution from Al backscattering atoms was undetectable in the overwhelming signals from outer-sphere Cd ions. This is due to the large difference in the atomic numbers of Al (Z ¼ 13) and Cd (Z ¼ 48) and associated reduction in backscattering amplitude. The observed decrease in Cd desorption is likely due to inter- and intra-particle diffusion of Cd ions, which should be enhanced at elevated temperatures due to a reduction in the solvated ion radius at higher temperatures. Subsequent desorption at room temperature and a re-expansion of the solvated ion radius would greatly retard the desorption of occluded Cd adions. Therefore, temperature and time appear to promote the sequestration of Cd at the kaolinite–water interface. This effect is likely


229

exacerbated in naturally occurring kaolinites, which exhibit lower crystalline states and greater amounts of external and internal lattice defects. The significant decrease in Pb desorption from kaolinite is due to the formation of polynuclear Pb–hydroxide complexes (with minor contribution from Cl), which with increasing aging period stabilize at the kaolinite surface. The increased stability was inferred from increased structural order observed by decreases of the Debye–Waller parameter. Polynuclear Pb–hydroxide complexes were also observed in 1 h–201C sample, which also showed to be bonded to the kaolinite surface. Increased stability, however, coincided with greater Pb–Pb edge sharing, suggesting that the polynuclear complexes increased in size with age and temperature. The desorption behavior shows that time was the more effective stabilizing factor than temperature, suggesting that polynuclear Pb complexes may undergo Ostwald ripening.

ACKNOWLEDGMENTS We gratefully acknowledge the financial support of the Australian Synchrotron Research Program. We thank Garry Foran at the Australian National Beamline Facility (20B) at the Photon Factory of the National Laboratory for High Energy Physics (KEK), Tsukuba, Japan, for his assistance and facility support for the Pb EXAFS experiments. We also thank Steve Heald for his help in conducting the Cd EXAFS experiments at the PNC-CAT beamline 20BM at the Advanced Photon Source, Argonne, IL, USA. PNC/ XOR facilities at the Advanced Photon Source, and research at these facilities, are supported by the US Department of Energy – Basic Energy Sciences, a major facilities access grant from NSERC, the University of Washington, Simon Fraser University, the Pacific Northwest National Laboratory and the Advanced Photon Source. Use of the Advanced Photon Source is also supported by the US Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. W-31-109-Eng-38. We would like to thank Navdeep Kaur for her assistance in EXAFS sample preparation. We also acknowledge the critical suggestions by Douglas Kent and three anonymous reviewers to improve the manuscript.

REFERENCES Ainsworth, C. C., Pilon, J. L., Gassman, P. L., & Van der Sluys, W. G. (1994). Cobalt, cadmium and lead sorption to hydrous iron oxide: Residence time effect. Soil Sci. Soc. Am. J., 58, 1615–1623.

230


Angove, M. J., Johnson, B. B., & Wells, J. D. (1997). Adsorption of cadmium(II) on kaolinite. Colloid Surf. A, 126, 137–147. Angove, M. J., Johnson, B. B., & Wells, J. D. (1998). The influence of temperature on the adsorption of cadmium(II) and cobalt(II) on kaolinite. J. Colloid Interface Sci., 204, 93–103. Backes, C. A., McLaren, R. G., Rate, A. W., & Swift, R. S. (1995). Kinetics of cadmium and cobalt desorption from iron and manganese oxides. Soil Sci. Soc. Am. J., 59, 778–785. Bargar, J. R., Brown, G. E. Jr., & Parks, G. A. (1997a). Surface complexation of Pb(II) at oxide–water interfaces: I. EXAFS and bond-valence determination of mononuclear and polynuclear Pb(II) sorption products on aluminium oxides. Geochim. Cosmochim. Acta, 61, 2617–2637. Bargar, J. R., Brown, G. E. Jr., & Parks, G. A. (1997b). Surface complexation of Pb(II) at oxide–water interfaces: II. XAFS and bond-valence determination of mononuclear and polynuclear Pb(II) sorption products and surface functional groups on iron oxides. Geochim. Cosmochim. Acta, 61, 2639–2652. Barrow, N. J., Gerth, J., & Bruemmer, G. W. (1989). Reaction kinetics of the adsorption and desorption of nickel, zinc and cadmium by goethite. II. Modeling the extent and rate of reaction. J. Soil Sci., 40, 437–450. Bruemmer, G. W., Gerth, J., & Tiller, K. G. (1988). Reaction kinetics of the adsorption and desorption of nickel, zinc, and cadmium by goethite. I. Adsorption and diffusion of metals. J. Soil Sci., 39, 37–52. Bunker, G. (2003). Overview of the Standard XAFS Data Analysis Procedure, http: //gbxafs.iit.edu/training/tutorials.html, accessed August 12, 2005. Chisholm-Brause, C. J., Roe, A. L., Hayes, K. F., Brown, G. E. Jr., Parks, J. A., & Leckie, J. O. (1990). Spectroscopic investigation of Pb(II) complexes at the g-Al2O3/water interface. Geochim. Cosmochim. Acta, 54, 1897–1909. Comans, R. N. J. (1987). Adsorption, desorption and isotopic exchange of cadmium on illite: Evidence for complete reversibility. Water Res., 21, 1573–1576. Davies, B. E. (1995). Lead. In: B. J. Alloway (Ed). Heavy Metals in Soils. Blackie Academic and Professional, Glasgow, UK, pp. 206–230. de Wolff, P. M. (1966). Crystal structure of g-Cd(OH)2. Acta Crystallogr., 21, 432–433. Eick, M. J., & Fendorf, S. E. (1998). Reaction sequence of nickel(II) with kaolinite: Mineral dissolution and surface complexation/precipitation. Soil Sci. Soc. Am. J., 62, 1257–1267. Eick, M. J., Peak, J. D., Brady, P. V., & Pesek, J. D. (1999). Kinetics of lead adsorption/desorption on goethite: Residence time effect. Soil Sci., 164, 28–39. Gao, Y., Kan, A. T., & Tomson, M. B. (2003). Critical evaluation of desorption phenomena of heavy metals from natural sediments. Environ. Sci. Technol., 37, 5566–5573. Garrett, R. F., Cookson, D. J., Foran, G. J., Sabine, T. M., & Wilkins, S. W. (1995). Powder diffraction using image plates at the australian national beamline facility at the photon factory. Rev. Sci. Instrum., 66, 1351–1353.


231

Glover, L. J. II, Eick, M. J., & Brady, P. V. (2002). Desorption kinetics of cadmium and lead from goethite: Influence of time and organic acids. Soil Sci. Soc. Am. J., 66, 797–804. Graf, D. L. (1961). Crystallographic tables for the rhombohedral carbonates. Am. Mineral., 46, 1283–1316. Gra¨fe, M., Singh, B., & Balasubramanian, M. (2007). Surface speciation of Cd(II) and Pb(II) on kaolinite by XAFS spectroscopy. J. Colloid Interface Sci., Manuscript Number JCIS-07-440, accepted for publication. Heald, S. M., Brewe, D. L., Stern, E. A., Kim, K. H., Brown, F. C., Jiang, D. T., Crozier, E. D., & Gordon, R. A. (1999). XAFS and micro-XAFS at the PNCCAT beamlines. J. Synchrotron Radiat., 6, 347–349. Helfferich, F. (1956). Kinetik des ionenaustansches. Angew. Chem. Int. Ed., 68, 693–698. Hill, R. J. (1985). Refinement of the structure of orthorhombic PbO (massicot) by rietveld analysis of neutron powder diffraction data. Acta Crystallogr., 41, 1281–1284. Ikhsan, J., Johnson, B. B., & Wells, J. D. (1999). A comparative study of the adsorption of transition metals on kaolinite. J. Colloid Interface Sci., 217, 403–410. Johnson, B. B. (1990). Effect of pH, temperature, and concentration on the adsorption of cadmium on goethite. Environ. Sci. Technol., 24, 112–118. Kabata-Pendias, A., & Pendias, H. (1992). Trace Elements in Soils and Plants. CRC Press, Boca Raton, FL. Katoh, M., Hori, Y., Kobayashi, Y., Sakanaka, S., Izawa, M., Haga, K., Honda, T., Mitsuhashi, T., Obina, T., Pak, C. O., Ohmi, K., Tsuchiya, K., Kanaya, N., Ueda, A., Tadano, M., Takiyama, Y., Asaoka, S., Nogami, T., Shioya, T., Mishina, A., Sato, Y., Tokumoto, S., Takahashi, T., Kasuga, T., Maezawa, H., Kobayashi, M., & Kobayakawa, H. (1998). Reconstruction for the brillianceupgrading project of the photon factory storage ring. J. Synchrotron Radiat., 5, 366–368. Lindsay, W. L. (1979). Chemical Equilibria in Soils. Wiley, NY. Martinetto, P., Anne, M., Dooryhee, E., Walter, P., & Tsoucaris, G. (2002). Synthetic hydrocerussite, 2PbCO3Pb(OH)2 by X-ray powder diffraction. Acta Crystallogr., 56, 82–84. McBride, M. B. (1994). Environmental Chemistry of Soils. Oxford University Press, New York, NY. McLaren, R. G., Backes, C. A., Rate, A. W., & Swift, R. S. (1998). Cadmium and cobalt desorption kinetics from soil clays: Effect of sorption period. Soil Sci. Soc. Am. J., 62, 332–337. McLaren, R. G., Lawson, D. M., & Swift, R. S. (1986). Sorption and desorption of cobalt by soils and soil components. J. Soil Sci., 37, 413–426. Moreno, A. M., Quintana, J. R., Pe´rez, L., & Parra, J. G. (2006). Factors influencing lead sorption–desorption at variable added metal concentrations in rhodoxeralfs. Chemosphere, 64, 758–763. Mustafa, G., Kookana, R. S., & Singh, B. (2006). Desorption of cadmium from goethite: Effects of pH, temperature and aging. Chemosphere, 64, 856–865.

232


Naidu, R., Kookana, R. S., Sumner, M. E., Harter, R. D., & Tiller, K. G. (1997). Cadmium sorption and transport in variable charge soils. A review. J. Environ. Qual., 26, 602–617. Parkman, R. H., Charnock, J. M., Bryan, N. D., Livens, F. R., & Vaughan, D. J. (1999). Reactions of copper and cadmium ions in aqueous solution with goethite, lepidocrocite, mackinawite, and pyrite. Am. Mineral., 84, 407–419. Puls, R. W., Powell, R. M., Clark, D., & Eldred, C. J. (1991). Effects of pH, solid/ solution ratio, ionic strength, and organic acids on lead and cadmium sorption on kaolinite. Water Air Soil Pollut., 57–58, 423–430. Randall, S. R., Sherman, D. M., & Ragnarsdottir, K. V. (2001). Sorption of As(V) on green rust (Fe-4(II)Fe-2(III)(OH)(12)SO4 3H(2)O) and lepidocrocite (gammaFeOOH): Surface complexes from EXAFS spectroscopy. Geochim. Cosmochim. Acta, 65, 1015–1023. Ressler, T. (1998). WinXAS: A program for X-ray absorption spectroscopy data analysis under MS-windows. J. Synchrotron Radiat., 5, 118–122. Sahai, N., Carroll, S. A., Roberts, S., & O’Day, P. A. (2000). X-ray absorption spectroscopy of strontium(II) coordination – II. Sorption and precipitation at kaolinite, amorphous silica, and goethite surfaces. J. Colloid Interface Sci., 222, 198–212. Sahl, K. (1974). Refining of crystalline structure of cerussite, PbCO3. Z. Kristallogr., 139, 215–222. Schindler, P. W., Leichti, P., & Westall, J. C. (1987). Adsorption of copper, cadmium and lead from aqueous solution to the kaolinite/water interface. Neth. J. Agric. Sci., 35, 219–230. Singh, B., & Gilkes, R. J. (1992). Properties of soil kaolinites from South-Western Australia. J. Soil Sci., 43, 645–667. Sokolova, E. V., & Egorovtismenko, Y. K. (1990). Realization of the spinel structural type in the structure of B-Cd2(OH)3Cl. Kristallographia, 35, 995–997. Spadini, L., Manceau, A., Schindler, P. W., & Charlet, L. (1994). Structure and stability of Cd2+ surface complexes on ferric oxides. 1. Results from EXAFS spectroscopy. J. Colloid Interface Sci., 168, 73–86. Srivastava, P. (2005). Competitive adsorption–desorption of Cd, Cu, Pb and Zn on kaolinite. PhD Thesis, University of Sydney, Australia. Srivastava, P., Singh, B., & Angove, M. (2005). Competitive adsorption behavior of heavy metals on kaolinite. J. Colloid Interface Sci., 290, 28–38. Stern, E. A. (1993). Number of relevant independent points in X-ray absorption fine structure spectra. Phys. Rev. B, 48, 9825–9827. Strawn, D. G., Scheidegger, A. M., & Sparks, D. L. (1998). Kinetics and mechanisms of Pb(II) sorption and desorption at the aluminium oxide–water interface. Environ. Sci. Technol., 32, 2596–2601. Thompson, H. A., Parks, G. A., & Brown, G. E. Jr. (2000). Formation and release of cobalt (II) sorption and precipitation products in aging kaolinite–water slurries. J. Colloid Interface Sci., 222, 241–253.


233

Venetopoulos, C. C., & Rentzeperis, P. J. (1975). Crystal-structure of laurionite, Pb(OH)Cl. Z. Kristallogr., 141, 246–259. Zabinsky, F., Rehr, J. J., Ankudinov, A., Albers, R. C., & Eller, M. J. (1995). Multiplescattering calculations of X-ray absorption spectra. Phys. Rev. B, 52, 2995–3009. Zelazny, L. W., He, L., & Vanwormhoudt, A. (1996). Charge analysis of soils and anion exchange. In: D. L. Sparks (Ed). Methods of Soil Analysis. Part III. SSSA Book Series 5. SSSA, Madison, WI, pp. 1231–1253. Zhang, J. (1999). Room-temperature compressibilities of MnO and CdO: Further examination of the role of cation type in bulk modulus systematics. Phys. Chem. Miner., 26, 644–648.


Chapter 9

Mechanism of Molybdenum Adsorption on Soils and Soil Minerals Evaluated Using Vibrational Spectroscopy and Surface Complexation Modeling Sabine Goldberg1,, Cliff T. Johnston2, Donald L. Suarez1 and Scott M. Lesch1 1

USDA-ARS, U.S. Salinity Laboratory, 450 W. Big Springs Road, Riverside, CA 92507, USA 2 Crop, Soils, and Environmental Sciences, 915 W. State Street, Purdue University, West Lafayette, IN 47907, USA

ABSTRACT Molybdenum adsorption on amorphous aluminum and iron oxides was investigated as a function of solution pH and solution ionic strength. In this study in situ Raman and Fourier transform infrared (FTIR) spectroscopic methods were combined with sorption techniques, electrophoretic mobility measurements, and surface complexation modeling to study the interaction of Mo with amorphous oxide surfaces. The speciation of Mo in aqueous solution was examined using Raman and attenuated total reflectance (ATR)-FTIR methods as a function of solution pH. Good agreement was found between the vibrational spectra of Mo in aqueous solution and those of Mo sorbed to amorphous Al oxide. The mechanisms of Mo sorption to these surfaces based on the spectroscopic, sorption, and electrophoretic mobility measurements are as follows: Mo forms predominantly innersphere surface complexes at low pH and predominantly outer-sphere surface complexes at high pH. These surface configurations were used to constrain the input parameters of the triple layer surface complexation model to describe Mo adsorption on soils. After applying the triple layer model to Mo adsorption on 36 soils, a general regression model was developed for predicting Mo surface complexation constants from 5 independently measured soil chemical characteristics: cation exchange capacity, organic carbon content, inorganic carbon Corresponding author. Tel.: 951-369-4820; Fax: 951-342-4962;

E-mail: [email protected] (S. Goldberg).

236

S. Goldberg et al.

content, aluminum oxide content, and iron oxide content. The triple layer model was well able to predict Mo adsorption on all the soils at all pH values. The surface speciation predicted by the model for soil was in agreement with that predicted from spectroscopy for Mo adsorption on amorphous Al oxide.

9.1. Introduction Molybdenum is an essential trace element for both plant and animal nutrition. Molybdenum deficiencies have been reported for many agronomic crops, especially on alkaline soils (Murphy and Walsh, 1972). Molybdenum occurs in anionic form in soil solution, is readily taken up by forage plants, and can accumulate to levels detrimental to grazing ruminant animals (Reisenauer et al., 1973). Molybdenum exerts its toxic effect on ruminants by inducing a copper deficiency that is especially pronounced in the presence of sulfur; this adverse effect can be mitigated by Cu supplementation (O’Connor et al., 2001). At present, our understanding of the biogeochemical factors that influence the bioavailability of Mo is limited. Therefore, careful quantification of soil solution Mo concentration and characterization of Mo adsorption reactions by soil minerals and soils is needed. Availability of Mo to plants is affected by a variety of soil factors including pH, temperature, texture, clay mineralogy, oxide, and organic matter content (Reisenauer et al., 1973). The active sites that bind Mo are pH dependent sites found on aluminum and iron oxides, clay minerals, and organic matter (Goldberg et al., 1996). Although Mo sorption has been investigated on a wide range of crystalline aluminum (e.g., Ferreiro et al., 1985; Spanos et al., 1990a,b; Vordonis et al., 1990; Spanos and Lycourghiotis, 1995; Goldberg et al., 1996, 1998; El Shafei et al., 2000; Vissenberg et al., 2000) and iron (Kyriacou, 1967; Reyes and Jurinak, 1967; McKenzie, 1983; Ferreiro et al., 1985; Zhang and Sparks, 1989; Goldberg et al., 1996, 1998; Goldberg and Forster, 1998; Lang et al., 2000; Lang and Kaupenjohann, 2003; Xu et al., 2006a,b) oxide minerals, only a few studies have dealt with Mo sorption on amorphous oxides (Jones, 1957; Reisenauer et al., 1962; Bibak and Borggaard, 1994; Goldberg et al., 1996, 1998; Gustafsson, 2003). The features of Mo sorption are consistent with a negatively charged oxyanion in solution where the protonation status of the 6 MoO2 4 and Mo7O24 species changes, along with the pH dependent behavior of variable charge sites. Maximum Mo sorption on soils and soil minerals is generally found in the pH range 4–5, and sorption decreases with increasing pH above pH 5 (Gonzalez et al., 1974; Mikkonen and Tummavuori, 1993; Goldberg et al., 1996, 1998).

Mechanism of Molybdenum Adsorption on Soils and Soil Minerals

237

Descriptions of Mo sorption behavior in natural systems require knowledge of the mode of bonding of the Mo anions on mineral surfaces. Insight into anion adsorption mechanisms can be provided by both macroscopic and microscopic experimental methods. Electrophoretic mobility, for example, measures the movement of charged particles in response to an applied electric field. When anions such as molybdate and phosphate form innersphere surface complexes through ligand exchange, the point of zero charge is lowered. The inner-sphere adsorption mechanism was suggested for molybdenum adsorption on crystalline Al and Fe oxides (Ferreiro et al., 1985). Molybdenum adsorption lowered the point of zero charge of crystalline Al and Fe oxides indicating specific adsorption (Goldberg et al., 1996). These shifts in PZC can be used as evidence of strong specific ion adsorption and inner-sphere surface complexation. Evaluation of the effect of changes in solution ionic strength on the extent of adsorption is another macroscopic method of inferring adsorption mechanisms (Hayes et al., 1988; McBride, 1997). Ions adsorbing as innersphere surface complexes show little ionic strength dependence in adsorption behavior. Such was the case for Mo adsorption on goethite (Zhang and Sparks, 1989; Goldberg et al., 1998), gibbsite, and amorphous Fe and Al oxides (Goldberg et al., 1998) suggesting formation of inner-sphere surface complexes. Investigations of anion desorption reactions have also been used to infer adsorption mechanisms. It was suggested that Mo desorption from goethites was limited due to Mo diffusion into micropores following adsorption (Lang and Kaupenjohann, 2003). Molybdenum desorption from soils was biphasic: the fast reaction was attributed to desorption from Fe oxides and the slow reaction to diffusion out of crystalline oxides (Lang and Kaupenjohann, 1999). Molybdenum adsorption on soils and soil minerals has been described using various surface complexation modeling approaches: constant capacitance model (Motta and Miranda, 1989; Goldberg et al., 1996, 1998, 2002; Saripalli et al., 2002), diffuse layer model (Dzombak and Morel, 1990; Stollenwerk, 1995; Gustafsson, 2003), triple layer model (Zhang and Sparks, 1989; Goldberg et al., 1998; Wu et al., 2001), and CD-MUSIC model (Bourikas et al., 2001; Gustafsson, 2003; Xu et al., 2006b). The advantage of surface complexation models over empirical adsorption models is that they define specific surface species, chemical reactions, mass balances, and charge balances and contain molecular features that can be given thermodynamic significance (Sposito, 1983). In a prior study (Goldberg et al., 2002), a general regression model was developed to obtain soil Mo surface complexation constants for use in the

238

S. Goldberg et al.

constant capacitance model to predict Mo adsorption. The constant capacitance model parameters were obtained from easily measured soil chemical properties: cation exchange capacity (CEC), organic carbon content (OC), inorganic carbon content (IOC), and free Fe oxide content. The prediction equations, when utilized in the constant capacitance model, provided good prediction of Mo adsorption behavior on 36 soils primarily from California. This comparison of prediction and observation resulted in a completely independent evaluation of the ability of the constant capacitance model to describe Mo adsorption. This approach avoids the necessity of performing time consuming detailed adsorption studies for each specific soil. Predictions of adsorption were obtained using both monodentate and bidentate surface configurations for molybdate and were of similar quality. This is fortunate since no spectroscopic observations have been carried out to date for molybdate adsorption on oxide minerals. Constant capacitance model results underpredicted the experimental Mo adsorption above pH 7.5, thus limiting the use of the model in arid zone soils. It is possible that molybdate forms both inner- and outer-sphere surface complexes on aluminum and iron oxides, as has been observed spectroscopically for the anions, sulfate and selenate (Peak et al., 1999; Wijnja and Schulthess, 2000). Monodentate innersphere surface complexes were observed to dominate at pH values below 6 and outer-sphere surface complexes dominated above pH 6–7 for sulfate and selenate adsorption on goethite and aluminum oxide. Simultaneous adsorption as both inner- and outer-sphere surface complexes cannot be described using the constant capacitance model. A three plane model, such as the triple layer model or the CD-MUSIC model, is required. Our study focuses on molybdate adsorption on amorphous Fe and Al oxides that, unlike crystalline oxides, such as goethite, have not been thoroughly characterized yet, especially with spectroscopic techniques. These materials constitute a major sink for ion adsorption in soils. A combination of macroscopic and microscopic experiments is appropriate to delineate the adsorption mechanism of molybdate. Our study contains the following objectives: (1) to determine Mo adsorption behavior on amorphous Al and Fe oxide as a function of solution pH and ionic strength and PZCs of amorphous Al and Fe oxide with and without molybdate; (2) to evaluate the ability of the triple layer model to describe molybdate adsorption on these oxides and on the soils studied by Goldberg et al. (2002); (3) to develop a set of regression model equations to predict triple layer model parameters and to subsequently predict Mo adsorption on the soils of Goldberg et al. (2002). Our goal was to establish a clear link between the known phase diagram of Mo and the aqueous speciation using Raman and attenuated total reflectance-Fourier transform infrared (ATR-FTIR) methods. In this context,


239

we were constrained to work at the higher concentrations to make a clear, unambiguous link between the speciation and the spectroscopy.

9.2. Experimental 9.2.1. Macroscopic Experiments Molybdenum adsorption behavior as a function of solution pH and ionic strength was studied on amorphous Al and Fe oxides synthesized using the method of Sims and Bingham (1968). Iron oxide was synthesized by neutralizing 100 mL of 1.5 M FeCl3 with 225 mL of 2.0 M NaOH. The sample preparation scheme for amorphous Fe oxide used in this paper was that used in Goldberg et al. (1996) and Goldberg (2002). In the Al oxide synthesis, 0.41 M AlCl3 was neutralized with an equal part of 1.1 M NaOH. The sample preparation scheme for amorphous Al oxide used in this paper and Goldberg (2002) was that of Sims and Bingham (1968), while the sample preparation scheme used in Goldberg et al. (1996) was that of McLaughlin et al. (1981). We changed procedures because the Sims and Bingham (1968) method gave much greater yield of solid. X-ray diffraction (XRD) analyses verified that the oxides were amorphous and contained no crystalline impurities detected by XRD. Surface areas (SAs) were determined from a single-point BET N2 adsorption isotherm. The SA was 65.6 m2 g1 for the Al oxide and 169 m2 g1 for the Fe oxide.

9.2.1.1. Electrophoretic Mobility Points of zero charge for the oxides were determined by microelectrophoresis. Electrophoretic mobility measurements of suspensions containing 0.02% oxide in 0.01 M NaCl were determined at various pH values in the presence of 0, 0.01, or 0.1 mM Mo from Na2MoO4 2H2O. Points of zero charge were obtained by linearly interpolating the data to zero mobility.

9.2.1.2. Adsorption Envelopes Molybdenum adsorption envelopes (amount of Mo adsorbed as a function of solution pH per fixed total Mo concentration) were determined in batch systems. Samples of 0.1 g of oxide were added to 50 mL polypropylene

240

S. Goldberg et al.

centrifuge tubes and equilibrated with 25 mL of a 0.01, 0.1, or 1.0 M NaCl solution by shaking for 20 h on a reciprocating shaker at 23.170.11C. Particle concentration of oxide was 4.0 g L1. The equilibrating solutions contained 1.0 mM Mo from Na2MoO4 2H2O and were adjusted to the desired pH values using 1.0 M HCl or NaOH. The samples were centrifuged and the decantates analyzed for pH, filtered through a 0.45 mm membrane filter, and analyzed for Mo concentration using inductively coupled plasma (ICP) emission spectrometry. Molybdenum adsorption envelopes on 36 soil samples were determined previously (Goldberg et al., 2002). Physical and chemical characteristics and experimental methods are provided in Goldberg et al. (2002). The soils had the following range of chemical characteristics: pH, 4.1–10.2; CEC, 27–467 mmolc kg1; SA, 21.2–286 m2 g1; IOC, 0.001–1.9%; OC, 0.11–3.2%; free Fe oxide (Fe), 0.17–4.9%; and free Al oxide (Al), 0.018–0.37%. Free oxides were determined using the method of Coffin (1963). Free oxides are defined as the oxides and other forms of Fe and Al found in soils but not as a part of the crystal lattice of other minerals present (Coffin, 1963). Molybdenum adsorption envelopes were determined on 5 g of soil equilibrated with 25 mL of a solution containing 0.292 mM Mo from Na2MoO4 2H2O for 20 h. All other steps in the adsorption procedure were as listed above for the amorphous oxides.

9.2.2. Vibrational Spectroscopy 9.2.2.1. ATR-FTIR Spectroscopy FTIR spectra were obtained with a Perkin-Elmer Model 2000/GX spectrometer and a horizontal ATR attachment (Pike Technologys) fitted with a ZnSe internal reflection element with a 451 angle corresponding to nine reflections in contact with the sample. The measured pathlength was 15 mm at 1,630 cm1 based on the molar absorptivity of water. The ZnSe internal reflection element did not permit observation of infrared (IR) bands below 700 cm1. Spectra were obtained at a resolution of 4 cm1 with each spectrum corresponding to the co-addition of 64 scans using a medium-band liquid N2 cooled MCT detector. In situ ATR-FTIR spectra were obtained by placing 1 mL of an aqueous suspension containing 2 mg of amorphous Al oxide into the ATR cell. The suspension was allowed to dry at room temperature and this resulted in a uniform dry deposit of the oxide on the ZnSe crystal. An aqueous solution of 0.01 M NaCl at pH 8 was then slowly passed over the oxide deposit using the ATR flow through system and a MasterFlex


241

computerized pump (Cole-Parmers) and FTIR spectra were recorded. Flow rate of the pump was maintained at 16 mL min1. Very little oxide was lost during the washing step with NaCl that took a few hours for equilibration. Then Mo was introduced to the flow at a concentration of 0.01 M in a 0.01 M NaCl background electrolyte at pH 8. ATR-FTIR spectra were recorded. The pH of the Mo solution was changed from pH 8 to 3.5 at 0.5 decrements using HCl after recording the spectra at each pH value. Spectra were analyzed using the GRAMS/32 AIs (Version 6.00) program from Galactic software.

9.2.2.2. Raman Spectroscopy Raman spectra were obtained on an Acton Research Corporation SpectroPro500s spectrograph. A Spectra-Physics Model 127 helium–neon laser with 632.8 nm wavelength and power output of 35 mW measured at the laser head was used as the excitation source. Raman scattered radiation was collected in a 901 scattering configuration. A 1/4 wave plate and polarization analyzer were used to measure Raman depolarization ratios. A calcite-wedge polarization scrambler was placed after the analyzer to minimize unwanted polarization effects in the spectrograph. The polarization discrimination of the instrument was checked by measuring the depolarization ratio for the 459 cm1 band of CCl4. The experimental value was 0.023 compared to a theoretical value of 0.01. The entrance slits to the spectrograph were set to 100 mm that corresponds to a resolution of 5 cm1. The spectrograph used a holographic grating with 600 grooves per mm with a blaze wavelength of 532 nm. The detector was a Princeton Instruments liquid N2 cooled CCD detector with an active array of 1,100 (h) 330 (v) pixels. The spectrograph was calibrated daily using a Ne–Ar calibration lamp based upon known spectral lines. Spectra were typically collected using 300 s of acquisition on the CCD array. Spectra were analyzed using the GRAMS/32 AIs (Version 6.00) program from Galactic software. Raman spectra of MoO4 were collected from 0.1 M solutions in 1 cm quartz cuvettes using a 901 backscattering geometry. 9.2.3. Surface Complexation Modeling A detailed discussion of the theory and assumptions of the triple layer surface complexation model is provided elsewhere (Goldberg, 1992). In the present application of the model to Mo adsorption, the following surface

242

S. Goldberg et al.

complexation constants were considered: þ SOHðsÞ þ Hþ ðaqÞ $SOH2ðsÞ

(9.1)

þ SOHðsÞ $SO ðsÞ þ HðaqÞ

(9.2)

þ þ SOHðsÞ þ Naþ ðaqÞ $SO NaðsÞ þ HðaqÞ

(9.3)

þ SOHðsÞ þ Hþ ðaqÞ þ ClðaqÞ $SOH2 ClðsÞ

(9.4)

SOHðsÞ þ H2 MoO4ðaqÞ $SHMoO4ðsÞ þ H2 O

(9.5)

SOHðsÞ þ H2 MoO4ðaqÞ $SOHþ 2 HMoO4ðsÞ

(9.6)

or 2 þ SOHðsÞ þ H2 MoO4ðaqÞ $SOHþ 2 MoO4ðsÞ þ HðaqÞ

(9.7)

where SOH(s) represents reactive surface hydroxyl groups on oxides and aluminol groups on clay minerals in the soils. Even though the surface complexation reactions are written starting with the completely undissociated acids, the model application contains the aqueous speciation reactions for Mo. Since the quality of model fit to experimental Mo adsorption data on soils was similar for monodentate and bidentate inner-sphere surface species (Goldberg et al., 2002), a monodentate Mo surface configuration was used. A predominantly monodentate surface configuration was found for sulfate adsorption by hematite (Lefevre and Fedoroff, 2006). An outer-sphere complex, Eq. (9.6) or Eq. (9.7), was added in an effort to improve the model fit at pH values above 7 and is consistent with spectroscopic observations for sulfate and selenate adsorption by Al and Fe oxides (Peak et al., 1999; Wijnja and Schulthess, 2000). Equilibrium constants for the surface complexation reactions are: ½SOHþ F co 2 Kþ ðintÞ ¼ exp (9.8) RT ½SOH½Hþ ½SO ½Hþ F co K ðintÞ ¼ exp RT ½SOH F ðcb co Þ ½SO Naþ ½Hþ exp KNaþ ðintÞ ¼ RT ½SOH½Naþ

(9.9)

(9.10)


F ðco cb Þ ½SOHþ Cl 2 KCl ðintÞ ¼ exp RT ½SOH½Hþ ½Cl

243

(9.11)

½SHMoO4 ½SOH½H2 MoO4

(9.12)

F ðco cb Þ ½SOHþ 2 HMoO4 exp RT ½SOH½H2 MoO4

(9.13)

2 F ðco 2cb Þ ½SOHþ 2 MoO4 exp ¼ ½SOH½H2 MoO4 RT

(9.14)

KisMo ðintÞ ¼ os K1Mo ðintÞ ¼

or os ðintÞ K2Mo

where F is the Faraday constant (C mol1 c ), c the surface potential (V), R the molar gas constant (J mol1 K1), T the absolute temperature (K), and square brackets indicate concentrations (mol L1). The exponential terms can be considered as solid-phase activity coefficients correcting for charge on the surface complexes. Mass balance of the surface functional group is: þ þ ½SOHT ¼ ½SOH þ ½SOHþ 2 þ ½SO þ ½SO Na þ ½SOH2 Cl þ 2 þ ½SHMoO4 þ ½SOHþ 2 HMoO4 or þ ½SOH2 MoO4

ð9:15Þ

and the charge balances are: so þ sb þ sd ¼ 0

(9.16)

þ þ so ¼ ½SOHþ 2 þ ½SOH2 Cl ½SO ½SO Na þ 2 þ ½SOHþ 2 HMoO4 or þ ½SOH2 MoO4

ð9:17Þ

þ sb ¼ ½SO Naþ ½SOHþ 2 Cl ½SOH2 HMoO4 or 2 2 ½SOHþ 2 MoO4

ð9:18Þ 1

where si has units of (molc L ). The computer code FITEQL 3.2 (Herbelin and Westall, 1996) was used to fit the Mo surface complexation constants to the experimental Mo adsorption data. The FITEQL program uses a nonlinear least squares optimization routine to fit equilibrium constants to experimental data and contains the triple layer model of adsorption. It can also be used as a chemical speciation program to evaluate predictions of Mo adsorption, such as those obtained using the regression model of Goldberg et al. (2002) to predict values of the Mo surface complexation constant from soil chemical properties.

244

S. Goldberg et al.

In our application of the triple layer model, the surface site density was fixed at a value of 2.31 sites nm2, as recommended by Davis and Kent (1990) for natural materials. Numerical values for the protonation constant, the dissociation constant, and the surface complexation constants for the background electrolyte were obtained from the literature. For the amorphous Al oxide and the soils we used logK+(int) ¼ 5.0, logK(int) ¼ 11.2, logKNa+(int) ¼ 8.6, and logKCl(int) ¼ 7.5 obtained by Sprycha (1989a,b) for g-Al2O3. For the amorphous Fe oxide we used logK+(int) ¼ 4.3, logK(int) ¼ 9.8, logKNa+(int) ¼ 9.3, and logKCl(int) ¼ 5.4 obtained by Zhang and Sparks (1990) for goethite. We used parameter values for crystalline oxides since values for amorphous oxides were not available. Molybdate surface complexation constants were fit simultaneously to the oxide adsorption data at three different ionic strengths using the innersphere, Eq. (9.5), and an outer-sphere, Eq. (9.6) or Eq. (9.7), adsorption mechanism. The capacitance values were fixed at C1 ¼ 1.2 F m2 and C2 ¼ 0.2 F m2 considered optimal for goethite by Zhang and Sparks (1990). When developing data bases of surface complexation constants, it is critically important to use consistent values of surface site density, capacitances, protonation-dissociation constants, and background electrolyte surface complexation constants. Constant values of these parameters are necessary to allow application of predictive equations to new soils.

9.3. Results and Discussion 9.3.1. Macroscopic Experiments 9.3.1.1. Electrophoretic Mobility Points of zero charge occurred at pH 7.8 for amorphous Fe oxide and at pH 9.4 for amorphous Al oxide. These PZCs are in excellent agreement with numerous literature values. Figure 9.1 presents electrophoretic mobility versus pH obtained upon adsorption of Mo onto amorphous Fe oxide (Fig. 9.1a) and amorphous Al oxide (Fig. 9.1b). The PZCs were shifted to lower pH value with increasing Mo concentration, characteristic of innersphere adsorption. This is clearly seen for the amorphous Fe oxide in the presence of Mo (Fig. 9.1a). The PZC of amorphous Al oxide decreased only slightly in the presence of increasing Mo concentrations (Fig. 9.1b) indicating the predominance of either an outer-sphere Mo surface complex, an innersphere Mo surface complex that did not change the surface charge, or a


Electrophoretic Mobility

3

3 (b)

(a)

2

2

1

1

0

0

-1

-1

245

0 Mo 0.01 mM Mo 0.1 mM Mo

pH

-2

-2 2

4

6

8

10

2

4

6

8

10

12

Figure 9.1: Electrophoretic Mobility of Amorphous Oxides as a Function of pH and Total Mo Concentration in 0.01 M NaCl Solution: (a) Fe Oxide; (b) Al Oxide. Circles Represent the Zero Mo Treatment. combination of both of these surface complexes. It is possible that the strange local maxima and minima in the Al oxide system are due to the predominance of inner-sphere Mo surface complexes at low pH and the predominance of outer-sphere Mo surface complexes at higher pH.

9.3.1.2. Adsorption Envelopes The effect of ionic strength on Mo adsorption on amorphous Al and Fe oxides is indicated in Fig. 9.2. Solution ionic strength varied by two orders of magnitude, from 0.01 to 1.0 M NaCl. Molybdate adsorption on amorphous oxides decreased with increasing solution pH. On amorphous Fe oxide, Mo adsorption exhibited no ionic strength dependence with increasing solution ionic strength (Fig. 9.2a). This behavior is indicative of an inner-sphere adsorption mechanism, in agreement with the mechanism inferred from PZC shift results for this material. On amorphous Al oxide, Mo adsorption exhibited decreasing adsorption with increasing ionic strength (Fig. 9.2b), indicative of an outer-sphere adsorption mechanism. This result is in agreement with the PZC shift results, which suggest that the majority of Mo adsorption may be outer-sphere. Our results indicate that Mo is more tightly

246

S. Goldberg et al.

5 0.01 M 0.1 M Mo adsorbed (mmol/g)

4

1.0 M

3

2

1 (a ) 0

1

2

3

4

5

6

7 pH

8

9

10

11

12

5 0.01 M 0.1 M

Mo adsorbed (mmol/g)

4

1.0 M

3

2

1 (b ) 0 3

4

5

6

7

8 pH

9

10

11

12

Figure 9.2: Fit of the Triple Layer Model to Mo Adsorption on Amorphous Oxides as a Function of Solution pH and Ionic Strength: (a) Fe Oxide; (b) Al Oxide. Particle Concentration ¼ 4.0 g L1. Solid Symbols Represent Experimental Data Points. Open Symbols Represent Model Fits using an Inner-Sphere and an Outer-Sphere Mo Surface Complex. Log KisMo ðintÞ ¼ 12:43; is 1os log K2os Mo ðintÞ ¼ 3:76 for Fe Oxide. Log KMo ðintÞ ¼ 10:55; log KMo ðintÞ ¼ 12:58 for Al Oxide.


247

bound to Fe oxide than to Al oxide, similar to our previous observations for arsenite adsorption on amorphous oxides (Goldberg and Johnston, 2001).

9.3.2. Vibrational Spectroscopy The ATR-FTIR spectra of a 2 mg deposit of amorphous Al oxide deposit on a ZnSe internal reflection element in the presence of 10 mM NaCl at pH 8 are shown in Fig. 9.3 (represented by a dashed line). The predominant band is found at 973 cm1 and weaker bands at 1,419 and 1,497 cm1. The 973 cm1 band corresponds to the Al–O–H bending vibration (Farmer, 1974; Wang and Johnston, 2000). The bands at 1,419 and 1,497 cm1 correspond to the n3 asymmetric stretch of sorbed carbonate (Su and Suarez, 1997). Since the PZC of this amorphous Al oxide is 9.4, it will have a net positive surface charge and attract aqueous carbonate species in solution at all pH values

Figure 9.3: ATR-FTIR Spectra of a 2 mg Deposit of Amorphous Al Oxide in a 10 mM NaCl Solution on a ZnSe Internal Reflection Element: (A) 10 mM NaCl Solution at pH 8; (B) 10 mM NaCl Solution and 10 mM Mo Solution at pH 8; (C) 10 mM NaCl and 10 mM Mo Solution at pH 6; (D) 10 mM NaCl and 10 mM Mo Solution at pH 4.

248

S. Goldberg et al.

used in this study. Molybdate was introduced into the ATR-FTIR cell as a 10 mM NaHMoO4 solution with a background electrolyte concentration of 10 mM NaCl adjusted to pH 8 using a peristaltic pump at a flow rate of 16 mLmin1. Based on prior work, aqueous molybdate species are known to interact with amorphous Al oxide to a greater extent at low pH (Goldberg et al., 1996). Thus, the experiment was started at pH 8 and the pH was lowered in 0.5 pH unit decrements until a pH value of 3.5 was reached. The ATR-FTIR spectrum of the Mo–Al oxide complex at pH 8 is shown as Spectrum B in Fig. 9.3. The presence of sorbed MoO4 is revealed by the shoulder at 838 cm1. The carbonate bands at 1,419 and 1,497 cm1 are more distinct at this high pH value indicating the presence of sorbed carbonate. The spectra of sorbed Mo did not change significantly until the pH of the aqueous solution was decreased too6 where ‘‘new’’ bands at 903 and 940 cm1 appeared (Spectra C and D in Fig. 9.3 were obtained at pH values of 6 and 4, respectively). To observe the spectral changes induced by Mo sorption, difference spectra were obtained by subtracting the ATR-FTIR spectrum of the amorphous Al oxide deposit from the spectra of the Mo–Al oxide complexes (Fig. 9.4). In this case, Spectrum A (Fig. 9.3) was subtracted from the Mo–Al oxide complexes. For example, the difference spectrum at pH 8 was obtained by subtracting Spectrum A from Spectrum B (Fig. 9.3). At pH 8, the difference spectrum clearly shows the presence of the sorbed Mo species by the bands at 838 and 1,020 cm1. Upon lowering the pH, two new bands appear at 903 and 940 cm1, with a concomitant decrease in the intensities of the bands at 838 and 1,020 cm1. The spectrum of the amorphous Al oxide did not change shape upon lowering the pH. The ATR-FTIR spectrum of the amorphous Al oxide deposited on the ATR cell at pH 8 (prior to the addition of Mo) was used as the reference file for all spectral subtractions. The observed Mo bands are not due to changes in bandshape of the Al oxide itself. Confirmation of this is found in the agreement of the Mo bands sorbed to the Al oxide surface with the spectra of Mo in aqueous solution. To better understand the solution speciation of Mo, ATR-FTIR and Raman spectra of aqueous Mo species were obtained as a function of pH and ionic strength and are shown in Fig. 9.5. The ATR-FTIR spectra of a 100 mM aqueous Mo solution in 10 mM NaCl are shown in Fig. 9.5a. The solution ATR-FTIR spectra show similar trends to the spectra of sorbed Mo: a strong feature at 837 cm1 along with a weak, broad band at 1,015 cm1 at pH 8–9 that ‘‘transitions’’ into bands at 904 and 941 cm1 at pH valueso6 (Weinstock et al., 1973; Jeziorowski and Knozinger, 1979; Tossell, 2005). Raman spectra provide complimentary speciation information because the selection rules that govern Raman- and IR-allowed


249

Figure 9.4: ATR-FTIR Difference Spectra of 10 mM Mo in 10 mM NaCl Sorbed to Al Oxide (the Difference Spectra shown here were Obtained from the Spectra shown in Fig. 9.3). Difference Spectra were Obtained by Subtracting the ATR-FTIR Spectrum of the Al Oxide Deposit in 10 mM NaCl at pH 8 from Each of the Spectra shown in Fig. 9.3. vibrational transitions are different (Herzberg, 1945; Long, 1977; Johnston and Aochi, 1996; Johnston and Wang, 2002). In the Raman spectra of the aqueous Mo species, the high pH (pH 8–9) spectra are dominated by a medium intensity band at 837 cm1 and a sharp, well-resolved band at 897 cm1 (Fig. 9.5b). The Raman spectra are strongly influenced by pH with the 837 and 897 cm1 bands diminishing in intensity with ‘‘new’’ bands appearing at 946 and 961 cm1. The aqueous speciation of Mo is shown in the inset of Fig. 9.5. At a concentration of 100 mM and at pH value> 5.8, the dominant species in 6 solution is MoO2 4 . Upon lowering the pH below 5.8, the polymer Mo7O24 becomes the predominant species in the narrow pH range of 5.8 to 5.5. The singly-protonated HMo7O5 24 species dominates from pH 5.5 to 4.2, with the doubly-protonated H2Mo7O4 24 species present at pH valueso4.2. Figure 9.6 shows the comparison of the ATR-FTIR spectrum of Mo sorbed to Al oxide at pH 8 (Spectrum C) to the Raman (Spectrum A) and ATR-FTIR (Spectrum B) spectra of a 100 mM Mo solution at pH 8. There is

250

S. Goldberg et al.

Figure 9.5: (a) ATR-FTIR Spectra of 100 mM Mo in Aqueous Solution as a Function of pH between pH 3 and 9 Plotted in the Region from 700 to 1,200 cm1, (b) Raman Spectra of 100 mM Mo in Aqueous Solution as a Function of pH between pH 3 and 9 in the Region from 700 to 1,200 cm1. Inset is from Bourikas et al. (2001).

good agreement between the ATR-FTIR spectra of Mo sorbed to Al oxide (Spectrum C) with that of the ATR-FTIR spectrum of Mo in aqueous solution at pH 8 (Spectrum B). According to the speciation diagram (Fig. 9.5, inset), the dominant solution species is MoO2 4 , consistent with the observed spectra. It is interesting to note that the 1,015 cm1 band is stronger when Mo is sorbed to the Al oxide surface. The corresponding spectra of Mo sorbed to Al oxide at pH 4 and the solution Raman and ATR-FTIR spectra of aqueous Mo at pH 4 are shown in Fig. 9.7. The ATR-FTIR spectrum of Mo sorbed to Al oxide at pH 4 (Spectrum C; Fig. 9.7) is characterized by three bands at 845, 903, and 940 cm1. The solution Raman (Spectrum A) and ATR-FTIR (Spectrum B) spectra of Mo in aqueous solution have essentially no intensity at 845 cm1 and are characterized by bands at 903 and 940–952 cm1.


251

Figure 9.6: Spectrum A Corresponds to the Raman Spectrum of a 100 mM Mo Solution at pH 8 in the 700–1,200 cm1 Region. Spectrum B was Obtained from the Same Solution as for Spectrum A using a 9-bounce ATRFTIR Solution Cell with a ZnSe IRE. Spectrum C Corresponds to the Difference ATR-FTIR Spectrum of a 10 mM Mo Solution Sorbed to Al Oxide at pH 8.

To the best of our knowledge, a detailed spectroscopic understanding of Mo speciation in aqueous media is not currently known. Based on recent experimental and theoretical work, however, the following conclusions can be drawn from the literature. At pH values>6, the dominant Mo species in aqueous solution is MoO2 4 (Bourikas et al., 2001). This species is characterized by Raman-active bands at 894, 837, and 318 cm1 (Weinstock et al., 1973). The band at 894 cm1 has been assigned to the n1 (a1) symmetric stretch, the band at 837 cm1 to the n3 (t2) asymmetric stretch, and the 318 cm1 band to the n4 (t2) bend. This is consistent with the bands we observed in the Raman spectrum of Mo at pH 9 (Fig. 9.5b). The symmetric Mo–O stretch has very high intensity in the Raman spectrum, whereas the asymmetric stretch has greater intensity in the FTIR spectrum, consistent

252

S. Goldberg et al.

Figure 9.7: Spectrum A Corresponds to the Raman Spectrum of a 100 mM Mo Solution at pH 4 in the 700–1,200 cm1 Region. Spectrum B was Obtained from the Same Solution as for Spectrum A using a 9-bounce ATRFTIR Solution Cell with a ZnSe IRE. Spectrum C Corresponds to the Difference ATR-FTIR Spectrum of a 10 mM Mo Solution Sorbed to Al Oxide at pH 4. with the selection rules. We also observed the depolarization ratio of the 897 cm1 band to be very low (0.02, not shown), confirming the assignment of the 897 cm1 band to the symmetric Mo–O stretch (Long, 1977). Only highly symmetric vibrations, such as the symmetric stretch of an XY4 molecule have very low depolarization ratios. In a related study (Goldberg and Johnston, 2001), we found similar results for the symmetric As–O stretch of the AsO3 species. Jeziorowski and Knozinger (1979) reported Raman 4 spectra of MoO2 (pH 11) and Mo7O6 4 24 (pH 6), but did not indicate the concentration of the Mo species in solution. According to a more recent speciation study (Bourikas et al., 2001), the distribution of MoO2 and 4 6 Mo7O24 would be present at approximately the same concentrations, so that


253

the Raman spectra collected at pH 6 would most likely contain contributions from both species. Upon decreasing the pH of the aqueous Mo solution to pH valueso6, the 837 and 897 cm1 bands diminish in intensity and ‘‘new’’ bands appear at 903 (R+IR active), 940 (IR active), and 952 (R active) cm1 (see Fig. 9.7). These bands have been assigned to the formation of a polymeric Mo7O6 24 species in aqueous solution. At pH values>6, the speciation is dominated by the MoO2 4 species where Mo is in tetrahedral coordination. In contrast, the Mo7O6 24 complex is made up of MoO6 octahedra (Knozinger and Jeziorowski, 1978; Jeziorowski and Knozinger, 1979). Thus, the spectral changes that occur upon lowering the pH are due to: (i) the formation of a polymeric species and (ii) the transition of Mo from a tetrahedrally coordinated species to an octahedrally coordinated species. More recently, there have been several matrix-isolation FTIR studies where Mo species were deposited on a cold window (10 K) in an inert gas matrix (e.g., Ar) (Bare et al., 1998; Wang and Andrews, 2006). This technique provides a very powerful tool for examining Mo species under very carefully controlled conditions. In our case, however, it is difficult to make a direct comparison between our observed Raman and ATR-FTIR spectra and matrix-isolated FTIR spectra of Mo species because our spectra, obtained in aqueous media/ suspensions, are often dominated by intermolecular interactions with water. Although the Mo7O6 24 species is the dominant form of Mo in aqueous solution at pH valueso6, Jeziorowski and Knozinger (1979) attributed sorption of Mo by alumina to the formation of an inner-sphere complex with MoO2 4 . Possible surface complexes of Mo sorbed to alumina catalysts included a bidentate surface complex and a two-dimensional polymeric surface species (Jeziorowski and Knozinger, 1979). It is important to note that all of the Mo surface species considered involve the formation of inner-sphere surface complexes. Based on the literature and our spectroscopic observations, we draw the following conclusions. At high pH, we have excellent agreement between the observed solution ATR-FTIR spectra and those of Mo sorbed to the amorphous Al oxide in aqueous suspension. We assign the 838 cm1 band of sorbed Mo to Al oxide as an outer-sphere surface complex of the MoO2 4 species. Assignment of the 838 cm1 band to MoO2 was based on agree4 ment with literature values and speciation studies, as well as the dependence of our ATR-FTIR and Raman spectra of aqueous Mo solutions as a function of pH and Mo concentration. At low pH, we observe the formation of a polymeric species in aqueous solution that is most likely Mo7O6 24 (or protonated version of this complex). Following the rationale of Jeziorowski and Knozinger (1979), we attribute the spectra of Mo sorbed to Al oxide at low

254

S. Goldberg et al.

pH to the formation of an inner-sphere surface complex that is characterized by the ATR-FTIR bands at 903 and 940 cm1. The specific identity of the surface complex (monodentate, bidentate, or ‘‘two-dimensional’’ polymeric surface species) is not known at this time. Based on the assignment of the 838 cm1 band to the outer-sphere MoO2 4 species and the 903 and 940 cm1 bands to some form of an inner-sphere surface complex, a spectroscopically derived surface complexation diagram is shown in Fig. 9.8. Although the spectra of Mo sorbed to Al oxide are similar to the spectra in solution, some important differences are apparent. Upon increasing the solution pH, the 837 and 897 cm1 bands have essentially no intensity at low pH. In contrast, the spectra of Mo sorbed to Al oxide reveal that the 838 cm1 band still has significant intensity at pH 4. We attribute this to the presence of some outer-sphere coordinated MoO2 4 .

Figure 9.8: The Relative Contribution of the Outer- and Inner-Sphere Surface Complexes Sorbed to Al Oxide as a Function of pH between pH 3.5 and 8. The 837–845 cm1 Band was used to Estimate the Proportion of Outer-Sphere Mo Complexed by the Surface and the Combined Intensity of the 904 and 940–947 cm1 Bands were used to Estimate the Proportion of Inner-Sphere Surface Complexed Mo.


255

In addition, the transformation of the octahedrally coordinated Mo species to the tetrahedrally coordinated complex occurs near pH 6. This pH value is shifted to 5 for Mo sorbed on the surface of the Al oxide. This result is most likely due to the fact that Al oxide has a positive surface charge at these pH values (o8). Thus, the local concentration of hydrogen ions [H+ aq] at the interface is lower than in the bulk solution, due to the accumulation of hydroxide ions that serve to compensate for the surface charge. To the best of our knowledge, ATR-FTIR spectra of Mo sorbed to a hydrous oxide in aqueous suspension have not been reported previously. We have shown that, in general, good agreement exists between the spectra of Mo in aqueous solution and those of Mo sorbed to Al oxide. Although a significant amount of effort has been extended to the study of Mo-based catalysts, relatively little is known about the sorption of Mo to environmentally relevant hydrous oxides in aqueous suspension. 9.3.3. Surface Complexation Modeling The ability of the triple layer model to describe Mo adsorption on amorphous Fe oxide is depicted in Fig. 9.2a. The model described the data quantitatively below pH 6.5 and above pH 9, with deviations in the intermediate pH range. In accordance with the experimental data, the model fits do not show ionic strength dependence, despite the consideration of the outer2 sphere surface complex, SOH+ 2 –MoO4 . Figure 9.2b shows triple layer model fits to Mo adsorption on amorphous Al oxide. While the model was able to fit the majority of the experimental data points, it did not show sufficient ionic strength dependence to provide a quantitative description, especially in the intermediate pH range and for the 1.0 M data. The decrease in quality of fit at 1.0 M may be due to the fact that this ionic strength is outside the applicable range of the Davies equation. The FITEQL program 2 was unable to fit the amorphous Al oxide data using the SOH+ 2 –MoO4 surface complex. Model convergence was obtained only after replacing this species with the outer-sphere surface complex, SOH+ 2 –HMoO4 . The Mo surface speciation predicted by the triple layer model is presented in Fig. 9.9. The inner-sphere surface complex dominates at low pH and the outer-sphere surface complex predominates at high pH. This is in agreement with our Raman and ATR-FTIR spectroscopic observations and those for sulfate and selenate adsorption (Peak et al., 1999; Wijnja and Schulthess, 2000). For amorphous Al oxide, the outer-sphere surface complex dominates above pH 8, while for amorphous Fe oxide, the outer-sphere surface complex does not become predominant until pH 9.5. These results are in agreement

256

S. Goldberg et al.

Mo species adsorbed (%)

100 80

60

SHMoO4 SOH2+-MoO42-

40 20 0 1

3

5

7

(a)

11

13

SHMoO4 SOH2+-HMoO 4-

100


9

pH

80 60 40 20 0 3 (b)

4

5

6

8

7

9

10

11

12

pH

Figure 9.9: Surface Speciation Predicted by the Triple Layer Model for Mo Adsorption on: (a) Amorphous Fe Oxide; (b) Amorphous Al Oxide. with our PZC shift observations. The crossover in dominance between innerand outer-sphere species for the IR (pH 5) and TLM (pH 7) occur within two pH units of each other. Considering that the IR experiments were done at 100 times the Mo concentration as the adsorption experiments, the agreement is quite reasonable.


Optimizing CCM 1.5

257

Optimizing TLM data model

Mo adsorbed (µmol/g)

1.0

0.5 (a)

(b) Prediction CCM

Prediction TLM

1.0

0.5 (c)

(d)

0.0 1

3

5 pH

7

9

2

4

6

8

10

pH

Figure 9.10: Surface Complexation Modeling of Mo Adsorption on Wyo Soil: (a) Constant Capacitance Model Monodentate Fit; (b) Triple Layer Model Inner- and Outer-Sphere Monodentate Fit; (c) Constant Capacitance Model Monodentate Prediction; (d) Triple Layer Model Inner- and OuterSphere Monodentate Prediction. Circles Represent Experimental Data. Model Fits are Represented by Solid Lines. Constant Capacitance Model Results from Goldberg et al. (2002).

Molybdenum adsorption as a function of solution pH had been determined previously for 36 different soil samples (Goldberg et al., 2002). Molybdenum adsorption was found to be maximal in the pH range 2–5 and to decrease rapidly with increasing pH above 5. The constant capacitance model was able to fit the Mo adsorption envelopes well on all of the soil samples, but increasing deviations occurred above pH 7.5 (see Fig. 9.10a). In an effort to improve the fit at high pH, the data were reanalyzed using the triple layer model containing one inner-sphere, SHMoO4, and one outer2 sphere, SOH+ 2 –MoO4 , Mo surface complexation constant. The improvement in the ability of the triple layer model approach to describe Mo adsorption on the Wyo soil is indicated in Fig. 9.10b. For reference, the original fit of the constant capacitance model to the experimental data is

258

S. Goldberg et al.

presented in Fig. 9.10a. The triple layer model was able to quantitatively describe the Mo adsorption data at all pH values. As in our previous investigation (Goldberg et al., 2002), the initial regression model was defined to be: gi ¼ b0 þ b1 ðln CECi Þ þ b2 ðln SAi Þ þ b3 ðln OCi Þ þ b4 ðln IOCi Þ þ b5 ðln Fei Þ þ b6 ðln Al i Þ þ i

ð9:19Þ

where g represents the surface complexation constant, log KiMo ðintÞ; bi the empirical derived regression coefficients, and e the residual error component, assumed to follow the usual regression modeling assumptions of normal errors. The optimization criterion used in each model building process was that the final selected prediction equation should exhibit the smallest possible jack-knifed prediction errors (Myers, 1986), on average. Jack-knifed prediction errors (also commonly referred to as PRESS residuals) can be conveniently computed in most standard regression software packages using the relationship: eî;i ¼

eî ð1 hii Þ

(9.20)

where eî represents the ordinary prediction error, hii the ith diagonal element of the regression model projection matrix, and eî;i the prediction error one obtains if the ith observation is not included in the model estimation process. Jack-knifing essentially represents a ‘‘one-at-a-time’’ deletion and re-estimation technique and is commonly used to develop robust, regressionbased prediction models (Myers, 1986). The PRESS statistic is defined to be the sum of the squared eî;i components and the jack-knifed mean square error (MSE) is simply defined to be the value of the PRESS statistic divided by the sample size. Model optimization was carried out using a backwards elimination, BWE, procedure, where the critical cut-off value used for parameter removal was set to a ¼ 0.15 (Myers, 1986). Based on the results from this BWE procedure, a reduced prediction equation was specified for each surface complexation constant. During the BWE procedure, the PRESS statistic was calculated after each parameter elimination step to verify that this statistic was consistently reduced. Upon termination of the BWE procedure, the remaining parameters left in the reduced prediction equations where then individually deleted and the PRESS statistic was recalculated. This last step was performed in order to verify that the BWE procedure had indeed also minimized the jack-knifed prediction variance. All regression model analyses were performed using the SAS REG procedure (SAS, 1999).


259

For the inner-sphere surface complexation constant, the initial estimate of Eq. (9.19) produced an R2 value of 0.486 and a jack-knifed MSE estimate of 0.519. An application of the BWE procedure removed all variables except lnIOC and lnAl; the corresponding R2 and jack-knifed MSE for this reduced model were 0.472 and 0.382, respectively. The PRESS statistic increased when either of these two remaining regression variables was removed. Hence, the final estimated prediction equation for log KisMo ðintÞ was defined to be: log KisMo ðintÞ ¼ b0 þ b1 ðln IOCÞ þ b2 ðln AlÞ þ

(9.21)

The corresponding parameter estimates for this model are shown in Table 9.1. The specific predicted and jack-knife predicted log KisMo ðintÞ constants for the soil samples are shown in Table 9.2. For the outer-sphere surface complexation constant, the initial estimate of Eq. (9.19) produced a slightly higher R2 value of 0.524 and a jack-knifed MSE estimate of 0.098. An application of the BWE procedure removed the lnIOC, lnAl, and lnSA regression variables; the corresponding R2 and jackknifed MSE for this reduced model were 0.487 and 0.080, respectively. Once again, the PRESS statistic increased when each of the three remaining regression variables were removed. Therefore, the final estimated prediction os equation for log K2Mo ðintÞ was defined to be: os ðintÞ ¼ b0 þ b1 ðln CECÞ þ b2 ðln OCÞ þ b3 ðln FeÞ þ log K2Mo

(9.22)

The corresponding parameter estimates for this model are shown in os ðintÞ Table 9.1, and the specific predicted and jack-knife predicted log K2Mo constants for the soil samples are shown in Table 9.2. Table 9.1: Regression Model Parameter Estimates and Standard Errors for the Triple Layer Model Surface Complexation Model Prediction Equations. Standard error

t-score

Prob4|t|

log KisMo ðintÞ parameter estimates Intercept 5.565 ln(IOC) 0.171 ln(Fe) 0.373

0.503 0.039 0.172

11.07 4.36 2.17

o0.0001 0.0002 0.0390

log K2os Mo ðintÞ parameter estimates Intercept 1.672 ln(CEC) 0.412 ln(OC) 0.131 ln(Fe) 0.178

0.448 0.088 0.068 0.079

3.74 4.70 1.92 2.26

0.0009 o0.0001 0.0654 0.0326

Parameter

Estimate

260

S. Goldberg et al.

Table 9.2: Triple Layer Model Surface Complexation Constants Obtained by Optimization and from the Prediction Equations. Soil

Altamont

Arlington Avon Bonsall Chino Diablo Fallbrook Fiander Haines Hanford Hesperia Holtville Imperial Lost Hills Nohili Pachappa

Porterville Reagan Ryepatch Sebree Wasco Wyo Yolo

Fitted Log KisMo

Fitted Log K2os Mo

Predicted Log KisMo

4.01 2.81 2.70 3.95 4.09 3.19 4.30 3.56 3.54 3.10 4.31 4.70 5.14 3.07 4.29 4.38 5.07 5.73 4.73 3.16 3.14 3.95 3.27 4.07 3.84 4.29 3.79 2.79 4.72 4.25

0.518 0.496 0.553 0.456 0.158 0.166 1.126 0.903 0.805 0.159 0.718 0.411 0.087 0.391 0.269 0.550 0.064 0.251 0.978 0.319 0.102 0.166 0.514 0.750 0.241 0.905 0.354 0.641 0.274 0.388

3.90 3.32 3.37 3.83 3.79 3.66 4.81 4.09 4.38 3.48 4.66 4.14 4.31 3.22 4.30 4.57 4.20 4.97 4.23 3.54 3.54 3.19 3.72 3.72 4.51 4.44 3.16 3.18 3.54 4.16

Predicted Jack-knife Jack-knife predicted predicted Log K2os Mo Log KisMo Log K2os Mo 0.332 0.449 0.506 0.386 0.477 0.077 0.881 0.650 0.536 0.402 0.733 0.396 0.024 0.192 0.331 0.674 0.777 0.478 0.545 0.012 0.012 0.302 0.513 0.513 0.353 0.866 0.026 0.437 0.324 0.362

3.90 3.37 3.44 3.83 3.78 3.69 4.91 4.12 4.43 3.55 4.70 3.95 4.20 3.24 4.30 4.59 4.15 4.59 4.21 3.56 3.56 3.10 3.75 3.69 4.61 4.45 3.08 3.23 3.42 4.15

0.303 0.447 0.503 0.383 0.528 0.065 0.825 0.626 0.504 0.421 0.736 0.392 0.265 0.164 0.338 0.689 0.933 0.641 0.508 0.067 0.031 0.335 0.513 0.496 0.360 0.853 0.115 0.408 0.330 0.359

Equations (9.21) and (9.22) were each derived independently, using the same initial set of optimized surface complexation constants. The jack-knifed prediction columns show the ability of the model to predict each individual soil surface complexation constant, without using this constant in the fitting process. The general excellent agreement between the predicted surface


261


110 90 70

SHMoO4 SOH2+-MoO 42-

50 30 10

1

2

3

4

5

6

7

8

9

10

pH

Figure 9.11: Surface Speciation Predicted by the Triple Layer Model for Mo Adsorption on Wyo Soil. complexation constants and the jack-knife estimates suggests that the regression prediction equations for both surface complexation constants should have good predictive capability. Figure 9.10d presents the ability of the triple layer model to predict Mo adsorption from chemical properties for one of the soils predicted previously using the constant capacitance model (Goldberg et al., 2002). While the prediction (Fig. 9.10d) is, of course, less good than the result of the model optimization (Fig. 9.10b), it is at least semi-quantitative at all pH values. The tremendous improvement in the quality of the triple layer model prediction at pH values above 7 (Fig. 9.10d) over the constant capacitance model prediction (Fig. 9.10c) is especially significant for prediction of Mo adsorption in arid zone soils where Mo can accumulate in forage plants to levels detrimental to the health of grazing ruminant animals. The surface speciation predicted for the Wyo soil is presented in Fig. 9.11. The outer-sphere Mo surface complex is predicted to dominate above pH 6. This is in agreement with the surface speciation predicted from spectroscopy for Mo adsorption on amorphous Al oxide.

9.4. Conclusions The results of all experimental methods, both macroscopic (PZC shifts and ionic strength effects) and microscopic (Raman and ATR-FTIR

262

S. Goldberg et al.

spectroscopies), provide self-consistent mechanisms for Mo adsorption on amorphous oxides. Molybdate forms predominantly inner-sphere surface complexes on both oxides at low pH and outer-sphere surface complexes dominate at high pH. The inner-sphere surface complex predominates to a greater extent for Mo adsorption on amorphous Fe oxide, while the outersphere surface complex is more significant for Mo adsorption on amorphous Al oxide. The triple layer model containing both an inner-sphere and an outer-sphere surface complex was able to describe and predict Mo adsorption on a large group of soils at all pH values investigated. The predominance diagram for the Mo surface species in soil was in complete agreement with that obtained spectroscopically for Mo adsorption on amorphous Al oxide.

ACKNOWLEDGMENTS Gratitude is expressed to Mr. H.S. Forster and Mr. G.S. Premachandra for technical assistance.

