supercritical fluid cleaning

SUPERCRITICAL FLUID CLEANING Fundamentals, Technology and Applications Edited by John McHardy Hughes Aircraft Company ...

Author: Samuel P. Haber

110 downloads 1377 Views 19MB 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

SUPERCRITICAL FLUID CLEANING Fundamentals, Technology and Applications

Edited by

John McHardy Hughes Aircraft Company El Segundo, California and

Samuel P. Sawan University of Massachusetts Lowell Lowell, Massachusetts

NOYES PUBLICATIONS Westwood, New Jersey, U S A .

Copyright 0 1998 by Noyes Publications No part of this book may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, recording or by any information storage and retrieval system, without permission in writing from the Publisher. Library of Congress Catalog Card Number: 97-44663 ISBN: 0-8 155- 14 16-6 Printed in the United States Published in the United States of America by Noyes Publications 369 Fairview Avenue, Westwood, New Jersey 07675 10 9 8 7 6 5 4 3 2 1

Library of Congress Cataloging-in-Publication Data Supercritical fluid cleaning : fundamentals, technology, and applications / edited by John McHardy and Samuel P. Sawan. p. cm. Includes bibliographical references and index. ISBN 0-8155-1416-6 1. Supercritical fluids. 2. Cleaning. 1. McHardy, John. 11. Sawan, Samuel P. TP156.E8S82 1998 667-dc21 97-44663 CIP

MATERIALS SCIENCE AND PROCESS TECHNOLOGY SERIES Series Editors Rointan F. Bunshah, University of California, Los Angeles Gary E. McGuire, Microelectronics Center of North Carolina Stephen M. Rossnagel, IBM Thomas J. Watson Research Center

Electronic Materials and Process Technology CHARACTERIZATIONOF SEMICONDUCTOR MATERIALS, Volume 1: edited by Gary E. McGuire CHEMICAL VAPOR DEPOSITION FOR MICROELECTRONICS: by Arthur Sherman CHEMICAL VAPOR DEPOSITION OF TUNGSTEN AND TUNGSTEN SILICIDES: by John E. J. Schrnitz CHEMISTRY OF SUPERCONDUCTOR MATERIALS: edited by Terrell A. Vanderah CONTACTS TO SEMICONDUCTORS: edited by Leonard J. Brillson DIAMOND CHEMICAL VAPOR DEPOSITION: by Huimin Liu and David S. Dandy DIAMOND FILMS AND COATINGS: edited by Robert F. Davis DIFFUSION PHENOMENA IN THIN FILMS AND MICROELECTRONIC MATERIALS: edited by Devendra Gupta and Paul S. Ho ELECTROCHEMISTRY OF SEMICONDUCTORS AND ELECTRONICS: edited by John McHardy and Frank Ludwig ELECTRODEPOSITION: by Jack W. Dini HANDBOOK OF CARBON, GRAPHITE, DIAMONDS AND FULLERENES: by Hugh 0. Pierson HANDBOOK OF CHEMICAL VAPOR DEPOSITION: by Hugh 0. Pierson HANDBOOKOF COMPOUND SEMICONDUCTORS: edited by Paul H. Holloway and Gary E. McGuire HANDBOOK OF CONTAMINATION CONTROL IN MICROELECTRONICS: edited by Donald L. Tolliver HANDBOOK OF DEPOSITION TECHNOLOGIES FOR FILMS AND COATINGS, Second Edition: edited by Rointan F. Bunshah HANDBOOK OF ION BEAM PROCESSING TECHNOLOGY: edited by Jerome J. Cuorno, Stephen M. Rossnagel, and Harold R. Kaufman HANDBOOK OF MAGNETO-OPTICAL DATA RECORDING: edited by Terry McDaniel and Randall H. Victora HANDBOOK OF MULTILEVEL METALLIZATION FOR INTEGRATED CIRCUITS: edited by Syd R. Wilson, Clarence J. Tracy, and John L. Freeman, Jr. HANDBOOKOFPLASMA PROCESSINGTECHNOLOGY: edited by StephenM. Rossnagel, Jerome J. Cuomo. and William D. Westwood HANDBOOK OF POLYMER COATINGS FOR ELECTRONICS, 2nd Edition: by James Licari and Laura A. Hughes