REFERENCES Bare, W. D., Souter, P. F., & Andrews, L. (1998). Reactions of laser-ablated molybdenum and tungsten atoms with dioxygen. Resolved infrared spectra of natural molybdenum and tungsten isotopic oxides in argon matrices. J. Phys. Chem. A, 102, 8279–8286. Bibak, A., & Borggaard, O. K. (1994). Molybdenum adsorption by aluminum and iron oxides and humic acid. Soil Sci., 158, 323–327. Bourikas, K., Hiemstra, T., & van Riemsdijk, W. H. (2001). Adsorption of molybdate monomers and polymers on titania with a multisite approach. J. Phys. Chem. B, 105, 2393–2403. Coffin, D. E. (1963). A method for the determination of free iron oxide in soils and clays. Can. J. Soil Sci., 43, 7–17. Davis, J. A., & Kent, D. B. (1990). Surface complexation modeling in aqueous geochemistry. Rev. Mineral., 23, 117–260. Dzombak, D. A., & Morel, F. M. M. (1990). Surface Complexation Modeling. Hydrous Ferric Oxide. Wiley, New York. El Shafei, G. M. S., Moussa, N. A., & Philip, C. A. (2000). Association of molybdenum ionic species with alumina surface. J. Colloid Interface Sci., 228, 105–113. Farmer, V. C. (1974). The layer silicates. In: V. C. Farmer (Ed). The Infrared Spectra of Minerals. Mineral Society, London, pp. 331–359.


263

Ferreiro, E. A., Helmy, A. K., & de Bussetti, S. G. (1985). Molybdate sorption by oxides of aluminium and iron. Z. Pflanzenernaehr. Bodenkd., 148, 559–566. Goldberg, S. (1992). Use of surface complexation models in soil chemical systems. Adv. Agron., 47, 233–329. Goldberg, S. (2002). Competitive adsorption of arsenate and arsenite on oxides and clay minerals. Soil Sci. Soc. Am. J., 66, 413–421. Goldberg, S., & Forster, H. S. (1998). Factors affecting molybdenum adsorption by soils and minerals. Soil Sci., 163, 109–114. Goldberg, S., Forster, H. S., & Godfrey, C. L. (1996). Molybdenum adsorption on oxides, clay minerals, and soils. Soil Sci. Soc. Am. J., 60, 425–432. Goldberg, S., & Johnston, C. T. (2001). Mechanisms of arsenic adsorption on amorphous oxides evaluated using macroscopic measurements, vibrational spectroscopy, and surface complexation modeling. J. Colloid Interface Sci., 234, 204–216. Goldberg, S., Lesch, S. M., & Suarez, D. L. (2002). Predicting molybdenum adsorption by soils using soil chemical parameters in the constant capacitance model. Soil Sci. Soc. Am. J., 66, 1836–1842. Goldberg, S., Su, C., & Forster, H. S. (1998). Sorption of molybdenum on oxides, clay minerals, and soils: Mechanisms and models. In: E. A. Jenne (Ed). Adsorption of Metals by Geomedia: Variables, Mechanisms, and Model Applications. Proc. Am. Chem. Soc. Symp., Academic Press, San Diego, CA, pp. 401–426. Gonzalez, B. R., Appelt, H., Schalscha, E. B., & Bingham, F. T. (1974). Molybdate adsorption characteristics of volcanic-ash-derived soils in Chile. Soil Sci. Soc. Am. Proc., 38, 903–906. Gustafsson, J. P. (2003). Modelling molybdate and tungstate adsorption to ferrihydrite. Chem. Geol., 200, 105–115. Hayes, K. F., Papelis, C., & Leckie, J. O. (1988). Modeling ionic strength effects on anion adsorption at hydrous oxide/solution interfaces. J. Colloid Interface Sci., 125, 717–726. Herbelin, A. L., & Westall, J. C. (1996). FITEQL: A Computer Program for Determination of Chemical Equilibrium Constants from Experimental Data. Rep. 96–01, Version 3.2, Department of Chemistry, Oregon State University, Corvallis, OR. Herzberg, G. (1945). Molecular Spectra and Molecular Structure: II. Infrared and Raman Spectra of Polyatomic Molecules. Van Nostrand Reinhold Company, New York. Jeziorowski, H., & Knozinger, H. (1979). Raman and ultraviolet spectroscopic characterization of molybdena on alumina catalysts. J. Phys. Chem., 83, 1166–1173. Johnston, C. T., & Aochi, Y. O. (1996). Fourier transform infrared and Raman spectroscopy. In: D. L. Sparks et al. (Eds). Methods of Soil Analysis: Part 3. Chemical Methods. Soil Science Society of America, Madison, WI. Johnston, C. T., & Wang, S. L. (2002). Application of vibrational spectroscopy in soil and environmental sciences. In: J. M. Chalmers, & P. R. Griffiths (Eds). Handbook of Vibrational Spectroscopy. Wiley, New York.

264

S. Goldberg et al.

Jones, L. H. P. (1957). The solubility of molybdenum in simplified systems and aqueous soil suspensions. J. Soil Sci., 8, 313–327. Knozinger, H., & Jeziorowski, H. (1978). Raman-spectra of molybdenum oxide supported on surface of aluminas. J. Phys. Chem., 82, 2002–2005. Kyriacou, D. (1967). The pH-dependence of adsorption of metallic oxyanions by ferric oxide powder. Surf. Sci., 8, 370–372. Lang, F., & Kaupenjohann, M. (1999). Molybdenum fractions and mobilization kinetics in acid forest soils. J. Plant Nutr. Soil Sci., 162, 309–314. Lang, F., & Kaupenjohann, M. (2003). Immobilisation of molybdate by iron oxides: Effects of organic coatings. Geoderma, 113, 31–46. Lang, F., Pohlmeier, A., & Kaupenjohann, M. (2000). Mechanism of molybdenum sorption to iron oxides using pressure-jump relaxation. J. Plant Nutr. Soil Sci., 163, 571–575. Lefevre, G., & Fedoroff, M. (2006). In situ Fourier-transform infrared spectroscopy studies of inorganic ions adsorption on metal oxides and hydroxides. Adv. Colloid Interface Sci., 107, 109–123. Long, D. A. (1977). Raman Spectroscopy. McGraw-Hill, New York. McBride, M. B. (1997). A critique of diffuse double layer models applied to colloid and surface chemistry. Clays Clay Miner., 45, 598–608. McKenzie, R. M. (1983). The adsorption of molybdenum on oxide surfaces. Aust. J. Soil Res., 21, 505–513. McLaughlin, J. R., Ryden, J. C., & Syers, J. K. (1981). Sorption of inorganic phosphate by iron and aluminium-containing components. J. Soil Sci., 32, 365–377. Mikkonen, A., & Tummavuori, J. (1993). Retention of molybdenum (VI) by three Finnish mineral soils. Acta Agric. Scand. B Soil Plant Sci., 43, 206–212. Motta, M. M., & Miranda, C. F. (1989). Molybdate adsorption on kaolinite, montmorillonite, and illite: Constant capacitance modeling. Soil Sci. Soc. Am. J., 53, 380–385. Murphy, L. S., & Walsh, L. M. (1972). Correction of micronutrient deficiencies with fertilizers. In: J. J. Mortvedt, P. M. Giordano, & W. L. Lindsay (Eds). Micronutrients in Agriculture. Soil Science Society of America, Madison, WI, pp. 347–387. Myers, R. H. (1986). Classical and Modern Regression with Applications. Duxbury Press, Boston, MA. O’Connor, G. A., Brobst, R. B., Chaney, R. L., Kincaid, R. L., McDowell, L. R., Pierzynski, G. M., Rubin, A., & Van Riper, G. G. (2001). A modified risk assessment to establish molybdenum standards for land application of biosolids. J. Environ. Qual., 30, 1490–1507. Peak, D., Ford, R. G., & Sparks, D. L. (1999). An in situ ATR-FTIR investigation of sulfate bonding mechanisms on goethite. J. Colloid Interface Sci., 218, 289–299. Reisenauer, H. M., Tabikh, A. A., & Stout, P. R. (1962). Molybdenum reactions with soils and the hydrous oxides of iron, aluminum, and titanium. Soil Sci. Soc. Am. J., 26, 23–37. Reisenauer, H. M., Walsh, L. M., & Hoeft, R. G. (1973). Testing soils for sulphur, boron, molybdenum, and chlorine. In: L. M. Walsh, & J. D. Beaton (Eds). Soil


265

Testing and Plant Analysis. Revised Edition. Soil Science Society of America, Madison, WI, pp. 173–200. Reyes, E. D., & Jurinak, J. J. (1967). A mechanism of molybdate adsorption on aFe2O3. Soil Sci. Soc. Am. J., 31, 637–641. Saripalli, P. K., McGrail, B. P., & Girvin, D. C. (2002). Adsorption of molybdenum on to anatase from dilute aqueous solutions. Appl. Geochem., 17, 649–656. SAS (1999). SAS/STAT Users Guide. Version 8. SAS Institute Inc., Cary, NC. Sims, J. T., & Bingham, F. T. (1968). Retention of boron by layer silicates, sesquioxides, and soil materials: II. Sesquioxides. Soil Sci. Soc. Am. Proc., 32, 364–369. Spanos, N., & Lycourghiotis, A. (1995). Codeposition of Mo(VI) species and Ni2+ ions on the g-alumina surface: Mechanistic model. J. Colloid Interface Sci., 171, 306–318. Spanos, N., Vordonis, L., Kordulis, C. H., Koutsoukos, P. G., & Lycourghiotis, A. (1990a). Molybdenum-oxo species deposited on alumina by adsorption. II. Regulation of the surface MoVI concentration by control of the protonated surface hydroxyls. J. Catal., 124, 315–323. Spanos, N., Vordonis, L., Kordulis, C. H., & Lycourghiotis, A. (1990b). Molybdenum-oxo species deposited on alumina by adsorption. I. Mechanism of the adsorption. J. Catal., 124, 301–314. Sposito, G. (1983). Foundations of surface complexation models of the oxide– aqueous solution interface. J. Colloid Interface Sci., 91, 329–340. Sprycha, R. (1989a). Electrical double layer at alumina/electrolyte interface. I. Surface charge and zeta potential. J. Colloid Interface Sci., 127, 1–11. Sprycha, R. (1989b). Electrical double layer at alumina/electrolyte interface. II. Adsorption of supporting electrolyte ions. J. Colloid Interface Sci., 127, 12–25. Stollenwerk, K. G. (1995). Modeling the effects of variable groundwater chemistry on adsorption of molybdate. Water Resour. Res., 31, 347–357. Su, C., & Suarez, D. L. (1997). In situ infrared speciation of adsorbed carbonate on aluminum and iron oxides. Clays Clay Miner., 45, 814–825. Tossell, J. A. (2005). Calculating the partitioning of the isotopes of Mo between oxidic and sulfidic species in aqueous solution. Geochim. Cosmochim. Acta, 69, A210. Vissenberg, M. J., Joosten, L. J. M., Heffels, M. M. E. H., van Welsenes, A. J., de Beer, V. H. J., van Santen, R. A., & van Veen, J. A. R. (2000). Tungstate versus molybdate adsorption on oxidic surfaces: A chemical approach. J. Phys. Chem. B, 104, 8456–8461. Vordonis, L., Koutsoukos, P. G., & Lycourghiotis, A. (1990). Adsorption of molybdates on doped g-aluminas in alkaline solutions. Colloids Surf., 50, 353–361. Wang, S. L., & Johnston, C. T. (2000). Assignment of the structural OH stretching bands of gibbsite. Am. Mineral., 85, 739–744. Wang, X. F., & Andrews, L. (2006). Contrasting products in the reactions of Cr, Mo, and W atoms with H2O2: Argon matrix infrared spectra and theoretical calculations. J. Phys. Chem. A, 110, 10409–10418.

266

S. Goldberg et al.

Weinstock, N., Schulze, H., & Muller, A. (1973). Assignment of n2 (E) and n4 (F2) of tetrahedral species by calculation of the relative Raman intensities: The vibra2 2 2 tional spectra of VO3 4 , CrO4 , MoO4 , WO4 , MnO4 , TcO4 , ReO4 , RuO4, and OsO4. J. Chem. Phys., 59, 5063–5067. Wijnja, H., & Schulthess, C. P. (2000). Vibrational spectroscopy of selenate and sulfate adsorption mechanisms on Fe and Al (hydr)oxide surfaces. J. Colloid Interface Sci., 229, 286–297. Wu, C.-H., Lo, S.-L., Lin, C.-F., & Kuo, C.-Y. (2001). Modeling competitive adsorption of molybdate, sulfate, and selenate on g-Al2O3 by the triple-layer model. J. Colloid Interface Sci., 233, 259–264. Xu, N., Christodoulatos, C., & Braida, W. (2006a). Adsorption of molybdate and tetrathiomolybdate onto pyrite and goethite: Effect of pH and competitive anions. Chemosphere, 62, 1726–1735. Xu, N., Christodoulatos, C., & Braida, W. (2006b). Modeling the competitive effect of phosphate, sulfate, silicate, and tungstate anions on the adsorption of molybdate onto goethite. Chemosphere, 64, 1325–1333. Zhang, P., & Sparks, D. L. (1989). Kinetics and mechanisms of molybdate adsorption/desorption at the goethite/water interface using pressure-jump relaxation. Soil Sci. Soc. Am. J., 53, 1028–1034. Zhang, P., & Sparks, D. L. (1990). Kinetics of selenate and selenite adsorption/desorption at the goethite/water interface. Environ. Sci. Technol., 24, 1848–1856.


Chapter 10

Blind Prediction and Parameter Uncertainty – A Sorption Test Case Anke Richter and Vinzenz Brendler Forschungszentrum Dresden-Rossendorf e.V., Institute of Radiochemistry, P.O. Box 51 01 19, D-01314 Dresden, Germany

ABSTRACT The effect of parameter consistency and uncertainty (protolysis constants, equilibrium constants) in surface complexation modeling (SCM) is illustrated. As a test case the blind prediction of Cu(II) sorption onto goethite was selected and the Diffuse Double Layer Model (DDLM) was applied. This electrostatic model requires the smallest set of parameters. An uncertainty analysis of the goethite protolysis constants (pK) distribution showed, that the experimentally determined conventional distribution coefficients (KD) could be predicted within one order of magnitude or better. Moreover, the uncertainty analysis concerning the formation constants of the Cu(II) surface species onto goethite was applied, again with encouraging results. We also investigated the effect of using parameters linked to electrostatic models other than DDLM. In the case of sparse SCM data matrices, even such inconsistencies may be tolerated. We conclude that the proposed methodology to cope with insufficient surface complexation datasets can help to estimate distribution coefficients for contaminants in well-defined mineral systems.

10.1. Introduction Worldwide there is a focus on the remediation of radioactively contaminated sites. One common aim is to deliver a more profound chemical base for risk assessment, namely all those physico-chemical phenomena governing the contamination plume development in time and space. Coupled transport Corresponding author. Tel.: ++49 351 260 2426; Fax: ++49 351 260 3553;

E-mail: [email protected] (A. Richter).

268

A. Richter and V. Brendler

codes that are able to tackle this challenge have to simplify the very complex reaction pattern. To do so in an adequate way, the knowledge about retardation and mobilization phenomena and the underlying basic processes and interactions, such as physisorption, chemisorption, co-precipitation, inclusion, diffusion, surface-precipitation, or the formation of solid solutions has to be extended. Whereas the simplest (and older) sorption models do not distinguish between the various processes contributing to the overall sorption, newer model approaches try to address all relevant processes separately. In a strict sense sorption models are usually grouped into two classes: the phenomenological models and the surface complexation models. Phenomenological adsorption models comprise different variants of the equilibrium distribution coefficient (KD) model. The KD framework is built on the concept of distribution (or retardation) coefficients. This is defined as the experimentally determined ratio of the sorption density (mol adsorbate/g adsorbent) to the concentration of the component in the aqueous phase under equilibrium conditions. The subsumption of many physico-chemical processes into one parameter is a severe weakness of the KD principle (Hayes et al., 1991; Wilhelm and Beam, 1999). Distribution coefficients are difficult to measure with good precision and accuracy. Even slight changes in one system parameter (say the EH, or the content of a major cation, or the occurrence of a new mineral phase, etc.) can drastically change them. It is impossible to measure the effect of all combinations of these parameters. That means, all KD values used nowadays for risk assessment or for other prognostic studies are just snapshots for specific locations of the site. They are only valid for the specific geochemical conditions at the time of the measurement. This in turn assigns them very large uncertainties. A much better strategy is the decomposition of the KD value into the main basic processes defining it. Such an approach will unfold the single distribution coefficient into a vector of parameters, such as pH, concentrations of the various components, binding site densities, surface areas, and temperature. This seems to be a step backwards, but it has the advantage that all vector parameters can be measured with more reliability and precision. Knowing the functional relationships between these processes and how they contribute to the KD allows a computation rather than measurement to occur. Moreover, simulations with variable parameter values, even for hypothetical conditions, may easily yield a KD surface as a function of the ‘‘primary’’ parameter vector: K D ¼ f ðpH; PCO2 ; T; . . .Þ: Usually, the function f can not be expressed as an explicit function, but contains implicit formulations accessible through numeric iterations. Also, some long-term effects, that may render conventional distribution coefficients meaningless

Blind Prediction and Parameter Uncertainty

269

(e.g., co-precipitation, diffusion of trace elements into crystal lattices), can be accounted for in a better way. Another application is expressing KD as a function of time, thereby relating it to better-defined time dependencies of other basic parameters. Furthermore, it becomes possible to identify those parameters affecting the KD strongest. Consequently, extra measurements can be designed efficiently to reduce its uncertainty. And last but not least, retaining the current KD paradigm renders it easy to couple a ‘‘smart KD’’ framework with already existing contaminant transport and risk assessment codes, enhancing its acceptance. The unfolding of KD values leads to modern concepts that treat surface reactions as complex formations analogous to such reactions in homogeneous aqueous solutions. Therefore these models are called Surface Complexation Models (SCM, for details refer to Stumm, 1992). The most important groups are the Diffuse Double Layer Model (DDLM), Constant Capacitance Model (CCM), Triple Layer Model (TLM), Basic Stern Model (BSM) and the (CD-)MUSIC Model. SCM describe adsorption in terms of ion complexation by specific surface sites. The activity coefficients of the surface species depend on the assumed structure of the electrical double layer and the manner in which ions interact with it. Therefore, postulated surface species are model dependent. There is considerable experimental data available for the sorption of metal ions in one-mineral-systems, including an accurate description of the data with SCM (e.g., Waite et al., 1994; Arnold et al., 2001). The application to mixtures of minerals in soils and sediments is more difficult. Generally, two approaches are discussed in the literature where the SCM concept to model adsorption to soils and sediments has been applied (Davis et al., 1998): the component additivity (CA) approach and the generalized composite (GC) approach. In the CA approach, the prediction of the adsorption on a complex mineral system is accomplished from surface complexation constants are determined for the individual mineral phases present in the soil or sediment (Honeyman, 1984; Jung et al., 1999; Arnold et al., 2001). In the GC approach, the modeler tries to quantify surface complexation behavior for the system as a whole (Davis et al., 1998; Payne et al., 2004). The SCM concept combined with a good sorption database allows a straightforward extension to rocks and soils composed of several minerals used in the CA approach. To support the above approach a digital thermodynamic database for surface complexation equilibria is essential. Therefore, a digitized version of a thermodynamic sorption database has been implemented as a relational database with the Microsoft Access software: ‘‘RES3T – Rossendorf Expert System for Surface and Sorption Thermodynamics’’ (Brendler et al., 2003). At present (as of March 2007) RES3T

270


contains 3,520 surface complexation constants, 1,270 protolysis constants, and 1,125 specific surface area values of 117 minerals, supplemented by more than 2,170 references. One of its major goals is to provide a sound foundation for an increasing use of SCM in risk assessment studies. As a first step toward KD prediction in geomedia, the RES3T database should help to derive recommended datasets for such SCM applications. During the next decade SCM will probably replace KD only in rather simple systems or systems dominated by just one mineral. However, SCM can help to verify measured KDs, to identify the most critical/sensitive experimental parameters, to assign uncertainty limits, to fill gaps difficult to access in sorption experiments, and to help to gain a better process understanding. Thus, SCM have a high potential to increase confidence in safety analysis and risk assessment studies. The transition process to a chemically more realistic approach to sorption seems to be a rather slow one, hampered by reservation toward SCM with regard to complexity, data availability, and trustworthiness (Lu¨tzenkirchen, 2002). Westall and Hohl (1980) compared different SCM in a quantitative way for the first time using the new fit program FITEQL and concluded that the different SCM submodels describe the experimental data equally well. A given set of data can be simulated by another model, sometimes with a different set of surface reactions. Nevertheless, the parameters derived from the particular models cannot be compared directly, i.e., intrinsic constants are dependent on the SCM submodel. Furthermore, the scope of the particular SCM submodel concerning the ionic strength of the system is different. Therefore, the high demands on the consistency of SCM datasets include the use of a uniform SCM submodel throughout a blind prediction task. To overcome bias toward SCM and the reluctance to apply this approach it is essential to develop a data supply strategy for modeling and to test whether SCM can be applied successfully for predicting KD values in performance assessment (PA). Such a general strategy for the numerical data inquiry has been proposed recently (Richter et al., 2005a). In this paper we present examples illustrating the current blind predictive capabilities of DDLM, which is the model requiring the smallest set of parameters. The system considered was Cu(II) sorption onto goethite. The predictions were compared with raw data from three independent experimental investigations. Where in most cases the model predictions represented the experimental sorption values for the sorbed amount of Cu(II), expressed as conventional distribution coefficients KD, within one order of magnitude or better. Uncertainty analyses are hardly ever a subject of sorption-related publications. However, in Richter et al. (2005b) the effect of parameter uncertainty (protolysis constants) in surface complexation modeling was


271

illustrated. There, as an example, the DDLM was applied for the blind prediction of Np(V) sorption onto hematite. Variations of pK values within two standard deviations consistently delivered predictions deviating less than one order of magnitude from the experimental distribution coefficients. The aim of the present paper is to illustrate the effect of uncertainty in blind predictions for another test case, based on the aforementioned data supply strategy. Also, we extended the scope from protolysis constants to surface complexation constants.

10.2. Methodology 10.2.1. Strategy of Data Compilation The strategy of data compilation is outlined in detail in a previous paper of the authors (Richter et al, 2005a). The following two steps summarize this approach. First, a literature survey supported by an appropriate database helps to define the chemical system, i.e., the mineral properties (specific surface area, site density, protolysis constants) as a function of ionic strength. Second, the set of surface species and surface complexation constants can be derived, preferably also based on a mechanistic database such as, e.g., provided by RES3T (Brendler et al., 2003). This MS Access implemented relational database is mineral specific and based on the concept of SCM. Both data categories are then treated in the same general way: 1. Selection of reliable reaction data. The consistency of the data with respect to model, mineral, and aquatic speciation is an important quality criterion. Yet, when only little data is available, a pragmatic way should be chosen. In most cases it is better to consider values with a large uncertainty than none at all, i.e., ignoring essential reactions. 2. Normalization. Usually, the reported data constants (protolysis constants, surface complexation constants) are related to different site densities GC and cannot be directly compared. Thus, it is necessary to convert them to a reference state to enable comparison and averaging (normalization). Kulik (2002) defined the standard state of a surface species as follow: One mole of species occupies all sites of the reference total density G0 ¼ 20 mmol/m2 (12.05 sites/nm2) on a surface of one mole of a sorbent suspended in 1 kg of the solution at a reference pressure of 1 bar and a reference temperature of 251C, in absence of external fields and at zero surface potential, c ¼ 0. This results in the conversion of the conventional reaction constants KC

272


formulated for the unreacted surface site being on the left hand side: Gc (10.1) log K 0 ¼ log K c þ log G0 Another site density of G0 ¼ 2.31 sites/nm2 for ferrihydrite was suggested by Dzombak and Morel (1990), while Davis and Kent (1990) proposed it as a sort of a ‘‘universally’’ recommended site density for all minerals. We decided to use this value of G0 ¼ 2.31 sites/nm2 consistently for all subsequent predictive modeling, because it comes close to the physical reality in the test case described later. 3. Extrapolation. All reaction constants must be extrapolated to infinite dilution by the Davies equation. Currently, there is no generally accepted convention for treating activity coefficients (f) of surface species. For dissolved species, the Davies equation (Davies, 1962) can be utilized for the calculation of activity coefficients in a range up to an ionic strength of 0.5 M: pffiffiffi I 2 p ffiffi ffi 0:3I (10.2) log f ¼ A z 1þ I where I denotes the ionic strength, z the charge of the ionic species and A is the Debye–Hu¨ckel parameter, 0.5093 (l/mol)1/2 for water at 251C. 4. Sparse data matrices. When key parameters are not available, various approximations can be utilized to derive them for SCM. This includes the estimation of protolysis constants based on crystallography and thermodynamics (Sverjenski and Sahai, 1996), or extrapolation from chemically similar systems (with regard to both mineral and sorbent) by applying the Linear Free Energy Relationships (LFER) (Dzombak and Morel, 1990). If such approaches fail, a simple transfer of data from chemically similar systems (with identical charge) is acceptable. As a last resort, parameters based on electrostatic terms different from the chosen SCM may also be taken into account, which reduces the internal data consistency. Preliminary uncertainty analysis showed that in most cases the sorption modeling error caused by totally omitting a surface reaction is much larger than that introduced by using a surface complex formation constant with large uncertainties (Richter et al., 2005b). 5. Comparison and averaging. After normalization, all data records related to the same reaction must be compared and judged to identify and exclude outliers and doubtful data points. The remaining selected thermodynamic data records sets are then averaged to obtain respective model parameters that are used in the blind prediction.


273

10.2.2. General Blind Prediction Modeling Predictive modeling of ion sorption to minerals, soils, and sediments can be used to better understand the fate of ions in the environment, in toxicity and risk assessment. In our understanding blind prediction means that each prediction is solely based on the specific experimental conditions, such as total concentration, ionic strength, surface area of soil, solid concentration, and pH range. Experimentally measured sorption data are not used in any way, hence, this is not a fitting procedure. It is not the intention of blind prediction to test several different datasets and to check which of them come closest to reality. The concept is designed for those cases where there are virtually no experimental data (to check against) available, possibly for reasons of difficulty of the chemical systems and the required environmental conditions or the high level of effort required. Several years ago the Nuclear Energy Agency (NEA) sorption project was initiated as a major international contribution toward demonstrating the consistency and applicability of SCM to support safety assessments of geological repositories. To enable an evaluation of the respective advantages and drawbacks of the different models, it was implemented in the form of a comparative modeling exercise based on selected datasets for radionuclide sorption by both simple and complex materials (Payne et al., 2004; Davis et al., 2005). Twenty international modeling teams, including the authors’ institution, took part. We applied the NEA criteria for the model quality assessment. Namely, a deviation in KD below one order of magnitude implies that the blind prediction for complex geomedia could be considered as quite good. For simple systems, a higher precision is sometimes realizable. One should, however, be aware that the quality of blind predictions will match the experimental uncertainty only in very exceptional cases. In the NEA blind prediction exercise, five out of the overall seven test cases dealt with rather simple systems. Nevertheless, the outcome (quality of blind predictions) was also on average of the order of one magnitude in KD. 10.2.3. Uncertainty Analysis The uncertainty of the calculated endpoints of an impact assessment model is composed of the uncertainty of scenario, of the model, and of the parameter. The uncertainty of the scenario description and the model uncertainty (lack of confidence about the mathematical model being a valid formulation of the assessment problem) are considered here only partially (restricted to the

274


chemical model). Parameter uncertainty considers the uncertainty about the true value of the parameters used in the model equations. This may be due to the lack of knowledge or to the random character of the parameter value. Reaction specific parameters can usually be taken from literature, ranging from large databases down to single values from a publication. Nevertheless, available data are often not sufficient in scope or quality (experimental conditions, reproducibility, internal consistency), and there are a variety of other problems, for example: – Many equilibrium constants are conditional ones. They are not converted to the standard state. – Some of the species reported in the literature are rather speculative. They are the result of best fits only and lack direct spectroscopic or other evidence. – It is common that uncertainties are not published, that their derivation is not clearly defined or that they are strongly correlated.

10.3. Modeling 10.3.1. Application Case Here, legacies of ore mining serve as a representative application case. Generally, the seepage water from ore mining piles poses a risk of contamination to ground and surface water with radionuclides, heavy metals, and arsenic. The seepage problem is especially pronounced in the case of acidic tailings, as the majority of contaminants involved are more mobile under acidic conditions. The tailings are dominated by granite and secondary iron phases. In tailings containing pyrite, acidic conditions automatically develop due to the inherent production of sulfuric acid, which increases the migration of contaminants to the environment. The problem of modeling the sorption processes in such complex systems is very diverse and difficult, starting with the proper definition of the chemical system (mineralogy and composition of the aqueous phase), its dominant reactions and processes, and the subsequent model parameterization. Figure 10.1 shows the number of surface complexation data records based on RES3T for the sorption of selected heavy metals, including uranium(VI), on secondary iron phases, quartz and muscovite as typical constituents of granite. It is quite evident that the distribution of the published data is very unbalanced. In Fig. 10.2 this data situation is shown for DDLM data only. It is obvious, that parameters are not available for each adsorbate/mineral


275

Figure 10.1: Number of Total logK Records Covering All SCM (Based on RES3T).

Figure 10.2: Availability of logK Records for DDLM Only (Based on RES3T) with Proposed Approximations to Close Data Gaps (Cross: Available Data, Dash: Missing Data, Open Arrow: Chemical Analogy of Adsorbate, Filled Arrow: Chemical Analogy of Mineral, SCM Acronyms: Acceptance of Electrostatic Inconsistencies, Gray Cell: Target System).

276


combination. The question is: ‘‘What can be done, if virtually no experimental data (to check against) are available?’’ Is it reasonable to apply data based on a different model (TLM, CCM), data based on the sorption onto another iron (hydr)oxide (chemical analogy of mineral) or on the sorption of another heavy metal (chemical analogy of adsorbate)? Figure 10.2 visualizes this for the sorption of Ni(II) onto hematite where no published DDLM data exist. Here, the sorption of Cu(II) onto goethite is taken as an example to illustrate the proceeding of blind prediction (gray cell in Fig. 10.2). The system Cu(II)/goethite (a-FeOOH) has been the subject of many experimental investigations (e.g., Balistrieri and Murray, 1982; Kooner, 1992; Ali and Dzombak, 1996a,b; Robertson and Leckie, 1998; Buerge-Weirich et al., 2002, 2003). The extent of metal uptake by soils is strongly influenced by several parameters: pH, ionic strength, metal concentration, mineral/sorbent ratio (i.e., solid concentration), reaction temperature, and time. Besides DDLM data there are also sufficient parameters for other electrostatic models available from the literature. Thus we chose this system to evaluate the effects arising from using data from other models and systems if sparse data matrices are a problem.

10.3.2. Blind Prediction Modeling Test Case The simplicity in blind prediction was a major goal, so we wanted to keep the number of parameters as small as possible. For pragmatic reasons we usually prefer the DDLM as the SCM variant, since this choice minimizes the number of parameters. The TLM may often come closer to the physical reality but needs more parameters and, probably even more crucially, always delivers parameters valid only for a distinct background electrolyte. Thus, it is very difficult to combine parameters from different background electrolytes to tackle applied problems with mixed complex background electrolytes. Moreover, many published datasets are based on the DDLM, and the parameter sets are not only valid for a specific background electrolyte. To keep the system as simple as possible, we did not distinguish between strong and weak binding sites on goethite, taking into consideration that there is no clear spectroscopic evidence for such a distinction so far. If complexation constants are published only for the sorption on weak and strong binding sites we took the values for the sorption on weak sites, because of their greater abundance, especially if the experimental contaminant concentration are not in the tracer region.


277

The predictive modeling was performed with the code FITEQL, Version 3.2 (Herbelin and Westall, 1996). The blind prediction example is based on experimental raw data published by Ali and Dzombak (1996b). In their CO2-free batch sorption experiments (1.6 g goethite/l) they obtained 64 data points with an experimental error of 4% by variation of pH

3:7 6:5

ionic strength

0:01=0:1 mol=L NaNO3

total CuðIIÞ concentration 2:3=23=98 mmol=L: The specific surface area depends on the sample preparation and cannot be generalized. As a result, the experimentally determined value for surface area should be used for the system definition in the blind predictions, too. The experimental value for the specific surface area of goethite, used for the predictive modeling, was 79.4 m2/g (Ali and Dzombak, 1996b). A search utilizing RES3T resulted in 16 independent data records for the goethite surface protolysis related to DDLM. The values of pK1 and pK2 for the two successive protolysis steps refer to the following deprotonation reactions: þ QFeOHþ 2 $ QFeOH þ H

pK 1 ¼ log

½ QFeOH½Hþ ½ QFeOHþ 2

(10.3)

QFeOH$ QFeO þ Hþ

pK 2 ¼ log

½ QFeO ½Hþ ½ QFeOH

(10.4)

After normalization of the surface site densities according to Eq. (10.1) and with G0 set to 2.31 sites/nm2, none of these pK values showed obvious inconsistencies or differed significantly enough from the bulk to become suspect. Thus, averaging seemed to be appropriate. Of course, for the blind prediction of the experimental data of Ali and Dzombak (1996b) the protolysis constants (and subsequently the complexation constants as well) derived by these authors were excluded – otherwise it would not have been a proper blind prediction. Table 10.1 (modified and updated after Richter et al., 2005a) shows the originally published protolysis constants of goethite as contained in the database, together with the values after normalization and extrapolation to infinite dilution, as described in the previous section. The selected values for the DDLM are pK1 ¼ 7.0070.35 and pK2 ¼ 9.3970.55. The errors given correspond to two standard deviations of the mean (95% level of significance).

Atkinson et al. (1967) Hayes et al. (1991) Mesuere and Fish (1992) Stone et al. (1993) van Geen et al. (1994) Lumsdon and Evans (1994) Turner and Sassman, (1996) Karltun (1997) Robertson and Leckie (1997) Robertson and Leckie (1997) Boult et al. (1998) Missana et al. (2003) Buerge-Weirich et al. (2003) Cohen and Waite (2004) Peacock and Sherman (2004) Naveau et al. (2005) Mean72rb a

Ionic strength (mol/L)

Ga (sites/ nm2)

pK1

pK2

pK1 (norm, IS ¼ 0)b

pK2 (norm, IS ¼ 0)b

0.1 0 0 0 0 0.01 0 0 0 0 0.01 0c 0.01 0 0c 0

0.56 10.00 1.50 6.50 2.31 2.74 2.31 4.2 7.00 2.31 2.27 2.20 1.3 10 6.02 1.8

6.99 7.10d 7.90 6.00 7.91 7.39 7.35f 6.30 6.91 7.72 6.2 7.20 5.6 6.4 6.78 6.53

8.40 10.24e 10.70 9.80 10.02 11.04 9.17g 8.92 10.80 10.09 8.2 10.00 8.9 7.6 10.10 7.54

6.480 7.74 7.71 6.45 7.91 7.51 7.35 6.56 7.39 7.72 6.24 6.96 5.40 7.04 7.20 6.42

9.12 9.60 10.89 9.35 10.02 11.01 9.17 8.66 10.32 10.09 8.25 10.24 9.19 6.96 9.68 7.65

6.8970.34

9.4770.56

7.0070.35

9.3970.55

G – original surface site density. Mean7standard deviation of the mean (significance level 95%), rounded to two decimal places. c Extrapolation to zero ionic strength assumed, not explicitly cited. d s ¼ 70.013. e s ¼ 70.033. f s ¼ 70.11 (95% confidence interval). g s ¼ 70.08 (95% confidence interval). b


Reference

278

Table 10.1: Protolysis Constants of Goethite Derived from RES3T for the DDLM (Original Values and Normalized to SSD ¼ 2.31 Sites/nm2 and Extrapolated to Zero Ionic Strength), Modified and Updated After Richter et al. (2005a).


279

After choosing appropriate protolysis constants a realistic set of surface species has to be built. There are several different chemical models discussed in the literature, many of them just containing one single species. Furthermore, not all surface species proposed in the literature are actually supported by spectroscopic evidence, many of them are mere results from best fits of sorption isotherms. The surface species QFeO–Cu+ and QFeO–CuOH have been reported for several iron(hydr)oxides and various SCM, and these species have been shown to exist spectroscopically (more details in Richter et al., 2005a), thus they were selected as the principal surface species. QFeOH þ Cu2þ $ QFeO Cuþ þ Hþ log K ¼FeOCuþ ¼

½ QFeO Cuþ ½Hþ ½ QFeOH½Cu2þ

(10.5)

QFeOH þ Cu2þ þ H2 O$ QFeO CuOH þ 2Hþ log K ¼FeOCuOH ¼

½ QFeO CuOH½Hþ 2 ½ QFeOH½Cu2þ

(10.6)

Below the averaged logK72s for the selected Cu(II) species sorbed onto goethite are given, normalized and extrapolated to infinite dilution (see Table 10.2 for the original data). Only one DDLM value was found in the literature (Robertson and Leckie, 1998) for the complexation constant of the Table 10.2: Surface Complexation Constants for the Cu(II) Sorption onto Goethite Taken from RES3T for the DDLM (Original Values and Normalized to 2.31 Sites/nm2 and Extrapolated to Zero Ionic Strength). Reference

Surface species

0b

Robertson and Leckie (1998)a Weirich (2000) Buerge-Weirich et al. (2003) Mean72rc Robertson and Leckie (1998)a a

logKc

logK (norm., IS ¼ 0)c

2.65

1.32

1.38

2.3 1.3

1.9 0.71

2.03 0.60

1.3170.69

1.3470.83

6.49

6.43

Ionic Original strength G (sites/ (mol/L) nm2)

0.01 0.01 QFeO–Cu+ QFeO–CuOH

b

0

2.65

Weak sites. Extrapolation to zero ionic strength assumed, not explicitly cited. c Mean7standard deviation of the mean (significance level 95%), rounded to two decimal places. b

280


Table 10.3: Aqueous Speciation of Cu(II) and Auxiliary Species (Smith and Martell, 1997). Reaction

logK

Cu2++H2O2CuOH++H+ Cu2++2H2O2Cu(OH)2+2H+ + 2Cu2++2H2O2Cu2(OH)2+ 2 +2H 2+ 2+ 3Cu +4H2O2Cu3(OH)4 +4H+ H2O2H++OH

7.5 16.2 10.6 20.8 14.00

surface species QFeO–CuOH: logK ¼FeO2Cuþ ¼ 1:34 0:83

logK ¼FeOCuOH ¼ 6:43

The aqueous speciation of Cu(II) is rather complex and will therefore significantly influence the sorption. Thus it has to be considered properly in all modeling efforts. All data for the aqueous speciation of Cu(II) and necessary auxiliary species (Table 10.3) are taken from the NIST database (Smith and Martell, 1997). 10.3.3. Uncertainty Analysis of the Test Case To prepare for a check of the effects of using inconsistent electrostatics, the above outlined methodology was also applied for CCM and TLM. Published individual uncertainties are not taken into account because they are only partially accessible, with some of the values not being accurately defined. The normalization, extrapolation, and averaging procedure led to the following averages of pK72s (95% level of significance) for the SCM submodels CCM, DDLM, and TLM (see also Fig. 10.3a and b for the pK1 and pK2 distributions), respectively. CCM DDLM TLM

pK1 ¼ 6.6270.39 pK1 ¼ 7.0070.35 pK1 ¼ 6.7770.48

pK2 ¼ 9.3070.49 pK2 ¼ 9.3970.55 pK2 ¼ 10.1570.47

The normalization and extrapolation procedure reduced the data scatter only slightly, the remaining uncertainty of both pK1 and pK2 is still remarkable. Thus it seemed possible that the influence of the data scatter may also introduce a large uncertainty in predicted KD values. This assumption has been tested through an uncertainty analysis.

pK2 for goethite surface protolysis


12.0

12.0

11.5

11.5

11.0

11.0

10.5

10.5

10.0

10.0

9.5

9.5

9.0

9.0

8.5

8.5

8.0

8.0

7.5 7.0 6.5

281

7.5 CCM

DDLM

TLM

7.0 6.5

Figure 10.3: Averaged pK172s (a) and pK272s (b) of Goethite (Stars), Original Values from RES3T (Closed Symbols, Sorted in Ascending Order) and After Extrapolation to Infinite Dilution and Normalization to the 2.31 sites/nm2 (Open Symbols). First, datasets with a correct probability distribution must be generated. Because the protolysis constants cannot be varied independently, the point of zero charge (PZC) and DpK are used instead as uncorrelated basic parameters. They are calculated according to the following equations

282


(error propagation):

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi s2pK 1 þ s2pK 2 pK 1 þ pK 2 PZC ¼ 2 ¼ 8:20 0:33 2 4 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi DpK ¼ pK 2 pK 1 2 s2pK 1 þ s2pK 2 ¼ 2:38 0:65

(10.7) (10.8)

The true density distribution function for the pK and logK values is unknown, but in most cases a Gaussian is a robust estimator for this function. Thus, this distribution was used and a finite number of grid points that are distributed according to a Gaussian was computed. 1 pK 1ij ¼ ðPZC þ 2sPZC RNi Þ ðDpK þ 2sDpK RNj Þ 2

(10.9)

1 (10.10) pK 2ij ¼ ðPZC þ 2sPZC RNi Þ þ ðDpK þ 2sDpK RNj Þ 2 The calculation of the grid points in turn requires the generation of (pseudo-)random numbers (RN), and here a routine was adapted from Press et al. (1992) applying a method recommended by Park and Miller (1988). A representative number of random numbers was generated (range 2.5y2.0), which conforms to the aforedefined Gaussian distribution (in our case 20 pK sets with 40 random numbers). In contrast to the protolysis constants, the particular complexation constants were varied independently according to the following equations: QFeO Cuþ : logK ¼FeOCuþ ;i ¼ logK ¼FeOCuþ þ ð2slogK ¼FeOCuþ RNi Þ (10.11) QFeO CuOH : logK ¼FeOCuOH;j ¼ logK ¼FeO2CuOH þ ð2slogK ¼FeOCuOH RNj Þ (10.12) Also here, the RN were computed following Press et al. (1992). Based on this, 30 logK datasets with the aid of 60 Gaussian-distributed random numbers were generated.

10.4. Results and Discussion A detailed description of the blind prediction modeling results is provided in our previous paper (Richter et al., 2005a). In Fig. 10.4 the predicted trend of Cu(II) sorption onto goethite is compared with the experimental values


283

(logKD and percentages, respectively) for three total Cu concentrations. The prediction quality is, in general, very satisfactory. Figure 10.5 shows the overall picture, i.e., the difference between experimental and predicted distribution coefficients logKD for 60 of the 64 data points. For the region with very high surface loadings (above 99.5%) small deviations between analytical determinations and model predictions are translated into high KD discrepancies. For this reason, surface loadings above 99.9% have not been taken into consideration, even though the prediction quality is nevertheless satisfactory, even in this region. A systematic overestimation of the logKD in particular at 98 mM may be either related to an inadequate description of the aqueous speciation of Cu(II) or problems with the correct determination of the surface site concentration. Some of the aqueous formation complexes are perhaps too weak (or missing). In addition, during the original parameter fitting the set of aqueous Cu(II) species differed among the authors. Concerning the number of surface binding sites, the surface area determined by N2-BET is perhaps higher than the effective surface really available for Cu(II) ions in aqueous medium. For more details see our previous paper (Richter et al., 2005a).

6

I=0.01 M NaNO3

[Cu T]=2.3 µM [Cu T]=23 µM [Cu T]=98 µM

5

100 80

3

% Cuadsorbed

log KD

4

2

60 40 20

1

0 3.5

4.0

4.5

5.0

5.5

6.0

pH

0 3.5

4.0

4.5

5.0

5.5

6.0

6.5

7.0

7.5

8.0

pH

Figure 10.4: Predicted logKD of Cu(II) Sorbed onto Goethite (Lines) Compared to the logKD of the Raw Data of Ali and Dzombak (1996b) (Symbols) at I ¼ 0.01 M NaNO3 (Inserted Figure: Predicted Percentages Compared to the Raw Data).

284


Figure 10.5: Differences in logKD between Measured and Predicted Cu(II) Sorption onto Goethite, Grouped by Experimental Conditions.

We applied the NEA Sorption Project criteria for the model quality assessment, namely, the deviation in KD should be below one order of magnitude to consider the blind prediction as reasonable for performance assessment software. As another, more condensed, expression of the prediction quality (QP), the sum of the absolute difference between predicted (result of FITEQL modeling) and experimental distribution coefficient has been computed. QP ¼

X jD logK D j n

¼ 0:232 ðn ¼ 60Þ

(10.13)

Figure 10.6 shows the results of the protolysis constants uncertainty analysis, based on the prediction quality computed for each of the 20 parameter variations. Based on the Gaussian distribution, the edges of the contour map are less occupied than the center, because here the number of supporting points is lower due to the lower probability. The respective values of QP are illustrated by the isolines of the plot, with the smallest value at 0.233 and the largest value at 0.461. Obviously, for the entire examined parameter space the blind prediction is still quite good. As in the former application case, Np(V) sorption onto hematite (Richter et al., 2005b), none of the randomly generated pK parameter sets delivered unacceptable blind predictions for the


285

Figure 10.6: Randomly Distributed pK of Goethite (Circle: Varied pK; Square, Diamond, Triangle: Experimental Mean pK Related to DDLM, CCM, or TLM, Respectively) Plotted in a Surface of the Resulting Sabs(Dlog KD)/n (see Legend). distribution coefficients. Thus the quite large spread of the pK values is actually not critical. What are the results when applying the mean pK for the other surface complexation models CCM or TLM? Even the use of inconsistent values from other electrostatic models will yield similar results (CCM: QP ¼ 0.294, TLM: QP ¼ 0.257). In conclusion, if the data situation is sparse, pK can still be used even though they were derived from calculations with other SCM. The uncertainty analysis results for the surface complexation constants of Cu sorption onto goethite are shown in Fig. 10.7. The supporting points were computed based on Eqs. (10.11) and (10.12). For Eq. (10.12) we used a standard deviation of 0.83 which is identical to that of the surface complexation constant of the QFeO–Cu+ surface complex. This decision was based on the fact that no information of the respective uncertainty was available since the logK for Eq. (10.12) was based on a single value. The randomly distributed surface complexation constants of Cu(II) onto goethite for the DDLM plotted in a surface of the resulting prediction quality QP (see numbers in the legend) with the smallest value at 0.219 and the largest value at 1.903. The predictions are more sensitive with respect to logK uncertainties. But most of the randomly generated logK parameter sets

286


Figure 10.7: Randomly Distributed logK of Goethite Surface Complexes (Circle: Varied logK; Square, Diamond, Triangle: Experimental Mean logK Related to DDLM, CCM, or TLM, Respectively; Star: Experimental Mean logK of Ferrihydrite Surface Complexes) Plotted in a Surface of the Resulting Sabs(DlogKD)/n (see Legend). delivered acceptable blind predictions with a prediction quality QP less than one (bold contour line). Using the inconsistent values from CCM (diamond) log K ¼FeOCuþ ¼ 2:08 2:00a a

log K ¼FeOCuOH ¼ 6:52b

Palmqvist et al. (1999); Weirich (2000). Palmqvist et al. (1999).

b

we obtain a similar result (QP ¼ 0.645). However, it must be noted that the data situation is very sparse in this case. A blind prediction applying the TLM values (triangle) log K ¼FeO2Cuþ ¼ 1:54 2:16a

log K ¼FeO2CuOH ¼ 7:55 0:82b

a

Balistrieri and Murray (1982); Coughlin and Stone (1995); Kooner et al. (1995); Jung et al. (1998); Robertson and Leckie (1998). b Balistrieri and Murray (1982); Jung et al. (1998); Robertson and Leckie (1998).

yields acceptable blind predictions (QP ¼ 0.232), too. As with the protolysis constants, the following can be concluded: If the data situation is sparse, the use of logK derived from calculations with other SCM types is possible.


287

What could be done if there would be no data for the system goethite/ copper available at all? In Fig. 10.7 the result for the Cu(II) sorption onto ferrihydrite (as a reasonable chemical analog to goethite) based on DDLM (star) is shown. log K ¼FeO2Cuþ ¼ 0:41 0:63a a

log K ¼FeO2CuOH ¼ 8:61b

Dzombak and Morel (1990); Karthikeyan and Elliott (1999); Swedlund and Webster (2001). Swedlund and Webster (2001).

b

The prediction quality QP ¼ 1.042 is still acceptable, but of course the result has to be treated with caution. Ferrihydrite is amorphous and has sorption characteristics (larger specific surface and more unsaturated surface groups causing a higher reactivity) different from the other ferric oxides.

10.5. Summary and Conclusions When focusing on the conventional distribution coefficient KD, the blind prediction of Cu(II) sorption onto goethite based on the DDLM showed that the simulation for all experimental data subsets are within one order of magnitude. This is reasonable for PA software. In a second step, the influence of varying goethite protolysis constants (DDLM) was investigated. Variations were based on the probability distribution found in the literature. The respective uncertainty analysis demonstrated that the blind prediction within a range of two standard deviations is still of good quality. The result is consistent with former investigations (Richter et al., 2005b) and increases confidence in the blind prediction approach as presented by the authors. The uncertainty analysis of the surface complexation constants (again for DDLM) demonstrated out that the predictions of the Cu(II) sorption onto goethite are more sensitive with respect to the logK uncertainties. However, most of the computations yielded a prediction quality less than 1.0. Going one step further, we addressed the problem of sparse SCM data matrices. A comparison of the predictions utilizing CCM and TLM parameter sets from the literature with the internally consistent case of DDLM showed that the prediction quality is still within the uncertainty range of the pure DDLM approach. Hence we might say that inconsistencies of parameters linked to electrostatic models other than DDLM may be tolerated. The use of SCM parameters from chemically similar/analogous systems is an alternative to a relaxation of electrostatic consistency. As an example,

288


ferrihydrite as a substitution for goethite was investigated. This proved not to be a promising approach. However, to generalize this statement more work involving other iron minerals is needed. This work, being the most comprehensive uncertainty analysis of surface complexation models so far, supplied rather promising results. Based on the current state of knowledge we summarize as follows: The often-encountered problem of sparse SCM data matrices, or the confrontation with parameters having large errors and deviating considerably from each other, can be easily circumvented if the electrostatic consistency is sacrificed. This may help with more realistic (thermodynamic) approaches to amend the KD concept. The SCM approach seems to be very promising for estimating KD values given well-defined mineral systems with a suitable database. The SCM database that is assembled within the RES3T project is able to provide the respective parameter sets following the stepwise strategy of species selection, data collection, normalization, and averaging. The use of spectroscopic tools such as EXAFS (extended X-ray absorption fine structure), NMR (nuclear magnetic resonance), FTIR (Fourier transform infrared), TRLFS (time-resolved laser-induced fluorescence spectroscopy), or LIPAS (laser-induced photoacoustic spectroscopy) are required to resolve questions of inconsistent species sets. Additionally, quantum chemical modeling may be helpful. This problem has been recognized by the scientific community, resulting in a steadily increasing number of papers dealing with structural investigations on surfaces and respective proofs of evidence for surface complexes.

ACKNOWLEDGMENTS The development of the mineral-specific sorption database RES3T was funded by the German Federal Ministry of Economics and Labour (BMWA) under contract No. PtWt+E 02E9471, which is gratefully acknowledged. We thank three anonymous reviewers and Mark Barnett, whose comments and suggestions improved the final version of this manuscript.

REFERENCES Ali, M. A., & Dzombak, D. A. (1996a). Competitive sorption of simple organic acids and sulfate on goethite. Environ. Sci. Technol., 30, 1061–1071. Ali, M. A., & Dzombak, D. A. (1996b). Effects of simple organic acids on sorption of Cu2+ and Ca2+ on goethite. Geochim. Cosmochim. Acta, 60, 291–304.


289

Arnold, T., Zorn, T., Za¨nker, H., Bernhard, G., & Nitsche, H. (2001). Sorption behavior of U(VI) on phyllite: Experiments and modeling. J. Contam. Hydrol., 47, 219–231. Atkinson, R. J., Posner, A. M., & Quirk, J. P. (1967). Adsorption of potentialdetermining ions at the ferric oxide-aqueous electrolyte interface. J. Phys. Chem., 71, 550–558. Balistrieri, L. S., & Murray, J. W. (1982). The adsorption of Cu, Pb, Zn, and Cd on goethite from major ion seawater. Geochim. Cosmochim. Acta, 46, 1253–1265. Boult, K. A., Cowper, M. M., Heath, T. G., Sato, H., Shibutani, T., & Yui, M. (1998). Towards an understanding of the sorption of U(VI) and Se(IV) on sodium bentonite. J. Contam. Hydrol., 35, 141–150. Brendler, V., Vahle, A., Arnold, T., Bernhard, G., & Fangha¨nel, T. (2003). RES3T – rossendorf expert system for surface and sorption thermodynamics. J. Contam. Hydrol., 61, 281–291. Buerge-Weirich, D., Behra, P., & Sigg, L. (2003). Adsorption of copper, nickel, and cadmium on goethite in the presence of organic ligands. Aquat. Geochem., 9, 65–85. Buerge-Weirich, D., Hari, R., Xue, H., Behra, P., & Sigg, L. (2002). Adsorption of Cu, Cd, and Ni on goethite in the presence of natural groundwater ligands. Environ. Sci. Technol., 36, 328–336. Cohen, D. R., & Waite, T. D. (2004). Interaction of aqueous au species with goethite, smectite and kaolinite. Geochem. Expl. Environ. Anal., 4, 279–287. Coughlin, B. R., & Stone, A. T. (1995). Nonreversible adsorption of divalent metal ions (Mn(II), Co(II), Ni(II), Cu(II), and Pb(II)) onto goethite: Effects of acidification, Fe(II) addition, and picolinic acid addition. Environ. Sci. Technol., 29, 2445–2455. Davies, C. W. (1962). Ion Association. Butterworths, Washington. Davis, J. A., Coston, J. A., Kent, D. B., & Fuller, C. C. (1998). Application of the surface complexation concept to complex mineral assemblages. Environ. Sci. Technol., 32, 2820–2828. Davis, J. A., & Kent, D. B. (1990). Surface complexation modeling in aqueous geochemistry. In: M. F. Hochella, & A. F. White (Eds). Mineral-Water Interface Geochemistry. Mineralogical Society of America, Washington, DC, pp. 177–258. Davis, J. A., Ochs, M., Olin, M., Payne, T. E., & Tweed, C. J. (2005). NEA Sorption Project Phase II. Interpretation and Prediction of Radionuclide Sorption onto Substrates Relevant for Radioactive Waste Disposal using Thermodynamic Sorption Models. NEA Report 5992, OECD Publishing, Paris, 286 pp. Dzombak, D. A., & Morel, F. M. M. (1990). Surface Complexation Modeling. Hydrous Ferric Oxide. Wiley, New York. van Geen, A., Robertson, A. P., & Leckie, J. O. (1994). Complexation of carbonate species at the goethite surface: Implications for adsorption of metal ions in natural waters. Geochim. Cosmochim. Acta, 58, 2073–2086. Hayes, K. F., Redden, G., Ela, W., & Leckie, J. O. (1991). Surface complexation models: An evaluation of model parameter estimation using FITEQL and oxide mineral titration data. J. Colloid Interface Sci., 142, 448–469.

290


Herbelin, A. L., & Westall, J. C. (1996). FITEQL -Version 3.2. Report 96-01, Department of Chemistry, Oregon State Univ. Honeyman, B. D. (1984). Cation and Anion Adsorption at the Oxide/Solution Interface in System Containing Binary Mixtures of Adsorbents: An Investigation of the Concept of Adsorptive Additivity. PhD Thesis, Stanford University, Palo Alto, CA. Jung, J. H., Cho, Y. H., & Hahn, P. (1998). Comparative study of Cu2+ adsorption on goethite, hematite and kaolinite: Mechanistic modeling approach. Bull. Korean Chem. Soc., 19, 324–327. Jung, J. H., Hyun, S. P., Lee, J. K., Cho, Y. H., & Hahn, P. (1999). Adsorption of on natural composite materials. J. Radioanal. Nucl. Chem., 242, 405–412. UO2+ 2 Karltun, E. (1997). Modelling SO2 4 surface complexation on variable charge minerals: I. H+ and SO2 4 exchange under different solution conditions. Eur. J. Soil Sci., 48, 483–491. Karthikeyan, K. G., & Elliot, H. A. (1999). Surface complexation modeling of copper sorption by hydrous oxides of iron and aluminum. J. Colloid Interface Sci., 220, 88–95. Kooner, Z. S. (1992). Adsorption of Cu(II) onto goethite in aqueous systems. Environ. Geol. Water Sci., 20, 205–212. Kooner, Z. S., Cox, C. D., & Smoot, J. L. (1995). Prediction of adsorption of divalent heavy metals at the goethite/water interface by surface complexation modeling. Environ. Toxicol. Chem., 14, 2077–2083. Kulik, D. A. (2002). Sorption modelling by gibbs energy minimisation: Towards a uniform thermodynamic database for surface complexes of radionuclides. Radiochim. Acta, 90, 815–832. Lumsdon, D. G., & Evans, L. J. (1994). Surface complexation model parameters for goethite (a-FeOOH). J. Colloid Interface Sci., 164, 119–125. Lu¨tzenkirchen, J. (2002). Surface complexation models of adsorption. In: A. Hubbard (Ed). Encyclopedia of Surface and Colloid Science. Marcel Dekker, New York, pp. 5028–5046. Mesuere, K., & Fish, W. (1992). Chromate and oxalate adsorption on goethite. 1. Calibration of surface complexation models. Environ. Sci. Technol., 26, 2357–2364. Missana, T., Garcı´ a-Gutie´rrez, M., & Maffiotte, C. (2003). Experimental and modeling study of the uranium(VI) sorption on goethite. J. Colloid Interface Sci., 260, 291–301. Naveau, A., Monteil-Rivera, F., Dumonceau, J., & Boudesocque, S. (2005). Sorption of europium on a goethite surface: Influence of background electrolyte. J. Contam. Hydrol., 77, 1–16. Palmqvist, U., Ahlberg, E., Lo¨vgren, L., & Sjo¨berg, S. (1999). Competitive metal ion adsorption in goethite systems using in situ voltammetric methods and potentiometry. J. Colloid Interface Sci., 218, 388–396. Park, S. K., & Miller, K. W. (1988). Random number generator: Good ones are hard to find. Commun. ACM, 31, 1192–1201.


291

Payne, T. E., Davis, J. A., Ochs, M., Olin, M., & Tweed, C. J. (2004). Uranium adsorption on weathered schist – intercomparison of modelling approaches. Radiochim. Acta, 92, 651–661. Peacock, C. L., & Sherman, D. M. (2004). Copper (II) sorption onto goethite, hematite and lepidocrocite: A surface complexation model based on ab initio molecular geometries and exafs spectroscopy. Geochim. Cosmochim. Acta, 68, 2623–2637. Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. (1992). Numerical Recipes in C – The Art of Scientific Computing. 2nd Ed., Cambridge University Press, Cambridge. Richter, A., Brendler, V., & Bernhard, G. (2005a). Blind prediction of Cu(II) sorption onto goethite: Current capabilities of diffuse double layer model. Geochim. Cosmochim. Acta, 69, 2725–2734. Richter, A., Brendler, V., & Nebelung, C. (2005b). The effect of parameter uncertainty on blind prediction of Np(V) sorption onto hematite using the diffuse double layer model. Radiochim. Acta, 93, 527–531. Robertson, A. P., & Leckie, J. O. (1997). Cation binding predictions of surface complexation models: Effect of pH, ionic strength, cation loading, surface complex, and model fit. J. Colloid Interface Sci., 188, 444–472. Robertson, A. P., & Leckie, J. O. (1998). Acid/base, copper binding, and Cu2+/H+ exchange properties of goethite, an experimental and modeling study. Environ. Sci. Technol., 32, 2519–2530. Smith, R. M., & Martell, A. E. (1997). NIST – Critically Selected Stability Constants of Metal Complexes Database-Version 4.0. Report, US Department of Commerce. Stone, A. T., Torrents, A., Smolen, J., Vasudevan, D., & Hadley, J. (1993). Adsorption of organic compounds possessing ligand donor groups at the oxide/ water interface. Environ. Sci. Technol., 27, 895–909. Stumm, W. (1992). Chemistry of the Solid-Water Interface. Wiley, New York. Sverjenski, D. A., & Sahai, N. (1996). Theoretical prediction of single-site surface protonation equilibrium constants for oxides and silicates in water. Geochim. Cosmochim. Acta, 60, 3773–3797. Swedlund, P. J., & Webster, J. G. (2001). Cu and Zn ternary surface complex formation with SO4 on ferrihydrite and schwertmannite. Appl. Geochem., 16, 503–511. Turner, D. R., & Sassman, S. A. (1996). Approaches to sorption modeling for highlevel waste performance assessment. J. Contam. Hydrol., 21, 311–332. Waite, T. D., Davis, J. A., Payne, T. E., Waychunas, G. A., & Xu, N. (1994). Uranium(VI) adsorption to ferrihydrite: Application of a surface complexation model. Geochim. Cosmochim. Acta, 58, 5465–5478. Weirich, D. (2000). Influence of Organic Ligands on the Adsorption of Copper, Cadmium, and Nickel on Goethite. PhD Thesis, ETH Zu¨rich. Westall, J. C., & Hohl, H. (1980). A comparison of electrostatic models for the oxide/solution interface. Adv. Colloid Interface Sci., 12, 265–294. Wilhelm, R. G., & Beam, P. (1999). Understanding Variation in Partition Coefficient, Kd, Values. EPA Report 402-R-99-004A, Washington.


Chapter 11

Biogeochemical Uranium Redox Transformations: Potential Oxidants of Uraninite Matthew Ginder-Vogel1,2 and Scott Fendorf1, 1

Department of Geological and Environmental Sciences, Stanford University, Stanford, CA 94305-2115, USA 2 Present Address: Department of Plant and Soil Sciences, University of Delaware, Newark, DE 19716, USA

ABSTRACT In aerobic environments, uranium is generally found in the hexavalent oxidation state, is quite soluble, and readily forms complexes with calcium and carbonate. However, under anaerobic conditions, common metal respiring bacteria can enzymatically reduce U(VI) to U(IV), resulting in the formation of sparingly soluble UO2 (uraninite). Uranium(VI) reduction, therefore, has a prominent role in natural attenuation of uranium and is being explored as a potential remediation option for this hazardous element. The stability of biologically precipitated uraninite is critical for determining the long-term fate of uranium and is not well characterized within soils and sediments. Given their environmental predominance, Fe(III) (hydr)oxides, which act as both U(VI) reductants and U(IV) oxidants, exert a prominent role in determining uranium chemical fate and associated mobility. Here, we examine several factors controlling the extent of uraninite oxidation by Fe(III) (hydr)oxides, including iron oxide and bicarbonate concentration, in addition to iron oxide type. Our analysis reveals that the extent of uraninite oxidation increases under conditions that increase the thermodynamic probability of the reaction; however, recrystallization of poorly crystalline Fe(III) (hydr)oxides to more crystalline forms may ultimately limit the uraninite oxidation reaction. Nevertheless, uraninite oxidation by Fe(III) (hydr)oxides may be a limiting factor on uraninite stability in the environment.

Corresponding author. Tel.: +1 650 723 5238; Fax: +1 650 725 2199;

E-mail: [email protected] (S. Fendorf).

294

M. Ginder-Vogel and S. Fendorf

11.1. Introduction The release of radionuclides, toxic heavy metals, and organic co-contaminants into the environment during the nuclear age poses a unique long-term environmental problem (Riley et al., 1992; Ginder-Vogel et al., 2005). Among these contaminants, uranium is of particular concern because of its carcinogenicity, long half-life, widespread distribution, and mobility (Riley et al., 1992). Uranium contamination of ground and surface waters has been detected at numerous sites throughout the world, including agricultural evaporation ponds (Bradford et al., 1990), nuclear weapons manufacturing areas, and mine tailings sites (Riley et al., 1992). The mobility of uranium within surface and subsurface environments is determined, in part, by its geochemical speciation. In oxic environments, uranium is generally present in the hexavalent oxidation state as the uranyl [U(VI)O2+ 2 ] species and under most environmental conditions is quite soluble – the solubility of U(VI) is particularly enhanced by complexation with carbonate, a common groundwater ligand (Grenthe et al., 1992). However, U(VI) also forms several sparingly soluble complexes with phosphate (Langmuir, 1978; Sandino and Bruno, 1992) and readily forms inner-sphere complexes with many transition metal hydroxides – absorption being most extensive at neutral pH (Barnes and Cochran, 1993; Barnett et al., 2000; Moyes et al., 2000; Bostick et al., 2002; Davis et al., 2004; Curtis et al., 2006). Nevertheless, U(VI) tends to be relatively soluble and thus subject to migration within ground and surface water. Conversely, U(IV) is sparingly soluble, even in the presence of common groundwater ligands such as carbonate, and thus tends to be relatively immobile. Therefore, the oxidation state of uranium, which may be controlled by a wide variety of biogeochemical processes (Fig. 11.1), will play an important role in determining its mobility in surface and subsurface environments. In fact, in situ transformation of mobile U(VI) species into immobile U(IV) uraninite is being explored as a possible uranium remediation technique (Wu et al., 2006a,b, 2007).

11.2. Uranium Oxidation–Reduction Reactions A host of reduction pathways exist for U(VI) and include both abiotic and biotic reactions. Abiotic (chemical) reduction of U(VI) may proceed via several pathways, albeit typically under a limited set of conditions, within low temperature geochemical systems (Fig. 11.1). Sulfide minerals, for

Biogeochemical Uranium Redox Transformations

295

Fe(III) Fe(II) Fe(III) Fe(II)

U(VI) (Soluble)

H2, OCreduced DMRB CO2, H2O, OCoxidized

Mn(IV)

H2, OCreduced SRB CO2, H2O, OCoxidized

Mn(II) 2-

0

S , SO4 HS-

-

NO, N2O, NO2

-

HS N2

2-

SO4

≡Fe(II)

T. denitrificans and others

Fe(III)

-

NO3

O2

AH2DS

N2 H2O

U(IV) (Uraninite)

AQDS

Figure 11.1: Biological and Abiotic Processes that Affect the Redox State of Uranium. Ovals Represent Biologically Catalyzed Processes and Include Uranium Reduction by Dissimilatory Metal-Reducing Bacteria (DMRB) and Sulfate-Reducing Bacteria (SRB). Other Abbreviations Include Surface Bound Iron (Fe), Reduced Organic Carbon (OCreduced), Oxidized Organic Carbon (OCoxidized), 9, 10-Anthraquinone-2, 6-Disulfonic Acid (AQDS), and Reduced AQDS (AH2DS). example, are commonly found in association with supergene uranium deposits, suggesting sulfide reduction of U(VI) (Langmuir and Chatman, 1980; Nash et al., 1981). Additionally, laboratory studies have demonstrated that partial U(VI) reduction by sulfide mineral surfaces occurs with the concomitant production of polysulfides (Wersin et al., 1994; Livens et al., 2004). Furthermore, complete reduction of dissolved U(VI) to microcrystalline uraninite by aqueous sulfide occurs at low pH and low bicarbonate concentrations, where the predominant species of U(VI) is UO2þ 2ðaqÞ – the most reactive uranyl species toward sulfide (Hua et al., 2006). Although Fe2þ ðaqÞ does not appear to be a facile reductant of U(VI), adsorbed Fe(II) and ferrousbearing minerals are kinetically viable reductants of U(VI). Uranium(VI) reduction by ferrous iron bound to the surface of microcrystalline hematite, goethite, smectite, and natural solids, has been observed at near-neutral pH (Liger et al., 1999; Jeon et al., 2005), as has reduction by hydroxysulfate

296


green rust (O’Loughlin et al., 2003). At low pH (o5), U(VI) is reduced to U(IV) in the interlayer of ferrous iron-bearing micas (Ilton et al., 2004, 2006), and partial uranyl reduction by magnetite is also observed (El Aamrani et al., 2000; Missana et al., 2003; Scott et al., 2005). In contrast to abiotic U(VI) reduction, numerous common dissimilatory metal-reducing bacteria (DMRB) and sulfate-reducing bacteria (SRB) couple the oxidation of organic matter and H2 to the reduction of U(VI), resulting in U(IV) and the subsequent precipitation of uraninite (UO2) (Gorby and Lovley, 1992; Fredrickson et al., 2000), a sparingly soluble phase. However, the formation of the tenary Ca2UO2(CO3)3(aq) species kinetically limits biological uranium reduction (Brooks et al., 2003; Stewart et al., 2007). Additionally, the presence of nitrate or Fe(III) (hydr)oxides as alternate electron acceptors and potential U(IV) oxidants impedes biological U(VI) reduction (Wielinga et al., 2000; Stewart et al., 2007). Although DMRB are capable of reducing solid Fe(III) (hydr)oxides, solid U(VI) is not available for biological reduction. The reduction of NaBoltwoodite (NaUO2SiO3OH 1.5H2O) by Shewanella oneidensis strain MR-1 requires the sequential coupling of U(VI) dissolution with microbial reduction (Liu et al., 2006). Sediment-bound U(VI) also appears to be unavailable for microbial reduction (Ortiz-Bernad et al., 2004). However, because U(VI) reduction occurs concurrently with Fe(III) and SO2 4 reduction, systems with active populations of DMRB and SRB should contain abundant Fe(II) and HS1, both of which can abiotically reduce U(VI) sorbed to synthetic and natural iron (hydr)oxides (Liger et al., 1999; Jeon et al., 2005). Bacterial uranium reduction may also occur via the reduction of electron shuttling compounds, such as quinones, which then reduce U(VI) to U(IV) (Fredrickson et al., 2000; Nevin and Lovley, 2000). In order to determine the viability of in situ biological uranium remediation and to discern the role of reductive processes in natural uranium cycling, it is critical to determine the stability of biologically precipitated uraninite, and in particular, it is important to identify environmentally relevant oxidative processes (Fig. 11.1). Potential UO2 oxidants include molecular oxygen, nitrate, nitrate reduction intermediates, Mn(IV) (hydr)oxides, and Fe(III) (hydr)oxides (Figs. 11.1 and 11.2). Additionally, UO2 oxidation may be catalyzed by biological activity. The oxidation of biologically precipitated uraninite by molecular oxygen remains poorly characterized; however, the oxidative dissolution of synthetic uraninite is relatively well characterized. In those studies, the uraninite grains examined were generally 100 mm in diameter (Torrero et al., 1997; Pierce et al., 2005), rather than 10 nm in diameter, as observed for biologically precipitated uraninite (Fredrickson et al., 2000; Singer et al., 2006). However, such


EH (pH = 7)

Oxidized

Reduced

Oxidized

Reduced

297

pε (pH = 7)

[volt] 1.0 15 O2 NO3-

0.8

H2O N2 10

0.6

MnO2(s) NO3-

0.4

MnCO3 NO25

0.2 0 -0.2

UO22+

UO2

2-

UO2 UO2

UO2(CO3)2 Ca2UO2(CO3)3

0 Fe(OH)3

Fe

2+

-5

-0.4 [U(VI)] = 1 x 10-6 M

[Fe(II)] = 1 x 10-5 M

Figure 11.2: Representative Redox Couples for Dominant Constituents within Soils/Sediments and Their Comparison to U(IV)/U(VI). studies still provide insight into the oxidation mechanism of, and possible geochemical limits on, uraninite oxidation by molecular oxygen. Below pH 7, the oxidation rate of uraninite increases as pH decrease, and an increase in either the bicarbonate or O2(aq) concentration results in an increase in the oxidation rate (Torrero et al., 1997; Peper et al., 2004; Pierce et al., 2005). Above pH 7 and in the presence of low bicarbonate concentrations, the uraninite oxidation rate decreases due to the precipitation of U(VI)-phases on the uraninite surface (Torrero et al., 1997; Peper et al., 2004; Pierce et al., 2005). Nitrate, a common co-contaminant with uranium (Riley et al., 1992), not only impedes biological uranium reduction (Finneran et al., 2002; Senko et al., 2002; Istok et al., 2004), but also may oxidize U(IV). Uraninite oxidation by nitrate is a thermodynamically favored process under environmental conditions (Fig. 11.2); however, it is rate-limited (Senko et al., 2005a,b). The biological transformation of NO 3 into NO2 , NO, and N2O increases the oxidation rate; however, the rate remains quite slow as compared to oxidation by Fe(III) and O2 (Senko et al., 2005a). Additionally, Thiobacillus denitrificans and Geobacter metallireducens are capable of catalyzing nitratedependent U(IV) oxidation, although it is currently not known if the bacteria obtain energy from this process (Finneran et al., 2002; Beller, 2005).

298


Furthermore, the indirect oxidation of U(IV) by nitrite, through production of Fe(III) (hydr)oxides, also increase this reaction rate (Senko et al., 2005a,b). In addition to their role as intermediates in U(IV) oxidation by nitrite, Fe(III) (hydr)oxide minerals have been implicated in U(IV) oxidation under sulfate-reducing (Sani et al., 2004, 2005) and methanogenic conditions (Wan et al., 2005). The redox couples for U(IV)/U(VI) and Fe(III) (hydr)oxide/ Fe(II) occur at similar potentials under common groundwater conditions (Fig. 11.3); therefore, small changes in aqueous and solid-phase chemistry can result in UO2 oxidation by Fe(III) (hydr)oxide oscillating between thermodynamic viability and nonviability (Fig. 11.3) (Ginder-Vogel et al., 2006). However, acidic, ferric iron-containing solutions rapidly oxidize and dissolve U(IV) from uranium ore minerals (Harrison et al., 1966; Vuorinen et al., 1985). This reaction is further enhanced by the activity of acidophilic, iron-oxidizing bacteria, such as Thiobacillus ferrooxidans (Dispirito and Tuovinen, 1982a,b). Iron(III) (hydr)oxides will likely play an important role in controlling the long-term stability of biologically precipitated uranium. Not only are they ubiquitous in soils and sediments (Cornell and Schwertmann, 2003), they accelerate uraninite oxidation by nitrite (Senko et al., 2005b) and are generated in environments undergoing redox cycling. However, the geochemical conditions conducive to uraninite oxidation by Fe(III) (hydr)oxides remain poorly constrained. In particular, the transformation of ferrihydrite into more thermodynamically stable Fe(III) (hydr)oxide minerals may ultimately limit uraninite oxidation. Accordingly, here we examine the potential impact of Fe(III) (hydr)oxide mineralogy on biogenic uraninite oxidation.

11.3. Experimental 11.3.1. Materials All acids were trace-metal grade, and chemicals were ACS grade or better. Anaerobic solutions used in this study were prepared using distilled deionized (DDI) water that had been treated to remove dissolved O2, by boiling, while purging with N2 gas that had been passed over hot Cu-metal filings. The water was cooled for 12 h, while being purged with N2 and immediately transferred to an anaerobic glovebox (Coy Laboratory Products) with a 95% N2 and 5% H2 atmosphere. All glassware and equipment were equilibrated in the anaerobic chamber for 24 h prior to use.


EH

Ox

(V)

[U(IV)] = 10-9 M [U(VI)] = 10-6 M

-0.10

UO2(CO3)2-2

Red

Ox

Red

299

pε

[Fe(II)] = 10-5 M

Fe(OH)3

UO2

Fe2+

-0.15

-2.5

+3

-0.20

UO2(CO3)2-2

U(OH)

-0.25

UO2(CO3)2-2

U(OH)4

UO2(CO3)2-2

U(CO3)4

γ-FeOOH

Fe2+

α-FeOOH

Fe2+

Fe2O3

Fe2+

-4

-0.30

-5.0

-0.35

-0.40

-0.45

-7.5 -6

UO2(CO3)2-2

U(CO3)5

UO2(CO3)2-2

U

-0.50

-0.55

-0.60

+4

-10.0

Figure 11.3: Representative Fe(III)/Fe(II) and U(VI)/U(IV) Redox Couples 6 at pH 7 with 3 103 M HCO M of Each U(VI) Species, 3 , 1 10 9 1 10 M of Each Dissolved U(IV) Species, and 1 105 M Fe(II).

300


11.3.2. Biogenic Uraninite Cell suspensions were prepared by growing Shewanella putrefaciens strain CN32 aerobically on Tryptic Soy Broth at 301C to late log phase. Cells were harvested by centrifugation (4,000 g, 5 min), washed twice in 100 mL of anaerobic bicarbonate buffer (24 mM KHCO3, pH 7), and resuspended in bicarbonate buffer. Uranium reduction was initiated by inoculating 1 L of anaerobic U(VI) reduction media (pH 7, 4 mM uranium acetate, 30 mM KHCO3, 10 mM PIPES, 3 mM NH4Cl, 40 mM lactate, and 10 mL Wolfe’s vitamins) with 100 mL of bacterial suspension (108 cells mL1). Media was then stirred continuously in an anaerobic glove box with an atmosphere of 95% N2 and 5% H2 (Coy Laboratory Products). After 4 days (d), the solids were collected and incubated in 10% NaOH for 3 d to digest cell material, washed three times in 24 mM KHCO3, and then washed twice in degassed, DDI water. X-ray diffraction (XRD) patterns were identical to those previously reported for biogenic UO2 (Fredrickson et al., 2000). The N2-BET surface area of the biogenic uraninite was 129 m2 g1.

11.3.3. Ferrihydrite Synthesis Two-line ferrihydrite was synthesized by rapidly titrating a ferric chloride solution with NaOH to a pH of 7.5 (Hansel et al., 2005). The Fe(III) (hydr)oxide flocs were washed by centrifugation twice with 1% HCl and three times with distilled H2O. The Fe(III) (hydr)oxide flocs were then resuspended and degassed by bubbling with N2 for 24 h. Oxide mineralogy and purity was confirmed with XRD. The N2-BET surface areas of the ferrihydrite, goethite, and hematite were 219, 80, and 60 m2 g1, respectively.

11.3.4. Uraninite Oxidation Experiments Ferrihydrite and uraninite were maintained as aqueous suspensions to avoid diminishing their reactivity and were added to each oxidation reaction as a slurry. Biogenic uraninite was used within 2 weeks, and ferrihydrite was used within 2 d of preparation. Changes in reactivity of the two solid phases were not observed over this time frame. A set of oxidation experiments was performed to examine the effect of ferrihydrite concentration on reaction extent and solid-phase Fe and U speciation during UO2 oxidation. These reactions were carried out in 1 L of


301

media in 1 L polypropylene Nalgene bottles with a single uraninite concentration of 30.7 m2 L1 and ferrihydrite concentrations of 23.5, 46.0, and 80.2 m2 L1. All reactions were continuously mixed, using an overhead stirrer at 75 rpm to avoid solid-phase abrasion, and, unless otherwise noted, were performed in 3 mM KHCO3 at pH 7.2. Ferrihydrite was added to the reaction media first, followed by uraninite, in less than 10 s for all reaction conditions studied. Initial and final pH of all reactions varied by less than 0.1 unit. The uraninite and ferrihydrite concentrations used in each experiment were determined by acidic dissolution of the reaction slurry and are denoted as m2 L1. 11.3.5. Sampling and Analytical Procedures The extent of biogenic UO2 oxidation by ferrihydrite was quantified, primarily using ferrozine extractable Fe(II) concentrations (Stookey, 1970). The system-partitioning coefficient for Fe(II) was used to determine total Fe(II), which was in excellent agreement with the stoichiometric amounts of U(VI) produced. Use of dissolved U(VI) to quantify the extent of reaction was less reliable because of the potential for variable U(VI) sorption on Fe(III) (hydr)oxides under the varying reaction conditions examined. Ferrozineextractable Fe(II) was used, rather than soluble or acid-extractable Fe(II), because (i) of the propensity for Fe(II) uptake by Fe(III) (hydr)oxides (Morrison et al., 1995; Moyes et al., 2000; Duff et al., 2002; Walter et al., 2003), and (ii) UO2(biogenic) oxidation by Fe(III) is more energetically favorable (Ginder-Vogel et al., 2006) and rapid under acidic conditions. In fact, after 24 h, we observe near-complete oxidation of 30.7 m2 L1 biogenic uraninite by 32.1 m2 L1 ferrihydrite in 0.5 M HCl. Extractable Fe(II) was determined by adding 1.5 mL of reaction slurry to 1.5 mL of ferrozine reagent and reacting for 20 s; the sample was then passed through a 0.2 mm polycarbonate filter and Fe(II) quantified by the absorbance at 562 nm. Prior to reaction, neither uraninite nor ferrihydrite slurries contained detectable quantities of Fe(II), as measured by this technique. Total iron and uranium were determined at the conclusion of each oxidation experiment by acidic dissolution of the reaction slurry with concentrated HNO3 and HCl and quantified with inductively coupled plasma-optical emission spectrometry (ICP-OES). Total dissolved iron and uranium were determined by passing 5 mL of reaction slurry through a 0.2 mm polycarbonate filter, which was then acidified and measured by ICP-OES. Soluble Fe(II) in the filtrate was measured by the ferrozine method, using 10 mM ferrozine, and soluble U(VI) was measured

302


spectroflourometrically. Uranium samples were diluted 1:30 in 10% phosphoric acid, and the fluorescence of the uranyl-phosphate complex was measured at 515.4 nm. All measurements were referenced to the fluorescence of the background matrix. 11.3.6. X-Ray Diffraction and X-Ray Absorption Spectroscopy XRD was used to identify crystalline iron phases after uraninite oxidation. XRD patterns were collected on beamline 11-3 of the Stanford Synchrotron Radiation Laboratory (SSRL) in transmission geometry, using monochromatic radiation (12732.137 eV) and a MAR 345 image plate. The resulting images were processed using FIT2D (Hammersley, 1997). The sampleto-detector distance and geometric corrections were calculated from the pattern of LaB6. After these corrections were applied, the 2D images were integrated radially to yield 1D powder diffraction patterns, which could then be analyzed using standard techniques. Peak identification and background correction, including removal of the scattering from the lexan window, were performed in JADE 6.5 (Materials Data, Inc., Livermore, CA). Samples were mounted in the anaerobic chamber between Lexan windows sealed with double-sided tape to limit sample oxidation during analysis. Extended X-ray absorption fine structure (EXAFS) spectroscopy was used to quantify the iron solid-phase distribution. Samples were powdered, using a mortar and pestle diluted with boron nitride, mounted on a Teflon plate, and sealed with Kapton polyamide film in an anaerobic glovebox, to prevent sample oxidation while minimizing X-ray absorption. Fluorescence data were collected at SSRL beamline 11-2, using a wide-angle (Lytle) fluorescence chamber. Incident and transmitted X-ray intensities were measured with in-line ionization chambers. The energy range studied was 200 to +1,000 eV around the Fe K-edge (7,112 eV). All samples were internally referenced to a Fe-metal standard, placed between the second and third in-line ionization chambers. Two to four individual spectra were averaged for each sample. EXAFS spectra were processed using the SixPACK (Webb, 2005) interface to IFEFFIT (Newville, 2001). After background subtraction and normalization, EXAFS data were extracted and k3-weighted. A set of reference standards for Fe was utilized to perform linear combination k3-weighted EXAFS spectral fitting, using SixPACK’s least-squares fitting module, which is a graphical interface to IFEFFIT’s minimization function (Newville, 2001). Linear combination fitting routines were used to reconstruct the


303

experimental spectrum and to determine the relative percentages of iron mineral phases (Hansel et al., 2003; Hansel et al., 2005). Each spectrum was fit using 2-line ferrihydrite, lepidocrocite, and goethite, which were detected in the XRD patterns for the 23.5 m2 L1 ferrihydrite experiment (Fig. 11.4A).

11.3.7. Thermodynamic Reaction Modeling Except for the calcium–uranyl–carbonate species, Gibbs free energies of formation for all uranium species were obtained from Guillaumont et al. (2003). Amorphous UO2 (UO2(am)) was chosen as the representative U(IV) species for all thermodynamic calculations because freshly bioreduced U(IV) is generally fine-grained and poorly crystalline (Giammar and Hering, 2001; Fredrickson et al., 2002; Suzuki et al., 2002). Amorphous uraninite (UO2(am)) is not thermodynamically well defined, but it provides a more realistic prediction of reactivity over shorter time-scales than crystalline UO2 (uraninite) (Wan et al., 2005). The Gibbs free energy of formation for Ca2UO2(CO3)3 and CaUO2(CO3)2 3 were calculated from stability constants provided in Dong and Brooks (2006). The value for hematite (Fe2O3) was obtained from Cornell and Schwertmann (2003), the values for lepidocrocite and goethite were obtained from Majzlan et al. (2003), and the value for 2-line ferrihydrite (Fe(OH)3) was obtained from Majzlan et al. (2004). All values were checked for internal consistency and are tabulated in Ginder-Vogel et al. (2006). The Gibbs free energy of reaction for specific conditions was calculated using standard convention at 298 K and noted in Table 11.1.

11.4. Results 11.4.1. Effect of Ferrihydrite Concentration Uraninite Oxidation Extent The ability of ferrihydrite to oxidize UO2 in the absence of calcium was investigated in a series of batch experiments in 3 mM HCO 3 at pH 7. After 48 h of reaction, the proportion of uraninite oxidized (calculated as the difference between the control and experimental dissolved U(VI) concentration) increases from 7.5% at 23.5 m2 L1 to 21.2% at 80 m2 L1 ferrihydrite (Fig. 11.5C).

304


Intensity (Counts)

chi(k)*k3

A

0

10

20

30

40

50

60

chi*k3

chi*k3

2 theta (degrees)

Figure 11.4: Iron K-Edge EXAFS Spectra (Solid Lines) and Linear Combination Fits (Dotted Lines) Used to Construct Fig. 11.6A–C. XRD Patterns Used to Identify Fe Minerals for Linear Combination Fits (Left Side of A); g ¼ Goethite, l ¼ Lepidocrocite, and u ¼ Uraninite.

Table 11.1: Gibb’s Free Energy of Reactions at Standard State Conditions (DGr1) and at Experimental Conditions Represented by pH 7, 1 106 M U(VI) Species, 5 107 M Fe2+, and 3 103 M HCO 3 (DGr ). Oxidation reaction

60.08 58.97 48.53 33.98 58.40 44.02 44.14 42.38 41.28 30.83 16.28 40.70 26.32 26.44 51.38 50.28 39.83 25.28 49.70 35.32 35.44 39.82 38.72 28.28 13.73 38.15 23.77 23.89

6.6 5.0 7.3 5.5 12.8 7.0 5.2 24.4 12.7 10.4 12.1 4.84 10.7 12.5 15.4 3.7 1.4 3.14 4.1 1.7 3.5 26.9 15.2 12.9 14.7 7.4 13.2 15.1

305

Thermodynamic data sources are described in the ‘‘Experimental’’ section. 1Standard state. Convention for noting the chemical gradients which are included in the thermodynamic value.

DGr (kJ mol1)


Fe(OH)3+0.5UO2+3H+2Fe2++0.5UO2+ 2 +3H2O 2+ +3H2O Fe(OH)3+0.5UO2+2.5H++0.5HCO 3 20.5UO2CO3+Fe 2 2+ +3H2O Fe(OH)3+0.5UO2+2H++HCO 3 20.5UO2(CO3)2 +Fe 4 2+ +3H2O Fe(OH)3+0.5UO2+1.5H++1.5HCO 3 20.5UO2(CO3)3 +Fe + 2+ Fe(OH)3+0.5UO2+1.5H +1.5HCO3 +Ca 20.5Ca2UO2(CO3)3+Fe2++3H2O 2+ 2+ 20.5CaUO2(CO3)2 +3H2O Fe(OH)3+0.5UO2+1.5H++1.5HCO 3 +0.5Ca 3 +Fe 2+ 20.25(UO ) CO (OH) +Fe +2.25H Fe(OH)3+0.5UO2+2H++0.25HCO 3 2 2 3 3 2O 2+ +2H2O a-FeOOH+0.5UO2+3H+20.5UO2+ 2 +Fe 2+ +2H2O a-FeOOH+0.5UO2+2.5H++0.5HCO 3 20.5UO2CO3+Fe 2 2+ 20.5UO (CO ) +Fe +2H2O a-FeOOH+0.5UO2+2H++1HCO 3 2 3 2 4 2+ +H2O a-FeOOH+0.5UO2+1.5H++1.5HCO 3 20.5UO2(CO3)3 +Fe 2+ 20.5Ca2UO2(CO3)3+Fe2++2H2O a-FeOOH+0.5UO2+1.5H++1.5HCO 3 +Ca 2+ 2+ 20.5CaUO2(CO3)2 +2H2O a-FeOOH+0.5UO2+1.5H++1.5HCO 3 +0.5Ca 3 +Fe 2+ +5/4H2O a-FeOOH+0.5UO2+2H++0.25HCO 3 20.25(UO2)2CO3(OH)3 +Fe 2+ +2H2O g-FeOOH+0.5UO2+3H+20.5UO2+ 2 +Fe + g-FeOOH+0.5UO2+2.5H +0.5HCO3 20.5UO2CO3+Fe2++2H2O 2 2+ +2H2O g-FeOOH+0.5UO2+2H++1HCO 3 20.5UO2(CO3)2 +Fe 4 2+ 20.5UO (CO ) +Fe +H2O g-FeOOH+0.5UO2+1.5H++1.5HCO 3 2 3 3 + 2+ g-FeOOH+0.5UO2+1.5H +1.5HCO3 +Ca 20.5Ca2UO2(CO3)3+Fe2++2H2O 2+ 2+ 20.5CaUO2(CO3)2 +2H2O g-FeOOH+0.5UO2+1.5H++1.5HCO 3 +0.5Ca 3 +Fe 2+ 20.25(UO ) CO (OH) +Fe +5/4H g-FeOOH+0.5UO2+2H++0.25HCO 3 2 2 3 3 2O 0.5Fe2O3+0.5UO2+3H+2Fe2++0.5UO2+ 2 +1.5H2O 2+ +1.5H2O+0.5UO2CO3 0.5Fe2O3+0.5UO2+2.5H++0.5HCO 3 2Fe 2+ 2 2Fe +1.5H 0.5Fe2O3+0.5UO2+2H++1HCO 3 2O+0.5UO2(CO3)2 2+ +1.5H2O+0.5UO2(CO3)4 0.5Fe2O3+0.5UO2+1.5H++1.5HCO 3 2Fe 3 2+ 2Fe2++1.5H2O+0.5Ca2UO2(CO3)3 0.5Fe2O3+0.5UO2+1.5H++1.5HCO 3 +Ca 2+ 2Fe2++1.5H2O+0.5CaUO2(CO3)2 0.5Fe2O3+0.5UO2+1.5H++1.5HCO 3 +0.5Ca 3 2+ +0.75H2O 0.5Fe2O3+0.5UO2+2H++0.25HCO 3 +0.25(UO2)2CO3(OH)3 +Fe

DGr1 (kJ mol1)

306


A

B

C

Figure 11.5: Percentage of Initial Uraninite Oxidized after 48 h as a Function of Fe(III) (Hydr)Oxide Type (A), Bicarbonate Concentration (B), and Ferrihydrite Concentration (C). Data Points for A and B are Calculated from Ginder-Vogel et al. (2006). Not Detected (ND).


307

11.4.2. Fe(III) (Hydr)Oxide Mineral Evolution In the presence of small amounts of Fe(II) (o0.67 mmol Fe(II) g1 ferrihdyrite), ferrihydrite transforms into lepidocrocite and goethite (Hansel et al., 2005). During 48 h of oxidation, dissolved Fe(II) remains below detection limits; extractable Fe(II) at all three ferrihydrite suspension concentrations reaches a maximum of 0.2 mmol Fe(II) g1 ferrihydrite (Fig. 11.6D), and dissolved U(VI) also plateaus at 75, 112, and 211 mM for ferrihydrite concentrations of 23.5, 46.0, and 80.2 m2 L1 ferrihydrite (data not shown). At all ferrihydrite suspension concentrations, the solid phase remains predominantly ferrihydrite over the first 2 h of the experiment (Figs. 11.4 and 11.6). Crystalline Fe(III) (hydr)oxides are not detected in XRD diffraction patterns prior to 5 h of reaction time (Fig. 11.4). As the reaction progresses, the goethite concentration continues to increase with detectable amounts of lepidocrocite accumulating only after 6 h of reaction. At the lowest ferrihydrite concentration, the solid-phase speciation stabilizes after 24 h of reaction at a distribution of 36% ferrihydrite, 19% lepidocrocite, and 44% goethite (Fig. 11.6A). The transformation of ferrihydrite is notably more complete at increasing ferrihydrite concentration, with only 8% of the ferrihydrite remaining in the 46 m2 L1 reaction and no ferrihydrite remaining in the 80.2 m2 L1 reaction after 48 h (Fig. 11.6B and C). Uraninite oxidation was confirmed by comparing the solid phase-uranium oxidation state of the last sample from each reaction to the oxidation state of biogenic uraninite. Uranium(VI) content of the biogenic uraninite is 5%, and the U(VI) content at the conclusion of each oxidation reaction is 10, 18, and 26%, respectively, for 23.5, 46.0, and 80.2 m2 L1 ferrihydrite (data not shown).

11.5. Discussion 11.5.1. Extent of Uraninite Oxidation The prevalence of materials capable of oxidizing biologically precipitated uraninite and geochemical characteristics will influence the long-term stability of biologically precipitated uraninite. Although microcrystalline uraninite has an estimated solubility of 108 to 1010 at pH 7 (Casas et al., 1998), it can be oxidized and remobilized by several common environmental constituents (Fig. 11.1). Given Fe(III) (hydr)oxide’s environmental

Figure 11.6: Transformation of Ferrihydrite by Fe(II) Generated during Uraninite Oxidation. Reactions were Conducted with 3.0 mM KHCO3, 30.7 m2 L1 UO2, and Either (A) 23.5, (B) 46.0, or (C) 80.2 m2 L1 Ferrihydrite in 3 mM HCO 3 at pH 7.2. Percentages (75%) were Determined from Linear Combination Fits of k3-Weighted Fe EXAFS Spectra (k ¼ 1–14) (Fig. 11.4). (D) Ferrozine Extractable Fe(II) for Each Reaction. 1 103 mmol Fe(II) m2 Ferrihydrite is Equivalent to 0.2 mM Fe(II) g1 Ferrihydrite.


309

prevalence and tendency to form in redox-active environments (Cornell and Schwertmann, 2003), it is crucial to develop a thorough understanding of the conditions that enhance uraninite oxidation by Fe(III) (hydr)oxides. The energetic favorability of uraninite oxidation by Fe(III) (hydr)oxides depends on Fe(III) (hydr)oxide type and abundance, in addition to geochemical conditions. For instance, we (Ginder-Vogel et al., 2006) observed that in 48 h 1.8% of the initial biogenic uraninite is oxidized by 2-line ferrihydrite in 3 mM HCO 3 at pH 7 (Fig. 11.5B); in contrast, uraninite exposure to goethite and hematite results in less, or no, uraninite oxidation (Fig. 11.5A). A similar increase in the amount of uraninite oxidized also occurs with increasing HCO 3 concentration, with 1.8% oxidized at 3 mM HCO 3 and nearly 10% oxidized at 100 mM HCO3 . However, increasing the ferrihydrite concentration has the most dramatic effect on the extent of uraninite oxidation, with 7.4% oxidized at 23.5 m2 L1 and >20% oxidized at 80.2 m2 L1 ferrihydrite. As opposed to changes in the Fe(III) (hydr)oxide type and HCO 3, changes in ferrihydrite concentration do not affect the thermodynamic favorability of the uraninite oxidation reaction. This suggests that the higher concentration of ferrihydrite is acting as a sink for Fe(II) and/or U(VI) produced during uraninite oxidation, which may otherwise inhibit uraninite oxidation. At 3 mM bicarbonate and 10 mM U(VI)(aq), the majority of uranium should be in solution and thus only slightly affected by changes in ferrihydrite concentration. However, uraninite oxidation appears to cease as extractable Fe(II) concentrations approach 1 103 mmol m2 ferrihdyrite, despite the accumulation of o1 mM dissolved Fe(II). Assuming that ferrihydrite has two surface sites per nanometer square (Davis and Kent, 1990), equivalent to 3.7 103 mmol surface site meter square, the 1.0 103 mmol m2 (0.2 mmol g1) extractable Fe(II) present after 10 h reaction time at all ferrihydrite concentrations (Fig. 11.6D) accounts for 25% of the total initial surface sites available and may prevent U(IV) from reaching reactive surface sites. Additionally, Fe(II) sorption to ferrihydrite may result in electron delocalization within the Fe (hydr)oxide structure, thereby lowering its redox potential (Williams and Scherer, 2004; Iordanova et al., 2005; Kerisit and Rosso, 2005; Larese-Casanova and Scherer, 2007). This suggests that 0.2 mmol g1 ferrihydrite is a thermodynamic threshold for the energetic favorability of uraninite oxidation. The alteration of ferrihydrite into more crystalline, lower surface area Fe(III) (hydr)oxide phases will further reduce the number of surface sites available for uraninite oxidation.

310


11.5.2. Evolution of Fe(III) (Hydr)Oxide Mineralogy In addition to limiting uraninite oxidation, Fe(II) generated during reaction will catalyze the transformation of ferrihydrite into an assemblage of more thermodynamically stable iron (hydr)oxide minerals. This mineral assemblage, depending on initial chemistry and Fe(II) concentration, is often dominated by lepidocrocite, goethite, or magnetite (Fredrickson et al., 1998; Benner et al., 2002; Hansel et al., 2003), all of which are less favorable oxidants than ferrihydrite (Cornell and Schwertmann, 2003). Indeed, as uraninite oxidation proceeds, Fe(II) induces the transformation of ferrihydrite into lepidocrocite and goethite (Fig. 11.6). Despite the similarity in the amount of Fe(II) generated relative to initial surface area, ferrihydrite conversion into lepidocrocite and goethite is noticeably more complete at 46.0 m2 L1 ferrihydrite (Fig. 11.6B) and proceeds to completion at 80.2 m2 L1 ferrihydrite (Fig. 11.6C). Although it is not possible to determine the exact cause of the observed differences in Fe(III) (hydr)oxide mineralogy from this study, the higher U(VI) concentrations produced in conjunction with the higher ferrihydrite concentration may consume dissolved bicarbonate through complexation reactions. The decrease in dissolved bicarbonate concentration may alter the reaction products produced during ferrihydrite recrystallization (Schwertmann and Cornell, 2000; Hansel et al., 2005); additionally, the increase in U(VI) concentration may, itself, alter the products of ferrihydrite transformation (Duff et al., 2002). Based upon thermodynamic considerations and previous studies, lepidocrocite accumulation should be observed prior to goethite accumulation (Hansel et al., 2005); however, goethite accumulates prior to lepidocrocite at all three ferrihydrite concentrations (Fig. 11.6). Since the oxidation of biogenic uraninite by lepidocrocite remains thermodynamically favorable at the low Fe(II)(aq) and U(VI)(aq) concentrations observed during the first 5 h of reaction, it may be a transient species that is rapidly consumed by uraninite oxidation. Indeed, uraninite oxidation by ferrihydrite, lepidocrocite, and goethite is no longer thermodynamically favorable at the reaction conditions observed after 10 h of reaction (100 mM U(VI)(aq), pH 7.2, 3 mM HCO 3 , and 1 mM Fe(II)(aq)).

11.6. Implications for Biogeochemical Uranium Cycling During uranium reduction in biostimulated environments, Fe(II) concentrations are frequently 10–40 mM (Wu et al., 2007), which will likely limit


311

uraninite oxidation by Fe(III) (hydr)oxides. However, intrusion of nitrate or molecular oxygen into a previously reduced (anaerobic) environment may consume Fe(II), resulting in the formation of poorly crystalline Fe(III) (hydr)oxides capable of oxidizing biologically precipitated uraninite. Consideration of this oxidation pathway is particularly important in the case of nitrate, a common co-contaminant with uranium, because abiotic and biotic oxidation of uraninite coupled to nitrate reduction occurs more slowly than oxidation by Fe(III) (hydr)oxides (Beller, 2005; Senko et al., 2005a). Additionally, the extent of biogenic uraninite oxidation by molecular oxygen increases below pH 7 in the presence of >20 mM Fe(II) (Zhong et al., 2005). Geochemical conditions play a critical role in determining the energetic favorability of the uraninite oxidation reaction with the favorability increasing with increasing carbonate and calcium concentrations and decreasing pH (Ginder-Vogel et al., 2006). However, the ratio of ferrihydrite to biogenic uraninite appears to be the most important factor affecting the extent of uraninite oxidation at neutral pH, with the amount of uraninite oxidized scaling linearly with ferrihydrite concentration (Fig. 11.5C); however, a 33-fold increase in the bicarbonate concentration only results in a 5-fold increase in the amount of uraninite oxidized (Fig. 11.5B). In batch systems, Fe(II) produced during the oxidation reaction inhibits uraninite oxidation at concentrations of 1.0 103 mmol m2 ferrihydrite; however, it is likely that transport of reaction products, including Fe(II) and U(VI), in groundwater systems would enhance the extent of uraninite oxidation. Additionally, Fe(II) catalyzes the transformation of ferrihydrite into thermodynamically more stable Fe(III) (hydr)oxide phases, which affects the viability and extent of uraninite oxidation (Figs. 11.3 and 11.5). The importance of considering Fe(III) (hydr)oxide crystallinity on the reoxidation of biogenic UO2 by Fe(III) (hydr)oxides is exemplified by several field and soil-column experiments (Fig. 11.7). Surprisingly, for all conditions reported, oxidation of biogenic UO2 by ferrihydrite is thermodynamically probable at Fe(II) concentrations of 20 mM. However, in the presence of 20 mM Fe(II), uraninite oxidation by lepidocrocite is thermodynamically favored for two conditions, and oxidation by goethite is only thermodynamically probable for one condition (Fig. 11.7). A host of U(IV) oxidants exist in surface and subsurface environments that include poorly crystalline Fe(III) (hydr)oxide phases. Therefore, while, iron(II) could serve to consume molecular oxygen in oxic waters, resulting ferric (hydr)oxides may continue to serve as oxidants of U(IV). However, long-term redox cycling promotes the conversion of poorly crystalline Fe(III) (hydr)oxide minerals into more crystalline forms, e.g., goethite or hematite (Thompson et al., 2006) and may be one method of limiting uraninite

312


0

UO2 Oxidation Favored

-1

log (HCO3 ) (M)

-2

e-

Senko

II)

Istok M 0µ

( Fe

Sani Wan-2

2

Anderson

it eth

II)

Go

µM

-3

( Fe

Wu-3

Wu-2 20 et i c Wu-1 o r

oc

pid

-4

Wan-1

II)

Le

0

ite dr

-5

-2

µM

( Fe

y rih

r Fe

2-

UO2(CO3)2 Reduction Favored

-6 5.5

6.0

6.5

7.0 pH

7.5

8.0

8.5

Figure 11.7: Thermodynamic Viability of UO2 (biogenic) Oxidation by Fe(III) (Hydr)Oxides for Conditions Reported for Various Field Sites and SoilColumn Experiments (Anderson et al., 2003; Istok et al., 2004; Sani et al., 2005; Senko et al., 2005a; Wan et al., 2005; Wu et al., 2006b). Conditions Representing Equilibrium with 0.126 mM UO2(CO3)2 2 (the Drinking Water Maximum) and 20 mM Fe(II) are Illustrated for Each Iron Oxide. Reoxidation is Viable at Bicarbonate and pH Conditions Plotting Above the Line and Nonviable for those Plotting Below. Reactions for Each Line are Listed in Table 11.1. Relevant Geochemical Conditions and References are Detailed in Ginder-Vogel et al. (2006). oxidation by Fe(III) (hydr)oxides. Nonetheless, oxidation by ferric (hydr)oxides is thermodynamically favorable and may limit uranium sequestration under mildly reducing conditions.

ACKNOWLEDGMENTS We would like to thank two anonymous reviewers and the editors for providing input that greatly improved the manuscript. We thank Kathleen Ginder-Vogel for her constructive input on the manuscript. We also thank Sam Webb, John Bargar, and Joe Rogers for their assistance in the collection of X-ray data at SSRL. This work was funded by the Office of Science


313

Biological and Environmental Research ERSD Program, U.S. Department of Energy (grant number ER63609-1021814) and the Stanford NSF Environmental Molecular Sciences Institute (grant number NSF-CHE-0431425). A portion of this work was conducted at Stanford Synchrotron Radiation Laboratory, a national user facility operated by Stanford University on behalf of the U.S. Department of Energy, Office of Basic Energy Sciences. The SSRL Structural Molecular Biology Program is supported by the Department of Energy, Office of Biological and Environmental Research, and by the National Institutes of Health, National Center for Research Resources, Biomedical Technology Program.

REFERENCES Anderson, R. T., Vrionis, H. A., Ortiz-Bernard, I., Resch, C. T., Long, P. E., Dayvault, R., Karp, K., Marutzky, S., Metzler, D. R., Peacock, A. D., White, D. C., Lowe, M., & Lovley, D. R. (2003). Stimulating the in situ activity of Geobacter species to remove uranium from the groundwater of a uranium-contaminated aquifer. Appl. Environ. Microbiol., 69, 5884–5891. Barnes, C. E., & Cochran, J. K. (1993). Uranium geochemistry in esturaine sediments: Controls on removal and release processes. Geochim. Cosmochim. Acta, 57, 555–569. Barnett, M. O., Jardine, P. M., Brooks, S. C., & Selim, H. M. (2000). Adsorption and transport of uranium(VI) in subsurface media. Soil Sci. Soc. Am. J., 64, 908–917. Beller, H. R. (2005). Anaerobic, nitrate-dependent oxidation of U(IV) oxide minerals by the chemolithoautotrophic bacterium Thiobacillus dentrificans. Appl. Environ. Microbiol., 71, 2170–2174. Benner, S. G., Hansel, C. M., Wielinga, B., Barber, T. M., & Fendorf, S. (2002). Reductive dissolution of iron hydroxide under dynamic flow conditions. Environ. Sci. Technol., 36, 1705–1711. Bostick, B. B., Fendorf, S., Barnett, M. O., Jardine, P. M., & Brooks, S. C. (2002). Uranyl surface complexes formed on subsurface media from DOE facilities. Soil Sci. Soc. Am. J., 66, 99–108. Bradford, G. R., Bakhtar, D., & Westcot, D. (1990). Uranium, vanadium, and molybdenum in saline waters of California. J. Environ. Qual., 19, 105–108. Brooks, S. C., Fredrickson, J. K., Carroll, S. L., Kennedy, D. W., Zachara, J. M., Plymale, A. E., & Fendorf, S. (2003). Inhibition of bacterial U(VI) reduction by calcium. Environ. Sci. Technol., 37, 1850–1858. Casas, I., dePablo, J., Gimenez, J., Torrero, M. E., Bruno, J., Cera, E., Finch, R. J., & Ewing, R. C. (1998). The role of pe, pH, and carbonate on the solubility of UO2 and uraninite under nominally reducing conditions. Geochim. Cosmochim. Acta, 62, 2223–2231.

314


Cornell, R. M., & Schwertmann, U. (2003). The Iron Oxides: Structure, Properties, Reactions, Occurrences and Uses. Wiley-VCH, Weinheim, Germany. Curtis, G.P., Davis, J.A., & Naftz, D.L. (2006). Simulation of reactive transport of uranium(VI) in groundwater with variable chemical conditions, Water Resour. Res., 42, W04404, doi: 10.1029/2005WR003979. Davis, J. A., & Kent, D. B. (1990). Surface complexation modeling in aqueous geochemistry. In: M. F. Hochella & A. F. White (Eds). Reviews in Mineralogy: Mineral–Water Interface Geochemistry. Mineralogical Society of America, Washington, D.C. Davis, J. A., Meece, D. E., Kohler, M., & Curtis, G. P. (2004). Approaches to surface complexation modeling of uranium(VI) adsorption on aquifer sediments. Geochim. Cosmochim. Acta, 68, 3621–3641. Dispirito, A. A., & Tuovinen, O. H. (1982a). Kinetics of uranous ion and ferrous iron oxidation by Thiobacillus ferrooxidans. Arch. Microbiol., 133, 33–37. Dispirito, A. A., & Tuovinen, O. H. (1982b). Uranous ion oxidation and carbon dioxide fixation by Thiobacillus ferrooxidans. Arch. Microbiol., 133, 28–32. Dong, W., & Brooks, S. C. (2006). Determination of the formation constants of ternary complexes of uranyl and carbonate with alkaline earth metals (Mg2+, Ca2+, Sr2+, and Ba2+) using anion exchange method. Environ. Sci. Technol., 40, 4689–4695. Duff, M. C., Coughlin, J. U., & Hunter, D. B. (2002). Uranium co-precipitation with iron oxide minerals. Geochim. Cosmochim. Acta, 66, 3533–3547. El Aamrani, F., Casas, I., de Pablo, J., Duro, L., Grive, M., & Bruno, J. (2000). Experimental and modeling study of the interaction between uranium(VI) and magnetite. J. Conf. Abstr., 5, 378. Finneran, K. T., Housewright, M. E., & Lovley, D. R. (2002). Multiple influences of nitrate on uranium solubility during bioremediation of uranium-contaminated subsurface sediments. Environ. Microbiol., 4, 510–516. Fredrickson, J. K., Zachara, J. M., Kennedy, D. W., Dong, H. L., Onstott, T. C., Hinman, N. W., & Li, S. M. (1998). Biogenic iron mineralization accompanying the dissimilatory reduction of hydrous ferric oxide by a groundwater bacterium. Geochim. Cosmochim. Acta, 62, 3239–3257. Fredrickson, J. K., Zachara, J. M., Kennedy, D. W., Duff, M. C., Gorby, Y. A., Li, S. M. W., & Krupka, K. M. (2000). Reduction of U(VI) in goethite (alphaFeOOH) suspensions by a dissimilatory metal-reducing bacterium. Geochim. Cosmochim. Acta, 64, 3085–3098. Fredrickson, J. K., Zachara, J. M., Kennedy, D. W., Liu, C. G., Duff, M. C., Hunter, D. B., & Dohnalkova, A. (2002). Influence of Mn oxides on the reduction of uranium(VI) by the metal-reducing bacterium Shewanella putrefaciens. Geochim. Cosmochim. Acta, 66, 3247–3262. Giammar, D. E., & Hering, J. G. (2001). Time scales for sorption-desorption and surface precipitation of uranyl on goethite. Environ. Sci. Technol., 35, 3332–3337.


315

Ginder-Vogel, M., Borch, T., Mayes, M. A., Jardine, P. M., & Fendorf, S. (2005). Chromate reduction and retention processes within arid subsurface environments. Environ. Sci. Technol., 39, 7833–7839. Ginder-Vogel, M., Criddle, C., & Fendorf, S. (2006). Thermodynamic constraints on the oxidation of biogenic UO2 by Fe(III) (hydr)Oxides. Environ. Sci. Technol., 40, 3544–3550. Gorby, Y. A., & Lovley, D. R. (1992). Enzymatic uranium precipitation. Environ. Sci. Technol., 26, 205–207. Grenthe, I., Fuger, J., Konings, R. J. M., Lemire, R. J., Muller, A. B., NguyenTrung, C., & Wanner, H. (1992). Chemical Thermodynamics of Uranium. NorthHolland Elsevier Science Publishers B.V., Amsterdam. Guillaumont, R., Fanghanel, T., Neck, V., Fuger, J., Palmer, D. A., Grenthe, I., & Rand, M. H. (2003). Update on the chemical thermodynamics of uranium, neptumium, plutonium, americium, and technetium. In: F. J. Mompean, M. Illemassene, C. Domenech-Orti, & K. Ben Said (Eds). Elsevier, Amsterdam, The Netherlands. Hammersley, A. P. (1997). FIT2D: An Introduction and Overview. European Synchrotron Radiation Facility, Grenoble, France. Hansel, C. M., Benner, S. G., & Fendorf, S. (2005). Competing Fe(II)-induced mineralization pathways of ferrihydrite. Environ. Sci. Technol., 39, 7147–7153. Hansel, C. M., Benner, S. G., Neiss, J., Dohnalkova, A., Kukkadapu, R. K., & Fendorf, S. (2003). Secondary mineralization pathways induced by dissimilatory iron reduction of ferrihydrite under advective flow. Geochim. Cosmochim. Acta, 67, 2977–2992. Harrison, V., Gow, W., & Hughson, M. (1966). Factors influencing the application of bacterial leaching to a canadian uranium ore. J. Metals, 18, 1189–1194. Hua, B., Xu, H., Terry, J., & Deng, B. (2006). Kinetics of uranium(VI) reduction by hydrogen sulfide in anoxic aqueous systems. Environ. Sci. Technol., 40, 4666–4671. Ilton, E., Haiduc, A., Moses, C., Heald, S. M., Elbert, D. C., & Veblen, D. R. (2004). Heterogeneous reduction of uranyl by micas: Crystal chemical and solution controls. Geochim. Cosmochim. Acta, 68, 2417–2435. Ilton, E., Heald, S. M., Smith, S. C., Elbert, D. C., & Liu, C. (2006). Reduction of uranyl in the interlayer region of low iron micas under anoxic and aerobic conditions. Environ. Sci. Technol., 40, 5003–5009. Iordanova, N., Dupuis, M., & Rosso, K. M. (2005). Charge transport in metal oxides: A theoretical study of hematite a-Fe2O3. J. Chem. Phys., 122, 1–10. Istok, J. D., Senko, J. M., Krumholz, L. R., Watson, D., Bogle, M., Peacock, A. D., Chang, Y.-J., & White, D. C. (2004). In situ bioreduction of technetium and uranium in a nitrate contaminated aquifer. Environ. Sci. Technol., 38, 468–475. Jeon, B.-H., Dempsey, B. A., Burgos, W. D., Barnett, M. O., & Roden, E. E. (2005). Chemical reduction of U(VI) by Fe(II) at the solid–water interface using natural and synthetic Fe(III) oxides. Environ. Sci. Technol., 39, 5642–5649.

316


Kerisit, S., & Rosso, K. M. (2005). Charge transfer in FeO: A combined moleculardynamics and ab initio study. J. Chem. Phys., 123, 1–10. Langmuir, D. (1978). Uranium solution–mineral equilibria at low temperature with applications to sedimentary ore deposits. Geochim. Cosmochim. Acta, 42, 547–569. Langmuir, D., & Chatman, J. R. (1980). Groundwater prospecting for sandstonetype uranium deposits: A preliminary comparison of the merits of mineral– solution equilibria and single element tracer methods. J. Geochem. Explor., 13, 201–219. Larese-Casanova, P., & Scherer, M. M. (2007). Fe(II) sorption on hematite: New insights based on spectroscopic measurements. Environ. Sci. Technol., 41, 471–477. Liger, E., Charlet, L., & Cappellen, P. V. (1999). Surface catalysis of uranium(VI) reduction by iron(II). Geochim. Cosmochim. Acta, 63, 2939–2955. Liu, C. G., Jeon, B.-H., Zachara, J. M., Wang, Z., Dohnalkova, A., & Fredrickson, J. K. (2006). Kinetics of microbial reduction of solid phase U(VI). Environ. Sci. Technol., 40, 6290–6296. Livens, F. R., Jones, M. J., Hynes, A. J., Charnock, J. M., Mosselmans, J. F., Hennig, C., Steefel, C. I., Collison, D., Vaughan, D. J., Pattrick, R. A. D., Reed, W. A., & Moyes, I. (2004). X-ray absorption spectroscopy studies of reactions of technetium, uranium and neptunium with machinawite. J. Environ. Radioact., 74, 211–219. Majzlan, J., Grevel, K.-D., & Navrotsky, A. (2003). Thermodynamics of iron oxides: part II. Enthalpies of formation and relative stability of goethite (a-FeOOH), lepidocrocite (g-FeOOH), and maghemite (g-Fe2O3). Am. Mineral., 88, 855–859. Majzlan, J., Navrotsky, A., & Schwertmann, U. (2004). Thermodynamics of iron oxides: part III. Enthalpies of formation and stability of ferrihydrite, schwertmannite, and e-Fe2O3. Geochim. Cosmochim. Acta, 68, 1049–1059. Missana, T., Garcia-Gutierrez, M., & Fernndez, V. (2003). Uranium(VI) sorption on colloidal magnetite under anoxic environment: Experimental study and surface complexation modelling. Geochim. Cosmochim. Acta, 67, 2543–2550. Morrison, S. J., Spangler, R. R., & Tripathi, V. S. (1995). Adsorption of uranium(VI) on amorphous ferric oxyhydroxide at high concentrations of dissolved carbon(IV) and sulfur(VI). J. Contam. Hydrol., 17, 333–346. Moyes, L. N., Parkman, R. H., Charnock, J. M., Vaughan, D. J., Livens, F. R., Hughes, C. R., & Braithwaite, A. (2000). Uranium uptake from aqueous solution by interaction with goethite, lepidocrocite, muscovite, and mackinawite: An X-ray absorption spectroscopy study. Environ. Sci. Technol., 34, 1062–1068. Nash, J. T., Granger, H. C., & Adams, S. S. (1981). Geology and concepts of genesis of important types of uranium deposits. Econ. Geol., 75, 63–116. Nevin, K. P., & Lovley, D. R. (2000). Potential for nonenzymatic reduction of Fe(III) via electron shuttling in subsurface sediments. Environ. Sci. Technol., 34, 2472–2478.


317

Newville, M. (2001). IFEFFIT: Interactive XAFS analysis and FEFF fitting. J. Synchrotron Radiat., 8, 322–324. O’Loughlin, E. G., Kelly, S. D., Cook, R. E., Csensits, R., & Kemner, K. M. (2003). Reduction of uranium(VI) by mixed iron(II)/iron(III) hydroxide (green rust): Formation of UO2 nanoparticles. Environ. Sci. Technol., 37, 721–727. Ortiz-Bernad, I., Anderson, R. T., Vrionis, H. A., & Lovley, D. R. (2004). Resistance of solid-phase U(VI) to microbial reduction during in situ bioremediation of uranium-contaminated groundwater. Appl. Environ. Microbiol., 70, 7558–7560. Peper, S. M., Brodnax, L. F., Field, S. E., Zehnder, R. A., Valdez, S. N., & Runde, W. H. (2004). Kinetic study of the oxidative dissolution of UO2 in aqueous carbonate media. Ind. Eng. Chem. Res., 43, 8188–8193. Pierce, E. M., Icenhower, J. P., Serne, J. N., & Catalano, J. G. (2005). Experimental determination of UO2(cr) dissolution kinetics: Effects of solution saturation state and pH. J. Nucl. Mater., 345, 206–218. Riley, R. G., Zachara, J. M., & Wobber, F. J. (1992). Chemical Contaminants on DOE Lands and Selection of Contaminant Mixtures for Subsurface Science Research. U.S. Department of Energy, Washington, D.C. Sandino, A., & Bruno, J. (1992). The solubility of (UO2)3(PO4)2 4H2O(s) and the formation of U(VI) phosphate complexes: Their influence in uranium speciation in natural waters. Geochim. Cosmochim. Acta, 56, 4135–4145. Sani, R. K., Peyton, B. M., Amonette, J. E., & Geesey, G. G. (2004). Reduction of uranium(VI) under sulfate-reducing conditions in the presence of Fe(III)(hydr)oxides. Geochim. Cosmochim. Acta, 68, 2639–2648. Sani, R. K., Peyton, B. M., Dohnalkova, A., & Amonette, J. E. (2005). Reoxidation of reduced uranium with iron(III) (hydr)oxides under sulfate-reducing conditions. Environ. Sci. Technol., 39, 2059–2066. Schwertmann, U., & Cornell, R. M. (2000). Iron Oxides in the Laboratory: Preparation and Characterization. Wiley-VCH, Weinheim, Germany. Scott, T. B., Allen, G. C., Heard, P. J., & Randell, M. G. (2005). Reduction of U(VI) to U(IV) on the surface of magnetite. Geochim. Cosmochim. Acta, 69, 5639–5646. Senko, J. M., Istok, J. D., Suflita, J. M., & Krumholz, L. R. (2002). In-situ evidence for uranium immobilization and remobilization. Environ. Sci. Technol., 36, 1491–1496. Senko, J. M., Mohamed, Y., Dewers, T., & Krumholz, L. R. (2005a). Role for Fe(III) minerals in nitrate-dependent microbial U(IV) oxidation. Environ. Sci. Technol., 39, 2529–2536. Senko, J. M., Suflita, J. M., & Krumholz, L. R. (2005b). Geochemical controls on microbial nitrate-dependent U(IV) oxidation. Geomicrobiol. J., 22, 371–378. Singer, D., Farges, F., & Brown, G. E. (2006). Biogenic UO2: Characterization and surface reactivity. Phys. Scr., 882, 277–279. Stewart, B. D., Neiss, J., & Fendorf, S. (2007). Quantifying constraints imposed by calcium and iron on bacterial reduction of uranium(VI). J. Environ. Qual., 36, 363–372.

318


Stookey, L. L. (1970). A new spectrophotometric reagent for iron. Anal. Chem., 42, 779–781. Suzuki, Y., Kelly, S. D., Kemner, K. M., & Banfield, J. F. (2002). Radionuclide contamination: Nanometer-size products of uranium bioreduction. Nature, 419, 134. Thompson, A., Chadwick, O. A., Rancourt, D. G., & Chorover, J. (2006). Ironoxide crystallinity increases during soil redox oscillations. Geochim. Cosmochim. Acta, 70, 1710–1727. Torrero, M. E., Baraj, E., de Pablo, J., Gimenez, J., & Casas, I. (1997). Kinetics of corrosion and dissolution of uranium dioxide as a function of pH. Int. J. Chem. Kinet., 29, 261–267. Vuorinen, A., Hiltunen, P., & Touvinen, O. (1985). Speciation of ferrous and ferric iron associated with the indirect bacterial leaching of uranium ore minerals. J. Ferment. Technol., 63, 337–342. Walter, M., Arnold, T., Reich, T., & Bernhard, G. (2003). Sorption of uranium(VI) onto ferric oxides in sulfate-rich acid waters. Environ. Sci. Technol., 37, 2898– 2904. Wan, J., Tokunaga, T. K., Brodie, E., Wang, Z., Zheng, Z., Herman, D., Hazen, T., Firestone, M. K., & Sutton, S. R. (2005). Reoxidation of bioreduced uranium under reducing conditions. Environ. Sci. Technol., 39, 6162–6169. Webb, S. M. (2005). Sixpack: A graphical user interface for XAS analysis using IFEFFIT. Phys. Scr., T115, 1011–1014. Wersin, P., Hochella, M. F., Persson, P., Redden, G., Leckie, J. O., & Harris, D. W. (1994). Interaction between aqueous uranium(VI) and sulfide minerals: spectroscopic evidence for sorption and reduction. Geochim. Cosmochim. Acta, 58, 2829– 2843. Wielinga, B., Bostick, B., Hansel, C. M., Rosenzweig, R. F., & Fendorf, S. (2000). Inhibition of bacterially promoted uranium reduction: Ferric (hydr)oxides as competitive electron acceptors. Environ. Sci. Technol., 34, 2190–2195. Williams, A. G. B., & Scherer, M. M. (2004). Spectroscopic evidence for Fe(II)Fe(III) electron transfer at the Fe oxide–water interface. Environ. Sci. Technol., 38, 4782–4790. Wu, W.-M., Carley, J., Fienen, M., Mehlhorn, T., Lowe, K., Nyman, J., Luo, J., Gentile, M., Rajan, R., Wagner, D., Hickey, R., Gu, B., Watson, D. B., Cirpka, O., Kitanidis, P., Jardine, P. M., & Criddle, C. (2006a). Pilot-scale in situ bioremediation of uranium in a highly contaminated aquifer: 1. Conditioning of a treatment zone. Environ. Sci. Technol., 40, 3978–3985. Wu, W.-M., Carley, J., Gentry, T., Ginder-Vogel, M., Fienen, M., Mehlhorn, T., Yan, H., Caroll, S., Pace, M., Nyman, J., Luo, J., Gentile, M., Fields, M. W., Hickey, R., Watson, D. B., Cirpka, O., Zhou, J., Fendorf, S., Kitanidis, P., Jardine, P. M., & Criddle, C. (2006b). Pilot-scale in situ bioremediation of uranium in a highly contaminated aquifer: 2. Geochemical control of U(VI) bioavailability and evidence of U(VI) reduction. Environ. Sci. Technol., 40, 3986–3995.


319

Wu, W.-M., Carley, J., Luo, J., Ginder-Vogel, M., Cardanans, E., Leigh, M. B., Hwang, C., Kelly, S. D., Ruan, C., Wu, L., Gentry, T., Lowe, K., Mehlhorn, T., Carroll, S. L., Fields, M. W., Gu, B., Watson, D., Kemner, K. M., Marsh, T. L., Tiedje, J. M., Zhou, J., Fendorf, S., Kitanidis, P., Jardine, P. M., & Criddle, C. (2007). In situ bioreduction of uranium(VI) to submicromolar levels and reoxidation by dissolved oxygen. Environ. Sci. Technol., 41, 5716–5723. Zhong, L., Liu, C., Zachara, J. M., Kennedy, D. W., Szecsody, J. E., & Wood, B. (2005). Oxidative remobilization of biogenic uranium(IV) precipitates: Effects of iron(II) and pH. J. Environ. Qual., 34, 1763–1771.


Chapter 12

Phosphate Interactions with Iron (Hydr)oxides: Mineralization Pathways and Phosphorus Retention upon Bioreduction Thomas Borch1, and Scott Fendorf2 1

Department of Soil and Crop Sciences and Department of Chemistry, Colorado State University, Fort Collins, CO 80523, USA 2 Department of Geological and Environmental Sciences, Stanford University, Stanford, CA 94305, USA

ABSTRACT Phosphorus is often the limiting nutrient in many terrestrial and aquatic/estuarian ecosystems, in part, due to strong sorption on Fe(III) (hydr)oxides. The fate of phosphate in anoxic (e.g., waterlogged) soils is often controlled by the extent of Fe(III) (hydr)oxide reductive transformation; however, the mechanisms of phosphorous retention and mobilization are not fully understood. Thus, the objectives of this study were to explore phosphate retention and transport upon bioreduction of ferrihydrite and to determine the resulting changes in both aqueous and solid phase constituents. Batch and column experiments were conducted under Fe(III) reducing and non-reducing (abiotic control) conditions to investigate phosphate retention mechanisms and ferrihydrite biomineralization pathways by dissimilatory Fe(III) reducing bacteria (Shewanella putrefaciens strain CN32 was used here) at different phosphate concentrations and at circumneutral pH. Phosphate sorption increases upon ferrihydrite bioreduction despite a concurrent decrease in specific surface area of the solid phases. Additionally, iron bioreduction under advective flow conditions results in decreased phosphate mobility compared to non-reducing environments due to sorption of phosphorus bearing secondary iron phases such as vivianite. The reduction of ferrihydrite is diminishing with increased phosphate surface coverage which is to some extent a result of phosphate-induced

Corresponding author. Tel.: +(970) 491-6235; Fax: +(970) 491-0564;

E-mail: [email protected] (T. Borch).

322

T. Borch and S. Fendorf

surface stabilization. Ferrihydrite biomineralization in the presence of phosphate results in formation of vivianite (ferrous phosphate) and green rust-like phases while inhibiting goethite and lepidocrocite production. Magnetite is formed ubiquitously although the quantity is related to both the ferrous iron concentration and adsorbed phosphate level. Our observations indicate that reductive dissolution of ferrihydrite can promote phosphate retention rather than desorption due to precipitation of ferrous bearing phosphate phases and/or increased sorption capacity of the secondary iron phases.

12.1. Introduction Iron minerals are ubiquitous in terrestrial and aquatic environments and control the fate and transport of many nutrients and contaminants via processes such as adsorption/desorption and oxidation/reduction, for example, (Lovley et al., 1989; Hansen et al., 1996; Scheinost et al., 2001; Zachara et al., 2001; Ler and Stanforth, 2003; Borch et al., 2005). For instance, Fe(III)- and Al (hydr)oxides are the primary sorbents for phosphate in the majority of soils (Pierzynski et al., 1990; Owusubennoah et al., 1997; Ruiz et al., 1997; Sallade and Sims, 1997; De Mello et al., 1998; Wilson et al., 2004). Unlike Al (hydr)oxides, however, Fe(III) (hydr)oxides can undergo reductive dissolution in anaerobic environments, releasing Fe(II), which may remain in solution, precipitate as ferrous bearing or mixed valent iron phases, or be oxidized into ferric phases such as ferrihydrite (Patrick and Henderson, 1981; Lovley et al., 1987; Stumm and Sulzberger, 1992; Roden and Zachara, 1996; Fredrickson et al., 1998; Lovley, 2000; Benner et al., 2002; Zachara et al., 2002; Burke et al., 2006). Reductive dissolution of Fe(III) (hydr)oxides can proceed by both biological (enzymatic) and abiotic means. Hydrogen sulfide controls the reductive dissolution and mineralization of Fe(III) (hydr)oxides under sulfidogenic conditions (Afonso and Stumm, 1992; Stumm and Sulzberger, 1992; Neal et al., 2001), while under non-sulfidogenic conditions dissimilatory iron reduction by bacterial genera, such as Shewanella and Geobacter, predominates (Lovley, 2000). Other sources of Fe(III) reduction include reaction with ascorbate (Afonso et al., 1990), cysteine (Cornell et al., 1989; Doong and Schink, 2002), humic acids (Kappler et al., 2004), and photolytically active reductants such as oxalate (Siffert and Sulzberger, 1991). Iron (hydr)oxide reduction within anaerobic soils and sediments may result in secondary iron phases such as magnetite (Fe3O4) (Lovley et al., 1987; X Xþ III Heron et al., 1994), green rust ð½FeII ½A2 Þ ð6XÞ FeðXÞ ðOHÞ12 X=2 yH2 O (Trolard et al., 1997), vivianite (Fe3(PO4)2 nH2O) (Postma, 1980; Zachara

Phosphate Interactions with Iron (Hydr)oxides

323

et al., 1998; Peretyazhko and Sposito, 2005), and siderite (FeCO3) (Postma, 1980; Coleman et al., 1993). The nature of the secondary iron minerals formed is influenced by numerous factors, but particularly the presence and type of adsorbates/ligands and Fe(II) concentration (Schwertmann and Taylor, 1972; Cornell et al., 1987; Fredrickson et al., 1998; Zachara et al., 2002; Hansel et al., 2003, 2005; Kukkadapu et al., 2004; Borch et al., 2007). Surface area, reactive surface site density, and mineralization products are important in that they influence the extent of Fe(III) (hydr)oxide reduction and/or overall thermodynamics of the bioreduction reaction (Roden and Zachara, 1996; Zachara et al., 1998; Liu et al., 2001; Cornell and Schwertmann, 2003; Neal et al., 2003; Roden, 2003). Ferrihydrite reductive transformation at high pH tends to increasingly favor the formation of magnetite, whereas neutral pH and high carbonate concentration favor siderite, and more acidic pH favors goethite/lepidocrocite (Zachara et al., 2002). Green rusts have also been reported as possible authogenic minerals in poorly drained soil horizons exhibiting redoxomorphic features (Genin et al., 1998), but also as biomineralization products of ferrihydrite reduction (Fredrickson et al., 1998; Kukkadapu et al., 2004; Ona-Nguema et al., 2004). Environments with concurrently high phosphate and Fe(II) concentrations may result in precipitation of vivianite (Postma, 1980). Vivianite has also been reported as a product of ferrihydrite reduction under conditions favoring high Fe(II) concentrations in the presence of aqueous phosphate (Roden and Edmonds, 1997; Fredrickson et al., 1998; Zachara et al., 1998; Glasauer et al., 2003; Kukkadapu et al., 2004; Borch et al., 2007). Many oxyanions (e.g., phosphate and arsenate) interact strongly with Fe(III) (hydr)oxides via surface complexation and may therefore affect Fe(III) reducibility and secondary mineralization pathways (Fredrickson et al., 1998; Galvez et al., 1999; Arai and Sparks, 2001; Dixit and Hering, 2003; Wilson et al., 2004; Shaw et al., 2005; Borch et al., 2007). Phosphate, for example, has a profound impact on the reduction and biotransformation of Si-ferrihydrite in the presence of anthraquinone-2,6-disulfonate (AQDS) (Kukkadapu et al., 2004); the extent of iron reduction and formation of carbonate green rust and vivianite, at the expense of magnetite, resulted upon phosphate addition (Kukkadapu et al., 2004). In contrast, phosphate inhibited or had no effect on the reducibility of synthetic and natural Fe(III) (hydr)oxides (Zachara et al., 1998), and magnetite was the sole ferrihydrite biomineralization product identified in a PIPES-buffered medium without AQDS (Fredrickson et al., 1998). Formation of vivianite has also been reported to out-compete cells for available P, while conditions low in P favored goethite formation over vivianite – an observation explained to be a consequence of insufficient Fe2+ and P to exceed the solubility of vivianite, and

324


insufficient Fe2+ for magnetite formation, as well as by P-limited cell growth (Glasauer et al., 2003). Studies in tropical and volcanic soils, however, have shown that strongly bound P is readily available to microorganisms (Olander and Vitousek, 2004). Phosphate has also been reported to retard the transformation of ferrihydrite into crystalline products such as goethite under abiotic alkaline conditions (pH>10.7) (Paige et al., 1997; Shaw et al., 2005), which is most likely due to surface stabilization of ferrihydrite (Biber et al., 1994; Galvez et al., 1999). Surface waters, such as lakes and streams, receive most of their P through particle transport (erosion) owing to the strong retention of phosphate on most soil and sediment solids (Correll, 1998). However, mobilization or leaching of P has been observed in anoxic water-logged soils and lake sediments, which can be a source of dissolved inorganic P in ground waters, aquifers, lakes, or coastal oceans (Holford and Patrick, 1979; Willett, 1985; Correll, 1998; De Mello et al., 1998; Szilas et al., 1998; Slomp and Van Cappellen, 2004; Peretyazhko and Sposito, 2005). Numerous studies have tried to link the mobility of phosphate under anoxic conditions to the reductive dissolution of Fe(III) (hydr)oxides; nevertheless, the dissolution and transport mechanisms of phosphate are not fully understood (Holford and Patrick, 1979; Ruiz et al., 1997; Sallade and Sims, 1997; De Mello et al., 1998; Szilas et al., 1998; Jensen et al., 1999; Young and Ross, 2001; Hutchison and Hesterberg, 2004; Peretyazhko and Sposito, 2005). Studies conducted under prolonged anaerobic conditions in the presence of high concentrations of Fe(III) (hydr)oxides and organic matter most often result in increased dissolved phosphate concentrations (Szilas et al., 1998; Hutchison and Hesterberg, 2004). However, an overall increase in phosphate dissolution was not observed after 209 days of continuous flooding of 7 Oxisols and 19 lowland soils. A transient increase of aqueous phosphate was, however, observed in the early stages of incubation, possibly due to the initial conversion of short-range order Fe phases into more crystalline materials (De Mello et al., 1998). In contrast to the above mentioned studies, iron-rich anaerobic sediments have also been reported to immobilize substantial amounts of phosphate under Fe(III) (hydr)oxide reducing and nonsulfidogenic conditions (Roden and Edmonds, 1997). Not withstanding appreciable work on P dynamics upon the onset of anaerobic conditions, and a general advancement in our comprehension of fate controlling processes, we have yet to fully resolve the impacts of Fe(III) reduction on dissolved concentrations of P. Accordingly, the objectives of our study were to (1) determine the impact of Fe(III) (hydr)oxide bioreduction on phosphate retention (or desorption), (2) identify the specific


325

mechanisms controlling the transport of phosphate upon biomineralization of ferrihydrite-coated quartz sand, and (3) examine the influence of sorbed phosphate on the mineralization pathway of ferrihydrite.