V

vi

Series

HANDBOOK OF REFRACTORY CARBIDES AND NITRIDES: by Hugh 0. Pierson HANDBOOK OF SEMICONDUCTOR SILICON TECHNOLOGY: edited by William C. O'Mara, Robert 8. Herring, and Lee P. Hunt HANDBOOK OF SEMICONDUCTOR WAFER CLEANING TECHNOLOGY: edited by Werner Kern HANDBOOKOF SPUTTER DEPOSITIONTECHNOLOGY: by KiyotakaWasa and Shigeru Hayakawa HANDBOOK OF THIN FILM DEPOSITION PROCESSES AND TECHNIQUES: edited by Klaus K. Schuegraf HANDBOOK OF VACUUM ARC SCIENCE AND TECHNOLOGY: edited by Raymond L. Boxman, Philip J. Martin, and David M. Sanders HANDBOOK OF VLSl MICROLITHOGRAPHY: edited by William B. Glendinning and John N. Helbert HIGH DENSITY PLASMA SOURCES: edited by Oleg A. Popov HYBRID MICROCIRCUIT TECHNOLOGY HANDBOOK, Second Edition: by James J. Licari and Leonard R. Enlow IONIZED-CLUSTER BEAM DEPOSITION AND EPITAXY: by Toshinori Takagi MOLECULAR BEAM EPITAXY: edited by Robin F. C. Farrow SEMICONDUCTOR MATERIALS AND PROCESS TECHNOLOGY HANDBOOK: edited by Gary E. McGuire ULTRA-FINE PARTICLES: edited by Chikara Hayashi, R. Ueda and A. Tasaki

Ceramic and Other Materials-Processing and Technology ADVANCED CERAMIC PROCESSING AND TECHNOLOGY, Volume 1: edited by Jon G. P. Binner CEMENTED TUNGSTEN CARBIDES: by Gopal S. Upadhyaya CERAMIC CUTTING TOOLS: edited by E. Dow Whitney CERAMIC FILMS AND COATINGS: edited by John 6. Wachtman and Richard A. Haber CORROSION OF GLASS, CERAMICS AND CERAMIC SUPERCONDUCTORS: edited by David E. Clark and Bruce K. Zoitos FIBER REINFORCED CERAMIC COMPOSITES: edited by K. S. Mazdiyasni FRICTION AND WEAR TRANSITIONS OF MATERIALS: by Peter J. Blau HANDBOOKOF INDUSTRIAL REFRACTORIES TECHNOLOGY: by Stephen C. Carniglia and Gordon L. Barna SHOCK WAVES FOR INDUSTRIAL APPLICATIONS: edited by Lawrence E. Murr SOL-GEL TECHNOLOGY FOR THIN FILMS, FIBERS, PREFORMS, ELECTRONICS AND SPECIALTY SHAPES: edited by Lisa C. Klein SPECIAL MELTING AND PROCESSING TECHNOLOGIES: edited by G. K. Bhat

Contributors

Shirish M. Chitanvis Los Alamos National Laboratory LQSAlamos,NM John L. Fulton Battelle Pacific Northwest Laboratories Richland, WA John E. Giles, Jr. Honeywell Space and Strategic Avionics Div. Clearwater, FL Michael 0.Hogan Battelle Pacific Northwest Laboratories Richland, WA Glen Huse Green Oaks, IL

Kevin Jackson Battelle Pacific Northwest Laboratories Richland, WA Kenneth E. Laintz Isotag L.L.C. Isotag Research Los Alamos, NM Gurusamy Manivannan BioPolymerix Tyngsboro, MA John McHardy Hughes Aircrafi Company El Segundo, CA Ramamurthy Nagarajan IBM San Jose, CA

x

Contributors

Chris W. Patterson Los Alamos National Laboratory Los Alamos, NM

Max R. Phelps Battelle Pacific Northwest Laboratories Richland, WA

Laura J. Silva Battelle Pacific Northwest Laboratories Richland, WA L. Dale Sivils Los Alamos National Laboratory Los Alamos, NM