12.2. Experimental 12.2.1. Preparation of Bacterial Cultures Shewanella putrefaciens strain CN32, a facultative, dissimilatory iron reducing bacterium (DIRB) that couples the incomplete oxidization of lactate to acetate with Fe(III) reduction (Fredrickson et al., 1998), was used in this study and maintained in frozen stock cultures containing tryptic soy broth (TSB) amended with 20% glycerol and stored at 801C. Cells were precultured in TSB (30 g/l) at 251C and 150 rpm on a horizontal shaker for 12 h, transferred (1 ml) into fresh TSB (100 ml) and grown again for 12 h under aerobic conditions. Cultures were harvested during late-log phase by centrifugation at 4,500 g for 10 min at 51C, washed once in 100 ml of anaerobic 10 mM PIPES (piperazine-1,4-bis(2-ethanesulfonic acid) buffered synthetic groundwater medium (SGM; pH 7.2), and resuspended in groundwater medium to obtain the desired cell concentration. Cell culture densities were determined using standard DAPI staining procedures. A 1 g sample of Fe-coated sand was added to 9 ml of 10 mM sodium pyrophosphate and subjected to sonication before DAPI staining. Standard procedures for culturing anaerobic bacteria and preparation of anoxic systems were used throughout. All experiments were conducted in an anaerobic glovebag (Coy Laboratories, Inc., Grass Lake, MI) containing an anoxic gas mixture (N2 95%; H2 5%).

12.2.2. Synthetic Groundwater Medium Synthetic groundwater medium (SGM) of the following composition was used in all experiments (in mg/l): KCl (5), MgSO4 (50), NaCl (30), NH4Cl (0.95), KH2PO4 (0.95; batch systems only), PIPES (3,020), 0.1 ml Wolfe’s mineral solution (Balch et al., 1979) and finally pH adjusted to 7.2 by addition of NaOH. Sodium lactate (3 mM) was added as the electron donor. Media and buffers were made anoxic by boiling and purging with (O2-free) N2 gas.

326


12.2.3. Ferrihydrite Coated Sand and Phosphate Sorption Ferrihydrite (2-line) was prepared by rapid titration (o10 min) of a FeCl3 suspension with NaOH (0.4 M) to a pH of 7.5 (Cornell and Schwertmann, 2003) and coated on quartz sand as described previously (Hansel et al., 2003). The concentration of Fe on the sand was 6 to 8 g/Kg (or 0.11 to 0.14 mol Fe/Kg sand). The presence and purity of 2-line ferrihydrite was verified using X-ray diffraction (XRD) and X-ray absorption spectroscopy (XAS). Specific surface area of the Fe-coated sand was 2.6 m2/g (uncoated quartz sand was 0.16 m2/g based on Kr-BET) as determined by N2/Kr-BET analysis using a Beckman–Coulter SA3100 analyzer. Ferrihydrite was reacted with PO3 4 (as Na2HPO4) at concentrations from 10 mM to 2,000 mM in synthetic groundwater medium for 3 days (d). Phosphate adsorption was well described by the Langmuir equation, and the monolayer sorption (Gmax) was 0.0035 mmol P/m2 solids (9 mmol P/Kg solids or 797 mmol P/Kg ferrihydrite) (Fig. 12.1).

12.2.4. Role of Sorbed P on Fe Mineralization Reactions were performed in 125 ml serum bottles containing 100 ml of sterile (autoclaved at 1211C for 15 min) synthetic groundwater medium and 1 g of ferrihydrite-coated sand. Phosphate was coated to various degrees (50, 81, and 100% of the adsorption maximum, Gmax) on the ferrihydrite-coated

Adsorbed P (mmol Kg-1)

12 10 8 6 4 2

Γmax = 9 mmol P Kg-1 ferrihydrite coated sand

0 0.0

0.2

0.4

0.6

0.8 1.0 1.2 Aqueous P (mmol L-1)

1.4

1.6

1.8

Figure 12.1: Phosphate Adsorption on Ferrihydrite-Coated Sand. The Adsorption Data were Best Fit using the Langmuir Isotherm.


327

sand; excess P was removed by replacing the aqueous solution with fresh synthetic groundwater medium (no significant desorption of P was observed). The role of phosphate on secondary mineralization was investigated by inoculation with S. putrefaciens to a cell density of 108 m/l (biotic) or by reacting 0.2 or 2 mM ferrous sulfate (abiotic) with the (P-)Fe-coated sand for 7 d. Triplicate samples were placed on a horizontal shaker at 80 rpm at 251C and sampled periodically for Fe(II) and pH with a sterile syringe or sacrificed for solid phase analysis. Flow experiments were conducted in columns (3.8 cm diameter (ID)) with 4 sampling ports located at 1.7, 5, 8.3, and 11.7 cm along the flow path, as described previously (Hansel et al., 2003). Ferrihydrite-coated sand was reacted with phosphate to defined proportions of Gmax, washed once in synthetic groundwater medium, and then inoculated with S. putrefaciens to a cell density of 8 108 g1 or no inoculation (abiotic control). Columns were wet-packed and prepared (Hansel et al., 2003) prior to initiating flow at a rate of 373 ml/d (2.9 pore volumes) upward through the column. The pore volume was calculated based on the total column volume (283.4 cm3) and the volume of sand mixture added (153.5 cm3; the sand mixture had a bulk density of 1.35 g/m3 and 207 g was added to the column). 12.2.5. Analytical Procedures Major solutes and pH were measured as a function of time in both batch and column experiments. Upon termination of the column experiments, solids were carefully extracted and homogenized at 3.1 cm intervals, rinsed in 10 ml anoxic DI water, and dried in the glovebag. Ferrous iron was determined spectrophotometrically using the Ferrozine method (Stookey, 1970). Aqueous and solid phase (after digestion in 6 M trace metal grade HCl for 24 h) concentrations of P and Fe were determined by inductively coupled plasma optical emission spectrometry (ICP-OES). Acetate and lactate concentrations were determined by ion chromatography (IC) with an IonPac ICE-AS6 column. Samples for X-ray absorption spectroscopic analyses were sonicated (o4 h) to remove the Fe from the sand, vacuum filtered onto a cellulose acetate filter to form a homogenously distributed Fe layer, dried (30 min), and then sealed between two pieces of Kapton polyimide film to prevent oxidation while minimizing X-ray absorption. The structural environment of Fe was determined using XRD and extended X-ray absorption fine structure (EXAFS) spectroscopy at the Stanford Synchrotron Radiation Laboratory (SSRL) on beamlines 2-3 (bend magnet), 11-2 (26-pole wiggler) and 11-3

328


(XRD), running under dedicated conditions. The XAS analytical procedures used here were similar to those described previously (Hansel et al., 2003). Energy selection was accomplished with a Si(220) monochromator and spectra were recorded by fluorescent X-ray production using a Lytle detector. A set of Fe reference compounds was used to perform linear combination k3-weightened EXAFS spectral fitting using the SIXPACK interface to IFEFIT (Webb, 2005). Linear combinations of the reference compounds were optimized and the only variable parameters were the fractions of each reference compounds. Reference compounds were chosen based on their likelihood of being a reaction product (including, e.g., criteria such as elemental composition), and were included in the fit only if they contributed with a fraction of 0.05 or more. Mass balance in Fe(II)/Fe(III) and Mo¨ssbauer spectroscopy were used to constrain the LC-XAFS fitting. Detection limit for minor constituents is approximately 5%. Mo¨ssbauer spectroscopy, with a detection limit of approximately 2 wt%, was used to confirm the EXAFS analysis as described earlier (Borch et al., 2007). XRD (whole) patterns were collected at SSRL beamline 11-3 using a Kappa diffractometer in transmission geometry with a monochromatic radiation of 12,732 eV and a Quantum 4 CCD detector. The resulting images were processed using FIT2D (Hammersley, 1997). The sample-to-detector distance and geometric corrections were calculated from the pattern of LaB6. The 2D images were integrated radially to yield 1D powder diffraction patterns that could be analyzed using standard techniques. Peak identification and background subtraction were accomplished using JADE 6.5 (Materials Data, Inc., Livermore, CA).

12.3. Results 12.3.1. Phosphate Retention upon Ferrihydrite Bioreduction Release of P from soils and sediments to the aqueous phase is a function of biogeochemical factors including pH, redox potential, temperature, and microbial activity. In iron rich soils, the shift from aerobic to anaerobic conditions in response to flooding causes the reduction of Fe(III) to Fe(II), which has been reported to result in P release due to dissolution of ferric phosphate and ferric (hydr)oxide phases (Moore and Reddy, 1994; Szilas et al., 1998) and, by contrast, P sequestration via ferrous phosphate precipitation (Roden and Edmonds, 1997). Here, decreased desorption of P is observed during Fe bioreduction compared to non-reducing (abiotic)


600

120

500

P and Br (µM)

329

80

400 30 µM

40

300 0 4

6

8

10 12 14 16

200

100

0

0

5 Br tracer

10 Time (d) Abiotic

15 Biotic

Figure 12.2: Effluent Concentrations of P in the Absence (Non-Reducing Conditions) and Presence of S. putrefaciens Strain CN32 (Reducing Conditions) from Column Experiments with Phosphate-Coated Ferrihydrite. conditions (Fig. 12.2). Aqueous or loosely bound P is flushed out of the columns within approximately 3–4 d of reaction time based on the drastic change in the concentration gradient (Fig. 12.2) – the conservative bromide tracer is fully eluted from the columns after approximately 1.5 d. Phosphorus desorption curves for the biotic and abiotic (no iron reduction) columns are consistent with each other until day 3 at which point they diverge for the remaining time of the experiment. The P effluent concentration is 20–30 mM lower (after 4 days of reaction) under bioreducing conditions compared to the abiotic control (inset in Fig. 12.2). In the timeframe from 4 to 17 d, approximately 0.26 mmol P is eluted under bioreducing conditions compared to 0.43 mmol in the absence of iron reducing bacteria. The concentration of P increases gradually downstream in the biotic column within the first 10 d. Interestingly, at times >5 d of reaction (flow), P concentrations reach a plateau within the flow field (Fig. 12.3a). The solid phase concentration of P after 17 d of reaction gradually increases downstream in the column, and the amount of residual P is approximately 30% higher in the biotic column compared to non-reducing conditions (Fig. 12.3b). The trend in solid phase P concentration mirrors that observed for aqueous phase P concentrations at day 5 and day 10 under Fe reducing conditions.


0.12

1d Flow Direction

P(aq) (mM)

0.10 0.08

a

1.0 0.8

5d 10d

0.06 0.04

1.2

0.6 0.4

17d

0.2

0.02

P(aq) (mM) for 1d

330

0.0 P mmol g-1 sand

0.007 b 0.006 0.005 Biotic

0.004

Abiotic

0.003 0

5 10 Distance (cm)

15

Figure 12.3: (a) Spatial and Temporal Distribution of Phosphate along Column Length Under Iron-Reducing Conditions (Presence of S. putrefaciens) and (b) Concentration of Solid Phase P in the Absence (Abiotic NonReducing Conditions) and Presence of S. putrefaciens Strain CN32 (Bioreducing Conditions) Within the Column at Day 17. To provide more detailed insight into temporal changes of secondary phases and the linkage between Fe mineralization and P sequestration, studies were conducted using batch reactors with ferrihydrite-coated sand having a P loading of 81% of Gmax; a series of reactors were run in parallel and sacrificed at specific times. The initial aqueous P concentration of 25 mM (or 2.5 mmol) was chosen to reflect the often low dissolved P concentrations found in natural soils and sediments (Newman and Pietro, 2001). Microbial reduction of Fe(III) to Fe(II) and concurrent bio-oxidation of lactate to acetate increases appreciably after a lag phase of about 2 d (Fig. 12.4a and inset in Fig. 12.4). The P concentration in both the solid and aqueous phases remains nearly unchanged during this period (Fig. 12.4a) and is followed by a drastic decrease from 31 mM P at day 2 to 2 mM (or 0.2 mmol) P at day 7 in the aqueous phase. The removal of P from the aqueous phase transpires with an accumulation of P in the solid phase and happens concurrently with the onset of high iron respiration. There is no detectable P in the aqueous phase after 11 d of reaction; however, the microbial respiration rate does not appear to decrease, indicating that the microorganisms are not P limited (Fig. 12.4a and inset in Fig. 12.4).


2.5

80

2.0

60 40

1.5

20

1.0

0 0

40

8 4 Time (d)

P(s) Fe(II) (µmole)

6

12

5

4

Fe(s)

30 3 20

Fe(aq)

P (µmole)

50

3.0

Acetate BET

100

BET surface area B (m2 g-1solid)

Acetate (µM)

120

331

2

P(aq) 10

1 a

15

75

10

50 Vivianite Magnetite

5

25

Green Rust

Ferrihydrite Residual (mole % Fe)

mole % Fe

Ferrihydrite

b 0

0 0

2

4

6 Time (d)

8

10

12

Figure 12.4: (a) Aqueous and Solid Phase Masses of Phosphate (P) and Ferrous Iron (Fe(II)) (Inset Shows the Formation of Acetate (Biooxidation Product of Lactate) and Changes in Specific Surface Area (BET)) and (b) Percents of Fe Phases (Mole-Basis) Resulting from Linear Combination Fits of k3-Weighted EXAFS Spectra (Data are 75% and the Detection Limit is 5 mole% Fe) as a Function of Time in Batch (Static Flow) Incubations with Phosphate-Coated (81% of Gmax) Ferrihydrite-Sand and S. putrefaciens.

332


Iron reduction results in 30% conversion of ferrihydrite into secondary Fe mineralization products after 11 d of reaction. Magnetite is the sole secondary Fe phase present after 4 d while both magnetite (15 mol% Fe) and a green rust-like phase (6 mol% Fe) are detected after 7 d (Fig. 12.4b). Precipitation of vivianite (7 mol% Fe) is observed after 11 d and is correlated with the high Fe(II) concentration and absence of aqueous P (Fig. 12.4). The specific surface area (BET) of the solids decreases from 2.6 m2/g to less than 1.5 m2/g solid during the course of ferrihydrite mineralization (inset in Fig. 12.4). 12.3.2. Impact of Adsorbed Phosphate on Ferrihydrite Reduction and Mineralization Pathway The effect of varying phosphate coverage under (a)biotic conditions was investigated in batch systems to improve our understanding of how strongly sorbing oxyanions such as phosphate (P) impact the reduction and mineralization pathway of ferrihydrite. Decreasing production of aqueous Fe(II) is observed with increasing P coverage during bioreduction of ferrihydrite by S. putrefaciens within a time period of 7 d (Fig. 12.5a). The formation of Fe(II)(aq) ceases after 6 days of reaction at 100% P coverage (Gmax). Bacterially induced Fe(III) reduction (R2 ¼ 0.993) and ferrihydrite transformation (R2 ¼ 0.998) are linearly correlated with phosphate coverage after 7 d of reaction, and a more than three-fold decrease in the extent of transformation is observed at P coverage equal to Gmax (Fig. 12.5b). The influence of adsorbed P on the mineralization pathway of ferrihydrite is substantial. Goethite, magnetite, and green rust are the major Fe biomineralization products in the absence of P (Table 12.1). When ferrihydrite is loaded with P at 50% of Gmax, only magnetite and green rust are observed, while a small amount of a green rust-like phase is the sole secondary Fe solid at 100% of Gmax. The role of microbial metabolism and P on the mineralization pathway of ferrihydrite was further evaluated by reacting 0.2 mM or 2.0 mM FeSO4 (Fe(II)) with ferrihydrite coated sand (Table 12.1). In the absence of P, reaction of Fe(II) with ferrihydrite yields lepidocrocite as the primary product at low FeSO4 concentration (0.2 mM), with small amounts of goethite and magnetite, while higher FeSO4 concentration (2.0 mM) yields goethite and magnetite as the main products. By contrast, goethite formation is clearly inhibited in the presence of surface-associated P, and the formation of green rust and vivianite seem to be favored (Table 12.1). Within continuous flow columns, effluent concentration profiles for acetate mirror that of lactate; Fe(II)aq concentrations contrast those of acetate and steadily increase throughout the experiment The concentration of


333

200 (of Γmax)

Fe2+ (aq) (µM)

150

0%P 50 % P

100

100 % P

50

0 2

0 a

4 Time (d)

6

8

at day 7 R2 = 0.998

95

40 85 30 20 75 Fe(II)total

10

Ferrihydrite Residual (%)

Fe(II) total (µmole)

50

Ferrihydrite 0 0 b

50 Percent of Adsorption Maximum P

65 100

Figure 12.5: (a) Aqueous Ferrous Iron Production as a Function of Time and (b) Total Ferrous Iron Production and Residual Ferrihydrite after 7 d of Reaction in a Batch System with Ferrihydrite-Coated Sand and Dissimilatory Iron Reducing Bacteria (S. putrefaciens Strain CN32) as a Function of Phosphate Coverage (% of Gmax). Aqueous Data are the Average of Triplicate Measurements and Error Bars Represent one Standard Deviation.

334

Phosphate coverage (%)a 0 50 100 0 100 0 50 100 a

S. putrefaciens S. putrefaciens S. putrefaciens 0.2 mM Fe(II) 0.2 mM Fe(II) 2.0 mM Fe(II) 2.0 mM Fe(II) 2.0 mM Fe(II)

% of Gmax. Valueso5 mol% Fe are not reported.

b

Mole% Feb

Biotic/abiotic reduction Ferrihydrite

Goethite

Lepidocrocite

Magnetite

Green rust

Vivianite

69 82 92 56 89 41 79 77

5 0 0 5 0 31 0 0

0 0 0 29 6 0 0 0

12 7 0 10 0 28 9 11

11 11 5 0 5 0 12 6

0 0 0 0 0 0 0 6


Table 12.1: Shewanella putrefaciens Strain CN32 (Biotic) and Fe(II)-Induced (Abiotic; using FeIISO4) Conversion of Ferrihydrite as a Function of Fe(II) Concentration or Phosphate Coverage Following 7 d of Reaction.


335

Fe(II)aq increases along the first 8 cm of the flow path, at which point its concentration reaches a maximum (indicating that production diminishes) (Fig. 12.6a). Only a minor fraction of Fe (3% or 0.6 mmol Fe) elutes from the column during the 17 d of reaction. Ferrous iron concentration in the solid phase increases within the first 8 cm of the flow path and then decreases at 12 cm (Fig. 12.6b). Mass balance between lactate oxidation (1.3 mmol lactate oxidized) and Fe(III) reduction (5.2 mmol Fe(III) reduced) is obtained, verifying that Fe(III) was the sole electron acceptor used in the respiration process. The number of bacterial cells (dead and alive) in the column 0.2

17d

mmol Fe(II) g-1 sand

Fe(II)aq (mM)

Flow

0.15

10d

a

0.1

5d

0.05

1d

0.030

b

0.025 0.020 0.015 0.010 0.005 Ferrihydrite

70

mole % Fe

60

c

50 Magnetite

40 30 20

Vivianite

10

Green Rust

0 0

5

10

15

Distance (cm)

Figure 12.6: (a) Spatial and Temporal Distribution of Aqueous Fe(II)(aq) along the Column Length. (b) Concentration of Solid Phase Fe(II), and (c) Percents of Fe Phases (Mole-Basis) Resulting from Linear Combination Fits of k3-Weighted EXAFS Spectra (Data are 75% and the Detection Limit is 5 mole% Fe).

336


6

Intens ity (%)

5 4 3 2 1 0 65-3487> Quartz S-alpha - Si O2

Quartz

64-6720> Magnetite - Fe2.939 O4

Magnetite

5.59

2.8

1.88

1.42

1.15

d-Scale (Å)

Figure 12.7: Synchrotron-Radiation Based X-ray Diffraction Pattern of Solids Resulting from Biomineralization of Phosphated Ferrihydrite 1.67 cm from the Column Inlet after 17 d of Reaction. was quite consistent throughout the column after 17 d of reaction with less than one order of magnitude of cells eluted from the column during the experiment (Borch et al., 2007). Ferrihydrite is primarily converted to magnetite throughout the column as revealed by EXAFS spectroscopy (Fig. 12.6c) and XRD (Fig. 12.7). However, linear combination fits of the EXAFS spectra also indicate the presence of vivianite- and green rust-like phases downstream in the column, with the highest concentration observed 8.3 cm from the inlet (Fig. 12.6c). The amount of Fe(II) estimated using linear combination fit of the solid-phase products throughout the column correlates well with the Fe(II) concentration measured by the ferrozine method (Fig. 12.6).

12.4. Discussion 12.4.1. Enhanced Phosphate Retention upon Ferrihydrite Biomineralization Bioreduction of ferrihydrite increases sequestration and decreases transport of phosphate relative to non-reducing conditions (Figs. 12.2–12.4). The lower effluent P concentration under iron reducing conditions is likely a


337

result of precipitation of ferrous bearing phosphate phases such as vivianite and retention on/in secondary mineralization products (Figs. 12.2 and 12.6c). Phosphorus attenuation upon Fe bioreduction (Fig. 12.2) is further illustrated by the gradually increasing solid phase P concentration in the downstream portions of the column (Fig. 12.3), a product consistent with high Fe(II) and P concentrations (Figs. 12.4 and 12.6). In contrast to this study, dissolution of Fe(III) (hydr)oxides in soils under anaerobic conditions has commonly been reported to cause increased desorption of phosphate (Sallade and Sims, 1997; Phillips, 1998; Hutchison and Hesterberg, 2004; Shenker et al., 2005). Thus, there is often a concern for increased loss of phosphate from waterlogged soils that contain high levels of P due to the more reduced conditions (Hutchison and Hesterberg, 2004). Increased retention of phosphate observed in this study is, however, similar to a select set of studies showing a decrease in dissolved phosphate upon the onset of anaerobic soil conditions (Khalid et al., 1977; Holford and Patrick, 1979). Phosphate leaching studies from intact soil columns in response to reducing conditions showed that amendments of high glucose concentrations (increased microbial metabolism) resulted in an increase in Fe(II) concentrations from 0.05 mM to 0.25 mM and significant fluctuations in the effluent phosphate concentration between 0.07 and 3.22 mmol P/l. The fluctuation in the P effluent concentration was ascribed, in part, to adsorption–desorption processes (Jensen et al., 1999). In addition, anoxic batch studies of loamy soils were incubated for 29 d and indicated that the concentrations of mobilized P was not correlated to total soil-P or Fe(II) concentration. The lack of proportionality between Fe(II) and P in solution was attributed to microbial uptake, re-adsorption, and precipitation of Fe(II)–P compounds such as vivianite (Jensen et al., 1998). The highest soluble P concentrations were observed in samples amended with low glucose amendments, likely resulting in Fe(II) and P concentrations limiting vivianite precipitation (Jensen et al., 1998). Slight to moderate reduction of a silt loam soil was illustrated to have minimal influence on labile phosphate and phosphate sorption; however, reduction to –150 mV caused marked increases in both labile P and the P sorption capacity of the soil (Holford and Patrick, 1979). The increased sorption capacity of the soil was explained by the hypothesized conversion of Fe(III) (hydr)oxides to secondary ferrous bearing phases with a more amorphous structure and higher sorption capacity than the native Fe(III) (hydr)oxides (Holford and Patrick, 1979). The differences in sorption and release of P between oxidized and reduced soils and sediments have previously been explained by the fact that ferric (hydr)oxides bind phosphate more strongly (higher sorption affinity) but have lower surface areas than hydrated ferrous-bearing phases (Patrick and Khalid, 1974).

338


We observed decreasing surface area concurrent with the formation of more crystalline secondary Fe phases such as magnetite upon the onset of anaerobic transformations (inset in Fig. 12.4 and Fig. 12.4b). Despite a general decrease in surface area and no clear formation of ferrous phosphate phases within the first 7 d of reaction, phosphate was retained more extensively than in abiotic counterparts. While we cannot rule out the possible formation of short-ordered ferrous (hydr)oxides with high phosphate sorption capacity, we have observed high phosphate sorption capacities (surface area normalized) of mixed ferrous–ferric iron phases (e.g., green rust and magnetite) (data not shown). Phosphate has also been reported to be retained in the interlayer of green rust, followed by a conversion into vivianite (Hansen and Poulsen, 1999). The presence of vivianite after 11 d of reaction corresponds with high Fe(II) concentration, in agreement with previous studies (Postma, 1980; Fredrickson et al., 1998; Zachara et al., 1998; Hansen and Poulsen, 1999; Glasauer et al., 2003; Kukkadapu et al., 2004; Borch et al., 2007), and results from oversaturation of Fe(II)–phosphate rather than conversion of green rust phases. 12.4.2. Impact of Phosphate on Ferrihydrite Reduction Inhibition of Fe(II) production is directly linked to phosphate loading of ferrihydrite, illustrating the impact of a stabilizing surface ligand on bioreduction (and resulting secondary mineralization). Surface complexes, especially those formed by the adsorption of oxyanions such as phosphate, arsenate, and borate are expected to inhibit dissolution, in part, because a higher activation energy is needed to destabilize and remove the metal centers of the solid surface (Willett, 1985; Biber et al., 1994; Paige et al., 1997; Galvez et al., 1999; Shaw et al., 2005). Although, ligand-promoted dissolution of ferrihydrite by tripolyphosphate has been reported (Lin and Benjamin, 1990), previous biological studies, however, showed that phosphate had only a minimal influence on reduction of natural Fe(III) (hydr)oxides (Zachara et al., 1998). Further, addition of phosphate to the aqueous phase stimulated bioreduction of synthetic Si-ferrihydrite coprecipitates in the presence of the soluble electron shuttle AQDS (Kukkadapu et al., 2004). In contrast, phosphate inhibited bioreduction and mineralization of Nisubstituted (5 mol%) ferrihydrite (Fredrickson et al., 2001). Long-term (80 d) batch experiments of S. putrefaciens and ferrihydrite in the presence of 20 mM lactate showed that initially high aqueous concentrations of phosphate (4.0 mM; the P surface-coverage was not determined) resulted in higher Fe dissolution (increased Fe(II)(aq)) relative to systems with low


339

aqueous concentrations of phosphate (0.4 mM) despite that the bacterial numbers were approximately equal in the two systems (Glasauer et al., 2003). Other long-term (75 d) studies of microbial ferrihydrite reduction were conducted in bicarbonate and PIPES buffered systems containing high lactate concentrations (30 mM) in the presence and absence of 4 mM P (phosphate surface-coverage was not determined but the aqueous P concentration was near analytical detection after 4 d of reaction) and did not show any influence of P on acid extractable Fe(II) (Fredrickson et al., 1998). The extent of iron reduction in our experiments is, however, similar to others with synthetic ferrihydrite (Zachara et al., 1998), showing that only 13% was bioreduced in the presence of 4 mM phosphate after 39 d of reaction with S. putrefaciens. Our studies (Table 12.1) are also in agreement with studies of P-coated ferrihydrite in flooded sand–rice straw mixtures, which show that the addition of phosphate at levels less than the sorption capacity to reduced systems containing ferrihydrite will result in P sorption and an increase in the stability of the oxide – as opposed to increase in the dissolution of ferrihydrite by vivianite formation (Willett, 1985). The different effects of phosphate (i.e., inhibited vs. enhanced Fe dissolution) observed among the various studies, including the present, may be explained by the way each study was conducted (e.g., mode of P addition and P surface/aqueous concentration), preparation of cell cultures, and ferrihydrite preparation and characteristics (aggregates vs. Fe-coated sand). Furthermore, iron bioreduction can differ in bicarbonate compared to PIPES buffer (Fredrickson et al., 1998; Hansel et al., 2005), low versus high electron donor to acceptor ratio (Fredrickson et al., 2003), and in the presence of readily available phosphate (Glasauer et al., 2003). This study, compared to other studies with S. putrefaciens, used 7–10 times lower lactate concentrations, and phosphate was sorbed to the ferrihydrite-coated sand followed by a rinse prior to inoculation of the samples, resulting in low aqueous P concentrations and a surface coverage below the oxide’s sorption capacity. 12.4.3. Impact of Phosphate on the Ferrihydrite (Bio)mineralization Pathway Phosphate adsorbed to ferrihydrite not only altered the extent of Fe reduction but also the biomineralization pathway. The extent of ferrihydrite transformation and Fe(II) production both decrease by a factor of four in the presence of phosphate at maximum surface coverage (Fig. 12.5b). In contrast to mineralization products in the absence of P, the presence of surface-sorbed phosphate prevents the production of goethite during biotic and abiotic mineralization of ferrihydrite (Table 12.1 and Fig. 12.8).

340


a

Ferrihydrite

b

P Ferrihydrite reduction

reduction

Fe(II)

Fe(II) ferrihydrite Fe(II) < 0.7 mmol/Kg

P ferrihydrite Fe(II)

Fe(II) Fe(II) > 0.7 mmol/Kg

goethite/ goethite/ lepidocrocite lepidocrocite

Fe(II) < 0.7 mmol/Kg

magnetite green rust (lepidocrocite)

Fe(II) > 0.7 mmol/Kg SIviv>0 green rust

magnetite vivianite ma

Figure 12.8: Ferrihydrite (Bio)mineralization Pathways Observed in this Study in the Absence (a) and Presence of Surface-Sorbed Phosphate (b) at Low and High Ferrous Iron Concentration (Modified from Hansel et al., 2005). The Rate and Extent of Ferrihydrite Reduction and Transformation is Higher in the Absence of Surface-Sorbed Phosphate (a). Formation of magnetite and green rust, although not the proportions, as mineralization products is in agreement with previous studies performed in a minimal growth medium (Glasauer et al., 2003); they differ, however, from the rather high quantities of vivianite and siderite formed under conditions where the growth medium is optimized to achieve more rapid and extensive reduction (Fredrickson et al., 1998). Lepidocrocite is the dominant secondary phase in the absence of phosphate at low Fe(II) concentrations under abiotic and ligand conditions of this study, while goethite and magnetite are formed in equal amounts at high concentrations of Fe(II) (2.0 mM) (Table 12.1 and Fig. 12.8), in agreement with previous studies (Hansel et al., 2005). While 44 and 59% of the ferrihydrite is being mineralized in the absence of phosphate, only 10–20% of the ferrihydrite is transformed in the presence of phosphate (Table 12.1). This result supports the argument that phosphate stabilizes the structure of ferrihydrite by hindering reductive dissolution, poisons goethite formation, thus promoting the formation of green rust and vivianite – the latter only at concurrently high phosphate concentrations (Fig. 12.8). Carbonate green rust was observed under biotic conditions in the absence of phosphate and is likely a result of microbially produced (and accumulated) bicarbonate alkalinity, which can reach high local concentrations under static flow conditions. Under dynamic flow conditions, the temporal variation in Fe(II)(aq) concentration differs between systems with and without phosphate. With phosphate present (100% of Gmax), Fe(II) gradually increases with time (Fig. 12.6a); in the absence of phosphate, high aqueous Fe(II) concentrations


341

form initially (to5 d) followed by a gradual decrease with time (Hansel et al., 2003). The fast initial reduction of ferrihydrite followed by a decrease in reduction rate in the absence of P results from rapid ferrihydrite transformation into highly crystalline secondary minerals such as goethite and magnetite. Ferrihydrite remains the dominant iron phase after 17 d of reaction with magnetite being the principal secondary phase; only small quantities of vivianite- and/or green rust-like phases form downstream in the column (Fig. 12.6c). In contrast to abundant formation of goethite in the absence of phosphate (in both column and (a)biotic batch studies) (Table 12.1), goethite was not formed in the presence of phosphate. The absence of goethite is most likely due to stabilization of ferrihydrite (and resulting decreases in dissolution) rather than by interfering with nucleation and growth of goethite (Biber et al., 1994; Paige et al., 1997; Galvez et al., 1999).

12.5. Conclusion and Implications Numerous environmental implications of Fe(III) (hydr)oxide biomineralization have been suggested but are primarily linked to changes in aqueous– solid phase partitioning of contaminants and nutrients (Lovley et al., 1989; Heijman et al., 1995; Roden and Edmonds, 1997; Fredrickson et al., 2000, 2001, 2004; Mccormick et al., 2001; Zachara et al., 2001; Roden et al., 2002; Jeon et al., 2005; Peretyazhko and Sposito, 2005; Stone et al., 2006; Wu et al., 2006). When evaluating the environmental implications of Fe(III) reduction, however, it is important to appreciate the biomineralization pathways (Fig. 12.8). For example, while Fe(III) (hydr)oxide reduction is generally perceived to mobilize phosphate under anaerobic conditions (Holford and Patrick, 1981; Sallade and Sims, 1997; De Mello et al., 1998; Jensen et al., 1998; Szilas et al., 1998; Young and Ross, 2001), we, as well as others (Khalid et al., 1977; Roden and Edmonds, 1997; Shenker et al., 2005; Murray and Hesterberg, 2006; Borch et al., 2007), demonstrate that it depends on the extent of iron biomineralization and secondary Fe–P phases. Thus, we must consider factors that impact microbial activity (e.g., temperature and moisture), the nature of the native Fe(III) (hydr)oxides, and concentration and type of soil organic matter (electron donor). In the case of phosphate, our studies, for example, illustrate slow and limited biomineralization of phosphate-coated ferrihydrite, even when supplied with an electron donor (lactate) concentration of 3 mM; ‘‘leaner’’ conditions common to most natural environments would likely exaggerate the limited reduction of phosphated ferrihydrite.

342


ACKNOWLEDGMENTS We thank Yoko Masue (Stanford University (SU)), Danny Richter (The Scripps Research Institute), Dr. Guangchao Li (SU), and Dr. Colleen Hansel (Harvard University) for their invaluable help on this project. The authors acknowledge three anonymous reviewers and the editors of this book for their helpful comments during manuscript review. This research was supported by U.S. Department of Energy’s ERSD program (grant number ER63609-1021814) and the Stanford NSF Environmental Molecular Sciences Institute (grant number NSF-CHE-0431425). Portions of this research were carried out at the Stanford Synchrotron Radiation Laboratory, a national user facility operated by Stanford University on behalf of the U.S. Department of Energy, Office of Basic Energy Sciences. Mo¨ssbauer analysis was performed in the Environmental Molecular Sciences Laboratory, a national scientific user facility sponsored by the Department of Energy’s Office of Biological and Environmental Research, located at Pacific Northwest National Laboratory. We are grateful to Ravi Kukkadapu for his help with the Mo¨ssbauer analysis.

REFERENCES Afonso, M. D., Morando, P. J., Blesa, M. A., Banwart, S., & Stumm, W. (1990). Reductive dissolution of iron oxides by ascorbate. The role of carboxylate anions in accelerating reductive dissolution. J. Colloid Interface Sci., 138, 74–82. Afonso, M. D., & Stumm, W. (1992). Reductive dissolution of iron(III) (hydr)oxides by hydrogen-sulfide. Langmuir, 8, 1671–1675. Arai, Y., & Sparks, D. L. (2001). ATR-FTIR spectroscopic investigation on phosphate adsorption mechanisms at the ferrihydrite–water interface. J. Colloid Interface Sci., 241, 317–326. Balch, W. E., Fox, G. E., Magrum, L. J., Woese, C. R., & Wolfe, R. S. (1979). Methanogens reevaluation of a unique biological group. Microbiol. Rev., 43, 260–296. Benner, S. G., Hansel, C. M., Wielinga, B. W., Barber, T. M., & Fendorf, S. (2002). Reductive dissolution and biomineralization of iron hydroxide under dynamic flow conditions. Environ. Sci. Technol., 36, 1705–1711. Biber, M. V., Dos Santos Afonso, M., & Stumm, W. (1994). The coordination chemistry of weathering: IV. Inhibition of the dissolution of oxide minerals. Geochim. Cosmochim. Acta, 58, 1999–2010. Borch, T., Inskeep, W. P., Harwood, J. A., & Gerlach, R. (2005). Impact of ferrihydrite and anthraquinone-2,6-disulfonate on the reductive transformation of 2,4,6-trinitrotoluene by a gram-positive fermenting bacterium. Environ. Sci. Technol., 39, 7126–7133.


343

Borch, T., Masue, Y., Kukkadapu, R. K., & Fendorf, S. (2007). Phosphate imposed limitations on biological reduction and alteration of ferrihydrite. Environ. Sci. Technol., 41, 166–172. Burke, I. T., Boothman, C., Lloyd, J. R., Livens, F. R., Charnock, J. M., Mcbeth, J. M., Mortimer, R. J. G., & Morris, K. (2006). Reoxidation behavior of technetium, iron, and sulfur in estuarine sediments. Environ. Sci. Technol., 40, 3529–3535. Coleman, M. L., Hedrick, D. B., Lovley, D. R., White, D. C., & Pye, K. (1993). Reduction of Fe(III) in sediments by sulphate-reducing bacteria. Nature, 361, 436–438. Cornell, R. M., Giovanoli, R., & Schindler, P. W. (1987). Effect of silicate species on the transformation of ferrihydrite into goethite and hematite in alkaline media. Clays Clay Miner., 35, 21–28. Cornell, R. M., Schneider, W., & Giovanoli, R. (1989). Phase transformations in the ferrihydrite/cysteine system. Polyhedron, 8, 2829–2836. Cornell, R. M., & Schwertmann, U. (2003). The Iron Oxides: Structure, Properties, Reactions, Occurrence and Uses. VCH, Weinheim, Germany. Correll, D. L. (1998). The role of phosphorus in the eutrophication of receiving waters: A review. J. Environ. Qual., 27, 261–266. De Mello, J. W. V., Barron, V., & Torrent, J. (1998). Phosphorus and iron mobilization in flooded soils from Brazil. Soil Sci., 163, 122–132. Dixit, S., & Hering, J. G. (2003). Comparison of arsenic(V) and arsenic(III) sorption onto iron oxide minerals: Implications for arsenic mobility. Environ. Sci. Technol., 37, 4182–4189. Doong, R. A., & Schink, B. (2002). Cysteine-mediated reductive dissolution of poorly crystalline iron(III) oxides by Geobacter sulfurreducens. Environ. Sci. Technol., 36, 2939–2945. Fredrickson, J. K., Kota, S., Kukkadapu, R. K., Liu, C. X., & Zachara, J. M. (2003). Influence of electron donor/acceptor concentrations on hydrous ferric oxide (HFO) bioreduction. Biodegradation, 14, 91–103. Fredrickson, J. K., Zachara, J. M., Kennedy, D. W., Dong, H., Onstott, T. C., Hinman, N. W., & Li, S.-M. (1998). Biogenic iron mineralization accompanying the dissimilatory reduction of hydrous ferric oxide by a groundwater bacterium. Geochim. Cosmochim. Acta, 62, 3239–3257. Fredrickson, J. K., Zachara, J. M., Kennedy, D. W., Duff, M. C., Gorby, Y. A., Li, S.-M., & Krupka, K. M. (2000). Reduction of U(VI) in goethite (a-FeOOH) suspensions by a dissimilatory metal-reducing bacterium. Geochim. Cosmochim. Acta, 64, 3085–3098. Fredrickson, J. K., Zachara, J. M., Kennedy, D. W., Kukkadapu, R. K., Mckinley, J. P., Heald, S. M., Liu, C., & Plymale, A. E. (2004). Reduction of TcO4 by sediment-associated biogenic Fe(II). Geochim. Cosmochim. Acta, 68, 3171–3187. Fredrickson, J. K., Zachara, J. M., Kukkadapu, R. K., Gorby, Y. A., Smith, S. C., & Brown, C. F. (2001). Biotransformation of Ni-substituted hydrous ferric oxide by an Fe(III)-reducing bacterium. Environ. Sci. Technol., 35, 703–712.

344


Galvez, N., Barron, V., & Torrent, J. (1999). Effect of phosphate on the crystallization of hematite, goethite, and lepidocrocite from ferrihydrite. Clays Clay Miner., 47, 304–311. Genin, J. M. R., Bourrie, G., Trolard, F., Abdelmoula, M., Jaffrezic, A., Refait, P., Maitre, V., Humbert, B., & Herbillon, A. (1998). Thermodynamic equilibria in aqueous suspensions of synthetic and natural Fe(II)-Fe(III) green rusts: Occurrences of the mineral in hydromorphic soils. Environ. Sci. Technol., 32, 1058–1068. Glasauer, S., Weidler, P. G., Langley, S., & Beveridge, T. J. (2003). Controls on Fe reduction and mineral formation by a subsurface bacterium. Geochim. Cosmochim. Acta, 67, 1277–1288. Hammersley, A. P. (1997). FIT2D: An Introduction and Overview, European Synchrotron Radiation Facility, Grenoble, France. Hansel, C. M., Benner, S. G., & Fendorf, S. (2005). Competing Fe(II)-induced mineralization pathways of ferrihydrite. Environ. Sci. Technol., 39, 7147–7153. Hansel, C. M., Benner, S. G., Neiss, J., Dohnalkova, A., Kukkadapu, R. K., & Fendorf, S. (2003). Secondary mineralization pathways induced by dissimilatory iron reduction of ferrihydrite under advective flow. Geochim. Cosmochim. Acta, 67, 2977–2992. Hansen, H. C. B., Koch, C. B., Nanckekrogh, H., Borggaard, O. K., & Sorensen, J. (1996). Abiotic nitrate reduction to ammonium: Key role of green rust. Environ. Sci. Technol., 30, 2053–2056. Hansen, H. C. B., & Poulsen, I. F. (1999). Interaction of synthetic sulphate ‘‘green rust’’ with phosphate and the crystallization of vivianite. Clays Clay Miner., 47, 312–318. Heijman, C. G., Grieder, E., Holliger, C., & Schwarzenbach, R. P. (1995). Reduction of nitroaromatic compounds coupled to microbial iron reduction in laboratory aquifer columns. Environ. Sci. Technol., 29, 775–783. Heron, G., Christensen, T. H., & Tjell, J. C. (1994). Oxidation capacity of aquifer sediments. Environ. Sci. Technol., 28, 153–158. Holford, I. C. R., & Patrick, W. H. (1979). Effects of reduction and pH changes on phosphate sorption and mobility in an acid soil. Soil Sci. Soc. Am. J., 43, 292–297. Holford, I. C. R., & Patrick, W. H. (1981). Effects of duration of anaerobiosis and reoxidation on phosphate sorption characteristics of an acid soil. Aust. J. Soil Res., 19, 69–78. Hutchison, K. J., & Hesterberg, D. (2004). Dissolution of phosphate in a phosphorus-enriched ultisol as affected by microbial reduction. J. Environ. Qual., 33, 1793–1802. Jensen, M. B., Hansen, H. C. B., Nielsen, N. E., & Magid, J. (1998). Phosphate mobilization and immobilization in two soils incubated under simulated reducing conditions. Acta Agric. Scand. B Soil Plant Sci., 48, 11–17. Jensen, M. B., Hansen, H. C. B., Nielsen, N. E., & Magid, J. (1999). Phosphate leaching from intact soil column in response to reducing conditions. Water Air Soil Pollut., 113, 411–423.


345

Jeon, B. H., Dempsey, B. A., Burgos, W. D., Barnett, M. O., & Roden, E. E. (2005). Chemical reduction of U(VI) by Fe(II) at the solid–water interface using natural and synthetic Fe(III) oxides. Environ. Sci. Technol., 39, 5642–5649. Kappler, A., Benz, M., Schink, B., & Brune, A. (2004). Electron shuttling via humic acids in microbial iron(III) reduction in a freshwater sediment. FEMS Microbiol. Ecol., 47, 85–92. Khalid, R. A., Patrick, W. H., & Delaune, R. D. (1977). Phosphorus sorption characteristics of flooded soils. Soil Sci. Soc. Am. J., 41, 305–310. Kukkadapu, R. K., Zachara, J. M., Fredrickson, J. K., & Kennedy, D. W. (2004). Biotransformation of two-line silica-ferrihydrite by a dissimilatory Fe(III)-reducing bacterium: Formation of carbonate green rust in the presence of phosphate. Geochim. Cosmochim. Acta, 68, 2799–2814. Ler, A., & Stanforth, R. (2003). Evidence for surface precipitation of phosphate on goethite. Environ. Sci. Technol., 37, 2694–2700. Lin, C. F., & Benjamin, M. M. (1990). Dissolution kinetics of minerals in the presence of sorbing and complexing ligands. Environ. Sci. Technol., 24, 126–134. Liu, C. X., Kota, S., Zachara, J. M., Fredrickson, J. K., & Brinkman, C. K. (2001). Kinetic analysis of the bacterial reduction of goethite. Environ. Sci. Technol., 35, 2482–2490. Lovley, D. R. (2000). Fe(III) and Mn(IV) reduction. In: D. R. Lovley (Ed). Environmental Microbe-Metal Interactions. ASM Press, Washington, DC. Lovley, D. R., Baedecker, M. J., Lonergan, D. J., Cozzarelli, I. M., Phillips, E. J. P., & Siegel, D. I. (1989). Oxidation of aromatic contaminants coupled to microbial iron reduction. Nature, 339, 297–299. Lovley, D. R., Stolz, J. F., Nord, G. L. J., & Phillips, E. J. P. (1987). Anaerobic production of magnetite by a dissimilatory iron-reducing microorganism. Nature, 330, 252–254. Mccormick, M. L., Bouwer, E. J., & Adriaens, P. (2001). Carbon tetrachloride transformation in a model iron-reducing culture: Relative kinetics of biotic and abiotic reactions. Environ. Sci. Technol., 36, 403–410. Moore, P. A., & Reddy, K. R. (1994). Role of Eh and pH on phosphorus geochemistry in sediments of Lake Okeechobee, Florida. J. Environ. Qual., 23, 955–964. Murray, G. C., & Hesterberg, D. (2006). Iron and phosphate dissolution during abiotic reduction of ferrihydrite-boehmite mixtures. Soil Sci. Soc. Am. J., 70, 1318–1327. Neal, A. L., Rosso, K. M., Geesey, G. G., Gorby, Y. A., & Little, B. J. (2003). Surface structure effects on direct reduction of iron oxides by Shewanella oneidensis. Geochim. Cosmochim. Acta, 67, 4489–4503. Neal, A. L., Techkarnjanaruk, S., Dohnalkova, A., McCready, D., Peyton, B. M., & Geesey, G. G. (2001). Iron sulfides and sulfur species produced at hematite surfaces in the presence of sulfate-reducing bacteria. Geochim. Cosmochim. Acta, 65, 223–235.

346


Newman, S., & Pietro, K. (2001). Phosphorus storage and release in response to flooding: Implications for everglades stormwater treatment areas. Ecol. Eng., 18, 23–38. Olander, L. P., & Vitousek, P. M. (2004). Biological and geochemical sinks for phosphorus in soil from a wet tropical forest. Ecosystems, 7, 404–419. Ona-Nguema, G., Carteret, C., Benali, O., Abdelmoula, M., Genin, J. M., & Jorand, F. (2004). Competitive formation of hydroxycarbonate green rust 1 versus hydroxysulphate green rust 2 in Shewanella putrefaciens cultures. Geomicrobiol. J., 21, 79–90. Owusubennoah, E., Szilas, C., Hansen, H. C. B., & Borggaard, O. K. (1997). Phosphate sorption in relation to aluminum and iron oxides of oxisols from Ghana. Commun. Soil Sci. Plant Anal., 28, 685–697. Paige, C. R., Snodgrass, W. J., Nicholson, R. V., Scharer, J. M., & He, Q. H. (1997). Effect of phosphate on the transformation of ferrihydrite into crystalline products in alkaline media. Water Air Soil Pollut., 97, 397–412. Patrick, W. H., & Henderson, R. E. (1981). Reduction and reoxidation cycles of manganese and iron in flooded soil and in water solution. Soil Sci. Soc. Am. J., 45, 855–859. Patrick, W. H., & Khalid, R. A. (1974). Phosphate release and sorption by soils and sediments – effect of aerobic and anaerobic conditions. Science, 186, 53–55. Peretyazhko, T., & Sposito, G. (2005). Iron(III) reduction and phosphorous solubilization in humid tropical forest soils. Geochim. Cosmochim. Acta, 69, 3643–3652. Phillips, I. R. (1998). Phosphorus availability and sorption under alternating waterlogged and drying conditions. Commun. Soil Sci. Plant Anal., 29, 3045–3059. Pierzynski, G. M., Logan, T. J., Traina, S. J., & Bigham, J. M. (1990). Phosphorus chemistry and mineralogy in excessively fertilized soils. Descriptions of phosphorus-rich particles. Soil Sci. Soc. Am. J., 54, 1583–1589. Postma, D. (1980). Formation of siderite and vivianite and the pore-water composition of a recent bog sediment in Denmark. Chem. Geol., 31, 225–244. Roden, E. E. (2003). Fe(III) oxide reactivity toward biological versus chemical reduction. Environ. Sci. Technol., 37, 1319–1324. Roden, E. E., & Edmonds, J. W. (1997). Phosphate mobilization in iron-rich anaerobic sediments: Microbial Fe(III) oxide reduction versus iron-sulfide formation. Arch. Hydrobiol., 139, 347–378. Roden, E. E., Leonardo, M. R., & Ferris, F. G. (2002). Immobilization of strontium during iron biomineralization coupled to dissimilatory hydrous ferric oxide reduction. Geochim. Cosmochim. Acta, 66, 2823–2839. Roden, E. E., & Zachara, J. M. (1996). Microbial reduction of crystalline iron(III) oxides: Influence of oxide surface area and potential for cell growth. Environ. Sci. Technol., 30, 1618–1628. Ruiz, J. M., Delgado, A., & Torrent, J. (1997). Iron-related phosphorus in overfertilized european soils. J. Environ. Qual., 26, 1548–1554.


347

Sallade, Y. E., & Sims, J. T. (1997). Phosphorus transformations in the sediments of delaware’s agricultural drainageways: II. Effect of reducing conditions on phosphorus release. J. Environ. Qual., 26, 1579–1588. Scheinost, A. C., Abend, S., Pandya, K. I., & Sparks, D. L. (2001). Kinetic controls on Cu and Pb sorption by ferrihydrite. Environ. Sci. Technol., 35, 1090–1096. Schwertmann, U., & Taylor, R. M. (1972). Influence of silicate on transformation of lepidocrocite to goethite. Clays Clay Miner., 20, 159–164. Shaw, S., Pepper, S. E., Bryan, N. D., & Livens, F. R. (2005). The kinetics and mechanisms of goethite and hematite crystallization under alkaline conditions, and in the presence of phosphate. Am. Mineral., 90, 1852–1860. Shenker, M., Seitelbach, S., Brand, S., Haim, A., & Litaor, M. I. (2005). Redox reactions and phosphorus release in re-flooded soils of an altered wetland. Eur. J. Soil Sci., 56, 515–525. Siffert, C., & Sulzberger, B. (1991). Light-induced dissolution of hematite in the presence of oxalate – a case-study. Langmuir, 7, 1627–1634. Slomp, C. P., & Van Cappellen, P. (2004). Nutrient inputs to the coastal ocean through submarine groundwater discharge: Controls and potential impact. J. Hydrol., 295, 64–86. Stone, J. J., Burgos, W. D., Royer, R. A., & Dempsey, B. A. (2006). Zinc and manganese inhibition of biological hematite reduction. Environ. Eng. Sci., 23, 851–862. Stookey, L. (1970). Ferrozine – a new spectrophotometric reagent for iron. Anal. Chem., 42, 779–781. Stumm, W., & Sulzberger, B. (1992). The cycling of iron in natural environments: Considerations based on laboratory studies of heterogeneous redox processes. Geochim. Cosmochim. Acta, 56, 3233–3257. Szilas, C. P., Borggaard, O. K., Hansen, H. C. B., & Rauer, J. (1998). Potential iron and phosphate mobilization during flooding of soil material. Water Air Soil Pollut., 106, 97–109. Trolard, F., Genin, J.-M. R., Abdelmoula, M., Bourrie, G., Humbert, B., & Herbillon, A. (1997). Identification of a green rust mineral in a reductomorphic soil by Mossbauer and Raman spectroscopies. Geochim. Cosmochim. Acta, 61, 1107–1111. Webb, S. M. (2005). SIXPack a graphical user interface for XAS analysis using IFEFFIT. Phys. Scr., T115, 1011–1014. Willett, I. R. (1985). The reductive dissolution of phosphated ferrihydrite and strengite. Aust. J. Soil Res., 23, 237–244. Wilson, G. V., Rhoton, F. E., & Selim, H. M. (2004). Modeling the impact of ferrihydrite on adsorption-desorption of soil phosphorus. Soil Sci., 169, 271–281. Wu, W. M., Carley, J., Gentry, T., Ginder-Vogel, M. A., Fienen, M., Mehlhorn, T., Yan, H., Caroll, S., Pace, M. N., Nyman, J., Luo, J., Gentile, M. E., Fields, M. W., Hickey, R. F., Gu, B., Watson, D., Cirpka, O. A., Zhou, J., Fendorf, S., Kitanidis, P. K., Jardine, P. M., & Criddle, C. S. (2006). Pilot-scale in situ bioremedation of uranium in a highly contaminated aquifer. 2. Reduction of

348


U(VI) and geochemical control of U(VI) bioavailability. Environ. Sci. Technol., 40, 3986–3995. Young, E. O., & Ross, D. S. (2001). Phosphate release from seasonally flooded soils: A laboratory microcosm study. J. Environ. Qual., 30, 91–101. Zachara, J. M., Fredrickson, J. K., Li, S. M., Kennedy, D. W., Smith, S. C., & Gassman, P. L. (1998). Bacterial reduction of crystalline Fe3+ oxides in single phase suspensions and subsurface materials. Am. Mineral., 83, 1426–1443. Zachara, J. M., Fredrickson, J. K., Smith, S. C., & Gassman, P. L. (2001). Solubilization of Fe(III) oxide-bound trace metals by a dissimilatory Fe(III) reducing bacterium. Geochim. Cosmochim. Acta, 65, 75–93. Zachara, J. M., Kukkadapu, R. K., Fredrickson, J. K., Gorby, Y. A., & Smith, S. C. (2002). Biomineralization of poorly crystalline Fe(III) oxides by dissimilatory metal reducing bacteria (DMRB). Geomicrobiol. J., 19, 179–207.


Chapter 13

Influence of Phosphate on Adsorption and Surface Precipitation of Lead on Iron Oxide Surfaces Liyun Xie and Daniel E. Giammar Department of Energy, Environmental & Chemical Engineering, Washington University in St. Louis, One Brookings Dr. St. Louis, MO 63130, USA

ABSTRACT Lead and phosphate sorption on goethite-coated and uncoated quartz sand was measured experimentally and modeled within a reaction-based framework. Single sorbate batch experiments and experiments with lead and phosphate present together were conducted. Adsorption of lead and phosphate to goethite-coated sand was dominated by adsorption to the goethite coating. A surface complexation model adapted from models for pure goethite successfully simulated lead and phosphate adsorption to goethite-coated sand over a broad range of pH and total sorbate concentrations; the inclusion of a surface complexation reaction for lead adsorption to the quartz surface was necessary to improve the model fit. Lead sorption on goethite-coated sand was enhanced by the presence of phosphate. The effect was most pronounced at low pH. The enhanced lead uptake was predicted by the combination of the single sorbate surface complexation models. The adsorption of phosphate at low pH decreased the surface charge and potential, which increases the extent of lead adsorption. No ternary surface complexes were needed to model the dual sorbate results. When a reaction for the precipitation of chloropyromorphite (Pb5(PO4)3Cl) was included, the model predicted precipitation only at the lowest pH and highest phosphate loading studied. Over most experimental conditions, including conditions that were initially supersaturated with respect to chloropyromorphite, the equilibrium model predicted that adsorption was the dominant mechanism of lead sorption; however, the actual mechanisms may be controlled by the relative rates of precipitation and adsorption reactions.

Corresponding author. Tel.: +314-935-6849; Fax: +314-935-5464;

E-mail: [email protected] (D.E. Giammar).

350

L. Xie and D. E. Giammar

13.1. Introduction The fate and transport of lead in soil and groundwater is strongly controlled by chemical reactions at the mineral–water interface. Surface reactions include adsorption and precipitation. Goethite (a-FeOOH), a common iron(III) oxyhydroxide mineral in natural systems, and other iron(III) oxides and oxyhydroxides are important environmental sorbents for heavy metals due to their high specific surface areas and reactive surfaces (Dzombak and Morel, 1990; Coston et al., 1995; Schwertmann and Cornell, 2000). Iron oxide coated sands prepared in the laboratory have previously been used as model materials for real soils and sediments that have iron oxide coatings (Gabriel et al., 1998; Cheng et al., 2004). Lead adsorbs to goethite through the formation of inner-sphere surface complexes with surface hydroxyl groups (RFeOH) as determined by extended X-ray adsorption fine-structure spectroscopy (EXAFS) (Roe et al., 1991; Bargar et al., 1997). Lead also adsorbs via inner-sphere complexation as a mixture of monodentate and bidentate complexes on ferrihydrite (Fe5HO8 4H2O) (Trivedi et al., 2003) and as mononuclear bidentate complexes on hematite (a-Fe2O3) (Bargar et al., 1997). Evidence for inner-sphere complexation is also provided by a lack of ionic strength effects on adsorption. The ionic strength of the bulk solution has less impact on inner-sphere complexation than on outer-sphere complexation (Hayes and Leckie, 1986). Adsorption of lead to goethite has been successfully interpreted through the inclusion of reaction 1 in a surface complexation model (SCM) (Hayes and Leckie, 1986). FeOH þ Pb2þ ¼ FeOPbþ þ Hþ

(13.1)

Other SCMs have included reaction 1 as well as two other reactions (2 and 3) (Gunneriusson et al., 1994). FeOH þ Pb2þ ¼ FeOHPb2þ

(13.2)

FeOH þ Pb2þ þ H2 O ¼ FeOPbOH þ 2Hþ

(13.3)

Surface complexation modeling considers the adsorption of inorganic species to specific reactive surface sites and includes terms to account for both the chemical and electrostatic energetics of adsorption (Dzombak and Morel, 1990). The model of Hayes and Leckie (1986) successfully predicted the pH dependence of lead adsorption. Although metal adsorption to quartz is not usually significant in soil and groundwater relative to adsorption to iron oxyhydroxides and clays, metal

Influence of Phosphate on Adsorption and Surface Precipitation

351

cations can still adsorb at the quartz surface. X-ray absorption spectroscopy investigations found inner-sphere lead surface complexes formed on quartz (Chen et al., 2006) and amorphous silica (Elzinga and Sparks, 2002). Lead sorption at the iron oxide–water interface can be enhanced by the presence of phosphate. Enhancement can occur as the result of: (a) changes to surface charge that make lead adsorption more favorable, (b) formation of ternary lead–phosphate–iron oxide surface complexes, (c) precipitation of lead phosphate solids, and (d) surface alteration from the formation of an iron(III) phosphate surface precipitate. Phosphate addition to leadcontaminated soils has been proposed as a means of reducing lead mobility and bioavailability (Ruby et al., 1994; Ma et al., 1995; Hettiarachchi et al., 2000; Yang et al., 2001; Cao et al., 2002; Ryan et al., 2004). Although remediation strategies usually suggest using phosphate addition to precipitate lead phosphate solids, enhancement of lead binding to iron oxyhydroxides may also increase lead sorption even without lead phosphate precipitation. Phosphate adsorbs to goethite through inner-sphere complex formation as observed directly with attenuated total reflectance Fourier transform infrared spectroscopy (Arai and Sparks, 2001). Atomic force microscopy (AFM) of phosphate adsorption to the (0 1 0) surface of a goethite single crystal observed phosphate adsorbed in a 1:1 ratio with the singly coordinated hydroxyl groups via monodentate complexes (Dideriksen and Stipp, 2003). A combination of three monodentate surface complexes (reactions 4–6) has been proposed for modeling phosphate adsorption to goethite (Nilsson et al., 1992). þ FeOH þ H2 PO 4 þ H ¼ FePO4 H2 þ H2 O

(13.4)

FeOH þ H2 PO 4 ¼ FePO4 H þ H2 O

(13.5)

2 þ FeOH þ H2 PO 4 ¼ FePO4 þ H þ H2 O

(13.6)

When metal ions (e.g., M2+) and anions (e.g., L) adsorb together on the oxide surface, ternary surface complexes can form by either metal-bridging (reaction 7) or ligand-bridging (reaction 8) reactions (Hering and Kraemer, 1994). SOH þ M2þ þ L ¼ SOM L þ Hþ

(13.7)

SOH þ L þ M2þ þ Hþ ¼ SLM2þ þ H2 O

(13.8)

Ternary surface complexation is an adsorption process with the simultaneous accumulation of metal cations and ligand anions at the solid– water interface without the development of three-dimensional structure

352


(Sposito, 1986). This adsorption process is in contrast to the formation of a precipitate at the surface that has longer range three-dimensional structure. Lead and phosphate can also precipitate at the goethite surface, most likely through heterogeneous nucleation of a lead phosphate precipitate on the goethite substrate. Once precipitation is initiated, the number of surface sites for lead or phosphate binding is no longer fixed by the goethite surface area. Surface precipitation is most likely at high sorbate loading. Potential lead-containing solids include lead hydroxide (Pb(OH)2) and the lead phosphates chloropyromorphite (Pb5(PO4)3Cl), hydroxypyromorphite (Pb5(PO4)3OH), and lead hydrogen phosphate (PbHPO4). Ler and Stanforth (2003) observed precipitation of lead and phosphate on goethite in a phase with the stoichiometry of chloropyromorphite or hydroxypyromorphite. The reaction of hydroxyapatite (Ca5(PO4)3OH) and goethite with adsorbed lead resulted in the formation of chloropyromorphite through homogeneous nucleation in the solution and possibly heterogeneous nucleation at the goethite surface (Zhang et al., 1997). In this study we examine the effects of phosphate on lead sorption to goethite-coated sand. The word sorption is used here to describe all processes (adsorption, surface precipitation, co-precipitation) that result in accumulation of lead at the goethite–water interface. The objectives of this project were to determine the utility of goethite-coated sand as a model porous medium with a reactive surface phase, distinguish the different mechanisms through which phosphate can increase lead sorption, and determine the geochemical conditions under which each mechanism is dominant. Once the interfacial chemical reactions on goethite-coated sand have been established, it can be a useful material for evaluating the effects of surface reactions on lead transport. The different mechanisms of lead sorption in the presence of phosphate may result in similar equilibrium partitioning of lead between the solid and dissolved phases, but the rates of transport are likely to be affected by differences in sorption mechanisms.

13.2. Experimental 13.2.1. Materials All chemicals used were certified ACS grade (Fisher Scientific) unless otherwise specified. The concentrated nitric acid was trace metal grade. Pure goethite was synthesized from an alkaline aqueous system according to methods of Schwertmann and Cornell (2000). A 100 ml volume of 1 M


353

Fe(NO3)3 was mixed with 180 ml of 5 M KOH in a 2 l polyethylene bottle and then immediately diluted with ultrapure water to 2 l. The bottle was sealed and heated in an oven at 651C for 60 h. An ocher precipitate formed during the heating process. The resulting suspension was washed free of excess dissolved ions, primarily K+ and NO 3 , by dialysis. The dialyzed suspension was freeze dried. The identity of the synthetic goethite (a-FeOOH) was confirmed by X-ray powder diffraction (XRD) (Rigaku Geigerflex D-MAX/A Diffractometer) with Cu–Ka radiation. Solid samples for XRD were ground to powders and loaded into glass sample holders for analysis. Quartz sand (U.S. Silica) with particle sizes of 212–355 mm was soaked in 1 M HCl (Fisher Scientific) overnight to remove surface impurities from the quartz. The sand was rinsed with ultrapure water to remove the acid and then dried at 1051C. Goethite-coated sand was then prepared by agitating a mixture of acid-washed quartz sand and goethite at pH 6.8 and 0.01 M ionic strength (NaNO3) for 24 h. These conditions had previously been determined to be optimal for goethite attachment based on the surface charges of goethite and quartz (Scheidegger et al., 1993). Unattached goethite was removed by repeatedly suspending the sand in 0.01 M NaNO3 solution, allowing the sand to settle, and decanting the supernatant. These steps were repeated until a clear supernatant was achieved. The iron content of the coated sand was determined by extraction in a solution of 0.3 M sodium dithionite, 0.3 M sodium citrate, and 0.2 M sodium bicarbonate. This solution effectively dissolves crystalline Fe(III) oxides and oxyhydroxides by reducing them to soluble Fe(II) species (Clark et al., 1996). The extraction mixture was shaken for 1 h, set overnight at room temperature for complete reaction, and the extract was then analyzed by inductively coupled plasmaoptical emission spectrometry (ICP-OES). The goethite-coated quartz sand contained 0.2% (w/w) iron; the value of 0.005% iron in uncoated sand was reported by the supplier (U.S. Silica). The typical crystalline morphology of goethite and of goethite present on goethite-coated sand was observed with a field emission scanning electron microscope (Hitachi S-4500) (Fig. 13.1). Dry samples were fixed to sample holders with double-sided tape. To improve the SEM images, gold coating was used on some samples to increase electrical conductivity. The quartz sand surface was partially covered by goethite particles but large portions of quartz remained exposed. Table 13.1 summarizes the properties of pure goethite, uncoated quartz sand, and goethite-coated sand. The size of uncoated and goethite-coated sand were the same, but the specific surface area of the coated sand was more than three times that of uncoated sand due to the high surface area goethite coating. Specific surface area (m2/g) was measured by

354


Figure 13.1: Scanning Electron Micrographs of Synthetic Goethite (Left) and Goethite-Coated Sand (Right).

Table 13.1: Sorbent Properties.

Size (mm) Iron (wt%) Specific surface area (m2/g)

Goethite

Quartz sand

Goethite-coated sand

1 62.8 33.6

212–355 0.005 0.05

212–355 0.2 0.16

BET-N2 adsorption (Quantachrome Autosorb AS-1). Prior to BET-N2 measurement, wet samples were freeze dried (Labconco, Freezone 4.5) and then loaded into sample cells for analysis. Based on the iron content of the goethite-coated sand and assuming the additional 0.11 m2/g of surface area for the coated sand is from goethite, the goethite in the coating is calculated to have a specific surface area of 35 m2/g, which is very close to the value of 33.6 m2/g measured for the pure goethite.

13.2.2. Recirculating Micro-Column Batch Adsorption Experiments Lead and phosphate sorption reactions were studied by continuously recirculating solution from a 50 ml reservoir through 1.8 ml columns that were filled with 2.5 g of either goethite-coated sand or uncoated sand. Columns with recirculation, instead of stirred batch reactors, were used because goethite detachment from quartz sand was observed when suspensions of goethite-coated sand were agitated on a rotary shaker or with a magnetic stir bar. The recirculation flow rate was controlled at 0.5 ml/min with a peristaltic pump. Samples were collected after 24 h, which allowed sufficient


355

residence times of both the column (3 min) and the reservoir (100 min) to allow the entire system to be considered as one well-mixed batch reactor. Lead adsorption to goethite occurs on time scales much faster than those of this experiment (Hayes and Leckie, 1986), and our preliminary experiments demonstrated that 24 h was also sufficient for equilibration. After the 24 h equilibration period, a 20 ml effluent sample was collected for pH and flow-rate measurement and another 10 ml sample was collected for dissolved lead and/or phosphate analysis. For many of the single sorbate experiments the influent reservoir was then refilled with 30 ml of solution to provide a new desired composition and the system was equilibrated for another 24 h until the next sampling time. By using this approach, multiple adsorption datapoints were generated while using a limited amount of solid material and maintaining a completely saturated environment in the columns. Single sorbate (lead-only or phosphate-only) and binary sorbate (lead and phosphate together) experiments were conducted (Table 13.2). Single sorbate experiments were conducted to evaluate the extent to which adsorption to goethite-coated sand could be explained by adsorption to pure goethite. The influent lead and phosphate concentrations were undersaturated with respect to lead hydroxide and iron phosphate precipitates. The range of total lead concentrations was 107–104 M and that of total phosphate concentrations was 107–5 105 M. Lead was added as Pb(NO3)2 (Acros Organic) and phosphate as KH2PO4 (Fisher Scientific). The desired pH of each adsorption experiment was from 4 to 8 and was controlled by MES (Sigma) or HEPES (Acros Organic) buffer solutions adjusted to pH 6–8 by addition of NaOH. These buffers were selected because of their established low affinities for metal complexation (Good et al., 1966; Soares et al., 1999). Buffer concentrations were 1 mM for lead adsorption and 10 mM for phosphate adsorption experiments. The buffer concentrations were selected to be at least an order of magnitude higher than the maximum total adsorbate concentration studied, and phosphate was studied at a higher total concentration than was lead. For control of the ionic strength, solutions also contained 1 mM NaNO3 for lead adsorption experiments and 10 mM with NaNO3 for phosphate adsorption experiments. When combined with the buffer species and associated NaOH was added, the total ionic strength of the solutions was 1.2–1.8 mM for lead experiments and 12–18 mM for phosphate experiments, depending on the pH. To examine the transition between adsorption and surface precipitation, binary sorbate experiments were conducted in which lead and phosphate were both present. The pH in these experiments was kept at 4–7 and the initial lead concentration was fixed at 5 107 M. To provide the desired

356

Experiment no.

A-1 A-2 B-1 B-2 C-1 C-2 a

Sorbent

Uncoated sand Goethite-coated sand Uncoated sand Goethite-coated sand Uncoated sand Goethite-coated sand

Solution composition pH

[Pb]init (M)

[P]init (M)

[Cl]init (M)

SICPY,

4–8 4–8

107–104a 107–104a

0 0

0 0

NA NA

4–8 4–8

0 0

107–5 105a 107–5 105a

0 0

NA NA

4–7 4–7

5 107 5 107

7.5 1010–103 7.5 1010–103

2.5 1010–104 2.5 1010–104

3, 1, 1, 5 3, 1, 1, 5

Single columns were used to evaluate adsorption over a broad concentration range at a given pH for lead-only and phosphate-only experiments.

init


Table 13.2: Conditions of Experiments Conducted.


357

initial saturation index with respect to chloropyromorphite, phosphate concentrations were varied from 7.5 1010 to 1.2 103 M and chloride concentrations from 2.5 1010 to 4 104 M (Table 13.3). These values were selected to approach relevant environmental concentrations while avoiding the use of initial solutions that were at even higher initial conditions of saturation. The initial saturation index (SI) (Eq. (13.9)) was examined at two conditions of undersaturation (3 and 1) and two of supersaturation (+1 and +5). The solubility product (Ksp) of chloropryromorphite is 1084.43 (Nriagu, 1973). 3 fPb2þ g5 fPO3 4 g fCl g SI ¼ log (13.9) K sp At negative values of SI, the precipitation of chloropyromorphite is not expected, and at positive SI values, chloropyromorphite precipitation is thermodynamically favorable. In sorbent-free control experiments, supersaturated conditions could be maintained for times of more than a day without loss of lead or phosphate from solution; however, at longer time scales, chloropyromorphite precipitation was observed for supersaturated conditions. Only the initial saturation indices were set at fixed values, and once lead and phosphate are taken up by the solid, either by adsorption or precipitation, the saturation index will decrease from its initial value. All experiments were run in duplicate. At the conclusion of each experiment, the sand in the columns was dried and kept for solid phase analysis. At most solution conditions, control experiments were performed with empty columns.

13.2.3. Analytical Methods Dissolved lead was analyzed by ICP-OES (Varian Liberty II) or inductively coupled plasma-mass spectrometry (ICP-MS, Agilent Technologies 7500ce). For lead, the detection limit of ICP-OES was 0.01 ppm and that of ICP-MS was 0.01 ppb. The acid matrix used for ICP-OES was 2% HNO3 and for ICP-MS it was 1% HNO3. The calibration standards for ICP analysis were trace ICP/ICP-MS grade (Fisher Scientific). Indium was used as an internal standard for ICP-MS analysis. Dissolved phosphate was measured by the ascorbic acid method (American Public Health Association, 1999). In this method ammonium molybdate and potassium antimonyl tartrate react in acid medium with orthophosphate to form phosphomolybdic acid, which can be reduced by

358 L. Xie and D. E. Giammar

Table 13.3: Composition of Solutions Used to Generate Saturation Indices at Different pH Values (Concentrations in M). pH

4 5 6 7

SI ¼ 3

SI ¼ 1

SI ¼ +1

SI ¼ +5

[Pb]init

[P]init

[Cl]init

[Pb]init

[P]init

[Cl]init

[Pb]init

[P]init

[Cl]init

[Pb]init

[P]init

[Cl]init

5 107 5 107 5 107 5 107

1.2 105 3.9 107 1.3 108 7.6 1010

4.1 106 1.3 107 4.4 109 2.5 1010

5 107 5 107 5 107 5 107

3.9 105 1.2 106 4.2 108 2.4 109

1.3 105 4.1 107 1.4 108 8.0 1010

5 107 5 107 5 107 5 107

1.2 104 3.9 106 1.3 107 7.6 109

4.1 105 1.3 106 4.4 108 2.5 109

5 107 5 107 5 107 5 107

1.2 103 3.9 105 1.3 106 7.6 108

4.1 104 1.3 105 4.4 107 2.5 108


359

ascorbic acid to molybdenum blue. The absorbance of the reacted sample solution was measured spectrophotometrically (Perkin-Elmer Lambda 2S) at 880 nm using a 10 cm pathlength quartz cuvette. The detection limit of phosphorus by this method was 1 ppb. Solution pH was measured with a glass pH electrode and a pH meter (Accumet Research AR25). 13.2.4. Geochemical Equilibrium Modeling The software program FITEQL 4.0 was used to determine chemical equilibrium constants of surface complexation reactions. The program uses a non-linear least-squares optimization method to determine parameters of equilibrium models that provide the best fits to experimental datasets (Herbelin and Westall, 1999). The program contains some of the most common models for accounting for electrostatic effects on surface complexation, including the constant capacitance, diffuse layer, and triple layer models. The constant capacitance model was used in this study to account for the electrostatic interactions between the charged solid surface and ionic adsorbates. Although the constant capacitance model is often used for systems of high ionic strength (Stumm, 1992), it was selected in this study because of its limited number of parameters and the availability of previously published single solute adsorption studies for comparison. Two types of adsorption sites were considered in the SCM: the surface hydroxyls associated with goethite (RFeOH) and with quartz (RSiOH). If goethite coating covers most of the surface of the quartz sand or if RSiOH sites do not have a significant effect on lead and phosphate adsorption, then only one type of adsorption site (RFeOH) would need to be considered. The equilibrium constants for the surface acid–base reactions are taken from the published literature for RSiOH sites (Sverjensky and Sahai, 1996) and RFeOH sites (Gunneriusson et al., 1994) (reactions S1–2 and S10–11 in Table 13.4). Surface site densities for RSiOH and RFeOH were set at the value of 2.3 sites/nm2 suggested by Davis and Kent (1990) for oxide minerals. Experimental data were also used in FITEQL to optimize surface site densities, but if optimized site densities were not substantially different from 2.3 sites/nm2, then the value of 2.3 sites/nm2 was used. The ability of equilibrium constants for single sorbate surface complexation reactions from references to simulate our experimental data was evaluated. If the constants from the literature did not provide a good fit, then they were further optimized in FITEQL. This optimization proceeded in two separate steps. First, the equilibrium constants for lead and phosphate surface complexation at

360


Table 13.4: Reactions Considered in Aqueous Lead Phosphate System.

P1 P2 P3 P4 S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12 S13

Reaction

logKa

Ref.b

Pb5(PO4)3Cl(s)Q5Pb2++3PO3 4 +Cl 2+ 3 Pb5(PO4)3OH(s)Q5Pb +3PO4 +OH + PbHPO4(s)QPb2++PO3 4 +H 2+ – Pb(OH)2(s)QPb +2OH + RFeOH+ 2 2RFeOH+H + RFeOH2RFeO +H RFeOH+Pb2+2RFeOPb++H+ RFeOH+Pb2+2RFeOHPb2+ + RFeOH+PO3 4 +3H 2RFePO4H2+H2O 3 RFeOH+PO4 +2H+2RFePO4H+H2O + 2 RFeOH+PO3 4 +H 2RFePO4 +H2O 2+ 3 + RFeOH+Pb +PO4 +H 2RFePO4Pb+H2O + RFeOH+Pb2++PO432RFeOPbPO 4 +H + + RSiOH2 2RSiOH+H RSiOH2RSiO+H+ RSiOH+Pb2+2RSiOPb++H+ + 2– RSiOH+PO3 4 +H 2RFePO4 +H2O

84.43 76.8 23.81 19.85 7.47 9.51 0.92 9.47 31.13 25.95 18.45

1 1 1 1 2 2 6 6 3 6 6

RFeOH site density ¼ 2.3 sites/nm2 RSiOH site density ¼ 2.3 sites/nm2 Capacitance ¼ 1.28 F/m2

c c

2.4 8.4 1.02

4 4 6

c

5 5 2

a

logK values for surface reactions are for the intrinsic equilibrium constants (Kint). 1: Schecher and McAvoy (1998); 2: Gunneriusson et al. (1994); 3: Nilsson et al. (1992); 4: Sverjensky and Sahai (1996); 5: Davis and Kent (1990); 6: This study. c Reactions were not needed for this study. b

RSiOH sites (reactions S12 and S13 in Table 13.4) were optimized using the datasets for lead and phosphate adsorption to uncoated sand. These values were then held constant as the equilibrium constants for complexation to RFeOH sites (reactions S3–S7 in Table 13.4) were optimized with the datasets for adsorption to goethite-coated sand. Finally, the site (RFeOH and RSiOH) densities and equilibrium constants from single sorbate experiments were combined to assess whether or not they effectively simulate the dual sorbate experimental data. If the combination of single adsorbate reactions did not sufficiently predict dual adsorbate adsorption, then ternary surface complexation reactions (reactions S8 or S9 in Table 13.4) could be added to the model and their equilibrium constants optimized in FITEQL. Precipitation reactions of cholorpyromorphite and lead hydroxide (reactions P1–P4 in Table 13.4) were also included to evaluate their impact on the simulation of lead uptake.


361

The software program MINEQL+ (Schecher and McAvoy, 1998) was used to calculate equilibrium speciation based on fixed reactions, constants, and total concentrations. Equilibrium constants for all aqueous phase reactions were those in the MINEQL+ database (Schecher and McAvoy, 1998).

13.3. Results and Discussion 13.3.1. Single Sorbate Adsorption 13.3.1.1. Effect of Goethite Coating on Adsorption The amount of adsorbed lead was always higher on the goethite-coated sand than on the uncoated sand (Fig. 13.2a). However, the iron oxide content was apparently not high enough for goethite to completely dominate adsorption of lead. As seen in SEM images, the sand surface was not completely covered by goethite particles. The quartz surface has sufficient adsorption capacity and affinity for lead that it must be considered in subsequent modeling. The amount of adsorbed phosphate was always higher on the goethitecoated sand than on the uncoated sand (Fig. 13.2b), which indicated that the goethite coating dominated phosphate adsorption. Phosphate adsorption to quartz was so insignificant that adsorption to the quartz was neglected in modeling. Both lead and phosphate adsorb to the goethite-coated sand with patterns typical of the Langmuir isotherm. Because of more favorable adsorption of phosphate than of lead at pH 6, the initial slope of the isotherm is steeper for phosphate than it is for lead.

13.3.1.2. Effect of pH on Lead Adsorption The adsorption of lead on goethite-coated sand was highly pH dependent (Fig. 13.3), increasing from minimal adsorption at pH 4 to nearly complete adsorption at pH 7. Similar pH edges have been observed for lead adsorption to pure goethite (Hayes and Leckie, 1986; Gunneriusson et al., 1994) and are typical of metal cation adsorption to iron oxides (Benjamin, 2002). The SCM, which was used to fit the experimental data of lead adsorption to goethite-coated sand, included two types of adsorption sites, quartz surface hydroxyl (RSiOH) and goethite surface hydroxyl (RFeOH) groups, which were each described by two surface acid–base reactions (S1, S2, S10,

362


Figure 13.2: Adsorption of (a) Lead and (b) Phosphate at pH 6, Solid:Liquid Ratio ¼ 50:1 g/l, and 24 h Contact Time. The Ionic Strength was Fixed by NaNO3 at 103 M and 103 M MES Buffer for Lead Adsorption and 102 M and 102 M MES for Phosphate Adsorption.

and S11 in Table 13.4). In the constant capacitance model, a capacitance of 1.28 F/m2 was used based on previous studies (Gunneriusson et al., 1994). The equilibrium constant for lead adsorption to RSiOH groups, with the site density fixed at 2.3 sites/nm2, was optimized by FITEQL as 101.02 using the dataset for lead adsorption to uncoated sand. The quality of the model fit is characterized by a weighted sum of squares divided by the degrees of freedom (WSOS/DF) value of 0.95. The constant for reaction S12 in Table 13.4 was then held constant in modeling lead adsorption to goethitecoated sand. A site density of 2.3 sites/nm2 was also used for RFeOH sites.


363

Figure 13.3: Adsorption Edges (a) and Surface Speciation (b) for Lead Adsorption to Goethite-Coated Sand as a Function of pH. Data are Shown with Points and Surface Complexation Modeling Simulations are Shown by the Solid Line. Reaction Conditions: I ¼ 1.2–1.8 103 M NaNO3, Contact Time ¼ 24 h, Solid:Liquid Ratio ¼ 50:1 g/l.

The optimal equilibrium constants for lead adsorption on goethite were determined to be 100.92 and 109.47 for reactions S3 and S4, respectively, with a WSOS/DF value of 2.7. The equilibrium constants are similar to those determined by Gunneriusson and coworkers (1994) for lead adsorption to pure goethite (100.17 and 108.20). Initial modeling was attempted using a single adsorption reaction to goethite, but the model with one reaction could not provide a good fit to the experimental data over the full range of total lead loading. Inclusion of a second reaction significantly improved the quality of the fit over a broader range of total concentration, although the model

364


still overpredicts adsorption for the highest lead concentration. The optimization sought the best fit for 5 107 M total lead, because this concentration was also studied in the presence of phosphate. The total site density of the goethite-coated quartz sand was set at 2.3 sites/nm2 for both goethite and quartz surfaces. This value was recommended by Davis and Kent (1990) for mineral oxide surfaces as one for which the results of multiple studies can be compared. When the goethite site density for phosphate adsorption to goethite-coated sand was simultaneously optimized with the equilibrium constant for reaction S7 in Table 13.4, a value of 2.2 sites/nm2 was determined (see next section), and so the value of 2.3 sites/nm2 was used. The site density for goethite is within the range of values used in previous work on lead adsorption to pure goethite of 1.7 sites/nm2 (Gunneriusson et al., 1994) and 7 sites/nm2 (Hayes and Leckie, 1986). The site density of 2.3 sites/nm2 used for quartz is lower than values of 5.9 sites/nm2 (Yee and Fein, 2003) and 5 sites/nm2 (Chen et al., 2006) used in previous studies. Without an independent measurement of site density, the determinations of the site density of surface groups and the equilibrium binding constants are not entirely independent in model derivation based on adsorption data sets. The use of a consistent site density is beneficial for comparison of equilibrium constants from different studies and is important for the development of a self-consistent thermodynamic database (Davis and Kent, 1990). Figure 13.3b shows the simulated surface speciation of lead adsorption to goethite-coated sand. The binding of lead to goethite is stronger than that to quartz and RFeOHPb2+ is the dominant surface species over the experimental pH range. At higher pH (>7.5), RFeOHPb2+ is surpassed by RFeOPb+ as the dominant species. The species RSiOPb+ is never dominant in this model, although its contribution increases with increasing pH.

13.3.1.3. Effect of pH on Phosphate Adsorption The adsorption of phosphate on goethite-coated sand was pH dependent (Fig. 13.4). The phosphate adsorption density decreased with increasing pH. These data were similar to previous observations of phosphate adsorption (Cheng et al., 2004). When the phosphate concentration was low ([P]tot ¼ 5 106 M), complete adsorption on goethite-coated sand was approached when the pH was less than 7. For a total phosphate concentration of 5 105 M, adsorbed phosphate was never more than 35% of the total. Adsorption of phosphate to uncoated quartz sand was insignificant compared with that to goethite-coated sand, therefore, the model used for


365

Figure 13.4: Adsorption Edges (a) and Surface Speciation (b) for Phosphate on Goethite-Coated Sand as a Function of pH. Data are Shown with Points and Surface Complexation Modeling Simulations are Shown by the Solid Line. Reaction Conditions: I ¼ 1.2–1.8 102 M NaNO3, Contact Time ¼ 24 h, Ptot ¼ 107 to 104 M, Solid:Liquid Ratio ¼ 50:1 g/l. phosphate adsorption to goethite-coated sand only considered adsorption to sites on goethite (RFeOH). The equilibrium constants of acid–base goethite surface reactions, the site density, and the capacitance were the same as those in the SCM for lead adsorption. Three reactions (S5, S6 and S7 in

366


Table 13.4) were included for phosphate adsorption, consistent with previous studies (Nilsson et al., 1992; Gao and Mucci, 2003). Initially the equilibrium constants for reactions S5 and S6 in Table 13.4 were set at values from a previous study of phosphate adsorption to goethite (Nilsson et al., 1992), and the constant for reaction S7 in Table 13.4 was determined by finding the value that provided the optimal agreement between the experimental data and the model simulation (Table 13.4). The optimized logK for reaction S7 in Table 13.4 was 18.45 and the goodness of fit was characterized by a WSOS/DF of 6.5. Because adsorption was already at the maximum for a given total phosphate concentration below pH 7, the quality of the fit was insensitive to the constants for reactions S5 and S6 in Table 13.4; however, optimization of the effect of phosphate adsorption on surface charge for dual sorbate experiments (discussed in the section ‘‘Co-Sorption of Lead and Phosphate in Combined Systems’’) required refinement of the constant for reaction S6 in Table 13.4 to a logK of 25.95. The values of 25.95 and 18.45 for reactions S6 and S7 in Table 13.4, respectively, are lower than the values of 26.38 and 20.61 used by Nilsson and coworkers. The species RFePO4H is dominant at lower pH and RFePO2 is 4 dominant at higher pH (Fig. 13.4b). Concentrations of RFePO2 increase 4 and those of RFePO4H decrease with increasing pH. The species RFePO4H2 is not significant over the experimental pH range. At higher is more total phosphate concentrations, the formation of RFePO2 4 significant (modeling data not shown). 13.3.2. Co-Sorption of Lead and Phosphate in Combined Systems When lead and phosphate were combined in the same systems, the presence of phosphate increased lead removal for all experiments. The effect was greatest at low pH. At pH 4, lead uptake increased from 6% to 40–54% when the initial conditions of the system were undersaturated with respect to pyromorphite and to 60–86% in systems with initial supersaturation. At higher pH (6 or 7), the presence of phosphate did not have as significant of an impact since lead removal was already approaching 100% in the absence of phosphate. Phosphate sorption in the dual sorbate experiments was measured and was generally consistent with the SCM. Three scenarios were considered as models for the enhancement of lead uptake by phosphate: (1) direct combination of binary surface complexation reactions; (2) combination of binary surface complexation reactions with an additional ternary surface complexation reaction (Table 13.4, reaction S8 or S9); (3) combination of binary surface complexation reactions with potential


367

Figure 13.5: Lead Sorption on Goethite-Coated Sand as a Function of pH and Solution Saturation with Respect to Pyromorphite. Fixed Reaction Conditions: I ¼ 1.2–1.8 102 M NaNO3, Contact Time ¼ 24 h, Solid:Liquid Ratio ¼ 50:1 g/l, Pbtot ¼ 5 107 M. Simulations Involve (a) Direct Combination of Binary Adsorption Reactions and (b) Binary Adsorption and Precipitation Reactions. precipitation of pyromorphite and lead hydroxide (Table 13.4, reactions P1 and P4). Figure 13.5a shows the simulation from directly combining surface complexation reactions from lead-only and phosphate-only systems. The model does predict that phosphate addition increases lead adsorption. The increase is caused by the lowering of the surface charge and surface potential (c) from the formation of RFePO4H and RFePO2 4 species, which consequently

368


Figure 13.6: Effect of Total Phosphate Loading on Goethite-Coated Sand Surface Charge at pH 4. Simulation Conditions: I ¼ 1.2–1.8 102 M NaNO3, Solid:Liquid Ratio ¼ 50:1 g/L, Pbtot ¼ 5 107 M. Total Phosphate Concentrations Varied from 109 to 103 M (Logarithmic Scale). increases lead adsorption as a result of a more favorable electrostatic component for adsorption (Fig. 13.6). The apparent equilibrium constant for reaction S3 in Table 13.4 (Eq. (13.10)) increases with decreasing surface potential (c) at pH 4 because the intrinsic constant remains fixed (Fig. 13.6). K app ¼ K int eF c=RT ¼

f FeOPbþ gfHþ g fFeOHgfPb2þ g

(13.10)

These simulations (Fig. 13.5a) show a good fit to experimental data at pH 4 and 5. Successfully fitting the lead sorption results at pH 4 did require model refinements beyond those initially attempted. Although only optimization of the constant for reaction S7 in Table 13.4 was necessary to model phosphate adsorption, the resulting effect of phosphate adsorption on surface charge overestimated the increase in lead sorption in the presence of phosphate. Optimization of both reactions S6 and S7 in Table 13.4 provided an equally good fit of phosphate sorption but then provided an improved fit for lead sorption in the dual sorbate experiments. The model predicts decreasing lead adsorption at pH 4.5–5 because the total phosphate concentrations at pH 4.5–5 are less than at pH 4 in order to maintain the same initial saturation indices with respect to pyromorphite (Table 13.3), although this effect remains to be tested experimentally.


369

The inclusion of reactions for pyromorphite and lead hydroxide precipitation (reactions P1 and P4 in Table 13.4) only predict higher uptake at the initial saturation index of +1 at pH less than 3.5 and at the initial SI of +5 at pH o 4.5. Pyromorphite was the solid predicted to form at these conditions. At lower initial saturation conditions and higher pH, adsorption was the dominant mechanism for lead uptake. Adsorption reduced dissolved lead and phosphate to concentrations that would no longer be supersaturated. The model simulations are based on equilibrium conditions, but in the actual experimental systems the sorption mechanism may be determined by the relative rates of adsorption and precipitation. The assumption of equilibrium also suggests that there would not be an effect of the order of sorbate addition in dual sorbate experiments; however, this is an effect that should be examined in future work. Inclusion of a ternary surface complexation reaction as a ligand-bridging surface complex (reaction S8 in Table 13.4) did not improve the model fit to the experimental data, nor were other reaction stoichiometries (e.g., metalbridging complexes) helpful for improving the fit. No values for a ternary surface complex improved the model fit. The combination of single-sorbate SCMs already fit the data or slightly overestimated lead uptake, and the inclusion of ternary surface complexes would only increase the predicted uptake. Previous studies of electrophoretic mobility and X-ray absorption spectroscopy show that adsorbed phosphate acts as a reactant to form lead phosphate surface phases that may be highly dispersed on goethite (Weesner and Bleam, 1998). In previous experiments by Ler and Stanforth (2003), the reaction stoichiometries of lead adsorption on phosphated goethite suggested that pyromorphite (Pb5(PO4)3Cl) may precipitate on the surface at conditions that are undersaturated with respect to pyromorphite solubility. The surface precipitation of lead phosphate on the goethite surface is most likely by formation of a new surface phase with the structure of pyromorphite. Analogous research on the co-sorption of zinc and arsenate, a good analog for phosphate, on goethite found that the presence of zinc increased arsenate adsorption and vice versa. The effect was most pronounced at high surface loadings and EXAFS data were used to determine that arsenate and zinc were present in zinc arsenate surface precipitates even at conditions that were undersaturated with respect to zinc arsenate solid phases (Gra¨fe et al., 2004; Gra¨fe and Sparks, 2005). Although these experiments showed no evidence of pyromorphite by SEM and XRD, precipitation of lead phosphate minerals cannot be ruled out. Since the maximum loading of lead on the solid is only 2 mg/kg, the formation of lead phosphate minerals may not be detectable by these

370


techniques. The low lead loading would even be challenging to detect by element-specific spectroscopic techniques such as EXAFS. Further investigation with goethite-coated sand on the nature of surface precipitation can focus on conditions of higher initial saturation index and lower pH. The low mass loading of lead on the solid is the result of working with goethite-coated sand, which was selected because of its potential utility as a model geomedium, and valuable information on mechanisms that will apply to this material can be gained from the numerous spectroscopic studies previously conducted with pure goethite.

13.4. Summary Goethite-coated sand can serve as a model porous medium to study the fate of lead in natural soil systems. Lead adsorption to goethite-coated sand can be simulated by a SCM that accounts for adsorption to goethite and quartz surface sites, and phosphate can be modeled using just goethite surface sites. The site densities, reaction stoichiometries, and equilibrium constants for goethite-coated sand are similar to those of pure goethite. The chemical equilibrium model developed in this study can be incorporated into reactive transport models for lead and phosphate transport through sand columns and ultimately to transport in soil and groundwater systems. Over a broad range of solution compositions, phosphate addition enhanced lead removal. Lead sorption can be predicted by the combination of lead and phosphate surface complexation reactions. Ternary surface complexation is not necessary for uptake simulation. The SCM suggests that the dominant sorption mechanism is adsorption, even when the system is initially supersaturated with respect to the lead phosphate mineral pyromorphite. Precipitation mechanisms will become significant at lower pH and higher supersaturation conditions. Since the loading of lead for this study is lower than the detection limits of most techniques for solid phase investigation, the possibility of pyromorphite formation cannot be ruled out based only on macroscopic sorption data.

ACKNOWLEDGMENTS Partial funding for this study was provided by the National Science Foundation (BES #0546219). Liyun Xie has been supported by fellowships through the Washington University School of Engineering and Applied Science, the Cecil Lue-Hing Scholarship, and the Lilia Abron Scholarship.


371

The authors acknowledge Claire Farnsworth’s contributions to lead adsorption experiments and Yu Wang’s to phosphate adsorption experiments. The insightful comments of Doug Kent and three anonymous reviewers assisted in the revision of this chapter.

REFERENCES American Public Health Association A. W. W. A. (1999). Water Environment Federation. Standard Methods for the Examination of Water and Wastewater. Arai, Y., & Sparks, D. L. (2001). ATR-FTIR spectroscopic investigation on phosphate adsorption mechanisms at the ferrihydrite-water interface. J. Colloid Interface Sci., 241 (2), 317–326. Bargar, J. R., Brown, G. E., & Parks, G. A. (1997). Surface complexation of Pb(II) at oxide-water interfaces. 2. XAFS and bond-valence determination of mononuclear Pb(II) sorption products and surface functional groups on iron oxides. Geochim. Cosmochim. Acta, 61 (13), 2639–2652. Benjamin, M. M. (2002). Water Chemistry. McGraw-Hill, New York, NY. Cao, X., Ma, L. Q., Chen, M., Singh, S. P., & Harris, W. G. (2002). Impacts of phosphate amendments on lead biogeochemistry at a contaminated site. Environ. Sci. Technol., 36 (24), 5296–5304. Chen, C. C., Coleman, M. L., & Katz, L. E. (2006). Bridging the gap between macroscopic and spectroscopic studies of metal ion sorption at the oxide/water interface: Sr(II), Co(II), and Pb(II) sorption to quartz. Environ. Sci. Technol., 40 (1), 142–148. Cheng, T., Barnett, M. O., Roden, E. E., & Zhuang, J. (2004). Effects of phosphate on uranium(VI) adsorption to goethite-coated sand. Environ. Sci. Technol., 39 (22), 6059–6065. Clark, S. B., Johnson, W. H., Malek, M. A., Serkiz, S. M., & Hinton, T. G. (1996). A comparison of sequential extraction techniques to estimate geochemical controls on the mobility of fission product, actinide, and heavy metal contaminants in soils. Radiochim. Acta, 74, 173–179. Coston, J. A., Fuller, C. C., & Davis, J. A. (1995). Pb2+ and Zn2+ adsorption by a natural aluminum- and iron-bearing surface coating on an aquifer sand. Geochim. Cosmochim. Acta, 59 (17), 3535–3547. Davis, J. A., & Kent, D. B. (1990). Surface complexation modeling in aqueous geochemistry. In: M. F. Hochella, & A. F. White (Eds). Mineral-Water Interface Geochemistry, Vol. 23. Mineralogical Society of America, Washington, DC, pp. 177–260. Dideriksen, K., & Stipp, S. L. S. (2003). The adsorption of glyphosate and phosphate to goethite: A molecular-scale atomic force microscopy study. Geochim. Cosmochim. Acta, 67 (18), 3313–3327. Dzombak, D. A., & Morel, F. M. M. (1990). Surface Complexation Modeling. Wiley-Interscience, New York, NY.

372


Elzinga, E. J., & Sparks, D. L. (2002). X-ray absorption spectroscopy study of the effects of pH and ionic strength on Pb(II) sorption to amorphous silica. Environ. Sci. Technol., 36 (20), 4352–4357. Gabriel, U., Gaudet, J.-P., Spadini, L., & Charlet, L. (1998). Reactive transport of uranyl in a goethite column: An experimental and modelling study. Chem. Geol., 151, 107–128. Gao, Y., & Mucci, A. (2003). Individual and competitive adsorption of phosphate and arsenate on goethite in artificial seawater. Chem. Geol., 199 (1–2), 91–109. Good, N. E., Winget, G. D., Winter, W., Connolly, T. N., Izawa, S., & Singh, R. M. M. (1966). Hydrogen ion buffers for biological research. Biochemistry, 5 (2), 467–477. Gra¨fe, M., Nachtegaal, M., & Sparks, D. L. (2004). Formation of metal-arsenate precipitates at the goethite-water interface. Environ. Sci. Technol., 38 (24), 6561–6570. Gra¨fe, M., & Sparks, D. L. (2005). Kinetics of zinc and arsenate co-sorption at the goethite-water interface. Geochim. Cosmochim. Acta, 69 (19), 4573–4595. Gunneriusson, L., Lovgren, L., & Sjoberg, S. (1994). Complexation of Pb(Ii) at the geothite (alpha-FeOOH) water interface – the influence of chloride. Geochim. Cosmochim. Acta, 58 (22), 4973–4983. Hayes, K. F., & Leckie, J. O. (1986). Mechanism of lead ion adsorption at the goethite-water interface. In: J. A. Davis, & K. F. Hayes (Eds). Geochemical Processes at Mineral Surfaces, Vol. 323. American Chemical Society, Washington, DC, pp. 114–141. Herbelin, A. L., & Westall, J. C. (1999). FITEQL – A Computer Program for Determination of Chemical Equilibrium Constants from Experimental Data. Oregon State University, OR. Hering, J. G., & Kraemer, S. (1994). Kinetics of complexation reactions at surfaces and in solution: Implications for enhanced radionuclide migration. Radiochim. Acta, 66/67, 63–71. Hettiarachchi, G. M., Pierzynski, G. M., & Ransom, M. D. (2000). In situ stabilization of soil lead using phosphorus and manganese oxide. Environ. Sci. Technol., 34 (21), 4614–4619. Ler, A., & Stanforth, R. (2003). Evidence for surface precipitation of phosphate on goethite. Environ. Sci. Technol., 37 (12), 2694–2700. Ma, Q. Y., Logan, T. J., & Traina, S. J. (1995). Lead immobilization from aqueoussolutions and contaminated soils using phosphate rocks. Environ. Sci. Technol., 29 (4), 1118–1126. Nilsson, N., Lovgren, L., & Sjoberg, S. (1992). Phosphate complexation at the surface of goethite. Chem. Speciation Bioavailability, 4 (4), 121–130. Nriagu, J. O. (1973). Lead orthophosphates – II. Stability of chloropyromorphite at 251C. Geochim. Cosmochim. Acta, 37, 367–377. Roe, A. L., Hayes, K. F., Chisholmbrause, C., Brown, G. E., Parks, G. A., Hodgson, K. O., & Leckie, J. O. (1991). In situ X-ray absorption study of lead-ion surface complexes at the goethite water interface. Langmuir, 7 (2), 367–373.


373

Ruby, M. V., Davis, A., & Nicholson, A. (1994). In-situ formation of lead phosphates in soils as a method to immobilize lead. Environ. Sci. Technol., 28 (4), 646–654. Ryan, J. A., Scheckel, K. G., Berti, W. R., Brown, S. L., Casteel, S. W., Chaney, R. L., Hallfrisch, J., Doolan, M., Grevatt, P., Maddaloni, M., & Mosby, D. (2004). Reducing children’s risk from lead in soil. Environ. Sci. Technol., 38 (1), 18A–24A. Schecher, W. D., & McAvoy, D. C. (1998). MINEQL+: A Chemical Equilibrium Modeling System. Version 4.5. Environmental Research Software. Scheidegger, A., Borkovec, M., & Sticher, H. (1993). Coating of silica sand with goethite: Preparation and analytical identification. Geoderma, 58, 43–65. Schwertmann, U., & Cornell, R. M. (2000). Iron Oxides in the Laboratory. Wiley-VCH, New York, NY. Soares, H. M. V. M., Conde, P. C. F. L., Almeida, A. A. N., & Vasconcelos, M. T. S. D. (1999). Evaluation of n-substituted aminosulfonic acid pH buffers with a morpholinic ring for cadmium and lead speciation studies by electroanalytical techniques. Anal. Chim. Acta, 394, 325–335. Sposito, G. (1986). Distinguishing adsorption from surface precipitation. In: J. A. Davis, & K. F. Hayes (Eds). Geochemical Processes at Mineral Surfaces, Vol. 323. American Chemical Society, Washington, DC, pp. 217–228. Stumm, W. (1992). Chemistry of the Solid-Water Interface. Wiley, New York, NY. Sverjensky, D. A., & Sahai, N. (1996). Theoretical prediction of single-site surfaceprotonation equilibrium constants for oxides and silicates in water. Geochim. Cosmochim. Acta, 60 (20), 3773–3797. Trivedi, P., Dyer, J. A., & Sparks, D. L. (2003). Lead sorption onto ferrihydrite. 1. A macroscopic and spectroscopic assessment. Environ. Sci. Technol., 37 (5), 908–914. Weesner, F. J., & Bleam, W. F. (1998). Binding characteristics of Pb2+ on anionmodified and pristine hydrous oxide surfaces studied by electrophoretic mobility and X-ray absorption spectroscopy. J. Colloid Interface Sci., 205 (2), 380–389. Yang, J., Mosby, D. E., Casteel, S. W., & Blanchar, R. W. (2001). Lead immobilization using phosphoric acid in a smelter-contaminated urban soil. Environ. Sci. Technol., 35 (17), 3553–3559. Yee, N., & Fein, J. B. (2003). Quantifying metal adsorption onto bacteria mixtures: A test and application of the surface complexation model. Geomicrobiol. J., 20 (1), 43–60. Zhang, P. C., Ryan, J. A., & Bryndzia, L. T. (1997). Pyromorphite formation from goethite adsorbed lead. Environ. Sci. Technol., 31 (9), 2673–2678.


Chapter 14

Uranium(VI) Release from Contaminated Vadose Zone Sediments: Estimation of Potential Contributions from Dissolution and Desorption Deborah L. Bond1,, James A. Davis1 and John M. Zachara2 1 2

U.S. Geological Survey, Menlo Park, CA 94025, USA Pacific Northwest National Laboratory, Richland, WA 99352, USA

ABSTRACT A key difficulty in developing accurate, science-based conceptual models for remediation of contaminated field sites is the proper accounting of multiple coupled geochemical and hydrologic processes. An example of such a difficulty is the separation of desorption and dissolution processes in releasing contaminants from sediments to groundwaters; very few studies are found in the literature that attempt to quantify contaminant release by these two processes. In this study, the results from several extraction techniques, isotopic exchange experiments, and published spectroscopic studies were combined to estimate the contributions of desorption and dissolution to U(VI) release from contaminated sediments collected from the vadose zone beneath former waste disposal ponds in the Hanford 300-Area (Washington State). Vertical profiles of sediments were collected at four locations from secondary pond surfaces down to, and slightly below, the water table. In three of the four profiles, uranium concentration gradients were observed in the sediments, with the highest U concentrations at the top of the profile. One of the vertical profiles contained sediments with U concentrations up to 4.2 107 mol g1 (100 ppm). U(VI) release to artificial groundwater solutions (AGWs) and extracts from these high-U concentration sediments occurred primarily from dissolution of precipitated U(VI) minerals, including the mineral metatorbernite, [Cu(UO2PO4)2 8H2O]. At the bottom of this profile, beneath the water table, and in all three of the other profiles, U concentrations were o5.88 108 mol g1

Corresponding author. Tel.: 650 329 4529; Fax: 650 329 4545;

E-mail: [email protected] (D.L. Bond).

376

D. L. Bond et al.

(14 ppm), and U(VI) release to AGWs occurred primarily due to desorption of U(VI). When reacted in batch experiments with AGWs with compositions representative of the range of chemical conditions in the underlying aquifer, all samples released U(VI) at concentrations greater than regulatory limits within few hours. A semi-mechanistic surface complexation model was developed to describe U(VI) adsorption on sediments collected from near the water table, as a function of pH, alkalinity, and Ca and U(VI) concentrations, using ranges in these variables relevant to groundwater conditions in the aquifer. Dilute (bi)carbonate solution extractions and uranium isotopic exchange methods were capable of estimating adsorbed U(VI) in samples where U(VI) release was predominantly due to U(VI) desorption; these techniques were not effective at estimating adsorbed U(VI) where U(VI) release was affected by dissolution of U(VI) minerals. The combination of extraction and isotopic exchange results, spectroscopic studies, and surface complexation modeling allow an adequate understanding for the development of a geochemical conceptual model for U(VI) release to the aquifer. The overall approach has generic value for evaluating the potential for release of metals and radionuclides from sediments that contain both precipitated and adsorbed contaminant speciation.

14.1. Introduction Uranium (U) is a pollutant of concern in the United States due to subsurface contamination at numerous sites (Crowley and Ahearne, 2002). At many of these sites, groundwater plumes with high concentrations of U(VI) and other contaminants have developed (Riley et al., 1992), some of which discharge to rivers. For example, a groundwater plume with elevated concentrations of dissolved U(VI) underlies the North and South 300-Area Process Ponds in the 300-FF-5 Operable Unit at the Hanford site in Washington State (Fig. 14.1). The infiltration basins at this site operated from 1943 to 1975, receiving various waste streams containing high concentrations of U, Cu, F, Al, and nitrate, and the pH of the wastewater in the ponds varied temporally from 2 to 11 (Zachara et al., 2005). Concentrations of U(VI) in the groundwater plume beneath the site have been persistently greater than expected (Zachara et al., 2005). Despite removal of highly contaminated pond bottom sediments (>4.2 mmol g1 or 1,000 ppm U), studies indicate that the 20–30 ft vadose zone beneath the infiltration ponds likely serves as a continuing source of U(VI) to the groundwater plume (Qafoku et al., 2005; Catalano et al., 2006). A geochemical model to evaluate the release of U(VI) from the sediments is needed as part of an overall conceptual model for the 300-Area site. The conceptual model will assist in understanding the expected longevity of the groundwater plume and its impact on the Columbia River, which is located 300 ft to the

Uranium(VI) Release from Contaminated Vadose Zone Sediments

377

North Process Pond

5

NPP1 NPP2 SPP1

South Process Pond

SPP2

10

20

Rive

10

30

mbia

50

Colu

0

r

Figure 14.1: Map of 300-Area at Hanford, showing the Locations of the North and South Process Pond Pits. Contour Lines Indicate U(VI) Concentrations in the Groundwater Plume Beneath the 300-Area in December 2002. east of the former infiltration ponds (Fig. 14.1). U(VI) is transported downward within the vadose zone as it is released from the contaminated sediments during the relatively wet season, and small amounts of drainage may reach the water table. The hydrologic model for the vadose zone is complex. At high stages of the Columbia River, the groundwater table rises into the lower vadose zone (Lindberg and Peterson, 2004), leading to a condition in which the deeper vadose zone sediments may serve as both a source and sink for U(VI) in the system (Qafoku et al., 2005). The geochemical conceptual model needs to be based on data from several different sources: (1) the conditions at the field site (groundwater compositions, redox status, etc.), (2) batch U(VI) water-sediment partitioning experiments, (3) long-term U(VI) release kinetic experiments, and (4) sediment characterization studies. Spectroscopic characterization of the vadose zone sediments has determined that U chemical speciation (Fig. 14.2)

378

D. L. Bond et al.

Original pond bottom

Excavated material and pond precipitates

Dispersed U(VI) coprecipitated in calcite

(Wang et al, 2005; Catalano et al, 2006)

D E P T H

Ground surface post-excavation

Upper Vadose zone sediments

Intermediate

Discrete uranyl phosphate precipitates (metatorbernite) Weak U(VI) adsorption complexes

Lower Groundwater Fines (colloidal particulates) Groundwater

Figure 14.2: Conceptual Diagram of a Depth Profile within a Pit at the 300-Area showing U(VI) Speciation from Previous Spectroscopic Studies. consists of: (a) U(VI) co-precipitated with calcite in the pond bottom and the uppermost vadose zone sediments (Wang et al., 2005; Catalano et al., 2006); (b) U(VI) precipitated as metatorbernite, [Cu(UO2PO4)2 8H2O] and other U(VI) minerals, at upper to intermediate depths in the vadose zone (Arai et al., 2007; Catalano et al., 2006); and (c) U(VI) adsorbed onto phyllosilicates at deeper depths in the vadose zone and extending into the saturated zone (Catalano et al., 2006). Hence, there is an important need at this site (and many other metal-contaminated sites) to have the capability to estimate the separate contributions of desorption and dissolution as release mechanisms. U(VI) desorption kinetics from the deeper vadose zone sediments are slow, contributing to significant retardation during transport and extensive tailing (Qafoku et al., 2005). Because of the presence of precipitated and co-precipitated U(VI) in the uppermost vadose zone sediments, the release of U(VI) to infiltrating precipitation can also be expected to be complex. In this chapter, we present the results of kinetic desorption and dissolution experiments conducted with depth sequences of vadose zone sediments collected from four pits excavated beneath the 300-Area infiltration ponds. Various extraction and uranium isotopic exchange techniques were employed to estimate the fraction of total sediment U that is available for potential release to the groundwater plume during vadose zone recharge or temporal flooding. When combined with previous spectroscopic studies


379

(Arai et al., 2007; Catalano et al., 2006), a conceptual geochemical model for the vadose zone sediments can be developed, with estimates of the potential contributions from dissolution and desorption to U(VI) release from the sediments in representative groundwater solutions. A semi-mechanistic surface complexation model is calibrated to describe U(VI) adsorption– desorption equilibria for the deeper vadose zone sediments as a function of pH, alkalinity, and Ca and U(VI) concentrations, similar to the approach used in other studies (Davis et al., 2002, 2004a). Variables were studied over ranges that are relevant to the groundwater in the aquifer beneath the site. The surface complexation model makes it possible to estimate separate contributions to U(VI) release from dissolution and desorption from the more highly contaminated sediments in the upper vadose zone.

14.2. Experimental 14.2.1. Site Description Samples were collected from the 300-Area North and South Processing Ponds at the Hanford site, two main disposal basins that overlay different concentration regions of the U(VI) plume (Fig. 14.1). The ponds served as liquid disposal units from 1943 to 1975, receiving cooling water and low-level liquid wastes from fuel fabrication facilities including U, copper, cobalt, and plutonium (Zachara et al., 2005). The pH of the waste varied widely due to the addition of acidic U(VI)/Cu(II) and basic sodium-aluminate solutions. Sodium hydroxide was added to the former waste streams to impede migration of copper through the aquifer to the Columbia River. Between 30,000 and 60,000 kg of U were disposed in the ponds (USDOE, 2005). During the period between 1948 and 1975, several unplanned releases of holding effluent occurred (USEPA, 1996). The result of these leaks and additional seepage into the vadose zone are evident in the U(VI) plume emanating from the South Processing Pond (Fig. 14.1). In 1996, following EPA recommendations (USEPA, 1996), 640,000 tons of contaminated soil were removed from the ponds to a waste disposal site. Additionally, in order to reduce the source of contamination from the ponds, several feet of sediment were scraped from the pond bottoms during 2001–2002, exposing a secondary surface. In 2003, two pits were excavated within each pond from the secondary surface down to the water table, 20 ft below ground surface (bgs). The pits are referred to as NPP1, NPP2, SPP1, and SPP2; locations are shown in Fig. 14.1. Sediments were collected by

380

D. L. Bond et al.

excavator from these pits at intervals of 2 or 4 ft. Sample designations identify the pit and depth below the secondary pit surface, e.g., North Processing Pit number 1, 16 ft bgs is designated NPP1-16. The water table was located at 22 ft bgs. The elevations of the samples relative to the water table are shown in Fig. 14.3. One sediment sample from each pit was collected from beneath the water table. Particle size in the samples ranged from clay-sized to cobbles, with river cobbles accounting for more than 65% of the mass. Samples identified as ‘‘groundwater fines’’ were collected separately as suspended material in groundwater that began to infiltrate the pits as excavation approached the water table. Generally, total U concentrations were at or below detection (o2 108 mol g1) in the size fraction ranging from 2.0 to 75 mm (Zachara et al., 2005). Sediments (19 samples from the 4 pits) were air-dried and sieved to o2 mm size fraction. North Pond Approximate original pond bottom

NPP 1

South Pond NPP 2

SPP 1

0

SPP 2

0 4 4 8 4 8

0 12 8 4

12 16

12 8

Annual water-table range

Depth below excavated ground surface (ft)

0

16 20

16 12

20 24

20 16

24 Fines = 157 ppm

Fines = 21 ppm 0

5

10

15

0

[U] (ppm) Excavated Material

40

80

Fines = 31 ppm 120

0

[U] (ppm) Sediment Sample

2

4

6

[U] (ppm)

Fines = 13 ppm 8

0

4

8

12

[U] (ppm)

Water table at time of sampling

Figure 14.3: Depth Sequence Profiles of the Four Sampling Pits in the North and South Process Ponds of the 300-Area at Hanford. Sample Depth below Excavated (or Secondary) Pond Surface is shown along with U Concentrations. Seasonal Water Table Variation is also shown. Note Different Scales for Total U Concentrations in the Sediments. 1 ppm U ¼ 4.2 nmol g1 U.


381

14.2.2. Surface Area Measurements Specific surface areas were determined by N2 adsorption at 77.351K (Micromeritics Tristar 3000) under atmospheric pressure. Samples were heated to 401C and degassed under N2 gas for 20 h. Measurements were made using the five-point method on 1 g sample masses. Surface areas of the groundwater fines were measured by Quantachrome Corporation (Boynton Beach, FL).

14.2.3. Total Uranium and Copper Content Total U content was measured for each sample (o2 mm) by nondestructive g-spectrometry. 238U was determined by measurement of the 234Th daughter 63 keV gamma ray emission line, assuming secular equilibrium in the sample (Davis and Curtis, 2003). 235U was determined from its 186 keV gamma ray after correction for the 226Ra contribution to this energy region. The 226Ra correction was based on measurement of its gamma-emitting daughters. Copper (Cu) content of sample material was obtained by energy dispersive X-ray fluorescence spectroscopy (KEVEX 0810A system) at Pacific Northwest National Laboratory (PNNL) utilizing the backscatter fundamental parameter approach. Solid samples were pelletized from 500 mg of dry material (o149 mm) into a 3 cm wafer. The X-rays were detected by a cryogenically cooled solid state lithium drifted silicon (Si(Li)) detector connected to a multi-channel analyzer. The method of peak analysis is described in Nielson (1978).

14.2.4. Hydroxylamine-Hydrochloride (HH) Extractions and Ammonium Oxalate Extractions Hydroxylamine-hydrochloride (HH) extractions were performed in duplicate to dissolve and estimate the abundance of poorly crystalline iron hydroxides (Chao and Zhou, 1983) in selected sediment samples. Two hundred grams of 0.25 M NH2OH HCl in 0.25 M HCl at 501C were added to 10 g of sediment in a 250 mL bottle and placed in a water bath at 501C. The bottles were shaken mechanically and sampled at 30 min. Ammonium oxalate (AMOX) extractions were performed as a second method to dissolve and estimate the abundance of poorly crystalline iron (oxy)hydroxides (Chao and Zhou, 1983). Two hundred milliliters of 0.12 M

382

D. L. Bond et al.

oxalic acid and 0.11 M AMOX (pH 3) were added to 5 g of sediment in foilwrapped 250 mL centrifuge bottles and shaken mechanically for 4 h at room temperature in the dark. For both extractions, an aliquot of the extract was transferred to a polycarbonate centrifuge tube and centrifuged to remove solids. From the supernatant, a subsample was converted to its nitrate salt by ad-mixing concentrated nitric acid and hydrogen peroxide followed by evaporation to dryness (Davis and Curtis, 2003). The solid was reconstituted in 0.15 M HNO3 and analyzed for U(VI) by kinetic phosphorescence analysis (KPA). The remainder of the supernatant was filtered through a 0.45 mm PVDF filter and diluted with 0.15 M HNO3 for analysis by inductively coupled plasmaatomic emission spectroscopy (ICP-AES). 14.2.5. Dithionite Citrate Bicarbonate Extractions Dithionite citrate bicarbonate (DCB) extractions were performed in duplicate to estimate the abundance of crystalline iron oxides in selected sediment samples (Chao and Zhou, 1983). Two hundred milliliters of 0.30 M sodium citrate, 0.20 M sodium bicarbonate, and 0.14 M sodium dithionite (pH 8.3) were added to 5 g of sediment in a 250 mL centrifuge bottle and placed in a water bath at 851C. The bottles were shaken mechanically and sampled after 0.5 h. The bottles were centrifuged, supernatant collected, and the extraction was repeated. The samples were centrifuged again and the leachates were combined. Samples were processed and analyzed by KPA and ICP-AES in the manner described above. The authors recognize that U(VI) released in these extractions is not necessarily associated with the dissolving crystalline iron oxide phases. However, useful comparative information can be obtained by putting the results of the extraction in context with the results from other extraction methods. 14.2.6. Formate Buffer Extractions Formate buffer extractions were conducted to estimate the quantity of U(VI) that was either sorbed U(VI) or present in mineral phases that could be dissolved by dilute acid, e.g., U(VI) co-precipitated within carbonates or poorly crystalline oxides. Sodium formate (0.5 M) was acidified to pH 3.5 with formic acid and added to sediments at a suspension density of 50 g L1 in 500 mL. We are unaware of a previous publication discussing the use of formate buffer as an extractant. It was chosen as an extractant here because


383

of its ability to: (1) dissolve poorly crystalline hydrous iron oxide and aluminum oxide phases (but not crystalline hydrous iron oxide phases) at pH 3.5 (Chao and Zhou, 1983), (2) expected complete desorption of U(VI) from oxide surfaces (Payne et al., 1998), and (3) dissolve carbonate mineral phases. The technique should be more quantitative at releasing U(VI) co-precipitated with carbonate mineral phases than acetate buffer extractions (pH 4.7), because any released U(VI) will likely not be re-adsorbed at pH 3.5 (Payne et al., 1998). Bottles were placed on an orbital shaker at room temperature and subsampled at various times up to 4 weeks. The pH varied minimally during the extractions, typically by only 0.05 pH units. Samples were processed and analyzed by KPA and ICP-AES in the manner described above. In a few samples (NPP2-2, NPP2-4, and NPP2-8), extracted copper concentrations were in excess of 1 mM. For these samples, the KPA method was less effective for measurement of U(VI) due to quenching effects. Dissolved U(VI) concentrations for these samples were measured instead by ICP-AES, as the concentrations were well within the detection limit for this instrument. 14.2.7. Dilute (Bi)Carbonate Extractions Dilute (bi)carbonate extractions were conducted to determine the amount of U(VI) released from the sediment samples under moderately alkaline conditions. A solution consisting of 1.44 102 M NaHCO3 and 2.8 103 M Na2CO3 at pH 9.45 was added to the sediments in a ratio of 50 g L1, following the approach of Kohler et al. (2004). The ionic strength and alkalinity of the solution are 0.022 M and 20 meq L1, respectively. The bottles were placed on an orbital shaker at room temperature and sampled at various times for at least 3 weeks. The extractions were performed in duplicate and pH values were measured at each time point. Samples were removed by allowing the suspension to settle and removing an aliquot of supernatant for centrifugation. For KPA analysis, a subsample of centrifugate supernatant was diluted with 0.15 M HNO3 for analysis. ICP-AES samples were filtered (0.45 mm) and diluted in 0.15 M HNO3 prior to analysis. For most samples, the pH of the extract initially dropped o0.5 pH units and remained higher than 8.8 throughout the extraction. However, during extractions of the groundwater fines samples, the pH dropped to 8.5 within the first hour of extraction and remained constant thereafter. As reported in Kohler et al. (2004), it is important to keep the pH above 8.8 during the extraction to avoid reaching conditions at which released U(VI) would begin to re-adsorb. Extractions of the groundwater fines samples were repeated at lower suspension densities (20 g L1) to maintain a pHZ8.8 and prevent

384

D. L. Bond et al.

U(VI) adsorption. Possible re-adsorption of extracted U(VI) was evaluated at the end of (bi)carbonate extractions by addition of 233U(VI) isotope and measurement of tracer activity for 24 h. No re-adsorption was measured for samples with pH above 8.6 (within 5% error). 14.2.8. Artificial Groundwater Experiments Sediments were reacted with artificial groundwater solutions (AGWs) with varying alkalinity and ionic strength to determine the amount of U(VI) released as a function of chemical conditions. The initial composition of the artificial groundwaters (Table 14.1) was based on the range of well water compositions in the 300-Area (Serne et al., 2003). The initial pH (Table 14.1) refers to the pH of each AGW prior to contact with the sediments. To study U(VI) desorption as a function of alkalinity, a series of groundwaters (AGW2-6, 12-13) was prepared with varying bicarbonate concentration. The concentrations of other major solutes were kept nearly constant and varied only to obtain waters with similar ionic strength. The composition of AGW2 was close to saturation with respect to calcite, while the other AGW solutions were significantly undersaturated (prior to sediment addition). Another series of groundwaters (AGW8-11) were synthesized to mimic AGW3-6, except with ionic strength increased to I ¼ 0.1 M by addition of sodium nitrate. Groundwaters were filtered (0.45 mm) prior to contact with sediments. The batch experiments were conducted with variable suspension densities (25 to 1,600 g L1) of sediment in polyethylene centrifuge tubes or bottles placed on a shaker table for up to 2 weeks. Some centrifuge tubes were also mixed on an end-over-end rotator (14 rpm) for method comparison and showed no statistical difference in results. Individual tubes were sacrificesampled during the course of the experiments, and the pH was measured immediately. The alkalinity of each sample was measured on a filtered, unacidified aliquot. Samples were then centrifuged (16,270 g RCF for 10 min) and an aliquot of supernatant was diluted with 0.15 M HNO3 for KPA analysis. Centrifugation was preferred as a method of phase separation to avoid U(VI) sorption onto filters. Sorption of U(VI) to sample-tubes was monitored with control samples of U(VI) spiked AGW and solution concentration remained constant over time. ICP-AES samples were filtered (0.45 mm) prior to dilution with 0.15 M HNO3 and analysis. The pH in all AGW experiments was well buffered by the carbonate alkalinity of the water and the buffer capacity of the sediments. There was some variability in pH, depending on the initial alkalinity, ionic strength, and solid:liquid ratio in each experiment. The range of pH observed among the

AGW 2 3 4 5 6 8 9 10 11 12 13

Ca2+

Mg2+

K+

Na+

HCO 3

SO2 4

NO 3

Alkalinity (meq L1)

Ionic strength

Initial pH

0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.6

0.2 0.2 0.1 0.4 0.5 0.2 0.1 0.4 0.5 0.1 0.5

0.1 0.1 0.1 0.5 0.5 0.1 0.1 0.5 0.5 0.1 0.2

15 7.0 8.0 6.0 5.5 97 98 96 96 8.4 8.9

10 2.0 4.0 1.0 0.5 2.0 4.0 1.0 0.5 6.0 8.0

1.8 1.8 1.2 2.0 2.1 1.8 1.2 2.0 2.1 0.9 0.8

3.0 3.0 3.0 3.5 3.5 93 93 94 94 2.1 1.7

10 2.0 4.0 1.0 0.5 2.0 4.0 1.0 0.5 6.0 8.0

19.3 11.3 11.4 11.4 11.4 101 101 102 101 11.5 13.2

8.30 8.40 8.65 8.07 7.85 8.26 8.30 8.12 8.05 8.59 8.55


Table 14.1: Composition of Artificial Groundwaters (Elemental Concentration in mmol L1), pCO2 ¼ 103.5.

385

386

D. L. Bond et al.

combinations of low ionic strength groundwater solutions and sediment samples for pits NPP1, SPP1, and SPP2 was 7.970.3. Sediment samples from the NPP2 pit yielded a lower pH range of 7.170.3. Experiments conducted with AGWs 8-11 (higher ionic strength) had slightly higher pH values. For example, sediments reacted with AGW9 equilibrated at pH 8.370.1.

14.2.9. U(VI) Isotopic Exchange Experiments Isotope exchange experiments were conducted following a modified approach of Kohler et al. (2004) to determine ‘‘labile’’ abundances of sediment U(VI). Variable suspension densities (25–200 g L1) were used to control the concentration of U(VI) in solution, in order to accurately measure 233U(VI) activity. Prior to the addition of the 233U(VI) isotope, the sediments were reacted with AGW4 (see section ‘‘Artificial Groundwater Experiments’’) for either: (1) 24 h, or (2) 1,260 h. The latter reaction time allowed the achievement of near steady-state dissolved U(VI) concentrations (o2% change per week) prior to the isotope addition. Subsamples were collected for pH, alkalinity, U(VI) concentration, and background activity (measured by LSC, liquid scintillation counting) prior to isotope addition, and then the suspensions were spiked with a 233U secondary stock to achieve a concentration of 4–10 109 M 233U(VI) (20–50 dpm mL1). The sediments were then shaken for up to 3,400 h, with aliquots removed periodically and measured for pH, alkalinity, 233U activity by LSC, and U(VI) concentration by KPA. Exchangeable (‘‘labile’’) U(VI) in the sediment was then determined by: ASystem C (14.1) A where CLabile is the concentration (mol L1) of the exchangeable U(VI), ASystem the total 233U activity (dpm L1) in the system, A the activity of dissolved 233U (dpm L1), and C (mol L1) the concentration of dissolved U(VI) (Kohler et al., 2004). Corrections were made for: (1) 233U and U(VI) removed during sampling and (2) the contribution of 233U to total system uranium concentration. Addition of the 233U(VI) spike increased the dissolved U(VI) concentrations in the experiments by 0.2–5.4% (average ¼ 1.0%). C Labile ¼

14.2.10. U(VI) Sorption Isotherms In order to develop a U(VI) sorption isotherm, experiments were conducted for selected deeper pit samples suspended in AGW4 with added U(VI).


387

Solutions were prepared by adding U(VI) (100 mg L1 in 2% HNO3) to AGW4 or AGW9, followed by a small adjustment of the pH to the original pH value. These solutions were stored overnight to allow equilibration of aqueous U(VI) speciation at the higher U(VI) concentration. Sediment samples in the experiments were pretreated by suspension in AGW4 or AGW9 for 72 h, after which the pH was measured and the tubes were centrifuged. Supernatant was collected to measure alkalinity, U(VI) concentration (by KPA), and water composition determined by ICP-AES. Following the pretreatment, weighed aliquots of U(VI)-spiked groundwater were added to each centrifuge tube. Tubes were sacrificed-sampled up to 100 h of reaction time. Alkalinities remained stable and pH in the experiments was relatively constant with time for each sediment sample, but varied from 7.9 to 8.25 among the sediment samples and at different solid:liquid ratios. Samples were centrifuged and processed for analysis by ICP-AES and KPA.

14.2.11. Modeling FITEQL 4.0 (Herbelin and Westall, 1999) was used for aqueous speciation and surface complexation modeling. The Davies equation was used for activity correction of aqueous species only. Thermodynamic data (Table 14.2) used in the modeling are consistent with the most recent NEA database for uranium (Guillaumont et al., 2003), except the aqueous ternary species, CaUO2(CO3)2 and Ca2UO2(CO3)03(aq) (Kalmykov and Choppin, 2000; 3 Bernhard et al., 2001), were also included. Calcite equilibrium in the experiments was not assumed; measured dissolved Ca values were used as FITEQL input and neither calcite precipitation nor dissolution was included in the calculations. FITEQL 4.0 was used to determine the best fit of various U(VI) surface reactions or combinations of reactions to experimental data in model calculations using a semi-mechanistic modeling approach (Davis et al., 2004a). Relative errors of 1% in the concentrations of surface sites, 3% in total U(VI), 4% in adsorbed U(VI), and 5% in log [H+] and log [H2CO3] were used as FITEQL input values. The purpose of the surface complexation model developed here is twofold: (1) to estimate U(VI) adsorption in samples from the sediment profile that may contain both adsorbed and precipitated U(VI), and (2) to provide a quantitative estimate of U(VI) adsorption– desorption equilibria for future reactive transport simulations of U fate and transport in the vadose and saturated zones of the aquifer underlying the former disposal ponds. As in previous work with this non-electrostatic

388

D. L. Bond et al.

Table 14.2: Formation Constants for U(VI) Solution Species. Reaction + + UO2+ 2 +H2O3UO2OH +H 2+ UO2 +2H2O3UO2(OH)2,aq+2H+ + UO2+ 2 +3H2O3UO2(OH)3 +3H 2+ 2 UO2 +4H2O3UO2(OH)4 +4H+ 3+ +H+ 2UO2+ 2 +H2O3(UO2)2OH 2+ + 2UO2 +2H2O3(UO2)2(OH)2+ 2 +2H 2+ 2+ 3UO2 +4H2O3(UO2)3(OH)4 +4H+ + + 3UO2+ 2 +5H2O3(UO2)3(OH)5 +5H + 2+ 3UO2 +7H2O3(UO2)3(OH)7 +7H + + 4UO2+ 2 +7H2O3(UO2)4(OH)7 +7H 2 UO2+ 2 +CO3 3UO2CO3(aq) 2+ 2 UO2 +2CO2 3 3UO2(CO3)2 2+ 2 UO2 +3CO3 3UO2(CO3)4 3 + 2 2UO2+ 2 +CO3 +3H2O3(UO2)2CO3(OH)3 +3H 2 2 Ca2++UO2+ 2 +3CO3 3CaUO2(CO3)3 2+ 2+ 2 2Ca +UO2 +3CO3 3Ca2UO2(CO3)3(aq) + UO2+ 2 +NO3 3UO2NO3 2+ + UO2 +Cl 3UO2Cl UO2+ 2 +2Cl 3UO2Cl2(aq) 2+ 2 UO2 +SO4 3UO2SO4(aq) 2 2 UO2+ 2 +2SO4 3UO2(SO4)2

Log K (I ¼ 0)a 5.25 12.15 20.25 32.4 2.70 5.62 11.90 15.55 32.20 21.9 9.94 16.61 21.84 0.855 25.64b 30.04c 0.3 0.17 1.1 3.15 4.14

a

Values from Guillaumont et al. (2003), unless otherwise indicated. Bernhard et al. (2001), with correction to be consistent with Guillaumont et al. (2003). c Kalmykov and Choppin (2000), with correction to be consistent with Guillaumont et al. (2003). b

modeling approach (Davis et al., 1998, 2004a), surface protonation and deprotonation reactions are not used.

14.3. Results 14.3.1. Sediment Characterization Mineralogical and particle size analysis of the bulk sediment and clay-sized fraction have been carried out in previous work (Serne et al., 2003; Qafoku et al., 2005; Zachara et al., 2005; Catalano et al., 2006). Diffraction analyses (Serne et al., 2003; Zachara et al., 2005) reveal nearly identical mineralogical assemblages in each of the 19 samples studied. Bulk XRD data of the


389

o2 mm fraction indicate that sediments in the NPP1 and SPP2 pits are primarily composed of quartz, plagioclase feldspar, muscovite, and hornblende. The vadose zone sediments contain little or no calcite (o0.01% inorganic carbon), whereas the pond sediments and sample NPP2-0.5 (not studied in this paper) contained abundant calcite (Qafoku et al., 2005; Catalano et al., 2006). Extracted iron varied from 77 to 300 mmol g1 in the deeper pit samples and fines (Table 14.3), presumably present primarily as hydrous iron oxides, with 30–60% of the iron oxides present as poorly crystalline iron (HH and AMOX compared with Fe in DCB extractions). Iron dissolved by the HH and AMOX methods varied significantly by method, suggesting that iron oxide phases of intermediate crystallinity were present. Microscopic studies of thin sections of the deeper vadose zone sediments (near the water table) showed that larger sediment grains were coated with thin layers of phyllosilicates (e.g., smectite, vermiculite, and chlorite) that were identified by X-ray diffraction of the clay-sized fraction (J.P. McKinley, personal communication). The poorly crystalline iron hydroxide phases likely result from the weathering of chlorite in the grain coatings. Surface area ranged from 15 to 27 m2 g1 among the pit samples; groundwater fines ranged from 41 to 54 m2 g1 (Table 14.4). Values of total U concentration in the o2 mm sediment samples ranged from 1.2 108 to 6.6 107 mol g1 (2.9–157 ppm) (Table 14.4). The average mass ratio of 235U to 238U for all samples was 0.008470.0011 in agreement with the natural mass abundance ratio of 0.0073. Total U values measured by g-spectrometry agree reasonably well with measurements made by X-ray fluorescence spectroscopy (J.M. Zachara, personal communication). NPP2 pit samples contained the highest U concentrations, nearly 10 times that of the other pit samples. In contrast to the other pits, the high U concentrations extended down to the water table (Fig. 14.3). With the exception of pit SPP1, total U concentrations decreased with depth. Excluding pit NPP2, total U concentrations ranged from 1.2 108 to 4.0 108 mol g1 (2.9–9.6 ppm) in sediments near the water table, compared to background values near 5.0 109 mol g1 (1.2 ppm) for Hanford sediments (Serne et al., 2003). The fines samples, collected from the water table, contained total U concentrations ranging from 5.5 108 to 6.6 107 mol g1 (13–157 ppm), with the NPP2-fines samples considerably greater than the other three pits. The groundwater fines were collected as suspended material in groundwater that infiltrated the pits during sampling of sediments beneath the water table. Dissolved U(VI) concentrations were determined in the extractions of selected deep pit samples that were conducted to dissolve iron oxide phases (Table 14.3). Eighty percent or more of the total U was dissolved or desorbed

390 D. L. Bond et al.

Table 14.3: Iron (mmol g1) and Uranium (% of Total U) Removed in Various Extractions. Sample NPP1-16 NPP1-20 SPP2-16 SPP2-18 NPP1-fines NPP2-fines SPP1-fines SPP2-fines

DCB Fe (mmol g1)a

HH Fe (mmol g1)b

AMOX Fe (mmol g1)c

DCB % of Utota,d

HH % of Utotb,d

AMOX % of Utotc,d

158 142 151 77 296 224 239 232

41 46 – 19 102 90 90 83

91 95 88 48 108 89 86 70

100 100 90.5 66.7 100 104 99.2 96.6

92.2 75.6 – 58.3 96.6 109 91.6 91.2

81.7 74.7 73.0 50.0 94.4 114 91.6 87.7

Values are the mean of two replicates, error based on replicate variation. a

DCB refers to dithionite citrate bicarbonate extractions. HH refers to hydroxylamine hydrochloride extraction for 0.5 h (dissolves poorly crystalline minerals). c AMOX refers to ammonium oxalate extraction (dissolves poorly crystalline iron oxyhydroxides). d % of Utot refers to percentage of total uranium dissolved by each extraction method. b


391

Table 14.4: Total Uranium and U/Cu Extracted by Sodium Formate (nmol g1). Sample

Surface area (m2 g1)

Total U measured by gspectrometrya

[U(VI)] Extracted by formateb

Total [Cu] measured by XRFc

[Cu] Extracted by formateb

NPP1-8 NPP1-12 NPP1-16 NPP1-20 NPP2-2 NPP2-4 NPP2-8 NPP2-12 SPP1-16 SPP1-18 SPP1-22 SPP2-8 SPP2-12 SPP2-16 SPP2-18 NPP1-fines NPP2-fines SPP1-fines SPP2-fines

19.970.04 27.270.05 27.270.01 17.570.06 17.870.23 21.870.03 18.570.04 14.770.05 21.270.04 22.070.09 25.870.63 17.770.11 15.970.04 15.570.05 15.370.19 46.972.3 N/A 53.972.7 40.572.0

44.073.2 58.974.2 40.473.2 26.372.1 444724 421721 167710 59.773.5 30.872.4 31.272.8 33.072.4 45.372.6 33.572.7 16.271.6 12.071.5 89.075.2 660734 13177.2 55.973.7

44.771.4 53.271.8 33.371.1 20.470.65 32275.8 34476.8 11473.3 63.970.73 31.471.0 24.570.77 22.770.73 41.571.3 33.571.1 15.170.45 7.2870.23 69.770.79 N/A 10371.2 42.471.4

9857110 1,2107120 1,1707120 1,0907110 64,80073,300 75,50073,800 20,20071,000 1,7707140 8777120 8977120 8157100 9037120 721798 845798 895798 2,6607160 43,00072,200 1,5207130 1,3307100

72.972.7 78.878.0 12373.5 151713 61,00074.4 69,10073,500 17,1007360 784736 34.370.44 34.670.44 14.670.90 31.274.5 18.271.3 28.470.59 40.171.8 31470.43 N/A 21573.6 10270.89


Sum of 238U and 235U. Formate extraction (72 h). c Total copper measured by X-ray fluorescence spectroscopy (XRF). Data collection and analyses by Steven Smith, PNNL. b

from NPP1 sediments in the HH and AMOX extractions, and essentially 100% in DCB extractions. SPP2 deep pit samples released a lower percentage of total U, ranging from 50 to 75% in the HH and AMOX extractions, and 70–100% during the DCB extractions. Between 92 and 96% of total U was dissolved from the groundwater fines samples during these extractions. The results do not mean that U(VI) released in these extractions was necessarily associated with the phases dissolved. For example, the low pH of the HH extraction can likely cause U(VI) desorption from any mineral phase in the sample.

14.3.2. Formate Buffer Extractions Formate extractions (pH 3.5) were conducted to measure elemental release as a function of time under dilute acidic conditions. U(VI) concentrations increased rapidly at the beginning of the extraction and appeared to reach steady-state at

392

D. L. Bond et al.

24 h. Formate-extractable U ranged from 7.3 109 to 3.4 107 mol g1 (Table 14.4), equivalent to 61–107% of the total U in the samples (Table 14.5). In addition to U, high concentrations of copper (Cu) were dissolved from samples NPP2-2 and 2-4 (Table 14.4), and the extractions removed more than 92% of total Cu in these samples. Catalano et al. (2006) suggested that around 50% of U in sample NPP2-4 was present as metatorbernite, and the high percentages of total U and Cu dissolved in this sample suggest that the formate extraction was efficient in dissolving metatorbernite. The pH and ionic strength conditions of the extraction should achieve essentially complete desorption of U(VI) from most mineral surfaces, either Table 14.5: Fractional Dissolution of Uranium in Extractions, Fraction of Total U Exchanged in Isotopic Exchange Experiments, and ModelEstimated Fraction of Total U Present as Adsorbed U(VI). Sample

NPP1-8 NPP1-12 NPP1-16 NPP1-20 NPP2-2 NPP2-4 NPP2-8 NPP2-12 SPP1-16 SPP1-18 SPP1-22 SPP2-8 SPP2-12 SPP2-16 SPP2-18 NPP1-fines NPP2-fines SPP1-fines SPP2-fines

Total U (ppm)a

Formate extraction % of Utotb

10.5 14.0 9.6 6.3 105.7 100.1 39.8 14.2 7.3 7.4 7.9 10.8 8.0 3.8 2.9 21.2 157.1 31.3 13.3

101.573.2 90.373.1 82.572.7 77.572.5 72.571.3 79.671.6 68.372.0 107.171.2 101.973.2 78.572.5 68.972.2 91.672.9 100.173.3 93.172.8 61.071.9 78.370.89 – 78.770.89 75.872.6

Bicarbonate Isotopic extraction exchange % % of Utotc of Utotd 45.070.2 42.771.0 38.373.3 29.770.45 29.872.8 29.471.3 37.671.1 56.171.8 54.770.99 36.070.97 35.470.08 43.872.5 57.671.2 41.071.9 18.971.3 41.470.77 41.471.5 58.070.13 39.970.68

46.073.3 44.773.6 44.671.9 35.870.70 46.472.2 88.278.6 56.372.5 61.272.0 55.172.5 35.870.74 37.171.4 44.774.9 54.772.7 42.274.0 21.170.3 47.772.1 – 56.177.0 52.671.0

Estimate of adsorbed U(VI) % of Utote 54 61 57 41 18 13 45 70 55 28 29 52 61 34 15 57 – 66 38


Sum of 238U and 235U (Uranium concentration in nmol g1 are listed in Table 14.4). Formate extraction (72 h). c Bicarbonate extraction, pH 9.45 (72 h). d Isotopic exchange, 336 h (after 1,260 h pre-equilibration in AGW4). e Estimate of adsorbed U(VI) in each sample from the semi-mechanistic surface complexation model calibrated with deep vadose zone samples. Adsorbed U(VI) in the sample calculated with the model by estimation of dissolved U(VI) concentration that would result from desorption after 96 h equilibration in AGW4 (see text). b


393

as surface complexes or from ion exchange sites in aluminosilicate minerals (Turner et al., 1996; Payne et al., 1998). In addition, poorly crystalline iron oxide/aluminosilicate phases and trace calcite may be dissolved under these conditions, releasing any co-precipitated U(VI). For sample SPP2-18, 39% of the total U was not extracted; this is equivalent to 4.6 109 mol g1 U (1.1 ppm), which is near the concentration of background U in Hanford sediments (Serne et al., 2003). Most of the background U is sequestered within the crystal matrices of silicate minerals, rather than present on mineral surfaces, and thus would not be expected to dissolve in chemical extractions other than hydrofluoric acid. 14.3.3. Dilute (Bi)Carbonate Extractions Kohler et al. (2004) found that a dilute (bi)carbonate extraction method could be used to estimate adsorbed U(VI) on aquifer sediments that were contaminated in the range of 1.3 108–3.4 108 mol g1 total U (3–8 ppm). The basis of the method is to desorb U(VI) by strong aqueous complexation with carbonate at pH 9.0–9.5, allowing for minimal dissolution of crystalline mineral matrices in comparison with harsher extraction methods. U(VI) speciation calculations with FITEQL (Herbelin and Westall, 1999) suggest that more than 98% of dissolved U(VI) is present as UO2(CO3)4 3 under the extraction conditions. Figure 14.4 shows the pH and concentrations of dissolved calcium and U(VI) in the dilute (bi)carbonate extraction of the deep vadose zone sample, NPP1-16, as a function of time. Calcium (Ca) concentrations dropped rapidly during the first 4 h of the extraction, and then stabilized, likely as a result of a small amount of calcite precipitation at the beginning of the extraction. Separate experiments have shown that calcite is not dissolved by the dilute (bi)carbonate extraction (D. Bond, unpublished results). U(VI) desorbs rapidly during the first 72 h of the extraction, but U(VI) concentrations continued to increase slowly for weeks thereafter (Fig. 14.4). The range of U(VI) released by the (bi)carbonate extraction was 19–58% of total U (Table 14.5).

14.3.4. Artificial Groundwater Extractions 14.3.4.1. Aqueous Compositions U(VI) dissolution and desorption from the sediment samples was studied in artificial groundwaters of varying composition (Table 14.1) in order to

D. L. Bond et al.

Dissolved Metal Concentration (M)

1.2E-6 U(VI) Ca*10-3 pH

1.0E-6

9.4

9.3

8.0E-7

9.1

6.0E-7

9.0

4.0E-7

8.9

2.0E-7

8.7

0.0E0

pH

394

8.6 1

10

100

Log Time (hr)

Figure 14.4: Uranium(VI) and Calcium Concentrations during a Dilute (Bi)Carbonate Extraction of Sediment Sample, NPP1-16 (50 g L1). Note Log Scale on X-Axis. Values are the Mean of Two Replicates. Error Bars based on Replicate Variation for all Figures. determine the effects of ionic strength, pH, alkalinity, Ca concentration, U(VI) concentration, and solid:water ratio on the rate and extent of U(VI) release to the aqueous phase. It was not possible to independently control these variables in the experiments, as the sediments exerted a strong influence on the aqueous composition, specifically the pH and Ca concentrations. Figure 14.5 shows the evolution in concentration of some major elements in the water during a typical extraction of a deep vadose zone sample (e.g., NPP1-16) with AGW4. Calcium, magnesium (Mg), and silicon are released from the sediments rapidly at first, and then continue to increase slowly in concentration for a few weeks. Sodium solution concentrations decrease with time due to ion exchange. One of the main differences in the initial compositions of the artificial groundwaters was dissolved bicarbonate (Table 14.1). The measured alkalinities of several artificial groundwaters during the reaction of sample NPP1-20 is shown as a function of time in Fig. 14.6. With the exception of AGWs 12 and 13, alkalinity generally remained constant during the experiments. For AGWs 12 and 13, alkalinity decreased rapidly during the first few hours of the experiments and then stabilized, suggesting that calcite precipitation may have occurred in the beginning of the experiments with


395

3.0E-4

8.5E-4

8.0E-4

7.5E-4

2.0E-4

7.0E-4

[Mg], [Si], [K] (M)

[Ca], [Na] (M)

2.5E-4 Ca Nax10-1 Mg Si K

1.5E-4 6.5E-4

6.0E-4

1.0E-4 1

10 Log Time (hr)

100

Figure 14.5: Aqueous Concentrations of Selected Elements during Reaction of Artificial Groundwater Solution 4 with Sediment Sample, NPP1-16 (100 g L1). Concentrations of Ca and Na( 0.1) are Plotted on the Left Y-Axis. All Other Elements are shown on the Right Y-Axis. Note Log Scale on X-Axis. AGWs 12 and 13, but not in the other AGW experiments. With the exception of AGWs 5 and 6, water was calculated to be oversaturated with respect to calcite for the duration of the experiments, and the degree of calcite saturation was well correlated with the alkalinity of the solution (Fig. 14.7). Partial pressures of carbon dioxide gas in the headspace of the experimental tubes after 96 h of reaction were calculated from the measured pH values and alkalinities (assuming carbonate species only contributed to alkalinity). The pH values and partial pressures of carbon dioxide gas ranged from 7.6 to 8.4 and 0.05–0.6% in the experiments, respectively, in AGW batch experiments with sediment samples collected near or beneath the water table. This range of values is representative of existing conditions in Hanford groundwaters beneath the 300-Area, including conditions expected from the mixing of regional groundwater with river water. Inhibition of calcite nucleation or crystal growth likely caused the observed oversaturation with respect to calcite in the batch experiments. Inhibition has been observed when solutions contain anions, such as phosphate or dissolved organic carbon (DOC) (Reddy and Wang, 1973; Inskeep and Bloom, 1986; Dove and Hochella, 1993; Lebron and Suarez, 1996). Calcite precipitation was greatly inhibited in the range of 0.02–0.15 mM

396

D. L. Bond et al.

12.0 GW GW GW GW GW GW

Alkalinity (mequiv/L)

10.0

8.0

6 5 3 4 12 13

6.0

4.0

2.0

0.0 0

100

200

300

400

Time (hr)

Figure 14.6: Measured Alkalinity as a Function of Reaction Time of Sample NPP1-20 (200 g L1) in Artificial Groundwater Solutions with Variable Bicarbonate Concentrations.

DOC (as C) with a saturation index (SI) equal to 0.95 (SI ¼ log [IAP/Ksp]) (Lebron and Suarez, 1996). In AGW4 experiments with samples NPP1-16 and SPP2-18, DOC values were measured at 0.13 and 0.025 mM (as C), respectively, with SIE0.4. Similar DOC values were measured in AGWs 5 and 13 for both sediment samples (SIE0.2 and 0.8). Calcite seeding after 72 h of reaction had minimal effects on the degree of oversaturation in the batch experiments, consistent with the hypothesis that inhibition of calcite nucleation or crystal growth caused the observed oversaturation. The speciation of dissolved U(VI) in the AGW is complex, and unfortunately, subject to some thermodynamic uncertainty. FITEQL was used to calculate the expected speciation of a 2 mM U(VI) solution in a closed system at pH 7.87 as a function of alkalinity (assuming only carbonate species contribute to alkalinity), with the aqueous phase assumed to be at equilibrium with calcite (Fig. 14.8). The predicted predominant species for these solutions is Ca2UO2(CO3)3(aq), except at the highest alkalinity values, where the UO2(CO3)4 species becomes important. The Ca2UO2 3 (CO3)3(aq) species is known to exist in Hanford groundwaters (Wang et al., 2004; Dong et al., 2005), but precise calculation of its concentration with


1.0

NPP 1-20 Other NPP SPP 1-18 Other SPP

0.8 0.6 Calcite Saturation Index

397

0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 0

2

4 Alkalinity (mequiv/L)

6

8

Figure 14.7: Calcite Saturation Index versus Alkalinity in Batch Reactions of Artificial Groundwater Solutions with Selected NPP and SPP Sediment Samples for 96 h. Saturation Index of Calcite was Calculated from FITEQLCalculated Aqueous Speciation using the Formula: SI ¼ (Log IAP, Ion Activity Product) – (8.48, the Log of the Solubility Product). thermodynamic calculations is subject to some uncertainty at present (Guillaumont et al., 2003). Despite the uncertainty, it is important to include the Ca2UO2(CO3)3(aq) species in calculations, because its formation has been shown to decrease U(VI) adsorption (Dong et al., 2005; Fox et al., 2006). The calculations shown in Fig. 14.8 and elsewhere in this paper assume a log K value of 30.0 for the Ca2UO2(CO3)3(aq) species, consistent with the calculations in Fox et al. (2006).

14.3.4.2. U(VI) Release to Solution Aqueous concentrations of U(VI) increased rapidly during the first 24 h of reaction between artificial groundwaters and Hanford sediments, as illustrated in Fig. 14.9 for the deep vadose zone samples. After 24 h, a steady slow release of U(VI) continued. The slow release of U(VI) did not appear complete after 7 weeks of reaction, at which point the experiments were ceased. Figs. 14.10 and 14.11 show that as the alkalinity of the artificial groundwaters increased, the release of U(VI) increased. The impact of ionic

398

D. L. Bond et al.

4

-log of concentration (moles/L)

pH = 7.87; calcite equilibrium assumed U(VI)tot = 2 x 10-6M

5

Ca2UO2(CO3)3o

UO2(CO3)34-

6 CaUO2(CO3)32-

7

UO2(CO3)228 UO2CO3o

9

(UO2)2CO3(OH)310

0

1

2

3 4 5 Alkalinity (mequiv/L)

6

7

8

Figure 14.8: Calculated Aqueous Speciation of a 2 mM U(VI) Solution at pH 7.87 and at Equilibrium with Calcite as a Function of Alkalinity. Total Alkalinity was Assumed to be Due to Bicarbonate and Carbonate Alkalinity Only. strength on U(VI) desorption was minimal, and desorption experiments with greater suspension density had lower concentrations of desorbed U(VI) per unit mass of sediment (Fig. 14.12). The dependence of desorption on suspension density is a consequence of the mass law for adsorption–desorption equilibrium; as the suspension density increases at constant aqueous conditions, the surface site concentration increases, which favors adsorption over desorption. As noted above, many of these experiments were carried out in solutions that were supersaturated with respect to calcite. Previous studies of uranyl co-precipitation within calcite report partitioning coefficients near 0.1 (Meece and Benninger, 1993; Reeder et al., 2001). Assuming that all of the initial bicarbonate ion lost from solution in AGWs 12 and 13 (Fig. 14.6) was precipitated as calcite, and that U(VI) was co-precipitated with calcite with a partitioning coefficient of 0.1, less than 1% of dissolved U(VI) would have been removed in the experiments by co-precipitation. Thus, the impact of any calcite precipitation on the evolution of dissolved U(VI) during the AGW experiments is expected to be within the experimental error of the measurements (0.2–8.6%, see Table 14.5).


399

3.0E-8 NPP 1-Fines SPP 2-Fines NPP 1-8

[U(VI)] released (mol/g)

2.5E-8

NPP 1-16 SPP 2-16 NPP 1-20

SPP 2-18

2.0E-8

1.5E-8

1.0E-8

5.0E-9

0.0E0 0

100

200

300 400 Time (hr)

1000

1200

1400

Figure 14.9: U(VI) Desorption from Various Samples (200 g L1) during Reaction with AGW4 (4 meq L1). 14.3.5. U(VI) Isotopic Exchange Experiments U(VI) isotopic exchange was used to assess the ‘‘labile’’ fraction of U(VI) distributed between the aqueous and sediment phases, after reaction with AGW4 for either 24 h or 1,260 h. In the case where sediment samples were reacted for 1,260 h, a relatively stable dissolved U(VI) concentration was obtained prior to addition of a 233U(VI) spike to the aqueous phase. In the case of 24 h for the initial reaction period, both total dissolved U(VI) and 233 U(VI) activity changed significantly as a function of time during isotopic exchange. Figure 14.13a shows the concentration of dissolved U(VI) and the activity of the 233U(VI) tracer in the isotopic exchange experiment conducted with the deep vadose zone sample, NPP1-16, after 24 h of pre-equilibration with AGW4. The 233U(VI) tracer activity in solution declined rapidly during the first 48 h, and then began to decline at a slower rate that appeared to continue after 1,350 h of exchange. Dissolved U(VI) increased rapidly during the first 200 h, and then reached a stable concentration near 0.8 mM after that. Fig. 14.13b shows the variables for the same sample, but in this case, the sample was first pre-equilibrated with AGW4 for 1,260 h. The decline in

D. L. Bond et al.

AGW AGW AGW AGW

1.6E-6

2 4 3 5

8.0E-9 44.0 % of Adsorbed U(VI) 6.0E-9

[U(VI)] (M)

1.2E-6

4.0E-9

8.0E-7 24.5 % of Adsorbed U(VI)

[U(VI)] (mol/g)

400

2.0E-9

4.0E-7 10.4 % of Adsorbed U(VI) 0.0E0

3.1 % of Adsorbed U(VI) 0

20

40

60

80

0.0E0

100

Time (h)

Figure 14.10: U(VI) Desorption from Sediment Sample, NPP1-16 (200 g L1), during Reaction with Artificial Groundwater Solutions of Varying Alkalinity. AGW2, Alkalinity ¼ 9 meq L1; AGW4 ¼ 4 meq L1; AGW3 ¼ 2 meq L1; and AGW5 ¼ 1 meq L1. Total Adsorbed U(VI) for the Sample was Estimated from the Amount of Labile U(VI) Determined in 336 h Isotopic Exchange Experiments. 233

U(VI) tracer activity exhibited similar kinetic behavior, but in this experiment, dissolved U(VI) had already reached the stable concentration near 0.8 mM before the isotopic exchange was initiated. Based on the fractional partitioning of 233U(VI) tracer between the solid and liquid phases, the dissolved and total U(VI) concentrations, and the solid:liquid ratio, one can calculate a ‘‘labile’’ fraction of U(VI) associated with the sediment (see Equation (14.1)) (Kohler et al., 2004). The term ‘‘labile’’ fraction can only be operationally defined because of the evolving 233 U(VI) tracer activity with time; calculated values as a function of time are shown in Fig. 14.13a and b. Labile U(VI) represents the mass of U(VI) that achieves isotopic equilibrium with the aqueous phase within the calculated time frame. 233U(VI) likely exchanges first with adsorbed U(VI) on mineral surfaces exposed to bulk solution, because of the fast chemical reaction and the lack of a diffusion barrier in the well-mixed bulk solution. Slower isotopic exchange then occurs with: (1) adsorbed U(VI) in micropores, due to


401

2.2E-8 NPP2-2 NPP1-16 SPP1-18 SPP2-16 NPP1-20 SPP2-18

[U(VI)] released (mol/g)

1.8E-8 1.4E-8 8.0E-9

6.0E-9

4.0E-9

2.0E-9

0.0E0 0

2

4

6

8

10


Figure 14.11: U(VI) Released from Various Samples (200 g L1) after 72 h of Reaction in Artificial Groundwater Solutions with Variable Alkalinity. 1.4E-8

[U(VI)] Desorbed (mol/g)

1.2E-8

NPP 1-16, GW 4, 4 meq/L, I = 0.01 M NPP 1-16, GW 9, 4 meq/L, I = 0.1 M

1.0E-8 8.0E-9 6.0E-9 4.0E-9 2.0E-9 0.0E0 0

50

100

150

200

250

300

350

400

450

Suspension Density (g/L)

Figure 14.12: U(VI) Desorbed from Sample, NPP1-16, after 72 h Reaction at Variable Solid:Water Ratios in AGW4 (I ¼ 0.01 M) and AGW9 (I ¼ 0.1 M).

402

D. L. Bond et al.

35

1.0E-7

25 6.0E-8

20 15

4.0E-8

Activity (dpm/mL)

and Dissolved [U(VI)] (M x 0.1)

Labile [U(VI)] in sediment (mol/g)

30 8.0E-8

10 2.0E-8 Calculated Labile U(VI) Isotope Activity Dissolved U(VI) (M) 0.0E0 0

500

1000 Time (hr)

(a)

1500

5 0 2000

35

1.0E-7

25 6.0E-8

20 15

4.0E-8

Activity (dpm/mL)



30 8.0E-8

10 2.0E-8 Calculated Labile U(VI) Isotope Activity Dissolved U(VI) (M) 0.0E0 0 (b)

500

1000

5 0 1500

Time (hr)

Figure 14.13: Dissolved U(VI) Concentration, 233U Activity in Solution, and Calculated Labile U(VI) from Isotopic Exchange Experiments Conducted in AGW4 with Sediment Sample, NPP1-16 (100 g L1), with (a) 24 h Pre-Equilibration Time, and (b) 1,260 h Pre-Equilibration Time.


403

mass transfer limitations, and (2) U(VI) present in precipitated or coprecipitated mineral phases, depending on kinetic constraints. Note that calculated labile U(VI) reached nearly the same value for sample NPP1-16 regardless of the time allowed for pre-equilibration in AGW4 (compare Fig. 14.13a and b). This near equality of the final labile U(VI) values can be expected in the case where the predominant form of labile U(VI) in the sediment is adsorbed U(VI). The kinetic behavior can be contrasted with the case in which U(VI) dissolution is also occurring on the same time scale as U(VI) desorption. For example, Fig. 14.14 shows similar data for sample NPP2-4, which contains a significant amount of precipitated U(VI). Because dissolution continues throughout the isotopic exchange experiment, the dissolved U(VI) concentration is much higher in the case of 1,260 h of pre-equilibration in AGW4 (at the time that 233U tracer was added) and continued to rise even after isotopic exchange was initiated. As a result, calculated labile U(VI) was much higher at any point in the isotopic exchange experiments with sample NPP2-4 if 1,260 h of pre-equilibration was allowed rather than 24 h (compare Fig. 14.14a and b). For sample NPP2-4, as precipitated U(VI), e.g., metatorbernite, dissolves during the preequilibration period, the mass of exchangeable U(VI) in the sample increases. Using an estimated solubility product of log K ¼ 45 for the mineral, torbernite, aqueous speciation calculations suggest that the solution was 10 orders of magnitude undersaturated with respect to torbernite during these experiments, even after long periods of equilibration. As will be discussed further below, we have assumed that the labile mass of U(VI) in the deeper pit samples (near the water table, e.g., NPP1-16, NPP1-20, SPP2-16, SPP2-18) is composed solely of adsorbed U(VI), with the definition of labile U(VI) based on 336 h of isotopic exchange (with 1,260 h of pre-equilibration in AGW4). Given that assumption, then the fractional amounts of U(VI) desorbed by various groundwater solutions can be calculated. Figure 14.10 illustrates that 44% of adsorbed U(VI) was desorbed in 96 h from sample NPP1-16 (200 g L1) by AGW2 (alkalinity ¼ 8 meq L1), whereas only 3% was mobilized by AGW5 (alkalinity ¼ 1 meq L1). As mentioned above, the fraction desorbed was also dependent on the sediment:water ratio; AGW2 desorbed 82% or 52% of adsorbed U(VI) from suspensions of sample SPP2-18 (25 or 200 g L1, respectively). Qafoku et al. (2005) found that 69% of labile U(VI) was released to an artificial groundwater solution (similar to AGW2) during slow passage of 100 pore volumes through a column packed with sample SPP2-18. The authors found a similar value for labile U(VI) for sample SPP2-18 (2.68 nmol g1) as that reported here (2.5370.3 nmol g1).

2.5E-7

10

2.0E-7

8

1.5E-7

6

1.0E-7

4

5.0E-8

2 Calculated Labile U(VI) Isotope Activity Dissolved U(VI) (M)

0.0E0 500

1000

1500

Time (hr)

6.0E-7

10

5.0E-7 and Dissolved [U(VI)] (M x 0.1)

0 2000

8

4.0E-7 6 3.0E-7 4 2.0E-7

1.0E-7

Calculated Labile U(VI) Isotope Activity Dissolved U(VI) (M)

0.0E0 0 (b)

500

1000

Activity (dpm/mL)

0 (a)


Activity (dpm/mL)


D. L. Bond et al.


404

2

0 1500

Time (hr)

Figure 14.14: Dissolved U(VI) Concentration, 233U Activity in Solution, and Calculated Labile U(VI) from Isotopic Exchange Experiments Conducted in AGW4 with Sediment Sample, NPP2-4, with (a) 24 h Pre-Equilibration Time (100 g L1), and (b) 1,260 h Pre-Equilibration Time (30 g L1).


405

14.3.6. U(VI) Sorption Isotherm U(VI) quickly sorbed to the sediment surfaces in these experiments (artificial groundwater spiked with 2 106–5 106 M U(VI)), reaching a steadystate dissolved U(VI) concentration within 24 h. Suspension density and concentration of U(VI) added were varied and combined with AGW4 desorption data (batch experiments without added U(VI)) to produce U(VI) isotherms for the deeper pit samples (near the water table, e.g., NPP1-16, NPP1-20, SPP2-16, SPP2-18) (Fig. 14.15). Ionic strength had a negligible effect on the amount of U(VI) adsorbed (AGW4 versus AGW9). For the deep pit samples it was observed that NPP samples adsorbed U(VI) more strongly than the SPP samples (Fig. 14.15). The bulk mineralogy of these samples, however, is nearly identical, as are the surface areas (average NPP ¼ 20.4 m2 g1 and SPP ¼ 19.1 m2 g1). The samples vary in particle size distribution, with a higher percentage of clay/silt in the NPP samples, and

-8.6

Log adsorbed U(VI), moles/m2

-8.8 -9.0 -9.2 -9.4

Sample NPP1-16 Sample NPP1-20 Sample SPP2-18 Sample SPP1-18 Sample SPP2-16 Model all data Model NPP data Model SPP data

-9.6 -9.8 -10.0 -10.2 -10.4 -7.2

-6.8

-6.4

-6.0

-5.6

-5.2

Log dissolved U(VI), moles/L

Figure 14.15: Uranium(VI) Adsorption Isotherms for Deeper Pit Sediment Samples Suspended in AGW4 or AGW9. Alkalinities in the NPP Experiments Ranged from 164 to 227 mg L1 as CaCO3 (Average of 188 mg L1), Causing Some of the Scatter in the Data. Alkalinities in the SPP Experiments Ranged from 168 to 226 mg L1 (Average of 190 mg L1). Solid Curves show the Fits to the Data with Surface Complexation Models Calibrated with All of the Data, or Separately with NPP or SPP Sediment Data.

406

D. L. Bond et al.

the greater abundance of fine-grain material may explain the observed difference in U(VI) adsorption.

14.3.7. Surface Complexation Model Based on the assumption that the labile mass of U(VI) in the deeper pit samples (NPP1-16, NPP1-20, SPP2-16, SPP2-18) is composed solely of adsorbed U(VI), a non-electrostatic surface complexation model for U(VI) adsorption on the sediments was developed using the generalized composite modeling approach (Davis et al., 2004a). The model provides a quantitative description of U(VI) sorption equilibria as a function of aqueous chemical conditions. All of the experimental data for the deeper pit samples were included in the model calibration. A total site density of 3.84 mmol m2 was used in the model (Davis and Kent, 1990). Based on previous work (Davis et al., 2004a), nine monomeric U(VI) surface reactions (Table 14.6) were considered to describe U(VI) sorption by the sediments. FITEQL calculations were first completed to determine which single surface reaction (Table 14.6) would provide the best fit to the experimental data. FITEQL output includes a goodness-of-fit parameter, WSOS/DF, the weighted sum of squares of the difference in value between model simulations and experimental data points, divided by the degrees of freedom (Herbelin and Westall, 1999). Lower values of WSOS/DF mean the proposed model is a better fit to the data; WSOS/DF is referred to as a ‘‘fit value’’ below. Representing the U(VI) adsorption data with a single reaction produced a reasonable fit (fit values ¼ 8.2–10.2), with the best fit values provided by Table 14.6: U(VI) Surface Reactions Considered for the Generalized Composite SCM. Number 1 2 3 4 5 6 7 8 9

Reaction + ¼ SOUO+ SOH+UO2+ 2 2 +H 2+ SOH+UO2 +H2O ¼ SOUO2OH+2H+ + SOH+UO2+ 2 +H2CO3 ¼ SOUO2HCO3+2H 2+ + SOH+UO2 +H2CO3 ¼ SOUO2CO3 +3H + 2 SOH+UO2+ 2 +H2CO3+H2O ¼ SOUO2OHCO3 +4H + 2+ SOH+UO2 +2H2CO3 ¼ SOUO2(HCO3)2 +3H 2 + SOH+UO2+ 2 +2H2CO3 ¼ SOUO2(CO3HCO3) +4H + 2+ 3 SOH+UO2 +2H2CO3 ¼ SOUO2(CO3)2 +5H + 4 SOH+UO2+ 2 +2H2CO3+H2O ¼ SOUO2OH(CO3)2 +6H


407

reactions 1, 2, 3, or 4 (Table 14.6). The second step in model development was to consider combinations of two reactions to represent the data. In many cases, the fit to the data was improved by adding a second reaction, with the best combinations being reactions 2 and 3 (fit value ¼ 7.3) or 2 and 6 (fit value ¼ 7.7). Adding a second surface site (strong versus weak) only improved the fit marginally, perhaps because the log–log isotherm data (Fig. 14.15) exhibit a slope near one, indicating a nearly linear isotherm. Very little improvement to the fit could be achieved by adding a third surface reaction; the best combination of three reactions was 2, 3, and 6, which only improved the fit from 7.29 to 7.27. This was not considered sufficient improvement to add reaction 6 to the model, and thus the recommended model to describe U(VI) adsorption in the 300-Area subsurface is a one-site model with reactions 2 and 3. Model parameters are given in Table 14.7. The surface complexation model describes the U(VI) sorption data reasonably well (Figs. 14.15 and 14.16) over the range of conditions considered in the experiments. However, the goodness-of-fit is clearly impacted by a difference in U(VI) adsorption between the NPP and SPP samples; a model calibrated with all data splits the two sets of experimental data. Models calibrated with data only from the NPP or SPP sediments are also shown in Figs. 14.15 and 14.16 in order to estimate the effect of the sediment heterogeneity on model parameter values (Table 14.7). Once the models were calibrated, the separate NPP and SPP models were used to estimate the amount of adsorbed U(VI) that was present on each Table 14.7: Surface Complexation Model Parameters. U(VI) Surface reactiona

Log Kf (I ¼ 0)

Average model based on all data + SOH+UO2+ 2 +H2O ¼ SOUO2OH+2H 2+ SOH+UO2 +H2CO3 ¼ SOHUO2CO3+2H+

5.152 0.833

NPP Sediment model + SOH+UO2+ 2 +H2O ¼ SOUO2OH+2H 2+ SOH+UO2 +H2CO3 ¼ SOHUO2CO3+2H+

4.722 0.895

SPP Sediment model + SOH+UO2+ 2 +H2O ¼ SOUO2OH+2H 2+ SOH+UO2 +H2CO3 ¼ SOHUO2CO3+2H+ a

Total site concentration equal to 3.84 mmol m2 of sediment surface area.

5.235 1.033

408

D. L. Bond et al.

2.75

Sample NPP1-16 Sample NPP1-20 Sample SPP2-18 Sample SPP1-18 Sample SPP2-16 Model based on all data Model based on all NPP data Model based on all SPP data

2.50

Log U(VI) Sorption Kd (L/kg)

2.25 2.00 1.75 1.50 1.25 1.00 0.75 0.50 0.25 0.00

0

1

2

3

4

5

6

7

8

9

10


Figure 14.16: Alkalinity Dependence of Log Kd Values for U(VI) Sorption for Deeper Pit Sediment Samples Equilibrated in Artificial Groundwater Solutions of Varying Composition. Solid Curves show the Fits to the Data with Surface Complexation Models Calibrated with all of the Data, or Separately with NPP or SPP Sediment Data. sample, based on the amount of U(VI) released in the first few hours of equilibration with AGW4. The individual models (NPP or SPP), not the composite, were used for estimating adsorbed U(VI) on samples from the two ponds. The inAn assumption of the estimation method is that only U(VI) desorption was significant during the first few hours of equilibration. Obviously, samples that contain both adsorbed U(VI) and precipitated U(VI) can release U(VI) either by dissolution or desorption, so the estimation method is, at best, an overestimate of desorbed U(VI). Using kinetic U(VI) desorption data for samples NPP1-16 (Fig. 14.10) and SPP2-18, it was estimated from the shapes of the curves that steady-state dissolved U(VI) concentrations in AGW4 were 1.5 times greater at equilibrium (after 96 h reaction) than after 4 h of U(VI) desorption from these samples. Using this ratio, dissolved U(VI) concentrations due to desorption alone in AGW4 were estimated for each sample based on the amount of U(VI) released after 4 h. The surface complexation model was used to calculate the amount of adsorbed U(VI) that would be in equilibrium with the dissolved U(VI) concentrations (specific for each sample), given the aqueous chemical conditions in the experiment (pH, dissolved carbonate, Ca concentration, etc.).


409

This allowed a calculation of adsorbed U(VI) for each sample prior to the reaction with AGW4; these quantities are given as a percentage of the total U for each sample in Table 14.5.

14.4. Discussion Catalano et al. (2006) used spectroscopic techniques to show that contaminant uranium in the depth sequence of vadose zone sediments is characterized by at least three types of uranium speciation: (1) U(VI) co-precipitated with carbonate minerals in the uppermost samples, close to the previous pond bottom, (2) U(VI) precipitated as metatorbernite and other phases (Arai et al., 2007) in samples 1 m below the top of the sequence, and (3) sorbed U(VI) in the intermediate and deeper depths of the sediment profile, near the water table. Wang et al. (2005) argued that a portion of the U(VI) in the pond bottom sediments was associated with carbonate mineral surfaces, both as a sorbed surface species and co-precipitated within the mineral structure. The combination of extraction and isotopic exchange results across the vertical profile of vadose sediments presented here allows a qualitative interpretation of the chemical forms of contaminant U(VI) in the vertical profile of sediments, and the development of a conceptual model that evaluates the potential release of U(VI) to the aquifer based on chemical speciation. The kinetics of U(VI) release from these samples in AGWs is complex, exhibiting both fast and slow release regardless of the depth of the sample in the profile (Figs. 14.9, 14.10, 14.13, and 14.14). The U(VI) release to AGW solutions cannot be separated into contributions from desorption and dissolution processes by simplistic analysis of the kinetic data; desorption of U(VI) can be a slow process and dissolution can be fast enough to contribute to early U(VI) release. Instead, the U(VI) chemical speciation as a function of depth in the subsurface needs to be inferred from the combination of published spectroscopic results and extraction and isotopic exchange results presented in this paper. Cu dissolution in the formate extractions (pH 3.5) illustrates clearly which samples contain U(VI) as precipitated metatorbernite. The greatest total Cu concentrations occurred in the NPP2-2, 2-4, and 2-8 samples (6.1 105, 6.9 105, and 1.7 105 mol g1, respectively), and the greatest percentages of total Cu dissolved by the formate extraction also occurred in these samples (94, 92, and 85%, respectively, Table 14.4). NPP2-4 is the sample in which Catalano et al. (2006) previously identified the occurrence of metatorbernite by spectroscopic methods. While some of the Cu dissolved in the

410

D. L. Bond et al.

formate extractions may have been adsorbed, the high total concentrations of Cu in these samples suggest that most of the Cu was present in precipitates. In contrast, deeper pit samples (e.g., NPP1-16, 1-20, and SPP216, 2-18) contained total Cu concentrations (8.5 107–1.2 106 mol g1) near that of background sediments, and the percentage of Cu dissolved by formate extraction was low (3–14%). The results suggest that the formate extraction was effective in dissolving contaminant Cu minerals from samples at the upper depths of the profile (including metatorbernite), and that metatorbernite was likely not present in the deep samples of the NPP1, SPP1, and SPP2 profiles, in agreement with the spectroscopic conclusions. Catalano et al. (2006) suggested that the predominant form of uranium speciation in the deeper samples in the profile was U(VI) sorbed onto phyllosilicate minerals. Desorption of U(VI) from most mineral phases is expected to be essentially complete under the conditions of the dilute (bi)carbonate extractions (Payne et al., 1998; Kohler et al., 2004). The dilute (bi)carbonate extractions exhibited a quick release of U(VI) in the first 72 h, followed by a slower release of U(VI) for several hundreds of hours (Fig. 14.4). For the deeper samples in the profiles (near or at the water table), the percentage of total U extracted in the first 72 h ranged from 19 to 56% (Table 14.5). For these samples, it can be argued that the (bi)carbonate extraction provides a good estimate of adsorbed U(VI), because there is excellent agreement with the amount of U(VI) that undergoes isotopic exchange under artificial groundwater conditions (Table 14.5). Samples that release U(VI) from both precipitated and adsorbed solidphase speciation exhibit different behaviors. For example, for the most contaminated samples (e.g., NPP2-2, 2-4, and 2-8), there is a lack of agreement between the (bi)carbonate extraction and isotopic exchange results (Table 14.5), suggesting that both desorption and dissolution are contributing to U(VI) release in the (bi)carbonate extraction, and that this extraction is not useful for estimates of adsorbed U(VI) in highly contaminated samples. The estimates of adsorbed U(VI) in each sample from the surface complexation model are in qualitative agreement with the extraction and spectroscopic results. For the samples where there was good agreement between the fraction of total U extracted by (bi)carbonate solution and undergoing isotopic exchange (samples NPP1-8, NPP1-12, NPP1-16, NPP1-20, NPP2-12, SPP1-16, SPP1-18, SPP1-22, SPP2-8, SPP2-12, SPP2-16, SPP218, NPP1-fines, and SPP1-fines), the estimates of adsorbed U(VI) by the surface complexation model were close to the values determined by (bi)carbonate extraction or isotopic exchange (Table 14.5). Those samples that contain metatorbernite and other U(VI) precipitates (i.e., NPP2-2 and NPP2-4) yielded estimates of adsorbed U(VI) from the surface complexation


411

model that were much lower than that measured in (bi)carbonate extractions and isotopic exchange measurements. The incomplete U extraction (19–55%) by the (bi)carbonate solution of the deep samples (and incomplete isotopic exchange) suggests that, in addition to adsorbed U(VI), there must be another type of uranium species in these samples. Table 14.3 compares the U extraction efficiencies of DCB, HH, AMOX, and formate extractions of deep vadose and groundwater fines samples, and shows that the DCB extraction was most effective at releasing U from these samples. DCB is effective at dissolving iron from crystalline matrices, and together with the high U extraction yields of the HH and AMOX extractions; this suggests that the other type of important U chemical species for these samples is U(VI) incorporated as a co-precipitate within mineral coatings (Payne et al., 1994). The co-precipitation within coatings likely occurred after U(VI) was transported to the lower depths of the sediment profiles and concentrated by adsorption at mineral surfaces. Although mineral coatings are often described in the literature as composed of poorly crystalline phases, other studies have shown that mineral coatings containing Fe, Al, and/or Si can form nanocrystalline phases that are resistant to dissolution in the milder HH and AMOX extractions (Banfield and Hamers, 1998; Davis et al., 1998). As in the analysis of the samples containing precipitated U(VI) (Fig. 14.14), significant release of U(VI) from the putative co-precipitated phases would have affected the kinetics of isotopic exchange in experiments with and without long pre-equilibration times. Therefore, it appears that the U(VI) that is incorporated in the mineral coatings in the deeper vadose zone samples is relatively resistant to release to the aquifer under oxidizing conditions. The lowest yield for the DCB extraction was 67% of total U for the SPP2-18 sample, which contained only 1.2 108 mol g1 U (2.9 ppm), suggesting that 4.2 109 mol g1 U (1 ppm) may be present as background U in crystalline matrices of uncontaminated Hanford silicate minerals. While Wang et al. (2005) found U(VI) incorporation in calcite and aragonite structures in the pond bottom precipitates, U(VI)-substituted carbonate phases are likely much less abundant in the vadose samples studied here. Total carbonate in the vadose samples was o0.1% by weight, whereas pond bottom precipitates contained 1–3% carbonate. The high levels of uranium in the pond precipitates (4.2 106–8.4 106 mol g1), as well as high calcium concentrations and alkalinity, increase the likelihood that U(VI)substituted carbonate phases formed within the pond or at the pond surface. The significant decrease in carbonate mineral content with depth suggests that carbonate minerals did not precipitate to a great extent within the vadose zone profile and have not been transported significantly as colloidal

412

D. L. Bond et al.

particles. In addition, Dong et al. (2005) have shown that calcite, when present, tends to actually lower U(VI) sorption in Hanford soils by blocking access to surfaces of higher sorption affinity.

14.5. Concluding Remarks The combined extraction, isotopic exchange, surface complexation modeling, and previous spectroscopic results allow a more detailed conceptual model to be developed for U(VI) release to the aquifer underlying the 300-Area at Hanford and demonstrate a generic approach for evaluating chemical speciation at other contaminated sites. Few studies have been published that attempt to estimate the separate contributions of desorption and dissolution to contaminant release. Without an approach for quantifying these separate contributions, it is very difficult to develop a useful conceptual model that can be applied in reactive transport modeling for the site. Using the results presented here, it can be concluded that the more contaminated sediments located near the original pond bottoms release U(VI) primarily by dissolution of U(VI)-bearing minerals to infrequent infiltrating precipitation (Catalano et al., 2006). The released U(VI) is transported downward in the sediment profile, but the transport is retarded by adsorption to mineral coatings on the sediments. However, not all of this U(VI) appears to make it to the water table. With wet/dry cycling, a fraction of the U(VI) deeper in the sediment profiles has been sequestered as a co-precipitate into coatings that have become resistant to dissolution as they have aged under oxidizing conditions. A fraction of sediment U(VI) near the water table is available for rapid U(VI) desorption in the capillary fringe zone, as the water table rises and falls with the variable stage of the Columbia River. The concentrations of total U in these sediments are low, but U(VI) desorption can still result in groundwater concentrations exceeding regulatory limits (for drinking water, 0.1 mM). The supply of adsorbed U(VI) in the sediments near the water table is presumably replenished over time by downward U(VI) migration from the more contaminated sediments above, although the magnitude of this flux is yet to be determined. If the groundwater table rose significantly and conditions became mildly reducing, a larger reservoir of sediment U could potentially be released by dissolution of iron oxide mineral coatings. The geochemical conceptual model described above is an obvious improvement over the constant Kd model that has previously been applied to describe U(VI) partitioning and release from the vadose zone sediments beneath the former waste disposal ponds in the 300-Area at Hanford (Zachara


413

et al., 2005). Although it is apparent that data obtained from chemical extractions and isotopic exchange experiments can be more meaningful when combined with spectroscopic studies of chemical speciation, few studies have applied a combination of these techniques to estimate the abundance of differing contaminant species in sediments. A better conceptual model for the site leads to a better understanding of the fundamental processes that drive U(VI) release to the aquifer, and the identification of critical fluxes in the system that need to be measured in situ (Davis et al., 2004b). If used carefully in combination with other methods, extractions and other types of carefully designed batch experiments can be useful investigative tools for field sites with mixtures of adsorbed, precipitated, and co-precipitated metal contaminants. The results have obvious relevance to the development of a better conceptual model for the 300-Area at Hanford and as input to reactive transport modeling simulations of processes occurring in the vadose zone at that site. In addition, if combined with spectroscopic studies, the experimental approach has relevance to other metal-contaminated sites where multiple processes may contribute to the overall release of contaminants to water and where the release is dependent on chemical conditions and aqueous metal speciation. In such cases, the release is poorly described by the constant Kd model or by linear sorption isotherms, which do not account for the effect of variable aqueous speciation on metal desorption or variable chemical conditions on metal dissolution.

ACKNOWLEDGMENTS Funding for this work was provided by an interagency agreement between the U.S. Geological Survey and Batelle Pacific Northwest National Laboratory #415428-A9E-P6490. The authors would like to thank Patricia Fox, Christopher Fuller, Matthias Kohler, and Steven Smith for technical assistance and insightful discussions.

REFERENCES Arai, Y., Marcus, M., Tamura, N., Davis, J. A., & Zachara, J. M. (2007). Spectroscopic evidence for uranium bearing precipitates in vadose zone sediments at the Hanford 300-Area Site. Environ. Sci. Technol., 41, 4633–4639. Banfield, J. F., & Hamers, R. J. (1998). Processes at minerals and surfaces with relevance to microorganisms and prebiotic synthesis. Geomicrobiology: Interactions between microbes and minerals. Rev. Mineral., 35, 81–122.

414

D. L. Bond et al.

Bernhard, G., Giepel, G., Reich, T., Brendler, V., Amayri, S., & Nitshce, H. (2001). Uranyl (VI) carbonate complex formation: Validation of the Ca2UO2(CO3)3 (aq) species. Radiochim. Acta, 89, 511–518. Catalano, J. G., McKinley, J. P., Zachara, J. M., Smith, S. C., & Brown, G. E. (2006). Changes in uranium speciation through a depth sequence of contaminated hanford sediments. Environ. Sci. Technol., 40, 2517–2524. Chao, T. T., & Zhou, L. (1983). Extraction techniques for selective dissolution of amorphous iron oxides from soils and sediments. Soil Sci. Soc. Am. J., 47, 225–232. Crowley, K. D., & Ahearne, J. F. (2002). Managing the environmental legacy of U.S. nuclear-weapons production. Am. Sci., 90, 514–523. Davis, J. A., Coston, J. A., Kent, D. B., & Fuller, C. C. (1998). Application of the surface complexation concept to complex mineral assemblages. Environ. Sci. Technol., 32, 2828. Davis, J. A., & Curtis, G. P. (2003). Application of Surface Complexation Modeling to Describe Uranium (VI) Adsorption and Retardation at the Uranium Mill Tailings Site at Naturita, Colorado. NUREG/CR-6708. U.S. Nuclear Regulatory Commission, Rockville, MD. Davis J. A., & Kent D. B. (1990). Surface complexation modeling in aqueous geochemistry: Mineral–water interface geochemistry. Rev. Mineral., 23, 177–260. Davis, J. A., Meece, D. E., Kohler, M., & Curtis, G. P. (2004a). Approaches to surface complexation modeling of uranium(VI) adsorption on aquifer sediments. Geochim. Cosmochim. Acta, 68, 3621–3641. Davis, J. A., Payne, T. E., & Waite, T. D. (2002). Simulating the pH and pCO2 dependence of uranium(VI) adsorption by a weathered schist with surface complexation models. In: P.-C. Zhang, & P. V. Brady (Eds). Geochemistry of Soil Radionuclides (Special Publication Number 59), Soil Science Society of America, Madison, WI, pp. 61–86. Davis, J. A., Yabusaki, S. B., Steefel, C. I., Zachara, J. M., Curtis, G. P., Redden, G. D., Criscenti, L. J., & Honeyman, B. D. (2004b). EOS, Assessing Conceptual Models for Subsurface Reactive Transport of Inorganic Contaminants, Vol. 85, pp. 449–455. Dong, W., Ball, W. P., Liu, C., Wang, Z., Stone, A. T., Bai, J., & Zachara, J. M. (2005). Influence of calcite and dissolved calcium on uranium(VI) sorption to a hanford subsurface sediment. Environ. Sci. Technol., 39, 7949–7955. Dove, P. M., & Hochella, M. F. J. (1993). Calcite precipitation mechanisms and inhibition by orthophosphate: in situ observations by scanning force microscopy. Geochim. Cosmochim. Acta, 57, 705–714. Fox, P. M., Davis, J. A., & Zachara, J. M. (2006). The effect of calcium on aqueous uranium(VI) speciation and adsorption to ferrihydrite and quartz. Geochim. Cosmochim. Acta, 70, 1379–1387. Guillaumont, R., Fanghanel, T., Neck, V., Fuger, J., Palmer, D. A., Grenthe, I., & Rand, M. H. (2003). Update on the Chemical Thermodynamics of Uranium, Neptunium, Plutonium, Americium, and Technetium. Elsevier, Amsterdam.


415

Herbelin, A. L., & Westall, J. C. (1999). FITEQL: A Computer Program for the Determination of Chemical Equilibrium Constants from Experimental Data. Chemistry Department, Oregon State University, Corvallis, OR. Inskeep, W. P., & Bloom, P. R. (1986). Kinetics of calcite precipitation in the presence of water soluble organic ligands. Soil Sci. Soc. Am. J., 50, 1167–1172. 2 complex Kalmykov, N., & Choppin, G. R. (2000). Mixed Ca2+/UO2+ 2 /CO3 formation at different ionic strengths. Radiochim. Acta, 88, 603–606. Kohler, M., Curtis, G. P., Meece, D. E., & Davis, J. A. (2004). Methods for estimating adsorbed uranium(VI) and distribution coefficients of contaminated sediments. Environ. Sci. Technol., 38, 240–247. Lebron, I., & Suarez, D. L. (1996). Calcite nucleation and precipitation kinetics as affected by dissolved organic matter at 251C and pH>7.5. Geochim. Cosmochim. Acta, 60, 2765–2776. Lindberg, J. W., & Peterson, R. E. (2004). 300-FF-5 Operable unit. In: M. J. Hartman, L. F. Morasch, & W. D. Webber (Eds). Hanford Site Groundwater Monitoring for Fiscal Year 2004. Pacific Northwest National Laboratory, Richland, Washington, PNNL-15070. pp. 2.12-1–2.12-31. Meece, D. E., & Benninger, L. K. (1993). The coprecipitation of Pu and other radionuclides with CaCO3. Geochim. Cosmochim. Acta, 57, 1447–1458. Nielson, K. K. (1978). Application of direct peak analysis to energy-dispersive X-ray fluorescence spectra. X-Ray Spectrom., 7, 15–22. Payne, T. E., Davis, J. A., & Waite, T. D. (1994). Uranium retention by weathered schists: The role of iron minerals. Radiochim. Acta, 66/67, 297–303. Payne, T. E., Lumpkin, G. R., & Waite, T. D. (1998). Uranium(VI) adsorption on model minerals: Controlling factors and surface complexation modeling. In: E. A. Jenne (Ed). Adsorption of Metals by Geomedia. Academic Press, San Diego, CA, pp. 75–97. Qafoku, N. P., Zachara, J. M., Liu, C., Gassman, P. L., Qafoku, O. S., & Smith, S. C. (2005). Kinetic desorption and sorption of U(VI) during reactive transport in a contaminated hanford sediment. Environ. Sci. Technol., 39, 3157–3165. Reddy, M. M., & Wang, K. K. (1973). Calcite crystal growth inhibition by phosphates. Desalination, 12, 61–73. Reeder, R. J., Nugent, M., Tait, C. D., Morris, D. E., Heald, S. M., Beck, K. M., Hess, W. P., & Lanzirotti, A. (2001). Coprecipitation of uranium(VI) with calcite: XAFS, Micro-XAS, and luminescence characterization. Geochim. Cosmochim. Acta, 65, 3491–3503. Riley, R. G., Zachara, J. M., & Wobber, F. J. (1992). Chemical Contaminants on DOE Lands and Selection of Contaminant Mixtures for Subsurface Science Research. United States Department of Energy, Subsurface Science Program, Washington, DC. DOE/ER–0547 T. Serne, R. J., Brown, C. B., Schaef, H. T., Pierce, E. M., Lindberg, M. J., Wang, Z., Gassman, P., & Catalano, J. (2003). 300 Area Uranium Leach and Adsorption Project. U.S. Department of Energy, Washington, D.C., PNNL-14022.

416

D. L. Bond et al.

Turner, G. D., Zachara, J. M., McKinley, J. P., & Smith, S. C. (1996). Surfaceadsorption of a subsurface smectite. Geochim. charge properties and UO2+ 2 Cosmochim. Acta, 60, 3399–3414. USDOE. (2005). Work Plan for Phase III Feasibility Study 300-FF-5 Operable Unit. United States Department of Energy, Richland, Washington. DOE/RL-2005-41. USEPA. (1996). Record of Decision for the Hanford 300-Area Site. Environmental Protection Agency, Washington, D.C., EPA/ROD/R10-96/143. Wang, Z., Zachara, J. M., McKinley, J. P., Smith, S. C., & Heald, S. M. (2005). Cryogenic laser induced U(VI) fluorescence studies of a U(VI) substituted natural calcite: Implications to U(VI) speciation in contaminated hanford sediments. Environ. Sci. Technol., 39, 2651–2659. Wang, Z., Zachara, J. M., Yantasee, W., Gassman, P. L., Liu, C., & Joly, A. G. (2004). Cryogenic laser induced U(VI) fluorescence characterization of U(VI) in hanford vadose zone pore waters. Environ. Sci. Technol., 38, 5591–5597. Zachara, J. M., Davis, J. A., Liu, C., McKinley, J. P., Qafoku, N., Wellman, D. M., & Yabusaki, S. (2005). Uranium Geochemistry in Vadose Zone and Aquifer Sediments from the 300 Area Uranium Plume. U.S. Department of Energy, Washington, D.C., PNNL-15121.


Chapter 15

Arsenic Speciation in Solid Phases of Geothermal Fields Marco A. Alsina1, Ian Saratovsky2,4, Jean-Franc- ois Gaillard3 and Pablo A. Pasteń1, 1

Departamento de Ingenierıá Hidraúlica y Ambiental, Pontificia Universidad Cato´lica de Chile, Santiago 6904411, Chile 2 Department of Chemistry, Northwestern University, Evanston, IL 60208, USA 3 Department of Civil and Environmental Engineering, Northwestern University, Evanston, IL 60208-3109, USA 4 Present address: Inorganic Chemistry Laboratory, Oxford OX1 3QR, UK

ABSTRACT The fate of As in geothermal systems is controlled by the interaction of several geochemical and biological processes, such as dissolution, precipitation, dilution, volatilization, sorption, biomineralization, and biocatalysis. Although these interactions remain poorly understood, the distinctive micro- and molecular-scale fingerprints of processes occurring at the solid–water interface are shedding new light on these complex systems. The microscopic and molecular speciation of As in geothermal deposits offers valuable information about its chemical evolution, the processes in which it is involved, and the reactivity of the solid materials formed. The amorphous and hydrous nature of most hot spring minerals renders the application of conventional identification methods based on X-ray or electron diffraction challenging. This chapter presents a brief review of As in geothermal deposits and some of our own recent results on hot spring deposits in the El Tatio geothermal field. We used X-ray absorption spectroscopy to investigate the speciation of As in sinter material and in biological mats. Our results suggest that Fe oxyhydroxides play an important role in controlling the concentration of As in both solid media. Although it is recognized that microbial populations in geothermal systems may catalyze the rapid oxidation of As(III) to As(V), it is still unclear whether microbial catalysis and/or biomineralization control the formation of Fe oxyhydroxides. Understanding the association between As and Fe Corresponding author. Tel.: +56-2-3545872; Fax: +56-2-3545876;

E-mail: [email protected] (P.A. Pasteń).

418

M. A. Alsina et al.

oxyhydroxides in geothermal systems is a key factor for evaluating the environmental impacts of geothermal energy exploitation as well as designing effective remediation approaches.

15.1. Introduction Geothermal systems commonly constitute an important source of As to watersheds, notably within water scarce regions. This combination can lead to negative impacts on human activities and development. For instance, average total dissolved arsenic (SAs) concentrations of 0.27 mM in the Los Angeles Aqueduct (California, USA) have been attributed to geothermal activity in the Long Valley area of Mono County (Wilkie and Hering, 1998). Hot spring discharges from the El Tatio geothermal field in northern Chile have SAs concentrations up to and exceeding 660 mM, contributing to As enrichment in the Loa River (Cusicanqui et al., 1976) – the main source of drinking and irrigation water for a population of 400,000 living in the Loa River basin. Therefore, understanding the fate of As in geothermal systems is relevant to evaluating the environmental impacts of human activities in geothermal fields (e.g., geothermal energy developments) and to formulate appropriate remediation technologies. Geothermal fields are recognized as modern analogous of fossil epithermal precious ore deposits (White, 1981) and, therefore, provide direct insight into the leading mechanisms of deposition of economic minerals. For example, evidence suggests that when significant sulfur concentrations occur, metals like copper and gold preferentially partition into the vapor phase, creating a pathway for ore epithermal deposition during condensation (Williams-Jones and Heinrich, 2005). In addition to metal precipitation, the main authigenic mineral in hot spring deposits is opaline silica (SiO2). Siliceous sinter diagenesis is still poorly understood but it contains relic information about various microbiological species, including life that inhabited the planet nearly 2,500 million years ago (Konhauser et al., 2003). Runoff channels of geothermal systems often support thriving mat-forming communities of extremophiles, which are sillicified as a result of the cooling and evaporation of silica supersaturated water. Consequently, geothermal systems offer unique natural research laboratories for sedimentologists, geomicrobiologists, and geochemists. Central to improving understanding of As biogeochemistry in geothermal systems is to evaluate its speciation in the solid phase. This chapter reviews the existing information on solid-phase speciation of As in geothermal systems, with special emphasis on hot spring deposits, and presents our recent

Arsenic Speciation in Solid Phases of Geothermal Fields

419

findings concerning As speciation in siliceous sinters and microbial mats of El Tatio geothermal field using X-ray absorption spectroscopy (XAS).

15.2. Arsenic in Geothermal Systems 15.2.1. Processes Controlling the Fluid Composition in Hot Springs – General Aspects Hot springs are defined as waters that emerge from the Earth’s crust at mean temperatures above the human body core temperature of 36.71C (Pentecost et al., 2003). Hot springs occur in areas of high geothermal gradients, where infiltrated meteoric waters are conductively heated in the high-temperature rock reservoir produced either by magmatic activity, metamorphic agents, faulting, or radioactivity (Webster and Nordstrom, 2003). In addition to geothermal fluid transport and leaching of rock constituents (Ellis and Mahon, 1964), intrusion of magmatic fluids and vapors also occur in some geothermal systems, particularly in those where high-sulfide epithermal ore deposits are formed (Williams-Jones and Heinrich, 2005). During the upward fluid movement and expansion through the vertical fractures of the host rock, the pressure drops and allows the separation of a vapor phase, a process called subsurface boiling. Both phases can reach the surface without further interaction, or get mixed with water from shallow aquifers. Direct discharge of hydrothermal fluids to the surface generally produces hot springs of circum-neutral pH and high concentrations of sodium-chloride and metals, which originate during the earlier high-temperature dissolution process. Fluid mixing with cooler shallow groundwater results in discharges of lower temperature and lower concentrations of dissolved species, whereas CO2 and H2S rich vapors in contact with shallow groundwater gives rise to sulfate and bicarbonate acidified hot springs. Hence, the interactions that occur at the solid–water interface, vapor–water interface, and during mixing with cooler groundwater govern the aqueous and solid-phase speciation and concentration in a geothermal system.

15.2.2. Arsenic Aqueous-Phase Speciation in Geothermal Waters Arsenic concentrations in geothermal reservoir fluids typically range from 1 to 600 mM (Table 15.1). Direct leaching from the reservoir rock has been

420

M. A. Alsina et al.

Table 15.1: As Concentrations (mmol) of Several Geothermal Systems: Well Fluids and Hot Springs. Field Geothermal well fluids Baca, New Mexico Broadlands, NZ Cerro Prieto, New Mexico El Tatio, Chile Kawerau, NZ Los Azufres, Mexico Orakei Korako, NZ Tongonan, Philippines Waiotapu, NZ Wairakei, NZ Hot springs Broadlands, NZ El Tatio, Chile Kamtchatka, Russia Kilahuea, Hawaii, USA Lesser Antilles, Dominica Orakei Korako, NZ Soda Dam, New Mexico Tipitapa, Nicaragua Tutum Bay, Papua NG Waiotapu, NZ Wairakei, NZ Yellowstone National Park, USA a

As (mmol kg1) 25.4–66.7 76.1–120.1 4–34 400.4–533.9 7.2–64.9 272.3–384.4 8.0–10.7 266.9–453.8 38.7–41.4 54.7–64.1 13.3 627.3 10.6–12.6 0.8–1.4 0.1–1.2 4.1–5.1 22.7 3.5 10.9–12.7 9.5–86.8 49.9–68.2 2.1–133.5

Reference White et al. (1984)a Ewers and Keays (1977) Manon et al. (1977)a Cusicanqui et al. (1976) Brown and Simmons (2003) Kruger et al. (1985)a Ellis and Mahon (1977) Kingston (1979) Ellis and Mahon (1977) Ellis and Mahon (1977) Ellis and Mahon (1977) Cusicanqui et al. (1976) Belkova et al. (2004) de Carlo and Thomas (1985) McCarthy et al. (2005) Papke et al. (2003) Reid et al. (2003) Lacayo et al. (1992) Pichler and Veizer (1999) Jones et al. (2001) McKenzie et al. (2001) Stauffer and Thompson (1984)

Cited by Ballantyne and Moore (1988).

proposed as the predominant mechanism for As release in the reservoir fluids (Ellis and Mahon, 1964; Webster, 1990). However, it is more likely that As in the geothermal fluid is controlled by a solid phase in thermodynamic equilibrium with the fluid (Ballantyne and Moore, 1988). Direct analysis of geothermal well brines (Heinrich and Eadington, 1986) along with high temperature solubility measurements of As oxide and sulfide minerals (Webster, 1990; Eary, 1992; Pokrovski et al., 1996) demonstrate that As(III) is predominant in the geothermal reservoir fluids. Thermodynamic calculations predict that under the conditions common to most geothermal fields the predominant aqueous species is likely to be AsðOHÞ03ðaqÞ ; whereas the predominance of HAsS02ðaqÞ complexes is restricted to reservoir solutions at


421

temperatures o1701C and with concentrations of H2S no less than 0.01 M (Ballantyne and Moore, 1988). Data for SAs concentrations of several hot springs are tabulated in Table 15.1. Most hot springs contain SAs in concentrations that range from 1 to 80 mM, with the exception of the El Tatio hot springs, which discharge As at concentrations greater than 600 mM, the largest reported value to date. Lower SAs concentration in the springs than in the reservoir fluids, such as the Orakei Korako spring (New Zealand), can be explained by dilution effects, accounted by the additional lower presence of other constituents (typically chloride or borate) in the hot springs compared to the reservoir fluids (Ballantyne and Moore, 1988). In contrast, higher SAs concentration in the springs than in the reservoir fluids, as in the case Waiotapu springs (New Zealand), could be explained by additional subsurface dissolution reactions, fluid evaporation due to subsurface boiling, or simply that spring waters are not derived from the sampled reservoir fluids (Ballantyne and Moore, 1988). A positive linear correlation has been traditionally shown between SAs and Cl concentrations in hot springs, and has been explained by the similar behavior of both constituents, staying primarily in the aqueous phase during subsurface boiling and consequent phase separation. Therefore, SAs and Cl concentrations are consistently higher in springs resulting from direct discharge of the reservoir fluid than those found in acidified discharges resulting from mixing of shallow groundwater with vapor phases (Webster and Nordstrom, 2003). This relationship has been used to track the effects of fluid dilution, evaporation, or mixing in a geothermal field. However, this correlation between SAs and Cl does not imply a direct chemical relationship, common origin, or similar behavior in the reservoir fluid (Ballantyne and Moore, 1988; Webster and Nordstrom, 2003). Based on thermodynamic considerations, the oxic and acidic-to-neutral conditions of most hot springs generally favor the oxidation of As(III) to As(V). For example, in oxic, aqueous solutions (251C and 1 atm), H2 AsO 4ðaqÞ predominates from pH 2–6.8, while HAsO2 4ðaqÞ is dominant at higher pH values. However, measurements of As aqueous speciation in hot springs have shown that oxidation to As(V) mostly occurs in acidic sulfate- or bicarbonate-containing hot springs, while As(III) is the predominant species in hot springs resulting from the direct discharge of the hydrothermal fluid (Criaud and Fouillac, 1989). Arsenic oxidation does not occur in sulfide-rich hot springs (Langner et al., 2001). Furthermore, slow rates of abiotic As(III) oxidation have been observed, under controlled laboratory experiments, in hydrothermal solutions with pH values ranging from 2 to 8 over a range of sulfate and bicarbonate concentrations (Wilkie and Hering, 1998; Gihring

422

M. A. Alsina et al.

and Banfield, 2001; Langner et al., 2001). The latter suggests that, despite thermodynamically favorable conditions, either abiotic or biogenic catalysis is required for quantitative oxidation of As(III) to As(V) in waters discharged from hot springs.

15.2.3. Arsenic Solid-Phase Speciation in the Geothermal Deposits In geothermal reservoir rocks, under temperatures ranging from 250 to 5001C, As is mostly found as arsenopyrite (FeAsS(s)), As-bearing pyrite (FeS2(s)), and other arsenides such as lollingite (FeAs2(s)) (Scott, 1983; Heinrich and Eadington, 1986; Pokrovski et al., 2002). At temperatures ranging from 150 to 2501C, As occurs predominantly as As-bearing pyrite, with As concentrations up to 3.7 wt.% (Ewers and Keays, 1977; Ballantyne and Moore, 1988), or in association with Fe oxides (Christensen et al., 1983). It should be noted that although pyrite has been implicated as a dominant As-bearing mineral in geothermal reservoir rocks, the exact chemical environment of As in pyrite is still a matter of debate (Simon et al., 1999; Kolker and Nordstrom, 2001). Spectroscopic evidence along with recalculation of thermodynamic data for arsenopyrite (Pokrovski et al., 2002) strongly support the formation of a solid solution between pyrite and arsenopyrite. Heinrich and Eadington (1986) and Ballantyne and Moore (1988) proposed a stoichiometric reaction for the formation of a solid solution between arsenopyrite and pyrite 4FeAsSðsÞ þ 4H2 SðgÞ þ 5O2ðgÞ þ 2H2 OðgÞ ¼ 4FeS2ðSÞ þ 4ðAsðOHÞ03ðaqÞ Þ (15.1) Currently, As concentrations in geothermal reservoir fluids that range from 4 to 600 mM can be explained by assuming that Eq. (15.1) is valid (Pokrovski et al., 2002). However, a decrease in As concentrations as the reservoir fluid temperature increases contrasts with experimental evidence for a positive correlation between As concentration and temperature in the geothermal reservoir fluids (Ballantyne and Moore, 1988). The most common As-bearing minerals encountered in epithermal deposits are orpiment (monoclinic As2S3(s)), its amorphous analog As2S3(s), realgar (monoclinic AsS(s)), and native arsenic (As0ðsÞ ) (Heinrich and Eadington, 1986; Ballantyne and Moore, 1988; Eary, 1992; Cleverley et al., 2003). Two explanations can account for the precipitation of native As in the epithermal deposits. The first is the thermodynamic stability of As sulfides and native As at temperatures above 1501C and hydrogen fugacities of 0.1 bar (Pokrovski et al., 1996),


423

while the second is the stability of native As after the fluid cools down on its way to the surface and in presence of pyrite (Heinrich and Eadington, 1986), an observation consistent with the presence of both minerals in epithermal deposits (Ballantyne and Moore, 1988). Similar thermodynamic considerations show that arsenolite (cubic As2O3(s)) and claudetite (monoclinic As2O3(s)) should not precipitate from hydrothermal fluids, even at low temperatures, owing to their high solubilities under the hydrothermal conditions. This thermodynamic prediction is consistent with the absence of arsenolite and claudetite within the reported As minerals in geothermal systems (Ballantyne and Moore, 1988).

15.3. Qualitative and Quantitative Characterization of Hot Spring Deposits Several difficulties arise when traditional approaches are used to identify the exact As mineralogy in hot spring deposits. Recently formed minerals tend to be significantly hydrated and often poorly crystalline, rendering them unsuitable for mineral identification through X-ray and electron diffraction. Additionally, after discharge from the hot spring vents hydrothermal fluids undergo cooling, evaporation, and CO2 degassing, becoming rapidly supersaturated with respect to SiO2(s) and CaCO3(s) solid phases (Guidry and Chafetz, 2002; Jones and Renaut, 2003). As a result, amorphous opaline silica and travertine commonly predominate in hot spring deposits, and greatly interfere with the correct identification of As-bearing minerals within solid matrices of hot spring discharges.

15.3.1. Microstructure and Molecular Structure/Mineralogy Electron microprobe analysis, X-ray diffraction (XRD), and selected area electron diffraction (SAED or ED) have been the traditional tools applied for mineral identification in hot spring deposits. As-bearing minerals identified using these techniques include: orpiment, realgar, As-rich marcasite (FeS2(s)), and As-rich metastibnite (AsxSb2xS3(s)) (Ewers and Keays, 1977; White, 1981; Ballantyne and Moore, 1988; Webster and Nordstrom, 2003). XAS studies have demonstrated the presence of As in smectite clays, occurring predominantly as occlusions of As(III) oxide and minor amounts of As(V) oxides (Pascua et al., 2005). However, the exact mineralogy of the As-bearing occlusions still remains unknown. Arsenic has also been found in

424

M. A. Alsina et al.

Fe oxides in the Roosevelt hot springs (Christensen et al., 1983), although it was considered to have resulted from a secondary deposition mechanism after As inclusion in pyrite. Nonetheless, recent studies support the role of Fe oxyhydroxides as primary As sequestration agents in hot spring deposits. For example, As-enriched hydrous Fe oxides have been reported in hydrothermal vents on the ocean floor at Tutum Bay, Florida, USA (Rancourt et al., 2001), and spectroscopic analyses of the mineralizations occurring in microbial mats from the Cinder Pool of Yellowstone National Park (YNP) demonstrate the presence of arsenate adsorbed onto ferric oxyhydroxides (Inskeep et al., 2004). The presence of Fe hydroxide minerals in hot springs has been linked to Fe enriched (19.7 wt.%) silica granules and spheroids (Ferris et al., 1986) and amorphous ferric hydroxides spheroids that occur as epicellular mineralizations of subaqueous extremophile communities in hot spring systems (Konhauser and Ferris, 1996; Belkova et al., 2004). Elevated As concentrations (9.7 wt.%) have also been measured in Fe-rich silica spheroids from epicellular sheaths of microbial mats from Cyanidium Creek, YNP (Ferris et al., 1986). Direct examination of naturally occurring benthic microbial mats (Konhauser et al., 1993), as well as controlled experiments of silica adsorption onto Bacillus subtilis (Fein et al., 2002) concluded that the association of silicate minerals with bacteria is produced by the reaction of silica with Al and Fe oxides adsorbed or precipitated onto the bacterial surface, rather than by direct silica mineralizations over the bacterial surface. Recent laboratory experiments that studied the immobilization of silica and Fe by B. subtilis confirm the strong affinity of bacterial cells for Fe, but also provided evidence that, under silica concentrations similar to those found in hot springs, both iron and silica immobilization occur predominantly as a non-biogenic process (Phoenix et al., 2003). Hence, specific studies under the biogeochemical conditions in hot springs are needed to assert the governing mechanisms (abiotic and biotic) that control silica and Fe deposition, and their possible influence over As immobilization in hot spring deposits. Uncommon As-Sb-sulfide mineralizations have been reported to occur on cell walls of rod-shaped bacteria recovered from the Champagne Pool, New Zealand (Phoenix et al., 2005). The hydrothermal waters of these systems are slightly acidic (pH 5.1), anoxic, and sodium-chloride dominated. Energy dispersive spectrometry (EDS) analyses of the mineralizations show significant concentrations of Sb (up to 60 wt.%), As (up to 25 wt.%), S (up to 20 wt.%), and silica (up to 20 wt.%). Other mineralizations on bacterial surfaces showed enrichment of Fe (up to 25 wt.%) and silica (up to 100 wt.%) and, therefore, the additional presence of Fe hydroxides cannot be ruled out. Thermodynamic calculations under the hydrothermal conditions


425

of the Champagne Pool (Phoenix et al., 2005) show that orpiment is the stable phase. This hypothesis is consistent with the As-S enrichment measured on bacterial walls, but has not been confirmed by SAED analyses, since no crystalline compounds were observed. Phoenix et al. (2005) additionally suggest that metal sorption by organic functional groups present on bacterial surfaces is a possible mechanism for epicellular As-Sb-S enrichment. Microstructural studies and EDS measurements of microbial mats from acid sulfate-chloride YNP hot springs (Langner et al., 2001) showed that yellow mat bacteria are closely associated with crystalline mineralizations of S0, but with no appreciable presence of As. In contrast, EDS measurements of reddish-brown mats showed that Fe and As were dominant but sulfur was absent. XRD measurements confirmed the presence of native sulfur in the yellow mats, but poor crystallinity of the Fe-As mineralizations in the reddish-brown mats. X-ray photoelectron spectroscopy (XPS) measurements of these reddish-brown mats showed two peaks at 712.5 and 45.2 eV, closely associated to Fe(III) and As(V), respectively. Further SEM-EDS analyses of these As-Fe bacterial filaments by Inskeep et al. (2004) showed that regardless of the location, the main constituents are O, Fe, and As, with As/Fe molar ratios of 0.7, likely for arsenate adsorbed to Fe hydroxides (Waychunas et al., 1993). Parallel Electron Energy Loss Spectroscopy (PEELS) confirmed the spatial correlation between these three elements, and provided strong evidence for the adsorption of arsenate on Fe oxyhydroxides. The Masutomi hydrothermal springs have a near neutral pH (6), are slightly oxidized, and are dominated by sodium, chloride, and bicarbonate ions. SEM, TEM, and EDS analyses of the reddish-brown microbial mats from these springs (Tazaki et al., 2003) showed that colloidal aggregates in bacillioform and coccoid type bacterial sheaths were enriched with Fe and As. XRD patterns collected from the mats only suggested the presence of poorly crystalline Fe hydroxides, but ED measurements showed that colloidal aggregates from the bacterial sheath contained calcite and poorly crystallized lollingite. 15.3.2. Evidence of Microbial Controls over Arsenic Speciation Prokaryotes that inhabit sediments, estuaries, lakes, mine tailings, and hot springs are known to reduce As(V) either as a detoxification pathway (via Ars operon) or for energy generation (Oremland and Stolz, 2003). Heterotrophic (Jackson et al., 2001) and chemolithoautotrophic prokaryotes (Santini et al., 2000; Donahoe-Christiansen et al., 2004) are known to oxidize As(III). Bacterial catalysis has also been proposed as the main mechanism

426

M. A. Alsina et al.

for As(III) to As(V) oxidation in hot spring environments. Wilkie and Hering (1998) found that the rapid in-stream oxidation of As(III) to As(V) (pseudo-first-order half-life of 0.3 h) in waters discharged from hot springs (Hot Creek Gorge) is primarily mediated by macrophyte-related bacteria and not by abiotic processes. Langner et al. (2001) found that As(III) discharged from acidic hot springs of YNP is rapidly oxidized to As(V) (half-life of 0.58 min) only after dissolved sulfide concentrations become less than 30 mM. In this sulfate-chloride dominated spring, microbes catalyze As(III) oxidation, particularly those that form the reddish-brown mat. Langer and coworkers also suggested that microbial catalysis would be responsible for Fe(II) to Fe(III) oxidation, since the rate of abiotic for Fe(II) oxidation under acidic conditions is extremely slow. Laboratory experiments demonstrated that active growing cells of Thermus termophilus and T. aquaticus were capable of mediating As(III) to As(V) oxidation (Gihring and Banfield, 2001). Bacterially mediated As oxidation was confirmed in the Butte Vista hot springs, which are basic waters (pH 8) with consistently low dissolved sulfide concentrations (2 mM). A Thermus strain has been isolated from the Growler hot springs (Gihring et al., 2001), exhibiting aerobic arsenite oxidizing rates that are 100-fold faster than under abiotic conditions. This Thermus strain is also capable of respiring arsenate under anaerobic conditions, suggesting that chemolithoautotrophic microbial activity is a determining factor controlling the As fate in hot springs and downstream flows. Microbes are known to influence the kinetics and course of reactions leading to mineral growth and dissolution. This influence helps them cope with aggressive environmental conditions (Ehrlich, 1996). Hence, it is logical to think that microbial activity controls As speciation in the solid phase, particularly in poorly crystalline Fe hydroxides mineralizations within silica granules and spheroids formed in epicellular depositions and bacterial sheaths (Ferris et al., 1986; Konhauser and Ferris, 1996; Fein et al., 2002). The affinity between As and Fe oxyhydroxides has been recognized in natural systems and extensively studied as an alternative for As removal. Table 15.2 summarizes current findings concerning As adsorption on Fe oxyhydroxides along with other reactive geomedia, such as Mn and Al oxides. Significant evidence for microbial control of As solid-phase speciation in hot springs is provided by the As-Sb-S enrichment observed at the external surfaces of microbes inhabiting the Champagne Pool of the Waiotapu geothermal field (Phoenix et al., 2005). Additional evidence has been provided by the identification of lollingite biomineralizations occurring in colloidal aggregates around the bacterial cell wall of reddish-brown microbial mats recovered from the Masutomi hot springs (Tazaki et al., 2003).


427

Table 15.2: Sorption Mechanisms for Arsenic Depending on the pH Values and the Type of Sorbent Material. pH

As oxidation state

Sorbent material

Sorption mechanism

3.0–5.5 3.0–8.0 5.0–8.0 5.0–8.0 5.0–8.0 5.0–9.0

III V V V V III

g-Al2O3 g-Al2O3 Mn-oxide Vernadite K-birnessite Al-oxyhydroxide

Bidentate Bidentate Bidentate Bidentate Bidentate Outer-sphere complex

5.0–9.0 5.3 5.5–6.5 5.5–6.5 5.5–6.5 5.5-8.0 6.0–9.0 6.4–8.6 6.5 6.5–8.2 7.6

V V V III V III V III V V III

Al-oxyhydroxide Fe-oxyhydroxide Lepidocrocite Lepidocrocite Goethite g-Al2O3 Goethite Goethite K-birnessite Gibbsite Ferrihydrite

Inner-sphere complex Inner-sphere complex Bidentate Bidentate Bidentate Outer-sphere complex Bidentate+monodentate Bidentate Bidentate Bidentate Bidentate

8.0

V

Ferrihydrite

Bidentate+monodentate

8.0 8.0 8.0 49.0 49.0 49.0 49.0 49.0 9.6

V V V III III V V V III

Goethite Akaganeite Lepidocrocite Goethite Lepidocrocite Goethite Lepidocrocite Maghemite Hematite

Bidentate Bidentate Bidentate Bidentate Bidentate Bidentate Bidentate Bidentate Bidentate

10.4 10.4

III III

Goethite Lepidocrocite

Bidentate+monodentate Bidentate+monodentate

Reference

Arai et al. (2001) Foster et al. (2003)

Goldberg and Johnston (2001) Morin et al. (2002) Farquhar et al. (2002)

Arai et al. (2001) Fendorf et al. (1997) Manning et al. (1998) Manning et al. (2003) Myneni et al. (1998) Ona-Nguema et al. (2005) Waychunas et al. (1993)

Manning et al. (2002)

Ona-Nguema et al. (2005)

15.4. XAS Analysis of Arsenic Solid-Phase Speciation in Hot Springs 15.4.1. Arsenic Solid-Phase Speciation using X-Ray Absorption Spectroscopy XAS is an element-specific technique widely used to determine the average oxidation state, local coordination environment, and the relative proportions

428

M. A. Alsina et al.

of species in a given sample. XAS is an excellent technique to investigate As solid-phase speciation in complex environmental matrices, particularly when poorly crystalline As-bearing phases are present at relatively low concentrations. For instance, XAS has been successfully applied to speciate the As solid phase in a variety of As-enriched systems, including mine tailings (Foster et al., 1998; Paktunc et al., 2003), soils contaminated by As-based pesticides (Cances et al., 2005), and As hyper-accumulating ferns (Webb et al., 2003). XAS is also a valuable tool for studying sorption/precipitation processes. Studies by Waychunas et al. (1993) and Sherman and Randall (2003) of the local coordination of As(V) adsorbed on Fe(III) oxyhydroxides showed the preferential formation of inner-sphere bidentate complexes, where the AsO3 4 tetrahedron share two vertex oxygen atoms of two FeO9 6 octahedra from the surface layer of ferrihydrite (bidentate corner fashion; As-Fe distance 3.27 A˚). XAS has also been used to determine the solid-phase speciation of As in subaqueous samples, such as sediments and microbial mats from Lost Lake, Minnesota, USA (Foster et al., 2001). The X-ray absorption near-edge structure (XANES) showed that the average oxidation state of As was +5 in these sediments, despite the predominance of As(III) species in the aqueous phase of Lost Lake. The extended X-ray absorption fine structure (EXAFS) spectra of the sediments closely resemble that of As(V) sorbed on the ferrihydrite model compound. This similarity is strong evidence for the presence of As(V) bound to Fe oxyhydroxides. Non-linear least squares fits of the As EXAFS spectra of the sediments identified a predominant bidentate corner-sharing surface complex, with an average number of 1.5 Fe atoms at a distance of 3.3 A˚. In addition, a minor fraction of As was bound in a monodentate edge-sharing geometry, yielding an average number of 0.5 Fe atoms at a distance of 2.85 A˚. Interestingly, As in the sediments and the mats possesses similar local coordination environments, as no spectroscopic differences were observed between both sample types. In geothermal environments, Pascua et al. (2005) applied XAS to probe the coordination environment of As-enriched smectite clays (with As concentrations up to 60 mmol kg1) recovered from sinters of geothermal springs in northwestern Japan. XRD measurements of the sinters revealed the presence of smectite, but found no appreciable As-containing mineral phases. In contrast, XANES revealed that As occurs mostly as As(III) (up to 90%) whereas As(V) species accounted for 10% of the total As. XRD, IR spectroscopy, and EDS chemical mapping did not detect the presence of Fe or sulfide near the As mineralizations. Non-linear least squares fitting of EXAFS spectra suggested the presence of oxygen ligands at 1.72 A˚,


429

consistent with the presence of As(III)-O bonds in greater proportion than As(V)-O bonds. However, the exact chemical nature of these As-containing minerals still remains unknown. Schwenzer et al. (2001) used XANES to examine sinters from the slightly acidic (pH 6) sodium-chloride hot springs in Weisbaden, Germany. They found that As(III) was predominant in the aqueous phase, but As(V) dominated the solid phase. Aeration experiments showed that As(III) oxidation took place concurrently with the removal of soluble Fe from solution, suggesting that As(V) sorbed onto Fe oxyhydroxides was the predominant As solid phase in the sinters. Inskeep et al. (2004) used XAS to probe the presence of As in biomineralizations of microbial mats from acidic (pH 3.1) sulfate-chloride hot springs from YNP. Synchrotron-based XRD patterns of the microbial mats showed the presence of amorphous silica deposits (mostly opal-A) but also identified poorly crystalline Fe hydroxides. XANES spectra of the samples confirmed the presence of As(V) in the microbial mats, while non-linear least squares fits of the EXAFS spectra showed that arsenate is most likely adsorbed onto ferric hydroxides as a bidentate complex that is either cornersharing or edge-sharing, as suggested by Fe-As bound distances of 3.33 and 2.82 A˚, respectively. No major differences in the EXAFS spectra were detected for microbial mat samples at different locations, suggesting that adsorption on ferric hydroxides is the main mechanism for As immobilization at the solid phase in these hot spring discharges. As-sulfide associations were not identified, either in the XANES or the EXAFS spectra of the microbial mats. 15.4.2. Preliminary Findings from El Tatio Geothermal Field We used XAS to probe the solid-phase speciation of As in siliceous sinters and microbial mats of a runoff channel from El Tatio geothermal field, northern Chile. The specific geothermal system selected for this study (22.34401 S; 68.01191 W, Fig. 15.1) comprises three geysers that feed a large hot spring pool, which discharges through a well-defined shallow channel, rich in siliceous sinter, and underlying reddish-orange mat-forming extremophiles. Springs from El Tatio are characterized by near neutral pH (6.7) and are dominated by sodium and chloride ions (Fig. 15.1). Dissolved SAs concentrations in El Tatio reach 660 mM, which is 5-fold greater than any other SAs concentration reported for a hot spring anywhere in the world. Reports about the aqueous geochemistry of El Tatio springs are scarce. However, recent measurements by Landrum et al. (2007) show that, in the same geyser system studied here, SAs is mostly As(III) after exiting

430

M. A. Alsina et al.

Figure 15.1: Hot Spring System of El Tatio Geothermal Field (Northern Chile). (a) Remnant of an Exploded Siliceous Sinter Cone and a Spring Pool Comprised by the Discharges of Three Hot Springs Surrounding the Cone; (b) Runoff Channel of the Spring System Discharge, Rich in Siliceous Sinter Deposition and Settlement of Reddish-Brown and Microbial Mats; (c) Micrograph of Tufted Microbial Mats Inhabiting the Channel (Scale Bar Represents 10 mm); (d) Micrograph of the Siliceous Sinter Beneath the Mats (Scale Bar Represents 2 mm).


431

the vents, but gets rapidly oxidized to As(V) in the runoff channel. Hence, despite its unique high As concentrations, it is most likely that El Tatio provides another example of the role exerted by microbes in As(III) oxidation. Microbial mats and the underlying siliceous sinter samples were collected at several 10 m intervals along a transect through the runoff channel that started at the rim of the main pool. Immediately after collection, samples were sealed in-situ between two layers of Kaptonr tape. XAS measurements were performed at DND-CAT, Sector 5 of the Advanced Photon Source, Argonne National Laboratory (IL, USA). Arsenic K-edge XANES spectra of sinters and mats are presented in Fig. 15.2a and b,

Figure 15.2: XAS Spectra of the Sinter and Microbial Mats Samples from El Tatio Geyser System. (a) XANES As K-Edge Spectra of Several As-Containing Minerals; (b) Comparison between XANES Spectra of Several Arsenate Minerals and the Spectra of As in the Sinter Material and Biological Mats; (c) k3-Weighted EXAFS Spectra of Several Sinters and Mats Collected for the Runoff Channel of El Tatio Spring System, along with the Spectra of As(V) Reference Compound Scorodite and Arsenate Sorbed onto Ferrihydrite (Amorphous Ferric Oxide).

432

M. A. Alsina et al.

along with some significant As reference compounds characterized by different oxidation states and local As coordination. The oxidation state of inorganic As compounds can be probed by the location of the edge energy, defined as the main inflection point in the XANES spectrum (Fig. 15.2a). Edge energies for As(III)-containing compounds are located ca. 11,870 eV, making it possible to further differentiate arsenious oxides from arsenious sulfides. Interestingly, the As(V) edge occurs at 11,874 eV, which is in excellent agreement with the edge energy observed in both siliceous sinters and microbial mats from the geyser system of El Tatio. Therefore, As(V) is the dominant form of As in El Tatio sinters and mats (Fig. 15.2b). The spectral similarity in the XANES suggests that As in the sinters and microbial mats has a coordination environment similar to As in scorodite (FeAsO4 2(H2O)(s)) or to As bound to Fe oxyhydroxides. Comparison of the EXAFS spectra shows that As coordination in the microbial mats and sinters is closer to As(V) sorbed onto Fe oxyhydroxides (Fig. 15.2c) than As in scorodite. EXAFS can be used to identify the chemical nature of the immediate atomic neighbors that surround As, but this is restricted to atoms with contrasting atomic number (i.e., at least five atomic numbers). Therefore, EXAFS does not allow unequivocal distinction between neighboring atoms such as Mn and Fe. In this instance, it is difficult to ascertain whether As is sorbed to Mn or Fe oxides based solely on EXAFS As K-edge measurements. Additional information is required in order to determine the identity of neighboring atoms that surround As. Fortunately, electronic transitions produce features in the XANES spectra that are markedly different between Fe and Mn. The As K-edge XANES spectrum of arsenate sorbed onto poorly crystalline Mn oxides shows a characteristic electronic transition peak around 11,890 eV, which is absent in scorodite and arsenate sorbed to poorly crystalline Fe hydroxides (Foster, 2001). The association between As and Fe in El Tatio is further supported by the significant concentrations of Fe found in sinters (on average 0.14 mol kg1) and microbial mats (on average 0.94 mol kg1) compared to the relatively low concentrations of Mn and Cu in sinters and mats (ranging from 0.5 to 10 mmol kg1 for both elements, Fig. 15.1). Non-linear least squares fits of the EXAFS spectra for mats and sinters are shown in Figs. 15.2c and 15.3b. The fits confirm the presence of As(V), with ca. four oxygen atoms found at an average distance of 1.69 A˚ from the central As, indicative of a tetrahedral configuration. Furthermore, a shell containing ca. two Fe atoms is located at 3.23 A˚, consistent with an inner sphere bidentate corner-sharing surface complex (Fig. 15.3c). This observation agrees with the results of Sherman and Randall (2003), predicting that arsenate bidentate corner-sharing surface complexes are


433

Figure 15.3: EXAFS Spectra of the Sinter and Microbial Mats Samples from El Tatio Geyser System. (a) Fourier Transform (FT) k3-Weighted EXAFS Spectra of Several As-Containing Minerals along with the Spectra of Mat and Sinter Samples; (b) FT k3-Weighted EXAFS Spectra of all Collected Sinter and Mat Samples from the Runoff Channel of El Tatio Spring System (Black Continuous Line) as well as the Non-Linear Least Squares Fitting of the Spectra (Black Dotted Line); (c) Representation of the Geometry of Sorption of Arsenate onto Fe Oxyhydroxides in a Bidentate Corner-Sharing Configuration for both the Sinter Material and Biological Mats. substantially more energetically stable than monodentate corner-sharing and bidentate edge-sharing configurations. XANES and EXAFS fingerprints showed no significant differences between the sinters and mats samples, suggesting that the same solid phases formed. Thus, the adsorption of As(V) onto Fe oxyhydroxides is probably the main mechanism accounting for the speciation of As in the sinters and the microbial mats of El Tatio.

434

M. A. Alsina et al.

15.5. Concluding Remarks Geothermal fields provide incomparable natural laboratories to explore the competitive mechanisms that control the fate of metals, and in particular sorption and precipitation reactions at abiotic and biogenic interfaces. Geothermal fields are also important natural sources of arsenic enrichment, compromising the quality of irrigation and drinking water at several locations around the world. XAS is an essential tool for probing As speciation in hydrous amorphous solid phases, but the reports that have used this technique to characterize geothermal deposits are still scarce. Available studies, including our own results on El Tatio geothermal field, point to As(V) as the dominant oxidation state for As in hot spring deposits. A review of the literature and our study show that the sorption of As by Fe oxyhydroxides controls solid-phase As speciation in a wide range of geothermal systems where Fe is present, such as hydrothermal waters that are either of circumneutral pH and sodium-chloride rich, or acidic and sulfate-chloride rich. Thus, Fe-containing sinters and mats need to be considered in conceptual and quantitative models for predicting the fate of As in geothermal systems. The intimate relationship between microbes and silica/iron mineralizations has been observed in several hot spring systems, and El Tatio is no exception. The epicellular and extracellular Fe-silica aggregates are capable of sorbing and immobilizing As(V). However, the microbial mediation of Fe oxyhydroxides formation remains unclear. As most geothermal systems contain considerable concentrations of dissolved ions, it is worthwhile to evaluate their influence on the fate of As in the solid phase. Relevant issues include the chemical species competing with As for sorption sites in the hot spring geomedia, the species that stimulate or inhibit the formation of Fe oxyhydroxides, and the species or energy sources (e.g., light) that are key to defining the biological control over As speciation in hot springs. Extensive laboratory and field experiments will be required to address these issues, but will greatly improve our conceptual and quantitative models of As speciation in geothermal systems.

ACKNOWLEDGMENTS We gratefully thank Scott Tyler, Doug Kent, and two reviewers for their contributions to improving this manuscript. This work was funded by Proyecto Fondecyt 1070737/2007, and Proyecto Bicentenario IICI-3. Portions of this work were performed at the DuPont-Northwestern-Dow Collaborative Access Team (DND-CAT) Synchrotron Research Center located at Sector 5


435

of the Advanced Photon Source. DND-CAT is supported by the E.I. DuPont de Nemours & Co., The Dow Chemical Company, the U.S. National Science Foundation through Grant DMR-9304725, and the State of Illinois through the Department of Commerce and the Board of Higher Education Grant IBHE HECA NWU 96. Use of the Advanced Photon Source was supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. W-31-109-Eng-38. PP gratefully thanks Drs. Phil Bennett and Annette Engel for fostering his interest in El Tatio geothermal field.

REFERENCES Arai, Y., Elzinga, E. J., & Sparks, D. L. (2001). X-ray absorption spectroscopic investigation of arsenite and arsenate adsorption at the aluminum oxide–water interface. J. Colloid Interface Sci., 235 (1), 80–88. Ballantyne, J. M., & Moore, J. N. (1988). Arsenic geochemistry in geothermal systems. Geochim. Cosmochim. Acta, 52 (2), 475–483. Belkova, N. L., Zakharova, J. R., Tazaki, K., Okrugin, V. M., & Parfenova, V. V. (2004). Fe-Si biominerals in the Vilyuchinskie hot springs, Kamchatka Peninsula, Russia. Int. Microbiol., 7 (3), 193–198. Brown, K. L., & Simmons, S. F. (2003). Precious metals in high-temperature geothermal systems in New Zealand. Geothermics, 32 (4), 619–625. Cances, B., Juillot, F., Morin, G., Laperche, V., Alvarez, L., Proux, O., Hazemann, J. L., Brown, G. E., & Calas, G. (2005). XAS evidence of As(V) association with iron oxyhydroxides in a contaminated soil at a former arsenical pesticide processing plant. Environ. Sci. Technol., 39 (24), 9398–9405. Christensen, O. D., Capuano, A., & Moore, J. N. (1983). Trace-element distribution in an active hydrothermal system, Roosevelt hot springs thermal area, Utah. J. Volcanol. Geotherm. Res., 16 (1–2), 99–129. Cleverley, J. S., Benning, L. G., & Mountain, B. W. (2003). Reaction path modeling in the As-S system: A case study for geothermal as transport. Appl. Geochem., 18 (9), 1325–1345. Criaud, A., & Fouillac, C. (1989). The distribution of arsenic (III) and arsenic (V) in geothermal waters: Examples from the Massif Central of France, the Island of Dominica in the Leeward Islands of the Caribbean, the Valles Caldera of New Mexico, U.S.A., and southwest Bulgaria. Chem. Geol., 76 (3–4), 259–269. Cusicanqui, H., Mahon, W. A. J., & Ellis, A. J. (1976). The Geochemistry of the El Tatio geothermal field, northern Chile. 2nd United Nations Geothermal Symposium Proceedings. Lawrence Berkeley Laboratory, University of California, Berkeley, CA. De Carlo, E. H., & Thomas, D. M. (1985). Removal of arsenic from geothermal fluids by adsorptive bubble flotation with colloidal ferric hydroxide. Environ. Sci. Technol., 19 (6), 538–544.

436

M. A. Alsina et al.

Donahoe-Christiansen, J., D’Imperio, S., Jackson, C. R., Inskeep, W. P., & McDermott, T. R. (2004). Arsenite-oxidizing hydrogenobaculum strain isolated from an acid-sulfate-chloride geothermal spring in Yellowstone National Park. Appl. Environ. Microbiol., 70 (3), 1865–1868. Eary, L. E. (1992). The solubility of amorphous As2S3 from 25 to 901C. Geochim. Cosmochim. Acta, 56 (6), 2267–2280. Ehrlich, H. L. (1996). How microbes influence mineral growth and dissolution. Chem. Geol., 132 (1–4), 5–9. Ellis, A. J., & Mahon, W. A. J. (1964). Natural hydrothermal systems and experimental hot-water/rock interactions. Geochim. Cosmochim. Acta, 28 (8), 1323–1357. Ellis A. J., & Mahon W. A. J. (1977). Chemistry and Geothermal Systems. Academic Press, New York, 392 pp. Ewers, G. R., & Keays, R. R. (1977). Volatile and precious metal zoning in the Broadlands geothermal field, New Zealand. Econ. Geol., 72 (7), 1337–1354. Farquhar, M. L., Charnock, J. M., Livens, F. R., & Vaughan, D. J. (2002). Mechanisms of arsenic uptake from aqueous solution by interaction with goethite, lepidocrocite, mackinawite, and pyrite: An X-ray absorption spectroscopy study. Environ. Sci. Technol., 36 (8), 1757–1762. Fein, J. B., Scott, S., & Rivera, N. (2002). The effect of Fe on Si adsorption by Bacillus subtilis cell walls: Insights into non-metabolic bacterial precipitation of silicate minerals. Chem. Geol., 182 (2–4), 265–273. Fendorf, S., Eick, M. J., Grossl, P., & Sparks, D. L. (1997). Arsenate and chromate retention mechanisms on goethite. 1. Surface structure. Environ. Sci. Technol, 31 (2), 315–320. Ferris, F. G., Beveridge, T. J., & Fyfe, W. S. (1986). Iron-silica crystallite nucleation by bacteria in a geothermal sediment. Nature, 320 (6063), 609–611. Foster, A. L. (2001). Synchrotron-based spectroscopic studies of metal species in solid phases: The case of arsenic. USGS Workshop on Arsenic in the Environment, February 21–22, Denver, CO, USA (abstract available at http://wwwbrr.cr.usgs. gov/Arsenic/finalabstracts.htm). Foster, A. L., Ashley, R. P., & Rituba, J. J. (2001). Direct and in-situ speciation of arsenic in microbial mats and sediments using X-ray absorption spectroscopy. USGS Workshop on Arsenic in the Environment, February 21–22, Denver, CO, USA (abstract available at http://wwwbrr.cr.usgs.gov/Arsenic/finalabstracts. htm). Foster, A. L., Brown, G. E., & Parks, G. A. (2003). X-ray absorption fine structure study of As(V) and Se(IV) sorption complexes on hydrous Mn oxides. Geochim. Cosmochim. Acta, 67 (11), 1937–1953. Foster, A. L., Brown, G. E., Tingle, T. N., & Parks, G. A. (1998). Quantitative arsenic speciation in mine tailings using X-ray absorption spectroscopy. Am. Mineral., 83 (5–6), 553–568. Gihring, T. M., & Banfield, J. F. (2001). Arsenite oxidation and arsenate respiration by a new Thermus isolate. FEMS Microbiol. Lett., 204 (2), 335–340.


437

Gihring, T. M., Druschel, G. K., McCleskey, R. B., Hamers, R. J., & Banfield, J. F. (2001). Rapid arsenite oxidation by Thermus aquaticus and Thermus thermophilus: Field and laboratory investigations. Environ. Sci. Technol., 35 (19), 3857–3862. Goldberg, S., & Johnston, C. T. (2001). Mechanisms of arsenic adsorption on amorphous oxides evaluated using macroscopic measurements, vibrational spectroscopy, and surface complexation modeling. J. Colloid Interface Sci., 234 (1), 204–216. Guidry, S. A., & Chafetz, H. S. (2002). Factors governing subaqueous siliceous sinter precipitation in hot springs: Examples from Yellowstone National Park, USA. Sedimentology, 49 (6), 1253–1267. Heinrich, C. A., & Eadington, P. J. (1986). Thermodynamic predictions of the hydrothermal chemistry of arsenic, and their significance for the paragenetic sequence of some cassiterite-arsenopyrite-base metal sulfide deposits. Econ. Geol., 81 (3), 511–529. Inskeep, W. P., Macur, R. E., Harrison, G., Bostick, B. C., & Fendorf, S. (2004). Biomineralization of As(V)-hydrous ferric oxyhydroxide in microbial mats of an acid-sulfate-chloride geothermal spring, Yellowstone National Park. Geochim. Cosmochim. Acta, 68 (15), 3141–3155. Jackson, C. R., Langner, H. W., Donahoe-Christiansen, J., Inskeep, W. P., & McDermott, T. R. (2001). Molecular analysis of microbial community structure in an arsenite-oxidizing acidic thermal spring. Environ. Microbiol., 3 (8), 532–542. Jones, B., & Renaut, R. W. (2003). Hot spring and geyser sinters: The integrated product of precipitation, replacement, and deposition. Can. J. Earth Sci., 40 (11), 1549–1569. Jones, B., Renaut, R. W., & Rosen, M. R. (2001). Biogenicity of gold- and silverbearing siliceous sinters forming in hot (751C) anaerobic spring-waters of Champagne Pool, Waiotapu, North Island, New Zealand. J. Geol. Soc., 158 (6), 895–911. Kingston, R. (1979). The Tongonan Geothermal Field Leyte Philippines: Report on Exploration and Development. Auckland, Kingston Reynolds Thom and Allardice Ltd., 31 pp. Kolker, A., & Nordstrom, D. K. (2001). Occurrence and Micro-Distribution of Arsenic in Pyrite. USGS Workshop on Arsenic in the Environment, February 21–22, Denver, CO, USA (abstract available at http://wwwbrr.cr.usgs.gov/ Arsenic/finalabstracts.htm). Konhauser, K. O., & Ferris, F. G. (1996). Diversity of iron and silica precipitation by microbial mats in hydrothermal waters, Iceland: Implications for Precambrian iron formations. Geology, 24 (4), 323–326. Konhauser, K. O., Fyfe, W. S., Ferris, F. G., & Beveridge, T. J. (1993). Metal sorption and mineral precipitation by bacteria in two Amazonian river systems: Rio Solimoes and Rio Negro, Brazil. Geology, 21 (12), 1103–1106. Konhauser, K. O., Jones, B., Reysenbach, A.-L., & Renaut, R. W. (2003). Hot spring sinters: Keys to understanding earth’s earliest life forms. Can. J. Earth Sci., 40 (11), 1713–1724.

438

M. A. Alsina et al.

Lacayo, M. L., Cruz, A., Calero, S., Lacayo, J., & Fomsgaard, I. (1992). Total arsenic in water, fish, and sediments from Lake Xolotlan, Managua, Nicaragua. Bull. Environ. Contam. Toxico., 49 (3), 463–470. Landrum, J. T., Bennett, P. C., Alsina, M. A., & Pasteń, P. A. (2007). Biogeohemical controls on arsenic: El Tatio geothermal field. 12th International Symposium on Water Rock Interaction. July 31–August 5, Kunming, China. Langner, H. W., Jackson, C. R., Mcdermott, T. R., & Inskeep, W. P. (2001). Rapid oxidation of arsenite in a hot spring ecosystem, Yellowstone National Park. Environ. Sci. Technol., 35 (16), 3302–3309. Manning, B. A., Fendorf, S. E., & Goldberg, S. (1998). Surface structures and stability of arsenic(III) on goethite: Spectroscopic evidence for inner-sphere complexes. Environ. Sci. Technol., 32 (16), 2383–2388. Manning, B. A., Fendorf, S. E., & Suarez, D. L. (2003). Arsenic(III) complexation and oxidation reactions on soil. Biogeochem. Environ. Imp. Trace Elements, 835, 57–69. Manning, B. A., Hunt, M. L., Amrhein, C., & Yarmoff, J. A. (2002). Arsenic(III) and arsenic(V) reactions with zero valent iron corrosion products. Environ. Sci. Technol., 36 (24), 5455–5461. McCarthy, K. T., Pichler, T., & Price, R. E. (2005). Geochemistry of Champagne hot springs shallow hydrothermal vent field and associated sediments, Dominica, Lesser Antilles. Chem. Geol., 224 (1–3), 55–68. McKenzie, E. J., Brown, K. L., Cady, S. L., & Campbell, K. A. (2001). Trace metal chemistry and silicification of microorganisms in geothermal sinter, Taupo Volcanic Zone, New Zealand. Geothermics, 30 (4), 483–502. Morin, G., Lecocq, D., Juillot, F., Calas, G., Ildefonse, P., Belin, S., Briois, V., Dillmann, P., Chevallier, P., Gauthier, C., Sole, A., Petit, P. E., & Borensztajn, S. (2002). EXAFS evidence of sorbed arsenic(V) and pharmacoside rite in a soil overlying the Echassieres geochemical anomaly, Allier, France. Bull. Soc. Geólog. France, 173 (3), 281–291. Myneni, S. C. B., Traina, S. J., Waychunas, G. A., & Logan, T. J. (1998). Experimental and theoretical vibrational spectroscopic evaluation of arsenate coordination in aqueous solutions, solids, and at mineral–water interfaces. Geochim. Cosmochim. Acta, 62 (19–20), 3285–3300. Ona-Nguema, G., Morin, G., Juillot, F., Calas, G., & Brown, G. E. (2005). EXAFS analysis of arsenite adsorption onto two-line ferrihydrite, hematite, goethite, and lepidocrocite. Environ. Sci. Technol., 39 (23), 9147–9155. Oremland, R. S., & Stolz, J. F. (2003). The ecology of arsenic. Science, 300 (5621), 939–944. Paktunc, D., Foster, A., & Laflamme, G. (2003). Speciation and characterization of arsenic in Ketza River mine tailings using X-ray absorption spectroscopy. Environ. Sci. Technol., 37 (10), 2067–2074. Papke, R. T., Ramsing, N. B., Bateson, M. M., & Ward, D. M. (2003). Geographical isolation in hot spring cyanobacteria. Environ. Microbiol., 5 (8), 650–659.


439

Pascua, C., Charnock, J., Polya, D. A., Sato, T., Yokoyama, S., & Minato, M. (2005). Arsenic-bearing smectite from the geothermal environment. Mineral. Mag., 69 (5), 897–906. Pentecost, A., Jones, B., & Renaut, R.W. (2003). What is a hot spring? Can. J. Earth Sci., 40 (11), 1443–1446. Phoenix, V. R., Konhauser, K. O., & Ferris, F. G. (2003). Experimental study of iron and silica immobilization by bacteria in mixed Fe-Si systems: Implications for microbial silicification in hot springs. Can. J. Earth Sci., 40 (11), 1669–1678. Phoenix, V. R., Renaut, R. W., Jones, B., & Ferris, F. G. (2005). Bacterial S-layer preservation and rare arsenic-antimony-sulphide bioimmobilization in siliceous sediments from Champagne Pool hot spring, Waiotapu, New Zealand. J. Geol. Soc., 162, 323–331. Pichler, T., & Veizer, J. (1999). Precipitation of Fe(III) oxyhydroxide deposits from shallow-water hydrothermal fluids in Tutum Bay, Ambitle Island, Papua New Guinea. Chem. Geol., 162 (1), 15–31. Pokrovski, G., Gout, R., Schott, J., Zotov, A., & Harrichoury, J.-C. (1996). Thermodynamic properties and stoichiometry of As (III) hydroxide complexes at hydrothermal conditions. Geochim. Cosmochim. Acta, 60 (5), 737–749. Pokrovski, G. S., Kara, S., & Roux, J. (2002). Stability and solubility of arsenopyrite, FeAsS, in crustal fluids. Geochim. Cosmochim. Acta, 66 (13), 2361–2378. Rancourt, D. G., Fortin, D., Pichler, T., Thibault, P. J., Lamarche, G., Morris, R. V., & Mercier, P. H. J. (2001). Mineralogy of a natural As-rich hydrous ferric oxide coprecipitate formed by mixing of hydrothermal fluid and seawater: Implications regarding surface complexation and color banding in ferrihydrite deposits. Am. Mineral., 86 (7–8), 834–851. Reid, K. D., Goff, F., & Counce, D. A. (2003). Arsenic concentration and mass flow rate in natural waters of the Valles Caldera and Jemez Mountains Region, New Mexico. New Mexico Geol., 25 (3), 75–81. Santini, J. M., Sly, L. I., Schnagl, R. D., & Macy, J. M. (2000). A new chemolithoautotrophic arsenite-oxidizing bacterium isolated from a gold mine: Phylogenetic, physiological, and preliminary biochemical studies. Appl. Environ. Microbiol., 66 (1), 92–97. Schwenzer, S. P., Tommaseo, C. E., Kersten, M., & Kirnbauer, T. (2001). Speciation and oxidation kinetics of arsenic in the thermal springs of Wiesbaden Spa, Germany. Anal. Bioanal. Chem., 371 (7), 927–933. Scott, S. D. (1983). Chemical behaviour of sphalerite and arsenopyrite in hydrothermal and metamorphic environments. Mineral. Mag., 47 (4), 427–435. Sherman, D. M., & Randall, S. R. (2003). Surface complexation of arsenic(V) to iron(III) (hydr)oxides: Structural mechanism from ab initio molecular geometries and EXAFS spectroscopy. Geochim. Cosmochim. Acta, 67 (22), 4223–4230. Simon, G., Huang, H., Penner-Hahn, J. E., Kesler, S. E., & Kao, L. S. (1999). Oxidation state of gold and arsenic in gold-bearing arsenian pyrite. Am. Mineral., 84 (7–8), 1071–1079.

440

M. A. Alsina et al.

Stauffer, R. E., & Thompson, J. M. (1984). Arsenic and antimony in geothermal waters of Yellowstone National Park, Wyoming, USA. Geochim. Cosmochim. Acta, 48 (12), 2547–2561. Tazaki, K., Rafiqul, I. A., Nagai, K., & Kurihara, T. (2003). FeAs2 biomineralization on encrusted bacteria in hot springs: An ecological role of symbiotic bacteria. Can. J. Earth Sci., 40 (11), 1725–1738. Waychunas, G. A., Rea, B. A., Fuller, C. C., & Davis, J. A. (1993). Surface chemistry of ferrihydrite: Part 1. EXAFS studies of the geometry of coprecipitated and adsorbed arsenate. Geochim. Cosmochim. Acta, 57 (10), 2251–2269. Webb, S. M., Gaillard, J. F., Ma, L. Q., & Tu, C. (2003). XAS speciation of arsenic in a hyper-accumulating fern. Environ. Sci. Technol., 37 (4), 754–760. Webster, J. G. (1990). The solubility of As2S3 and speciation of As in dilute and sulphide-bearing fluids at 25 and 901C. Geochim. Cosmochim. Acta, 54 (4), 1009–1017. Webster, J. G., & Nordstrom, D. K. (2003). Geothermal arsenic. In: A. H. Welch, & K. G. Stollenwerk (Eds). Arsenic in Ground Water: Geochemistry and Occurrence. Kluwer Academic Publishers, Boston, pp. 101–125. White, D. E. (1981). Active geothermal systems and hydrothermal ore deposits. 75th Anniversary Volume (1905–1980), Econ. Geol., 392–423. Wilkie, J. A., & Hering, J. G. (1998). Rapid oxidation of geothermal arsenic(III) in streamwaters of the eastern Sierra Nevada. Environ. Sci. Technol., 32 (5), 657–662. Williams-Jones, A. E., & Heinrich, C. A. (2005). 100th anniversary special paper: Vapor transport of metals and the formation of magmatic-hydrothermal ore deposits. Econ. Geol., 100 (7), 1287–1312.


Chapter 16

Reactive Transport and Residence Times in Unsaturated Fractured Rocks from Field-Scale Experiments Eric Pili1,, Sarah Bureau1,5, Fre´de´ric Perrier2, Delphine Patriarche1,6, Laurent Charlet3, Pierre M. Adler4 and Patrick Richon1 1

Commissariat a` l’Energie Atomique, De´partement Analyse Surveillance Environnement, CEA/DASE/SRCE, BP 12, 91680 Bruye`res-le-Chaˆtel, France 2 ´ Equipe de Geómagne´tisme, Universite´ Paris 7 Denis Diderot, Institut de Physique du Globe de Paris, 4 place Jussieu, 75252 Paris Cedex 05, France 3 Laboratoire de Geóphysique Interne et Tectonophysique, Maison des Geósciences, BP 53, 38041 Grenoble Cedex 9, France 4 UPMC-Sisyphe, 4 place Jussieu, 75252 Paris Cedex 05, France 5 Present address: Institute of Isotope Geochemistry and Mineral Resources, Department of Earth Sciences, ETH Zentrum, Claussiusstrasse 25, NW D84, 8092 Zurich, Switzerland 6 Present address: Gaz de France, 361 avenue du Pre´sident Wilson, 93211 Cedex Saint-Denis La Plaine, France

ABSTRACT Reactive transport mechanisms and transit times in unsaturated fractured crystalline rocks have been determined from field-scale observations using highresolution monitoring over a three-year period. In the Roselend natural laboratory (French Alps), a tunnel provided access to the heart of the fractured-rock unsaturated zone, at 55 m depth, where two sampling sites provided access to dripwater samples from zones with contrasting contributions from matrix and fracture flow. Solute concentrations and fluid fluxes were determined, taking advantage of two events during which water with elevated concentrations of chloride (Cl) and other solutes infiltrated into the fractured medium. In one event, approximately 500,000 l of water with high NaCl concentrations flooded an area of 10–100 m2 above the sampling sites in the tunnel. The other event involved sampling after a Cl-rich rainfall event, which is unusual at this location. Elevated Corresponding author. Tel.:+33 1 69 26 50 11; Fax: +33 1 69 26 70 65;

E-mail: [email protected] (E. Pili).

442

E. Pili et al.

Na concentrations from the NaCl spill triggered cation exchange reactions, which apparently occurred mainly in the soil above the bedrock, releasing Ca, Mg, and K. Concentrations of Cl and the exchanged cations during transport through the fractured rock were controlled primarily by dilution, although Na–Mg exchange occurred to a limited extent. Chloride concentrations significantly above background were observed one day after the beginning of the NaCl spill. Average travel times estimated from moment calculations of breakthrough curves from the two events ranged from 25 to 100 days at the sampling site dominated by flow through major fractures and 200–300 days at the site dominated by matrix or sub-order fracture flow. Comparison with results of flow-model simulations conducted using reconstructed fracture networks indicated that rapid flow occurs at nearsaturation conditions in the fractures and that slower flow rates occurred at lower degrees of saturation. Thus, it is important to consider both variable saturation and fracture versus matrix flow in order to understand fluid fluxes and reactive transport in unsaturated, fractured media. This study illustrates how long-term, high-resolution monitoring of solute concentrations and fluid fluxes can provide valuable insight into flow and transport in these complex systems.

16.1. Introduction Adsorption to soils and rocks is a key complex process that determines both the natural baseline composition of groundwater and the fate of contaminants in the environment. Despite great improvements in the understanding of these processes (e.g., Jenne, 1998), new approaches to validate modeling at the field scale are still needed, especially in the unsaturated zone. Indeed, solute transport in fractured rocks under unsaturated conditions has been the focus of relatively few studies despite its important role in generating chemical anomalies in groundwater and other phenomena (e.g., Nativ et al., 1995; Pili et al., 2004). Evans et al. (2001) and a panel (National Research Council, 2001) provided a review of concepts for flow through unsaturated fractured media. Although the use of tracers is very useful in assessing the behavior of fractured porous media, solute transport has been the subject of many fewer studies than water flow. Su et al. (1999, 2001) reported laboratory work and modeling on the 20–30 cm scale using an epoxy replica of a Stripa-granite fracture. Dahan et al. (2001) described field tracer experiments in fractures on a vertical scale of 1 m in chalk. Faybishenko et al. (2001) reported on field tracer experiments on the 1-m, 10-m, and 50-m scales in fractured basalts. From their field experiments on the 20-m scale in the tuff of Yucca Mountain, where water with added tracers was injected into a single, unsaturated fault, Salve et al. (2004) emphasized the importance of

Reactive Transport and Residence Times in Unsaturated Fractured Rocks

443

understanding fracture networks. Reactive transport and water–rock interactions in the fractured vadose zone are rarely addressed. The present work is focused on a shallow but large (0–55 m depth) natural fractured system in crystalline rocks with intense and highly variable unsaturated flow and transport phenomena due to both natural and artificial causes. As part of a multidisciplinary research project aiming at understanding the impact of mechanical, meteorological, hydrogeological, and geochemical stimuli on transport in unsaturated fractured media (Provost et al., 2004), an experiment was set up based on long-term, high-resolution monitoring of a variety of physical and chemical parameters. This research program is performed in a tunnel that provides access to variably saturated fractures 55 m below the soil-fractured-rock interface. Fracture occurrences and dripwater fluxes were measured along the entire tunnel. For more than three years, water chemistry was monitored with sampling times varying from 36 h to 4 days, and flow rates were determined hourly. Major and trace elements were analyzed in water samples representative of different contributions of matrix porosity and fracture networks. The goal of the present paper is to derive a conceptual model for the transport of ideal (which here means conservative and non-reactive) and reactive tracers in unsaturated fractured rocks by using field-scale experiments. The structure of the system, a description of various events that impact the modeled system, and the principal hydrodynamic and geochemical processes are presented. The study focused on the system hydrogeochemical response in two distinct settings involving ideal and reactive tracers infiltrating from the ground surface 55 m above the tunnel: (i) natural chloride brought by rainfall events monitored during year 2002; and (ii) flooding with ca. 500,000 l of a sodium chloride solution starting with a salinity close to that of seawater, infiltrated in spring 2003 and monitored for more than two years. We describe how water moves through the system and what reactions control the major cation concentrations. Geochemical reactions, mass balance and transit times are determined experimentally. Transit times are also assessed using simple modeling where permeabilities are calculated from reconstructed fractured media.

16.2. Setting The study site is located in the French Alps, 25 km southwest from Mont Blanc, close to and above the artificial Roselend Lake on its western shore. The dead-end horizontal tunnel, at an altitude of 1,576 m, is 128 m long and

444

E. Pili et al.

ca. 2.4 m in diameter. The tunnel is hosted in gneiss and micaschists and capped with the same rocks with increasing thickness from 7 m at the entrance to 55 m at its closed end (Fig. 16.1a). In the tunnel, the average air temperature is 6.870.21C (1s) over the whole year, and relative humidity remains near 100%. The local meteorology is characterized by contrasting precipitation regimes with alternating snow (winter), rain (spring and fall), and drought (summer) periods. The section of the tunnel between 110 m and 128 m from entrance (Fig. 16.1) was chosen for dripwater monitoring for the following reasons. This section presents the largest thickness of rock overburden (55 m). Over small distances, water fluxes are highly variable due the heterogeneity of major fractures but lithological heterogeneity is limited. Moreover, the corresponding ground surface (Fig. 16.2) is nearly flat, at the border of an abandoned quarry, and easily accessible. These features are favorable for conducting tracer experiments. The flat area corresponds to the top of an embankment on which a long-abandoned dirt road had been built. Dispersed grass grows on the dirt road and the quarry, while a light mixed forest of broad-leaved trees and conifers grows on the hillslopes. As determined from seven shallow boreholes (I2–I8, Fig. 16.2) and four trenches dug down to the top of the bedrock, the embankment is less than 1 m thick, including a less than 10 cm-thick layer of soil developed at the top. The embankment is composed of gneiss blocks (size>10 cm, 20 wt%) and the same lithology in the form of gravels (2 cmosizeo10 cm, 40 wt%), a medium fraction (2 mmosizeo2 cm, 26 wt%), and a finer fraction (sizeo2 mm 14 wt%). Two trenches have been used for systematic sampling along a vertical profile (Table 16.1). Rocks from the trenches were selected manually and representative hand specimens were sampled for analysis. Fractures were carefully mapped along the tunnel walls. Their features are statistically similar to external outcrops within kilometers on the same massif. Some fractures are filled in part or totally with quartz (7chlorite7 Ca-Mg-Fe carbonate7pyrite) while others show no secondary mineral. The extent of fracturing along the tunnel is heterogeneous, with 21 large fractures entirely intersecting the tunnel (Fig. 16.1a), and 172 small ones with only partial intersection. Fracture densities rs and rL for small and large fractures, respectively, were calculated and amount to 6.1 102 and 4 103 fractures/m3 (Patriarche et al., 2008). Percolation through the massif is evidenced by water dripping from the tunnel roof with varying fluxes; null water fluxes coincide with the lack of large fractures (Fig. 16.1a,b). Seepage fluxes were measured along the tunnel when snowmelt was ending in April 2003. These fluxes were measured by counting drop impacts within 2-min intervals using a 1.4 m wide by 2.7 m long plastic sheet moved along


445

Figure 16.1: (a) Cross Section of the Roselend Tunnel Along its N 651E Axis, Showing the Local Topography and the Location of the Two Experimental Zones (HGZ and FZ1) Inside the Tunnel where Dripping Waters are Collected. Orientations of Major Fractures (Full Intersections) Observed in the Tunnel are Superimposed. The Two Types of Tracer Infiltration Used in this Study are Illustrated: Limited Surface Area for NaCl Spilled from Tank (Thick Arrow) and Extensive Surface Area for Rain-Derived Chloride (Array of Gray Arrows). (b) Profile of Dripwater Flux Along the Tunnel on April 25, 2003. (c) Profile of Specific Conductivity Measured on November 24, 2003, after NaCl Flooding (May 27, 2003). ‘‘Error Bars’’ in Profiles Refer to Sections in the Tunnel from where Drops were Collected. Most of these Sections have the Same Size or are Smaller than the Symbols.

446

E. Pili et al.

Figure 16.2: Map View of the Area over Which Tracer Infiltration Occurred with the Tunnel Lying 55 m Underneath, Showing the Locations of the Two Dripwater Monitoring Zones, HGZ and FZ1, at 125 m and 115 m from the Tunnel Entrance, Respectively. Stars Labeled I2–I8 are Shallow Boreholes. The Dashed Star Shows where NaCl-Rich Water Fell on the Ground During Overflow of the Tank. The Dashed Line Delineates the Zone Flooded when the Tank was Drained through its Three Bottom Holes. the axis of the tunnel and a mean drop volume of 0.165 ml. Numerical simulations indicate that seepage occurs along networks combining large and small fractures but not in networks made exclusively of small fractures (Patriarche et al., 2008). Thus, the rare large fractures in the system are necessary for percolation to occur. Fracture networks within a matrix with limited porosity form at least two distinct pore spaces (Pili et al., 2004). This could also be viewed as a multiple continuum of fracture networks within an impermeable matrix. In what follows, we will use the concept of fractures and matrix, although what we call matrix could be a sub-order fracture network.

16.3. Experimental Setup, Tracer Introduction and Recovery Two experiments were used to study reactive transport and solute residence times; each involved tracers infiltrating from the ground surface above the

Nature of samples

Sample type (depth for profile in trench)

Soil (sievedo2 mm)

Total CEC

CaCEC

MgCEC

K-CEC

NaCEC

Trench T2 0–5 cm Trench T4 0–15 cm

12.0 10.2

8.6 6.2

1.5 1.5

1.5 1.2

0.4 0.4

4.84 5.77

6.1 5.7

Quartz, plagioclase, microcline, mica, chlorite, sepiolite.

Embankment, (size fractiono2 mm)

Trench Trench Trench Trench Trench

12.2 9.0 8.0 7.7 15.4

8.8 2.4 4.1 6.2 13.5

0.7 0.6 1.3 2.0 1.5

2.1 0.7 1.2 0.8 1.1

0.3 0.3 0.4 0.4 0.3

6.68 15.34 9.12 5.78 2.95

5.9 5.3 5.5 6.2 6.3

Quartz, plagioclase, microcline, mica, chlorite, sepiolite.

Blocks of rocks from embankment (crushedo200 mm)

Dark gneiss Light gneiss Weathered dark gneiss Weathered dark gneiss with shear bands Mica-rich layer in gneiss

2.9 2.1 1.7 2.3

2.5 0.4 1.0 0.3

1.1 0.8 0.4 1.1

0.6 – – 0.2

0.8 0.3 0.1 0.4

o0.10 o0.10 0.32 3.92

– – – –

Quartz, plagioclase, microcline, mica, chlorite7pyrite7 Fe-oxides.

2.6

0.7

1.1

0.1

0.2

1.46

T2 T2 T2 T4 T4

5–45 cm 50–95 cm 95–100 cm 15–45 cm 45–65 cm

Specific surface area

pH

– 2

Note: Cation Exchange Capacity and exchangeable cations in millimoles of charge per 100 g. Specific surface area in m /g.

Main mineralogy


Table 16.1: Main Geochemical Properties of Media.

447

448

E. Pili et al.

Table 16.2: Characteristics of Infiltration in the Two Tracer Experiments. Infiltration Occupied Max water type & area volumea nature Flooding (NaCl spill) Rain (4 Clrich rainfall events) a

Limited (10– 100 m2) Extensive (several km2)

Max water flux (L/m2/ d)

Max Cl concentration (mmol/l)

Duration & dates of infiltration

500750 m3

3,000

540

8 L m2

5

0.447

14-day period (May 27–June 10, 2003) 4 individual days (April 7, 11, 14, 26, 2002)

Unknown infiltrated volumes due to unknown run-off or catchment surface areas.

tunnel (Fig. 16.1a). The main characteristics of the experiments are summarized in Table 16.2. The concentrations of tracers were measured in water samples collected in sampling devices in the tunnel, 55 m below the source. In order to highlight the main locations where breakthrough of tracers occurred in the tunnel, a profile of conductivity in dripwaters was measured (Fig. 16.1c). Breakthrough of tracers was vertically aligned with injection locations 55 m above. Thus, these experiments can be considered to be giant 1-D vertical column experiments conducted under unsaturated conditions.

16.3.1. Meteorological and Meteoric Infiltration Data Rainfall, snow pack height, air temperature and humidity, solar irradiation, wind speed, and atmospheric pressure data were collected at local weather stations. For periods during which precipitation occurs as rainfall, daily infiltration (Fig. 16.3a) is calculated in the following manner. The hourly potential evapotranspiration is computed according to the Penman– Monteith method for grass-covered ground (Allen et al., 1998). Then the daily water balance (Thornthwaite and Mather, 1957) is obtained assuming a null run-off and considering a soil available water capacity of 75 mm. For periods with snow, snowmelt flux at the snow pack base is calculated by using the CROCUS model, which simulates the snow pack height and density by the mean of energy and mass budgets (Brun et al., 1992), based on meteorological information. The model takes into account compaction, sublimation and snow metamorphosis. Snowmelt flux is calibrated by reproducing the daily snow pack height given by the meteorological station. Then, infiltration is calculated as above for rainfall.


449

Figure 16.3: (a) Infiltration and Snow Pack Height from Local Meteorological Stations. (b) Dripwater Fluxes. HGZ Data Smoothed with a Moving Window Using a Time Step of 48 h. (c)–(e) Concentrations of Cl, Ca, and Mg in Dripwater over Time. Dashed Line: Beginning of Flooding with NaCl Solution on May 27, 2003. Dotted Line: Return of Infiltration on September 1, 2003 after Drought. Inset in (c): Calculated Input Concentration (mmol /l) of Chloride during Flooding.

450

E. Pili et al.

16.3.2. Selection and Instrumentation of Two Dripwater Sampling Sites Two zones were selected near the end section of the tunnel (Figs. 16.1a and 16.2) within the same lithological unit composed of gneissic rocks (Pili et al., 2004). They represent different contexts in terms of structures and water fluxes. In our reference host gneiss zone (HGZ, 123–125 m from the tunnel entrance), the dripping water fluxes are low (2 ml/h/m2). In contrast, a highly fractured zone (FZ1, 114–116 m from the tunnel entrance) has a significantly higher water flux (250 ml/h/m2). Both zones were equipped in June 2002 with a 2.5 m2 polymer sheet hung near the roof in such a way that dripping water is collected at one point. For continuous flow-rate measurements, water is collected separately from the HGZ and FZ1 plastic sheets into rain gauges. FZ1 provides water mostly contributed from the fractures, with hydrogeological and geochemical characteristics of the fracture system. HGZ provides water with geochemical characteristics dominated by the host rock (Pili et al., 2004). However, water dynamics in HGZ is still partially controlled by fractures. Flow rates measured in rain gauges are obtained from the counts recorded with a time step of 1 h. They are given as daily water fluxes for each monitoring zone in Fig. 16.3b. Data presented for HGZ are averaged with a moving window with a time step of 48 h. Water fluxes are used in order to describe the hydrodynamic state of the sampled fractured networks and to establish mass budgets and calculate volume-weighted average concentrations. Rain gauge calibrations are performed once a year. All the data, including precipitation data from the outdoor weather station, are recorded on the same DELTA-T DL-2 data logger for synchronous time series. After leaving the rain gauges, water flows to a customized ISCO 3700 automated sampler whose pump was bypassed. Each drop of water flows from the ceiling of the tunnel to the sample bottle by gravity through the entire collecting system in less than 1 h for FZ1 and 10 h for HGZ. In each sampling zone, twenty-four 125 ml plastic bottles are filled one after the other with time steps up to 4 days. Each bottle freely overflows and stays open until being manually capped with gloved hands when the samplers are serviced. Residence times of water samples in the sampling device and bottles for FZ1 and HGZ are of the order of 10 min and 20 h, respectively. Indeed, as shown by an analogue experiment conducted in the laboratory using tracer-tagged solutions, the evolution of concentrations in the sampling device and in each bottle can be described using a box-model with a residence time given by the total volume of water stored in the device divided by the flow rate. During chloride breakthrough, the sampling time step was set to 36 h and the samplers


451

were serviced at least every 30 days. Each new set of bottles was rinsed with acetone then with MilliQ water, bathed for 48 h, rinsed again in MilliQ water, dried, and labeled. Precipitation of calcium carbonate, with purity greater than 99%, is observed in the sampling device (plastic sheets, rain gauges).

16.3.3. Analytical Methods Chloride concentrations were determined by ion chromatography (DIONEX DX-500); the analytical uncertainty is 0.5%. Cation concentrations (Na, K, Mg, and Ca) were determined by ICP-OES (PERKIN ELMER OPTIMA 3300DV); the analytical uncertainty is 2%. Analytical instruments were calibrated daily; standards were renewed weekly. Concentrations of Cl, Na, Ca, Mg, and K in dripwater are presented in Figs. 16.3–16.5. Cation exchange capacity and exchangeable cation concentrations were determined at BRGM (Orleáns, France) following standard method NF X 31-130 on soil samples sieved to 2 mm and rock samples ground to 200 mm. Specific surface areas were determined by the N2-BET method at BRGM. Soil pH was determined following standard method NF ISO 10390 with pure water. Mineralogy was determined on soil samples using X-ray diffraction (at BRGM) and on thin sections prepared from rock samples using a petrographic microscope. Results are presented in Table 16.1.

16.3.4. Sodium Chloride Flooding Event An area that we were preparing for a tracer test experienced an accidental spill of large volume of water with elevated concentrations of NaCl. Chloride is an ideal tracer while Na can undergo ion exchange reactions (e.g., Appelo and Postma, 1999). This spill was used to study residence times and reactive transport. A farmer placed a used tank at a location shown in Fig. 16.2. The tank was intended to hold water temporarily before being piped further down to a watering trough. However, the 30 m3 tank had previously been used to truck NaCl (purity greater than 99.99%) and was still coated with salt. The tank started to overflow on May 27, 2003 at a rate of ca. 30 m3 per day. Overflow mainly occurred in a ca. 10 m2 area at the southwest of the tank (Fig. 16.2), killing plants in this area. Overflowing water either infiltrated or ran off. The tank was drained completely on June 10, 2003 by opening the three drain holes (Fig. 16.2). A total volume of 500750 m3 of saline water was released. On draining, ca. 30 m3 of water flooded an area of

452

E. Pili et al.

Figure 16.4: (a) Breakthrough of Chloride in FZ1 and HGZ Dripwaters after NaCl Spill. Box-Model Calculated Dilution for FZ1 with a Residence Time of 85 Days Fits the Data for the First Part of Breakthrough. A Residence Time of 200 Days and a Lower Initial Concentration are Needed to Account for the Second Part of Breakthrough, Both in FZ1 and HGZ (Solid Line). The Quality of the Fit Between the Modeled Curves and the Data is Evaluated by the Misfit Factor Fm, the Smaller the Better (Eq. (16.3)). (b)–(e) Breakthrough of Na, K, Ca, and Mg (mmol /l) Measured in FZ1 Dripwaters. Dripwater Concentrations for Na and K in HGZ are Also Given Though no Breakthrough was Detected. As for Cl in FZ1, a Two-Part Box-Model Dilution with Residence Times of 85 Days and 200 Days Fits the Data for Na. Dilution for Ca, Mg, and K can be Modeled with a Single Residence Time of 85 Days.


453

Figure 16.5: Temporal Evolution of Chloride Concentrations Recorded in Rainwater at Le Casset (See Text), and in FZ1 and HGZ Dripwaters. The Dashed and Solid Lines Depict the Background Level 71s, Respectively. Dots for the Calculated Times of Center of Mass are Set at the VolumeWeighted Centers of Mass of the Peaks. ca. 100 m2, (Fig. 16.2). Elevated concentrations of NaCl were detected in only three of our shallow boreholes (I2, I3, I4), corroborating the indicated area impacted by the spill. A mass of NaCl, estimated to be 50 kg from a description provided by the farmer, fell on the ground from the three bottom holes, and completely dissolved in a matter of hours. An increase in water flux at FZ1 was recorded over approximately 1 month (Fig. 16.3b). Concentrations in Cl, Na, Ca, Mg, and K in dripwater in response to the NaCl spill are given in Figs. 16.3 and 16.4. Based on the assumed 50 m2 NaCl-contaminated region of the tank, an assumed NaCl thickness of 1 cm, and a bulk density of 2 g/cm3 for the NaCl we estimate that the mass of NaCl in the tank was 1,000 kg. This value is consistent with known quantities of salt that cannot be removed from such tanks indicated by the manufacturer. Assuming 50 kg of undissolved NaCl fell from the tank during draining, 950 kg of NaCl were dissolved during overflow of the tank. The salinity of water at the beginning of overflow may have reached 32 g/l ([Cl] ¼ 0.54 mol/l). From the above information, the NaCl concentration in overflowing water was calculated using a box model (Fig. 16.3c) (e.g., Albare`de, 1995): t CðtÞ ¼ ðC init C input Þ exp þ C input t

(16.1)

454

E. Pili et al.

where C(t) is the concentration measured at time t in water overflowing the tank; Cinput the NaCl concentration in water flowing into the tank from the nearby reservoir (0.05 mmol/l); Cinit the NaCl concentration in the tank just before the beginning of overflow (0.54 mol/l). t is the residence time in the tank: t¼

V Q

(16.2)

with V, the volume of water in the tank (30 m3), and Q the flow rate (30 m3/d). The box model indicates that the NaCl concentration in overflowing water decreased to that in the incoming water within 12 days (Fig. 16.3c). Thus, the period during which elevated concentrations of NaCl infiltrated was short compared to breakthrough period. The input of NaCl resulting from draining the tank 14 days later can be neglected.

16.3.5. Elevated-Chloride in Rainfall Events Chloride in precipitation at Roselend has only been determined sporadically, with concentrations in the range 4–15 mmol/l in rain and 5–25 mmol/l in snow. We benefited from data collected at a monitoring station at Le Casset, a similar alpine area 80 km south from Roselend, at an altitude of 1,750 m (http://www.emep.int and http://www.nilu.no/projects/ccc/emepdata.html). Chloride concentrations in most precipitation events in 2002 at Le Casset were lower than 22 mmol/l (average is 4.7 mmol/l). However, there were four rainfall events in April 2002 during which chloride concentrations reached 447 mmol/l (Fig. 16.5). These four rainfall events, which accounted for o8 mm of the 850 mm of precipitation recorded at Le Casset during 2002, accounted for 25% of the total rainfall-derived Cl mass. Back-tracking meteorological modeling at the scale of Europe (A. Bellivier, personally written communication, June 2006) predicts that such rare and spike-like events are likely due to air masses derived from the English Channel, 700 km northwest from Roselend. Elevated chloride concentrations in these air masses are likely caused by industrial areas of the UK and Normandy. These air masses further split over the Alps and deliver Cl-bearing precipitation to Roselend and Le Casset with only few-hour time lag. Chloride concentrations in dripwater from HGZ and FZ1 collected in the Roselend tunnel from mid-June 2002 to mid-February 2003 show broad breakthrough curves (Fig. 16.5). Both maximum Cl concentration and water flux were greater at large-fracture-dominated FZ1 than in the matrixdominated HGZ zone. A background Cl concentrations of 0.1170.01 (2s)


455

mmol/l was calculated from the volume-weighted average of the first 4 samples collected at FZ1 and the first 13 samples collected at HGZ (Fig. 16.5). From mid-December 2002 to February 2003, concentrations in FZ1 were lower than the estimated background concentration. This is attributed to dilution by meteoric water infiltrating fractures during highflow periods (Pili et al., 2004). Presumably, these low concentrations are compensated by higher ones over the entire water year. The volume-weighted average Cl concentrations in dripwater from HGZ and FZ1 during the sampling period are 137 and 131 mmol/l, respectively. These weightedaverage Cl concentrations show a striking similarity considering that concentrations of other solutes differ by a factor 1.5–2 between FZ1 and HGZ (Table 16.3 and Pili et al., 2004). The observed similarity in Cl concentrations at FZ1 and HGZ during this period is consistent with a Cl source located above the rock system, for which rain comprises the principal contribution. Given that the Roselend and Le Casset stations have similar geographical and meteorological features, it is reasonable to assume that Cl concentrations in rain are likely to be similar and to show synchronous rare spike-like events. Therefore, the increase in Cl concentrations observed at Roselend can be attributed to rare rainfall events such as those monitored at Le Casset. These high-Cl rain events provide estimates of Cl residence times in FZ1 and HGZ owing to uniform aerial recharge as compared to those obtained from recharge of NaCl-spiked water over a small area (Fig. 16.1a).

16.4. Reactive Transport 16.4.1. Identification and Quantification of Cation Exchange Infiltration of water from the NaCl spill triggered cation exchange reactions. This is demonstrated by the breakthrough of elevated concentrations of Ca, Mg, and K, and the attenuated concentrations of Na (compared to Cl) during Cl breakthrough (Figs. 16.3 and 16.4). No breakthrough of elevated concentrations of SO2 4 , HCO3 , or H4SiO4 was observed (data not shown). In addition, the cumulative amount of charge associated with Na+, Ca2+, Mg2+, and K+ breakthrough (determined from concentrations and water fluxes measured in FZ1) balances the cumulative charge associated with Cl breakthrough (Fig. 16.6a). The mass balance illustrated in Fig. 16.6a involved more than 8,500 l of dripwater flowing through the 55 m of fractured rock and soil over a period of two years (May 2003–May 2005). This computation assumes a constant background concentration of these ions, but the choice of concentrations has a negligible effect on the mass balance.

Tracer Tfirst (first Tmax (maximal Tcenter appearance) concentration) (center of mass)

tdilution

Background concentration (mmol/l)

Peak/ Recovered Background mass fraction concentration (%)

FZ1 (fracture-dominated zone) Rain Cl 67 Flooding Cl 171

97 27

108 88–118a

– 8575 – 200720b

0.1170.01 0.1270.06

2.7 139

Na

171

27

99–154a

8575 – 200720b

0.0670.01

133

Ca Mg K

673 673 673

27 21 24

65 91 81

8575 8575 8575

1.3770.15 0.4970.03 0.1070.01

2.9 2.2 2.7

HGZ (matrix-dominated zone) Rain Cl 109 Flooding Cl 1173

147 99

184 168–250a

– 200720

0.1170.01 0.1370.02

1.7 18

Na

1773

99

150–300a

–

0.1470.01

4.7

Ca Mg K

– – –

– 106 –

– 179–257a –

– – –

1.3570.09 0.7370.03 0.0570.01

– 1.5 –

– (7.571.5)102 – (8.271.8)102a (2.970.6)102 – (3.370.7)102a – – – – (1774)105 – (2375)105a (1.370.3)105 – (2.470.5)105a – – –

Note: Uncertainties of residence times are 72 days, unless otherwise stated. a Calculated from May 2003 to May 2004 and from May 2003 to May 2005, respectively. The first value better represents fracture flow; the second better represents matrix flow. b Determined from June 2003 to January 2004 and from January 2004 to May 2005, respectively.

E. Pili et al.

Infiltration

456

Table 16.3: Synthesis of Residence Times (in Days) and Other Breakthrough Parameters Determined Experimentally in Two Observation Zones.


457

For breakthrough at FZ1, cation exchange mainly occurred in the soil and the embankment as compared to the fractured medium underneath. The CEC values of soil and embankment samples greatly exceed those of rock samples (Table 16.1), even though we selected weathered rocks and the samples were crushed to less than 200 mm. Clay minerals, the most likely exchangers, were identified in the soil and the embankment samples but were not detected in samples of the fractured rock. The fractions of total charge accounted for by Ca, Mg, and K in FZ1 dripwater samples were 71%, 26%, and 3%, respectively, and were nearly constant (Fig. 16.6a). These fractions are similar to the fractions of exchangeable cations in soil and embankment samples (75%, 12%, and 13%, respectively, Table 16.1). Calcium concentrations in dripwater samples are likely underestimated due to CaCO3 precipitation in the sampling device. The discrepancy between the fraction of charge in dripwater samples and exchangers in the soils might have resulted from K uptake by plants. The greater contribution of Mg to the charge in dripwater samples as compared to exchangers in the soil may indicate additional input of Mg during transport through the fractured rock. A calculation of the theoretical composition of the exchanger from the dripwater concentrations using the literature Gaines–Thomas coefficients relative to Na+ (Appelo and Postma, 1999) gives fractions for Ca2+ and Mg2+ of 80% and 10%, respectively, similar to those measured in soil and embankment samples (Table 16.1), while the calculated fraction for K+ is 10 times lower than that measured in the soil and embankment samples. No retardation of Ca, Mg, and K was observed during breakthrough at FZ1 (Fig. 16.4), as might be expected if cation exchange reactions occurred all along the flow path. From these arguments, we infer that cation exchange occurred mainly in the soil and embankment above the fractured rocks in the FZ1 zone. Some ion exchange in the rock matrix during transport through the HGZ zone may have occurred. During breakthrough at HGZ, the Mg signal was similar to that of Cl (Fig. 16.3c and e) and Na was strongly attenuated (Fig. 16.4b). The ratio of the recovered masses of Cl and Na in FZ1 samples is 2.5 while that in HGZ samples is 10 (Table 16.3). This disparity might have resulted from cation exchange of Na for Mg during transport through the rock matrix prior to breakthrough at HGZ. A hypothesis whereby both Cl and Na invaded the matrix porosity and Na reacted inside the rock to produce exchanged Mg, would account for the observed breakthrough. The possible exchangers in the rock are unknown. Apart from Mg that might be stored along grain boundaries, the main Mg-bearing minerals in the rocks are chlorite and carbonates, although the presence of a small quantity of clay minerals in the weathered granite cannot be excluded. The solution charge balance calculation is presented in Fig. 16.6b for HGZ. At HGZ the

458

E. Pili et al.

Figure 16.6: Cumulative Amounts of Cl, Na+, Ca2+, Mg2+, and K+ (Expressed in Moles or Millimoles of Charge) Determined from Concentrations and Water Fluxes Measured in (a) FZ1 and (b) HGZ. In FZ1, the Sum of Cations Balances the Amount of Cl within 5% of Uncertainty (Due to Analytical Methods and Determination of the Baseline) Demonstrating Cation Exchange in Response to NaCl Spill. Exchange Reactions Occurred in the Soil and Embankment Above the Bedrock (See Text). In HGZ, Despite Excess Ca2+ Brought by a Recharge Event, the Cations Eventually Balance Cl. A Higher Fraction of Mg2+ in HGZ Compared to FZ1 Might be Attributed to Cations Released from the Host Rock. Computation from May 2003 to May 2005 Assuming a Constant Baseline Determined as the Average of Concentrations from Mid-April to May 26, 2003.


459

cumulative mass of solutes that broke through is three orders of magnitude lower than at FZ1 (Fig. 16.6a), primarily because the total dripwater volume was two orders of magnitude lower. Differences between the contributions of cations, particularly Mg, to the charge balance at HGZ and FZ1 (Fig. 16.6) support the hypothesis that additional exchange of Na for Mg occurred along the flow paths in the HGZ zone. The total mass of exchanged cations (Ca+Mg+K) that broke through at FZ1 represented 7.5 moles of charge (Fig. 16.6a), while that at HGZ was 55 mmol (Fig. 16.6b). Thus, breakthrough at HGZ comprised a negligible fraction of the total mass solute breakthrough resulting from the NaCl spill. Assuming that cation exchange occurred primarily on the o2 mm fraction in the soil and embankment, with a measured total CEC of ca. 0.1 molC/kg (Table 16.1), 75 kg of exchanger would be required. As the o2 mm fraction represents 100% and 14% of the soil and embankment, respectively, the exchanged cations could have come from 75 kg or 500 kg of soil and embankment, respectively. Given a bulk density of 2,000 kg/m3 for these materials, this represents a volume of 400 l of soil or 2,500 l of embankment. Assuming that the spill had a footprint of 10 m2, a 4-cm-thick layer of soil or a 25-cm-thick layer of embankment would account for the observed cation exchange. However, this calculation assumes that the NaCl-spiked water flowed vertically when, in all likelihood, some lateral spreading occurred during infiltration. Also, these calculations only account for the portion of the NaCl-spiked water that discharged in the tunnel. It will be shown later that only 0.08% of that water discharged in the tunnel (Table 16.3). If we extrapolate these numbers to the total mass of 1,000 kg of NaCl, then 500 m3 of soil or 3,600 m3 of embankment would be required to account for the observed extent of cation exchange. This suggests a much larger spreading area for the NaCl spill than suggested by the footprint of the spill at the ground surface (Fig. 16.2). Elevated conductivity values for dripwater collected in the tunnel starting at 95 m from the tunnel entrance (Fig. 16.1) supports the assumption of lateral spreading of the NaCl plume. 16.4.2. Modeling Transport Associated with Cation Exchange No retardation was observed for the exchanged cations Ca, Mg, K (Figs. 16.3d,e and 16.4c–e); ratios of their concentrations remained nearly constant during breakthrough (Fig. 16.6). This suggests that cation exchange occurred nearly instantaneously in the soil and embankment prior to entering the fractured medium. As a result, the decrease in concentrations for Cl, Na, and the exchanged cations observed after peak breakthrough can be described

460

E. Pili et al.

using a model that assumes dilution of water with elevated solute concentrations derived from the NaCl spill and water with background concentrations of the solutes. Since no natural recharge occurred during the first 3 months of the flooding experiment (Fig. 16.3a), dilution water came from the fractures, the matrix, or rapidly infiltrating overflow from the tank. The trailing edges of the breakthrough curves were simulated using a box model Eq. (16.1), which provides residence times Eq. (16.2) reported as tdilution in Table 16.3. Values for Cinit in Eq. (16.1) for Cl, Na, Ca, Mg, and K were taken as the peak concentration in the breakthrough curve for each solute at FZ1 (Figs. 16.3c–e and 16.4). Values for Cinput in Eq. (16.1) were taken as the calculated background concentrations reported in Table 16.3. Values for the residence time Eq. (16.1) and (16.2) were determined by fitting the trailing edge of the breakthrough curves. Equation (16.1) was used in its logarithmic form and the best slope was determined. Associated uncertainties describe possible variations due to the scatter in the data. The quality of the fit is expressed using the misfit factor Fm defined as: N 1X C computed C observed 2 (16.3) Fm ¼ C observed N 1 where N is the number of observations and C the concentration of the element of interest. The model curves and the misfit factors are shown in Fig. 16.4. As shown in Fig. 16.4a, the box model fits satisfactorily the initial part of the trailing edge of the Cl breakthrough curve at FZ1 with a residence time of 8575 days (until mid-January 2004). A residence time of 200720 days and a lower initial concentration are needed to fit the second part of trailing edge (from mid-January 2004) at FZ1. The long residence time determined for the tail in FZ1 also satisfactorily describes the trailing edge of the Cl breakthrough curve at HGZ (Fig. 16.4a). Thus, we attribute a residence time of 200720 days to transport controlled by flow through the matrix and a residence time of 8575 days to transport controlled by flow through major fractures. The same residence times of 85 days and 200 days fit the trailing edge of the Na breakthrough curve at FZ1 (Fig. 16.4b). Sodium concentrations during breakthrough in HGZ were too close to background values (Fig. 16.4b) to obtain a reliable fit. A single residence time of 85 days fits the trailing edges of breakthrough curves for Ca, Mg, and K at FZ1 (Fig. 16.4c–e). The fact that we can obtain satisfactory agreement between the dilution model and the observed concentrations suggests that exchange


461

reactions did not have a major impact on solute transport along most of the flow path. Residence times calculated for mono- and divalent exchanged cations are identical (8575 days), further suggesting rapid initial exchange rather than progressive ion exchange during transport.

16.5. Tracer Transport In addition to reactive transport, the breakthrough curves can be used to estimate residence times, especially for the elevated concentrations of Cl resulting from the NaCl spill and the rain events. 16.5.1. Methods for Temporal Analysis Three different types of arrival times were calculated from breakthrough curves to estimate flow velocities, as described by Schulz (1998). Results reported in Table 16.3 include the time of first appearance of concentrations above background (Tfirst), the time of appearance of the maximum concentration (Tmax), and the arrival time of the center of mass (Tcenter) of concentrations above background. Tfirst is defined as the time of the first arrival of a solute at a concentration higher than the average background value (Table 16.3; Fig. 16.3) plus 1 standard deviation. Tcenter is the mean travel time, modified from Roberts et al. (1986): P ðC ij C i0 Þtj j (16.4) T center ¼ P ðC ij C i0 Þ j

where Cij is the concentration of solute i at time tj and Ci0 the background concentration of solute i. For a given transport distance (here 55 m), Tfirst yields the maximum flow velocity. Tmax yields a good approximation of the dominant flow velocity, especially in the case of steep and narrow breakthrough curves with a welldefined maximum concentration. Since the monitored zones yield water with contributions from various fractures and the matrix, the breakthrough curves record contributions from these zones that have to be unraveled. In the fracture-dominated zone (FZ1), Tcenter will be higher than that expected for water flowing through the dominant fractures because of contributions from the matrix. On the other hand, water that has traveled through major fractures will shift residence times toward lower values at HGZ. In addition

462

E. Pili et al.

to the residence times, background concentrations, the ratio of peak over background concentrations, and the recovered mass fractions are also given in Table 16.3. 16.5.2. Breakthrough after NaCl Spill At FZ1, breakthrough curves for Cl and the exchanged cations Ca, Mg (Fig. 16.3c–e), Na, and K (Fig. 16.4b–c) are similar. Breakthrough curves are not symmetrical, showing short rise times and long tails. Values of Tfirst for Cl and Na reflect almost immediate arrival (171 days). First arrival times of the exchanged cations are delayed somewhat (Tfirst ¼ 673 days), probably reflecting cation exchange that occurred along the flow path (peak/background concentrations for these solutes are less than 3 as compared to 139 and 133 for Cl and Na, respectively). As reported in Table 16.3, Tmax for Cl, Na, and Ca at FZ1 are identical (27 days) and similar to those for Mg and K (21 and 24 days, respectively). Because of significant tail of the breakthrough curves, values of Tcenter depend on what integration window is used in the calculation. Using the one-year period from May 2003 to May 2004, which includes the peaks in the breakthrough curves, and taking into account seasonal variations, Tcenter values are 88 days for Cl and 99 days for Na. Using the two-year period from May 2003 to May 2005, which includes the breakthrough-curve tails, Tcenter values are 118 days for Cl and 154 days for Na (Table 16.3). The larger Tcenter value for Na than for Cl is consistent with retardation of Na resulting from ion exchange reactions. However, Cl and Na have the same time of first appearance at FZ1. This is attributed to Na being present at concentrations in excess exchange capacity, and, therefore, a portion of the Na would be expected to behave as a non-reactive tracer. This is consistent with our assumption that where fracture flow dominates (FZ1), Na undergoes minimal reactions after leaving the soil and embankment above. The recharge event on September 1, 2003 following the drought period of June–August 2003 (Fig. 16.3a) is responsible for accelerated dilution of Cl (Fig. 16.3c). Recharge is also associated with a slight increase in Ca concentrations (Fig. 16.4d) and a minor increase in Mg concentrations (Fig. 16.4e) followed by dilution. This probably results from dissolution of calcium carbonate upon recharge either due to CO2-rich water infiltrating through the soil or to H+ produced by oxidative pyrite dissolution. For the breakthrough of Cl and Na at HGZ, Tfirst equals 1173 and 1773 days, respectively. Possible retardation of Na is consistent with its


463

involvement in exchange reactions. Values of Tmax for Cl and Na are identical (99 days, similar as Tcenter in FZ1) while the Tmax for Mg is 106 days. Although it is more difficult to interpret in terms of water velocity than when calculated from sharply defined peaks, Tcenter values calculated for the time windows of May 2003 to May 2004 and May 2003 to May 2005, respectively, are 168 and 250 days for Cl, 150 and 300 days for Na, and 179 and 257 days for Mg (Table 16.3). Since arrivals at HGZ likely are dominated by matrix flow, it is more appropriate to include the breakthrough-curve tails in the integration. Therefore, we consider residence times in the range 250–300 days to be characteristic of matrix-dominated flow.

16.5.3. Chloride Breakthrough Curves from Rain Chloride breakthrough curves from the rain event at FZ1 and HGZ show broad Cl peaks with maxima on August 1, 2002 and September 20, 2002, respectively (Fig. 16.5). We assume that high-Cl occurred on April 27, 2002, as recorded at the Le Casset weather station. The Tfirst value is 6772 days in FZ1 and 10972 days in HGZ. Maximum Cl concentrations are reached after about 3 months (97 days) at FZ1 and 5 months (147 days) at HGZ. These residence-time estimates are consistent with those obtained using H–O stable isotopes of less than 6 months (Pili et al., 2004). Values of Tcenter are 108 days at FZ1 and 184 days at HGZ (Table 16.3). Differences in residence times calculated from the rain event (especially Tfirst and Tmax) determined at the fracture-dominated site FZ1 and the matrix-dominated site HGZ are lower than those calculated from the NaCl spill event. During the spill event, a large water volume was spread over a limited surface area. This difference is interpreted as resulting from moderate fracture flow and high matrix– fracture interactions in fractures naturally recharged by rain because saturation is not reached. With natural tracer techniques but with larger water volumes, eventually resulting in water saturation in fractures, the residence time of water was estimated from observations of the beginning of snowmelt to the beginning of dilution in dripwater or, alternatively, from the end of snowmelt to the end of dilution (Pili et al., 2004). Values range from 25 to 60 days, as compared to Tmax values of 97 and 147 days determined from the rain event (Table 16.3). Higher saturation during recharge of snowmelt results in faster transport, as was also observed for Cl from the flooding event versus Cl in the rain event.

464

E. Pili et al.

16.6. Fracture Networks and Flow Simulations The Tfirst value of 1 day for Na and Cl at FZ1 after the NaCl flooding event is strikingly short considering (i) the presence of the low permeability matrix and (ii) longer transit times, including those for first appearance, are observed depending on the saturation state of the fractures (Table 16.3). Tracer breakthrough analysis clearly showed that FZ1 is much less influenced by the matrix than HGZ. Additional work on fractures logged in the tunnel provides complementary information on fast flows occurring through the fracture networks. Using a stereological analysis of the Roselend fracture system (Patriarche et al., 2008), 3-D fracture networks, consistent with the observations, were generated. Using numerical simulations, the percolation properties of such networks were calculated and the macroscopic permeability of the medium was estimated. Flow simulations were performed under saturated conditions. Results of these simulations can be compared with water fluxes recorded on April 25, 2003 (Fig. 16.1b), which corresponds to the end of snowmelt, because the fracture network during that period could be considered to be nearly saturated. The vertical hydraulic conductivity, KZ, at the end section of the tunnel is estimated to be 1.9 106 m/s (Patriarche et al., 2008). This value is intermediate between the hydraulic conductivity (K ¼ 2.9 107 m/s) determined from a single-well pump test performed in the vicinity of the tunnel entrance and the hydraulic conductivities for weathered granites, which range from 3.3 106 to 5.2 105 m/s (Domenico and Schwartz, 1998). Water travel time through the medium can be estimated using an assessment of kinematic porosity. The estimate of fracture porosity o (i.e., kinematic porosity as matrix was considered impermeable in the flow simulations) is based on both small and large fracture densities (cf. the section ‘Setting’) using the equation: o ¼ osmall þ olarge ¼ rs pðhRs iÞ2 bs þ rL pðRL Þ2 bL

(16.5)

where hRs i ¼ 0:5513 m is the mean radius of small fractures, RL ¼ 5 m the radius of large fractures (Patriarche et al., 2008), and bs and bL apertures of small and large fractures, respectively. Since fracture apertures were not measured systematically, bs and bL correspond to rough estimates of the largest apertures for small and large fractures; they are assumed equal to 2 103 and 102 m, respectively. Kinematic porosity is then equal to 0.33%. With a vertical hydraulic gradient equal to 1 in the column of saturated medium, the Darcy velocity equals the hydraulic conductivity KZ, which, divided by porosity, gives the filtration velocity. Travel time through


465

the medium is then calculated by dividing the medium thickness by the filtration velocity. In this way, we obtain a water transit time through the 55 m-thick medium of 1.1 days. This calculated water travel time is in good agreement with those calculated from the first appearance of elevated Na and Cl concentrations following the NaCl spill (approximately 1 day, Table 16.3), which, presumably resulted in nearly saturated conditions.

16.7. Conclusions Sampling and monitoring techniques developed at the Roselend natural laboratory allow for the use of tracers in highly variably saturated fractured media. High-resolution monitoring conducted over three years and fieldscale experiments in unsaturated fractured rocks provided new insights into flow and reactive transport processes. This work also shed light on the controls on temporal and spatial geochemical variability of water in the vadose zone. From the data collected, a conceptual model for flow and transport processes in the systems was developed and quantitative estimates of some important properties were obtained. These results provide a promising basis for subsequent quantitative modeling of these processes. Concentrations of Cl, a conservative, non-reactive (‘‘ideal’’) tracer, were monitored following two events with contrasting modes of infiltration. In one event, artificial flooding with ca. 500,000 l of water with highly elevated NaCl concentrations over an area of 10–100 m2 provided an opportunity to observed flow and transport under intense infiltration conditions. In the other event, breakthrough of elevated Cl concentrations from spatially uniform infiltration of water from rare Cl-rich rainfall was observed. Tracers introduced at the ground surface were detected in water dripping from fractures in a tunnel 55 m below. Two observation sites provided water samples with contrasting contributions of matrix and fracture flow. The tracers were detected in a section of the tunnel nearly vertically aligned with the infiltration zone. Elevated Na concentrations owing to the NaCl spill triggered cation exchange reactions that likely occurred mainly in the clay-mineral rich soil and embankment, releasing Ca, Mg, and K before the water entered the fractured medium. Assuming that minimal ion exchange occurred once the water entered the fractured rocks, concentrations of Cl, Na, and the exchanged cations during transport were primarily controlled by dilution. Results suggest that a limited amount of exchange of Na for Mg occurred in the matrix. In the coming years one would expect fewer fluctuations in cation concentrations in dripwater because of the dominance of spill-water derived Na on the exchange sites.

466

E. Pili et al.

Various methods were applied to determine solute residence times from breakthrough curves and to unravel salient features associated with transport through major fracture networks and matrix or, possibly, sub-order fracture networks within an impermeable matrix. Three main behaviors and associated residence times appear characteristic of the variably saturated fractured media. (1) Quasi-instantaneous flow and transport occur in some fractures with transit times over the 55-m-vertical distance of approximately 1 day. This mainly resulted from forced artificial infiltration associated with flooding or from large natural infiltration events, such as snowmelt-water infiltration. Here, fracture flow occurs at its highest possible velocity, i.e., at water saturation, in agreement with flow simulations performed with reconstructed fracture networks. (2) Fast flow and transport, with transit times of 25–100 days over the 55-m-veritical distance, were observed in the majorfracture networks. This likely is associated with an intermediate state of water saturation in the fractures. (3) Slow flow and transport, with transit times of 200–300 days or greater over the 55-m vertical distance, likely occurred in the matrix or sub-order fracture networks. Transit times for the matrix only are difficult to obtain because water sampling from purely unfractured media is not feasible. The ability of water with tracers to find its way to the fracture networks is a limiting factor. The flooding event released large volumes of water over a surface area large enough for rapid flow paths to be entered, filled with water, possibly up to saturation, thus resulting in the shortest observed residence times. Long-term and high-resolution monitoring of at least two zones with contrasting behaviors, i.e., fracture- and matrix-dominated, is mandatory for the understanding flow and transport in complex fractured media. Simultaneous determinations of water fluxes and chemical concentrations not only provide valuable information on the variable water saturation states of the system and allow for mass balance calculations, but also open the field to the study of hydrogeochemical coupling. The use of tracers with differing reactivity, different infiltration conditions, and several methods of data analysis in a complex fractured medium under variable water saturation contributes to quantifying processes characteristic of the different components of the system.

ACKNOWLEDGMENTS We thank Jean-Pierre Blanc for cooperation upon reconstruction of the NaCl flooding process, Christian Le Mouellic at Me´taux Spećiaux SA for valuable information on the tank coating and for providing analyses and


467

samples of trucked salt. The technical help of Bertrand Menotto and Patrick Carrera is acknowledged. We thank Patrice Coddeville and EMEP (www.emep.int) for the amounts and chemical compositions of precipitations at Le Casset. We thank Axel Bellivier for meteorological simulations with the ERA40 data from the European Centre for Medium-Range Weather Forecasts (www.ecmwf.int). Thanks go to EDF for providing us with some local meteorological data, to Me´teó France for providing the CROCUS code, and to the City of Beaufort for access to the Roselend tunnel. Editorial handling by Douglas Kent and remarks from two anonymous reviewers helped to improve the manuscript. This is IPGP contribution number 2229.

REFERENCES Albare`de, F. (1995). Introduction to Geochemical Modeling. Cambridge University Press, Cambridge, UK, 543 pp. Allen, R. G., Pereira, L. S., Raes, D., & Smith, M. (1998). Crop Evapotranspiration (Guidelines for Computing Crop Water Requirements). FAO, Rome, Italy, 290 pp. Appelo, C. A. J., & Postma, D. (1999). Geochemistry, Groundwater and Pollution. 4th Ed., Balkema, Rotterdam, 536 pp. Brun, E., David, P., Sudul, M., & Brunot, G. (1992). A numerical model to simulate snow-cover stratigraphy for operational avalanche forecasting. J. Glaciol., 38 (128), 13–22. Dahan, O., Nativ, R., Adar, E. M., & Berkowitz, B. (2001). Water flow and solute transport in unsaturated fractured chalk. In: D. D. Evans, T. J. Nicholson, & T. C. Rasmussen (Eds). Flow and Transport through Unsaturated Fractured Rock, 2nd Ed., Vol. 42, Geophysical Monograph Series, AGU, Washington, D.C., pp. 183–196. Domenico, P. A., & Schwartz, F. W. (1998). Physical and Chemical Hydrology. 2nd Ed., Wiley, New York, 506 pp. Evans, D. D., Nicholson, T. J., & Rasmussen, T. C. (2001). Flow and transport through unsaturated fractured rock: An overview. In: D. D. Evans, T. J. Nicholson, & T. C. Rasmussen (Eds). Flow and Transport Through Unsaturated Fractured Rock, 2nd Ed., Vol. 42, Geophysical Monograph Series, AGU, Washington, D.C., pp. 1–18. Faybishenko, B., Witherspoon, C. P. A., Doughty, C., Geller, J. T., Wood, T. R., & Podgorney, R. K. (2001). Multi-scale investigations of liquid flow in a fractured basalt vadose. In: D. D. Evans, T. J. Nicholson, & T. C. Rasmussen (Eds). Flow and Transport Through Unsaturated Fractured Rock, 2nd Ed., Vol. 42, Geophysical Monograph Series, AGU, Washington, D.C., pp. 161–182. Jenne, E. A. (1998). Adsorption of Metals by Geomedia: Variables, Mechanisms, and Model Applications. Academic Press, New York, 583 pp.

468

E. Pili et al.

National Research Council (2001). Conceptual Models of Flow and Transport in the Fractured Vadose Zone. National Academy Press, Washington, D.C., 392 pp. Nativ, R., Adar, E., Dahan, O., & Geyh, M. (1995). Water recharge and solute transport through the vadose zone of fractured chalk under desert conditions. Water Resour. Res., 31, 253–262. Patriarche, D., Pili, E., Adler, P. M., & Thovert, J. -F. (2008). Fracturation analysis and flow simulations in the Roselend fractured granite. In: A. Soares, M. J. Pereira, & R. Dimitrakopoulos (Eds). geoENV VI – Geostatistics for Environmental Applications, Vol. 15. Quantitative Geology and Geostatistics, Springer, Berlin. Pili, E., Perrier, F., & Richon, P. (2004). Dual porosity mechanism for transient groundwater and gas anomalies induced by external forcing. Earth Planet. Sci. Lett., 227 (3–4), 473–480. Provost, A.-S., Richon, P., Pili, E., Perrier, F., & Bureau, S. (2004). Fractured porous media under influence: The roselend experiment. EOS Trans. AGU, 85 (12), 113. Roberts, P. V., Goltz, M. N., & Mackay, D. M. (1986). A natural gradient experiment on solute transport in a sand aquifer 3. Retardation estimates and mass balances for organic solutes. Water Resour. Res., 22 (13), 2047–2058. Salve, R., Liu, H. H., Cook, P., Czarnomski, A., Hu, Q. H., & Hudson, D. (2004). Unsaturated flow and transport through a fault embedded in fractured welded tuff – art. no. W04210. Water Resour. Res., 40 (4), 4210. Schulz, H. D. (1998). Evaluation of breakthrough curves. In: W. Ka¨ss (Ed). Tracing Techniques in Geohydrology. Balkema, Rotterdam, pp. 366–376. Su, G. W., Geller, J. T., Pruess, K., & Hunt, J. R. (2001). Solute transport along preferential flow paths in unsaturated fractures. Water Resour. Res., 37 (10), 2481–2491. Su, G. W., Geller, J. T., Pruess, K., & Wen, F. (1999). Experimental studies of water seepage and intermittent flow in unsaturated, rough-walled fractures. Water Resour. Res., 35 (4), 1019–1037. Thornthwaite, C. W., & Mather, J. R. (1957). Instructions and Tables for Computing Potential Evapotranspiration and the Water Balance. Drexel Institute of Technology, Centerton, 311 pp.

Subject Index Acid sulfate-chloride Yellowstone National Park hot springs 425 Acid washed kaolinite 208 Adsorbed phosphate impact on ferrihydrite reduction and mineralization pathway 332–336, F12.5, F12.6, F12.7, T12.1 Adsorption contact method 68–69, 90 envelopes 240, 245–247, F9.2 goethite and quartz 361 modeling 359–361 of lead 350 (see also phosphate, study of effects on lead adsorption) Aging temperature effect on Cd binding on goethite 192–200 effect on Cd desorption and binding on kaolinite 208–213 effect on Pb desorption and binding on kaolinite 208–213 time effect on Cd binding on goethite 189, 194, 199–200 effect on Cd desorption and binding on kaolinite 209, 226 effect on Pb desorption and binding on kaolinite 207, 209, 229 Al-substituted massicot 213

Alumina, chromate adsorption 115–118 Andosols 134 Anoxic environments dissimilatory iron reduction 322 phosphate 324–325, 327, 337 Apatite 188 Aqueous lead phosphate system, reactions in 360 Aqueous-phase speciation of As, in geothermal fields 419–422 Arsenate (As(V)) As-EXAFS, on Fe(III) oxyhydroxide F15.3 As-EXAFS, scorodite F15.3 As-EXAFS, sodium arsenate F15.3 As-XANES, As2O5 F15.2 As-XANES, scorodite F15.2 bond valence analysis of 33 bonding and structure of surface complexes on oxide minerals T15.2 changes in surface speciation on nano-scale goethite with aging 6.3, 157, 161–168, F6.1, F6.2, F6.5 impact on XRD pattern, nano-scale goethite F6.3 linkages in mineral structures 34–38, F2.1, T2.1 rate of adsorption on nano-scale goethite 164, 166–167, 176–177 sorption on hematite surface, GIEXAFS analysis of 46–50, F2.4, F2.5, F2.6

470

Subject Index

Arsenic (As) precipitation, in geothermal deposits aqueous-phase speciation 419–422 controlling in fluid composition in hot springs 419 quantitative and qualitative characterization of deposits 423–427 solid-phase speciation 422–423, 427–433 Arsenite (As(III)) As-XANES, As2O3 F15.2 bonding and structure of surface complexes on oxide minerals T15.2 oxidation 421 Arsenopyrite (FeAsS(s)) 422 As-Sb-sulfide mineralizations 424 Atacamite 213, 225 Atomic force microscopy (AFM) 100 analysis 11, 16, 19, 23 ATR-FTIR spectroscopy 240–241, 247–255, F9.3, F9.4, F9.5a, F9.6, F9.7 Australian National Beamline Facility (ANBF) 210 Bacillus subtilis 424 Bacterial cultures, preparation of 325 Biogenic uraninite 300 Biogeochemical uranium cycling, implications for 310–312, F11.7 Blind prediction of Cu(II) sorption onto goethite, for SCM 276–280, 282–287, F10.4, F10.5, T10.1, T10.2, T10.3 Bond valence analysis 7 Bond valence sum analysis (BVSA) 189 Bovine spongiform encephalopathy (BSE) 126 Cadmium contamination with 188 EXAFS, on goethite vs. aging time F7.2, F7.4, T7.2

EXAFS, on kaolinite vs. aging time T8.1 local coordination on kaolinite and desorption behavior 225–226 on goethite, surface complexation states of T7.1, T190 K-edge data collection 192–193 data analysis 193–197 Calcite 188 Calf serum protein 143 CCM 269, 276, 280–281, 285–287, F10.3, F10.6, F10.7 Cd-substituted goethite 193 Cerussite 213 Chromate, interaction with aluminum oxide 115–118 Chronic wasting disease (CWD) 126 Clioquinol (CQ) 138 Clustering, of surface silicate (see polymerization, of surface silicate) Cobalt hydrotalcite-like precipitates 207 Constant Capacitance Model (see CCM) Contact angle measurements 100 Copper activity of free ion in soils 132–133 complexation in prions 143–144 copper (I) changes in surface speciation on goethite with aging 157, 161–168, F6.1, F6.2, F6.3, F6.5 rate of adsorption on goethite 164, 166–167, 176–177 soils, soluble and free copper in 132–133 soils, total in copper 130 Coprecipitation contact method 68–69, 90 Crystal truncation rod (see CTR)

Subject Index

CTR analysis of goethite-water interface 11–15, F1.4 of hematite-water interface 16–17 of magnetite-water interface 19–21, F1.7 data analysis and model comparison 6–9, F1.2 diffraction 4 measurement and instrumentation 6 observations and GIEXAFS analysis 56–57 of arsenate and silicate sorption 52–56, F2.9, F2.10, F2.11 DDLM 269–271, 274–281, 285–287, F10.2, F10.3, F10.6, F10.7, T10.1 Debye–Waller factor 161 Density functional theory (see DFT) Desorption rate, Cd on goethite 188–189, 191, 200 DFT analysis of goethite-water interface 13–15 of hematite-water interface 17 of magnetite-water interface 19–21 Diffuse Double Layer Model (see DDLM) Dimeric complexes 175 Dissimilatory iron reducing bacterium 325 impact of adsorbed phosphate 332–336 products 323, 328, 332, 337, F12.4 El Tatio geothermal field 429–433 El Tatio hot springs 421 Electron paramagnetic resonance (EPR) 144 cu-polypeptide complexes 144 Electrophoretic mobility 239, 244–245, F9.1

471

Ellipsometry 100–102 EXAFS (see also GIEXAFS) arsenic in sodium arsenate, scorodite, orpiment, microbial mats, siliceous sinters F15.3 As(V) on nano-scale goethite 166–167, 176–180 Cd on goethite vs. aging time F7.2, F7.4, T7.2 Cd on kaolinite vs. aging time T8.1 Cu(II) on nano-scale goethite 166–167, 176–180 Fe in Pb-Fe coprecipitates F3.3, F3.4 Hg(II) on nano-scale goethite 173–180, F6.10, T6.2 model building for lead and iron 82–84, F3.5, T3.4 model for the lepidocrocite sheet structure 79–82, F3.5, F3.6, T3.3 modeling with molecular moiety 76, T3.1, T3.2 Pb in Pb-Fe coprecipitates F3.3, F3.4, F3.7 Pb on kaolinite vs. aging time T8.1 spectra for ferrihydrite, 2line 75, F3.3 spectra for iron, linear combination fitting of Pb-Fe oxyhydroxide co-precipitates 76–79, F3.3 spectra for lepidocrocite 75, F3.3 spectra for PbFe coprecipitate 75–79, F3.3 spectroscopy 207, 350 Cd 189, 192 Cd single and double edge-sharing complexes F7.3 Zn(II) on nano-scale goethite 167–168, 170–173, 176–180, F6.7, T6.1 Extended X-ray absorption fine structure (see EXAFS)

472

Subject Index

Fe oxides, in the Roosevelt hot springs 424 Fe(III) (hydr)oxide mineral evolution, in uraninite oxidation by ferrihydrite 307, 310, F11.4, F11.6 FEFF 7.02 program 213 Ferrihydrite (FeOOH H2O) 156, 163, 166, 350 biomineralization, phosphate retention on 336–338, F12.2, F12.3, F12.6 bioreduction, phosphate retention on 328–332, F12.2, F12.3, F12.4 coated sand and phosphate sorption 326, F12.1 concentration, effect on uraninite oxidation by ferrihydrite 303, F11.5 mineralization pathway, phosphate impact on 339–341, F12.8, T12.1 reduction, phosphate impact on 338–339 synthesis 300 Ferrihydrite, 2-line EXAFS spectra for 75–76, 78–79, F3.3 Fe local structure 79–81 HRTEM-AEM spectra for 72–75, F3.2 XRD pattern for 72, F3.1 Field-scale reactive transport 455–461 FITEQL 4.0 software program 359 Fourier transforms 168, 173–174 Fractured crystalline rocks, NaCl concentration reactive transport and transit times in estimation of breakthrough curves 462–463 fracture networks and flow simulations 464–465 identification and quantification of cation exchange 455–459

modeling transport associated with cation exchange 459–461 temporal analysis of transit time 461–462 Galena 188 General blind prediction modeling 273 Geochemistry of scrapie hotspots 129–130 Geothermal fields, arsenic (As) aqueous-phase speciation 419–422 controlling in fluid composition in hot springs 419 solid-phase speciation 422–423, 427–433 Gibbs free energy of formation, for uranium species 303, T11.1 GIEXAFS analysis and CTR observations 56–57 As-GIEXAFS of arsenate sorption on hematite surface 46–50, F2.4, F2.5, F2.6 for low atomic number sorbate 58–59 Si-GIEXAFS of silicate sorption complex on hematite surface 50–52, F2.7, F2.8 Goethite arsenate adsorption 161–164, 166–167, 176–177, F6.1, F6.2, F6.3, F6.5 Cd desorption 188–189, 191, 200 Cd-substituted 193 Cd surface speciation 188–200 copper (I) adsorption 161–162, 164, 166–167, 176–177, F6.1, F6.2, F6.3, F6.5 crystals 159 impact of adsorption on aging of nano-particles 154–180, F6.1, F6.2, F6.3, F6.4, F6.5, F6.6, F6.7, F6.8, F6.9, F6.10, F6.11, T6.1, T6.2

Subject Index

mercury (II) adsorption 159, 161–162, 165–167, 173–180, F6.3, F6.5, F6.11, T6.2 zinc adsorption 159, 161–162, 165–168, 173, 176–180, F6.6, F6.7, F6.8, F6.9, F6.10, T6.1 Goethite (a-FeOOH) 4, 156, 189, 323, 332, 339–340, 350, F12.8, T12.1 Goethite-water interface, structure and reactivity of 9–15, F1.3, F1.4, F1.5 Gouy-Chapman model 105–107, 114–115 Grazing incidence extended X-ray absorption fine structure (see GIEXAFS) Green rust 323, 332, 336, 338, 340, F12.4, F12.6, F12.8, T12.1 Hamamatsu photon multiplier tube 98 Hematite (a-Fe2O3) 4 oxidation of biogenic uraninite by F11.3, F11.5, T11.1 silicate polymerization on F2.9, F2.11 surface complexation of As(V) on F2.5 surface complexation of silicate on F2.8, F2.9, F2.10 surface, structure of 44–46, F2.3 Hematite-water interface, structure and reactivity of 15–17, F1.6 High-resolution transmission and analytical electron microscopy (see HRTEM-AEM) Horizontal scrapie infectivity 145 Hot springs 419–420 HRTEM-AEM 72–75, F3.2 Hydro-cerussite 213 Iceland soils 134 Imogolites 134

473

Ion binding chromate interaction with aluminum oxide 115–118 manganese binding 111–115 Ion exchange field-scale observations 451, 455–459 Ion-selective electrode (ISE) 132 Iron oxyhydroxide nanoparticle growth, associations between 175–179 macroscopic uptake of As(V), Hg (II), Cu(II), and Zn(II) on 161–162 nanoparticle preparation 156 synchrotron X-ray microdiffraction, aged with As(V), Hg(II), Cu(II), and Zn (II) 162–166 Zn(II) and Hg(II) X-ray absorption spectroscopy 166–175 Isotopic exchange U(VI), with sediments 378, 386, 399–404, F14.5 Itai-itai disease 188 Italian soils 134 Kaolinite 141, 143, 207 Langmuir adsorption isotherm 115 Laurionite 213 Lead adsorption 350 at the iron oxide–water interface 351 study of, in the presence of phosphate 351–370 synergistic effects of phosphate adsorption on F13.5 Pb-EXAFS on kaolinite vs. aging time T8.1 PbFe coprecipitation atomic environment of iron in 75–76, 88, T3.5 atomic environment of lead in 75–76, 88, T3.5 EXAFS spectra 75–79, F3.3

474

Subject Index

HRTEM-AEM of 72–75, F3.2 particle structure 88–90, F3.8 with hydrous ferric oxide 70–72 X-ray absorption modeling of 75–84, F3.4, F3.5, F3.6, F3.7, T3.1, T3.2, T3.3, T3.4, T3.5 XRD pattern for 72, F3.1 Pb local coordination on kaolinite and desorption behavior 226–228 solid solution in hydrous ferric oxide 70–72 Lepidocrocite (g-FeOOH) 189, 323, 336, 340, F12.8 T12.1 EXAFS spectra for 75–76, 78–79, F3.3 Fe local structure 79–81 HRTEM-AEM spectra for 3.2, 72–75 sheet structure, EXAFS model for 79–82, F3.5, F3.6, T3.3 XRD pattern for 72, F3.1 Magnetite 4, 323, 332, 338, 340, F12.6, F12.7, F12.8, T12.1 Magnetite-water interface, structure and reactivity of 17–21, F1.7, F1.8, F1.9 Manganese bioavailability in soils 130–131, 133–136 complexation in prions 138–139, 145 soils, total in manganese 130 Manganite (Mn3O4) 189 Massicot 213 Masutomi hot springs 426 Mercury Hg-EXAFS on nano-scale goethite 173–180, F6.10, T6.2 mercury(II) changes in surface speciation on nano-scale goethite with aging 157, 159–161, 173–180, F6.1, F6.2, F6.3, F6.5, F6.10, F6.11, T6.2

rate of adsorption on nano-scale goethite 165–168, 173–180 Metatorbernite 378, 392, 403, 409–410 Microbial control, of As solid-phase speciation 426 Microbial mats arsenic EXAFS in F15.3 arsenic XANES in F15.2 MINEQL+ software program 361 Mineral structures arsenate anion linkages in 34–38, F2.1, T2.1 silicate anion linkages in 39–44, F2.2, T2.2 Model spinel clusters 168 Molecular dynamics (MD) simulations 139 Molybdenum adsorption, on amorphous Al and Fe oxides adsorption envelopes in 240, 245–247, F9.2 electrophoretic mobility in 239, 244–245, F9.1 microscopic experiments of 239–240, 244–247, F9.1, F9.2 surface complexation modeling of 241–244, 255–261, F9.9, F9.10, F9.11, T9.1, T9.2 BWE procedure 258–259 computer code FITEQL 3.2 in 243, 255 on Wyoming soil 257–258, F9.10, F9.11, T9.2 PRESS statistic 258–259 surface complexation constants 242, 244, 260–261, T9.2 surface complexation reactions 242–243 vibrational spectroscopy of 240–241, 247–255, F9.3, F9.4, F9.5, F9.6, F9.7, F9.8 ATR-FTIR spectroscopy 240–241, 247–255, F9.3, F9.4, F9.5a, F9.6, F9.7

Subject Index

Raman spectroscopy 241, 248–254, F9.5b, F9.6, F9.7 Molybdenum, in soils, characteristics and availability of 236–238 Monteponite 193, 196, 199, 213 Montmorillonite 139, 141, 143–144 interactions with prion proteins 138–139, 144–145 NaCl, reactive transport and transit times in unsaturated fractured crystalline rocks estimation of breakthrough curves 462–463 identification and quantification of cation exchange 455–459 modeling transport associated with cation exchange 459–461 temporal analysis of transit time 461–462 Nanogoethite 157, 162, 168, 173 preparation 156 Nonlinear optical (NLO) techniques 96–97 Optical parametric amplifiers (OPA-800CF) 97–98 Orakei Korako spring 421 Organic adlayer 101 Orpiment (As2S3), arsenic EXAFS in F15.3 Orpiment (As2S3), arsenic XANES in F15.2 Otavite 193, 196, 213 Oxidation–reduction reactions, of uranium 294–298, F11.1, F11.2 Oxyanions interaction, with iron (hydr)oxides 323 Oxygen chemical potential 14 Peyer’s patches 145 Phosphate adsorption 351 on ferrihydrite-coated sand F12.1

475

impact on ferrihydrite mineralization pathway 339–341, F12.8, T12.1 reduction 338–339 retention on ferrihydrite during biomineralization 336–338, F12.2, F12.3, F12.6 bioreduction 12.3, 12.4, 328–332, F12.2 sorption and ferrihydrite coated sand 326, F12.1 Phosphate, study of effects on lead adsorption co-sorption of lead and phosphate in combined systems 366–370 effect of goethite coating on adsorption 361 effect of pH on lead adsorption 361–364 effect of pH on phosphate adsorption 364–366 geochemical equilibrium modeling 359–361 micro-column batch adsorption experiments 354–357 PNC-CAT 192, 210 Polymerization, of surface silicate 58 Prion disease ‘‘Hotspot’’ geographical distribution 127 Prion disease transmission: the soil hypothesis 126–127 Prion proteins (PrPs) 127, 129 Cu and Mn complexation in 138–139, 145 double nature of 136–139 interactions with clay minerals 139–145 Prion sorption and transformation on clays modes of sorption 139–143 speciation change upon adsorption 143–145 Prokaryotes 425

476

Subject Index

PrPSc 129, 139 Pyromorphite (Pb5(PO4)3Cl) 369 Pyrophyllite 141 Quartz Mn(II) interactions with carboxylate groups in adlayer on F4.6, F4.9, F4.10, F4.11 sand 353 Radial structure function (RSF) 193, 199, 211 Raman spectroscopy 241, 248–254, F9.5b, F9.6, F9.7 Raninite oxidation by ferrihydrite 307 Reactive transport of NaCl in unsaturated fractured crystalline rocks, study fracture networks and flow simulations 464–465 identification and quantification of cation exchange 455–459 modeling transport associated with cation exchange 459–461 Rhodocrocite (MnCO3) 134 Roosevelt hot springs 424 Rossendorf Expert System for Surface and Sorption Thermodynamics 69–71, 274–275, 288, F10.1, F10.2, T10.1, T10.2 Sarcocystis parasite infection 136 Scanning probe microscopies (SPM) 3–4 Scherrer equation 159 SCM Constant Capacitance Model in (see CCM) Diffuse Double Layer Model in (see DDLM) for blind prediction of Cu(II) sorption onto goethite 276–280, 282–287, F10.4, F10.5, T10.1, T10.2, T10.3

for uncertainty analysis of Cu(II) sorption onto goethite 280–282, 284–287, F10.3, F10.6, F10.7 Triple Layer Model in (see TLM) Scorodite (FeAsO4 2H2O) arsenic EXAFS in F15.3 arsenic XANES in F15.2 Scrapie 126–127, 129–130 Scrapie-prone farm 132 Second harmonic generation (see also SHG) 96, F4.1 Shewanella putrefaciens 325, T12.1 SHG experiments laser system and signal detection 97–98 sample cell and flow system 98–99 SHG experiments, results of w(3) measurements charge screening 105–106 surface acid–base titration 106–110 uncertainty in the pKa values 110–111 Silicate anion bond valence analyis 38 linkages in mineral structures 39–44, F2.2, T2.2 relevant solution analysis 38–39 Silicate sorption on hematite surface, GIEXAFS analysis of 50–52, F2.7, F2.8 Siliceous sinter diagenesis 418 Silicon GIEXAFS (see silicate) SixPACK software package 160 Sodium dodecyl sulfate (SDS) 143 Soil-borne microbes (Acinetobacter) 127, 129 Solid-phase speciation of As, in geothermal solids 422–423, 427–433 Solid solution, lead in hydrous ferric oxide 70–72 Soller slits 160

Subject Index

Sorption models phenomenological models 68–69 surface complexation models (see SCM) Stanford Synchrotron Radiation Laboratory (SSRL) 327–328 Structure and reactivity of goethite-water interface 9–15, F1.3, F1.4, F1.5 of hematite-water interface 15–17, F1.6 of magnetite-water interface 17–21, F1.7, F1.8, F1.9 Structure-based models, application of 21–23, F1.10 Sum frequency generation (SFG) 96, 102–103, F4.4 Superoxide dismutase (SOD) 129 Surface complexation model (SCM) 350 Surface complexation modeling 350 of molybdenum adsorption on amorphous Al and Fe oxides 241–244, 255–261, F9.9, F9.10, F9.11, T9.1, T9.2 BWE procedure in 258–259 computer code FITEQL 3.2 in 243, 255 on Wyoming soil 257–258, F9.10, F9.11, T9.2 PRESS statistic 258–259 surface complexation constants in 242, 244, 260–261, T9.2 surface complexation reactions in 242–243 Surface free energies 14 of goethite surface 13–14, F1.5 of hematite surface 17, F1.8 Surface functional groups on Fe oxides and oxyhydroxides 23–24 Surface hydroxyl (SOH) groups 207 Surface polymerization, silica on hematite 58–59

477

Surface speciation, impact of aging time and temperature 188–200 Talc 143 Ternary surface complexation 351 Thermodynamic reaction modeling, of uraninite oxidation by ferrihydrite 303, T11.1 Thermus aquaticus 426 Thermus termophilus 426 Time of flight-secondary ion mass spectrometry (ToF-SIMS) 102 TLM 269, 276, 280–281, 285–287, F10.3, F10.6, F10.7 Transit times of NaCl in unsaturated fractured crystalline rocks, study estimation from breakthrough curves 462–463 fracture networks and flow simulations 464–465 temporal analysis 461–462 Transmissible spongiform encephalopathies (TSEs) 127 Triple Layer Model (see TLM) U(VI) release, from contaminated sediments aqueous concentrations 397–399 concentrations in the groundwater site 376 co-precipitation with carbonate minerals 409 dilute (bi)carbonate extractions 383–384, 393 dithionite citrate bicarbonate (DCB) extractions 382 estimates of adsorbed U(VI) 410 formate buffer extractions of 382–383, 391–393, 411 hydroxylamine-hydrochloride (HH) extractions and ammonium oxalate extractions 381–382 in the pond bottom precipitates 411

478

Subject Index

isotope exchange experiments 386, 399–404 modeling 387–388, 406–409 precipitation as metatorbernite 409 reaction in artificial groundwater solutions (AGWs) 384–386, 393–397, 409 total uranium and copper content 381 U(VI) sorption isotherm 386–387, 405–406 Ultra high vacuum (UHV) analysis 3 Uncertainty analysis 273–274 of Cu(II) sorption onto goethite, for SCM 280–282, 284–287, F10.3, F10.6, F10.7 Uraninite oxidation by ferrihydrite effect of ferrihydrite concentration on 303, F11.5 Fe(III) (hydr)oxide mineral evolution in 307, 310, F11.4, F11.6 thermodynamic reaction modeling of 303, T11.1 X-ray diffraction and absorption spectroscopy 302–303 Uranium cycling, biogeochemical, implications for 310–312, F11.7 oxidation-reduction reactions of 294–298, F11.1, F11.2 Variant Creutzfeld Jacob disease (vCJD) 126 Vibrational Spectroscopy, of molybdenum adsorption on amorphous Al and Fe oxides 240–241, 247–255, F9.3, F9.4, F9.5, F9.6, F9.7, F9.8 ATR-FTIR spectroscopy 240–241, 247–255, F9.3, F9.4, F9.5a, F9.6, F9.7

Raman spectroscopy 241, 248–254, F9.5b, F9.6, F9.7 Vibrational sum frequency generation 102–104 Vivianite 323, 332, 337–340, F12.4, F12.6, F12.8, T12.1 Waiotapu springs 421 Waipiti (Cervus elaphus nelsoni ) 126 Wild deer (Odoco leus spp.) 126 WinXAS 2.1 193 for data analysis 211 XANES 168 arsenic in As2O3, As2O5, orpiment, scorodite F15.2 X-ray absorption modeling, of PbFe coprecipitate 75–84, F3.4, F3.5, F3.6, F3.7, T3.1, T3.2, T3.3, T3.4, T3.5 X-ray absorption near-edge structure (see XANES) X-ray diffraction and absorption spectroscopy, of uraninite oxidation by ferrihydrite 302–303 X-ray scattering CTR technique 4–9, F1.1, F1.2 Zinc changes in surface speciation on nano-scale goethite with aging 157, 159–161, 168–173, 176–180, F6.6, F6.7, F6.8, F6.9, T6.1 rate of adsorption on nano-scale goethite 168–173, 176–180 Zn-EXAFS on nano-scale goethite 167–168, 170–173, 176–180, F6.7, T6.1

Adsorption of Metals by Geomedia II: Variables, Mechanisms, and Model Applications (Developments in Earth and Environmental Sciences 7)

Developments in Earth & Environmental Sciences, 7

Physical Processes in Earth and Environmental Sciences

Neutron applications in earth, energy and environmental sciences

Climate, Environment, and Society in the Pacific during the Last Millennium, Volume 6 (Developments in Earth and Environmental Sciences) (Developments in Earth and Environmental Sciences)

Evolution of Earth and its Climate: Birth, Life and Death of Earth (Developments in Earth and Environmental Sciences 10)

Bioinformatics: applications in life and environmental sciences

The Diatoms: Applications for the Environmental and Earth Sciences

The Diatoms: Applications for the Environmental and Earth Sciences

Geosciences, Environment and Man (Developments in Earth and Environmental Sciences, Volume 1)

Catalysis and Adsorption by Zeolites

Global Warming and Global Cooling, Volume 5: Evolution of Climate on Earth (Developments in Earth and Environmental Sciences)

Complex Variables and Applications

Complex Variables and Applications

Global Warming and Global Cooling, Volume 5: Evolution of Climate on Earth (Developments in Earth and Environmental Sciences)

Oil Pollution and its Environmental Impact in the Arabian Gulf Region, Volume 3 (Developments in Earth and Environmental Sciences)

Complex Variables and Applications

Complex variables and applications

Developments and Applications in Solubility

Adsorption by Powders and Porous Solids: Principles, Methodology and Applications

Mechanisms, Transmissions and Applications

Earth sciences and mathematics

Earth sciences and mathematics

Immunotoxicology Of Environmental And Occupational Metals

Antarctic Climate Evolution, Volume 8 (Developments in Earth and Environmental Sciences)

Partition and Adsorption of Organic Contaminants in Environmental Systems

Persistent Organic Pollutants in Asia, Volume 7: Sources, Distributions, Transport and Fate (Developments in Environmental Science) (Developments in Environmental Science)

Complex Variables: Introduction and Applications

Recent Advances in Adsorption Processes for Environmental Protection and Security

Principles of adsorption and adsorption processes

Adsorption of Metals by Geomedia II: Variables, Mechanisms, and Model Applications (Developments in Earth and Environmental Sciences 7)

Developments in Earth & Environmental Sciences, 7

Physical Processes in Earth and Environmental Sciences

Neutron applications in earth, energy and environmental sciences

Climate, Environment, and Society in the Pacific during the Last Millennium, Volume 6 (Developments in Earth and Environmental Sciences) (Developments in Earth and Environmental Sciences)

Evolution of Earth and its Climate: Birth, Life and Death of Earth (Developments in Earth and Environmental Sciences 10)

Bioinformatics: applications in life and environmental sciences

The Diatoms: Applications for the Environmental and Earth Sciences

The Diatoms: Applications for the Environmental and Earth Sciences

Geosciences, Environment and Man (Developments in Earth and Environmental Sciences, Volume 1)

Catalysis and Adsorption by Zeolites

Global Warming and Global Cooling, Volume 5: Evolution of Climate on Earth (Developments in Earth and Environmental Sciences)

Complex Variables and Applications

Complex Variables and Applications

Global Warming and Global Cooling, Volume 5: Evolution of Climate on Earth (Developments in Earth and Environmental Sciences)

Oil Pollution and its Environmental Impact in the Arabian Gulf Region, Volume 3 (Developments in Earth and Environmental Sciences)

Complex Variables and Applications

Complex variables and applications

Developments and Applications in Solubility

Adsorption by Powders and Porous Solids: Principles, Methodology and Applications

Mechanisms, Transmissions and Applications

Earth sciences and mathematics

Earth sciences and mathematics

Immunotoxicology Of Environmental And Occupational Metals

Antarctic Climate Evolution, Volume 8 (Developments in Earth and Environmental Sciences)

Partition and Adsorption of Organic Contaminants in Environmental Systems

Persistent Organic Pollutants in Asia, Volume 7: Sources, Distributions, Transport and Fate (Developments in Environmental Science) (Developments in Environmental Science)

Complex Variables: Introduction and Applications

Recent Advances in Adsorption Processes for Environmental Protection and Security

Principles of adsorption and adsorption processes

Recommend Documents