Robert J. Purtell IBM Thomas J. Watson Research Center Yorktown Heights, NY Samuel P. Sawan Polymer Science/plasticsEngineering program University of Massachusetts, Lowell Lowell, MA Yeong-Tarng Shieh Polymer Science/PlasticsEngineering Program University of Massachusetts, Lowell Lowell, MA

Charles W. Smith Fairview, PA W. Dale Spall Isotag L.L.C. Isotag Research Los Alamos, NM

Jan-Hon Su Polymer SciencdPlasticsEngineering Program University of Massachusetts, Lowell Lowell, MA Robert G. Honeywell, Jr. Honeywell Inc. Glendale, AZ

NOTICE To the best of our knowledge the information in this publication is accurate; however the Publisher does not assume any responsibility or liability for the accuracy or completeness of, or consequences arisingfrom, suchinformation. Thisbookisintended for informational purposes only. Mention of trade names or commercial products does not constitute endorsementor recommendation for use by the Publisher. Final determination of the suitability of any information or product for use contemplated by any user, and the manner of that use, is the sole responsibility of the user. We recommend that anyone intending to rely on any recommendation of materials or procedures mentioned in this publication should satisfy himself as to such suitability, and that he can meet all applicable safety and health standards.

Preface

Although supercritical fluid (SCF)technology is now widely used in extraction and purification processes (in the petrochemical, food and pharmaceuticals industries), this book is the first to address the new application of cleaning. Our objective was to provide a roadmap for readers who want to know whether SCF technologycan meet their own processing and cleaning needs. It should be particularly helpfbl to those striving to balance the requirements for a clean product and a clean environment. The interdisciplinary subject matter will appeal to scientists and engineers in all specialties ranging from materials and polymer sciences to chemistry and physics. It should also be useful to those developing new processes for other applications, and referencesgiven at the end ofeach chapter provide links to the wider body of SCF literature. The book is organized with topics progressing from the fundamental nature of the supercritical state, through process conditions and materials interactions, to economic considerations. Practical examples are included to show how the technology has been successhlly applied. The first four chapters consider principles governing SCF processing, detailing issues such as solubility, design for cleanability, and the dynamics of particle removal. The next three chapters discuss surfactants and microemulsions, SCF interaction with polymers, and the use of supercritical carbon dioxide (CO,) as cleaning solvent. The closing chapters focus on more practical considerations such as scaleup, equipment costs, and financial analysis. Many contributors to this book belong to the “Joint Association for the Advancement of SCF Technology” (JUST), a consortium established

vii

viii Preface to fosterthe developmentand application of SCFs. Members of J U S T meet several times each year to compare results and exchange information on their research and development activities. A primary motivation for the formation of J U S T was the growing worldwide need to replace ozonedepleting compounds (ODCs) and smog-forming volatile organic compounds (VOCs) in manufacturing processes. Although aqueous cleaning has been adopted successfully for many applications, water is not a panacea and SCF technology has emerged as a leading alternative where aqueous processes cannot be used. A substance is said to be in the supercritical state when its temperature and pressure are raised abovethe critical point. The state is characterized by low viscosities, low surface tensions, high diffisivities and near-liquid molar volumes, resulting in rapid wetting and excellent penetrationcharacteristicsthat are highly desirable for good cleaning performance. The critical point for CO, occurs at 31.3"C and 7.4 MPa, conditions readily attained with commercially available equipment. Carbon dioxide (CO,) is also inert, ozone-friendly, noncombustible, naturally occumng, inexpensive, and does not contribute to smog. Thus, for many applications, supercritical CO, will predominate as the solvent of choice. Indeed the work of most authors has been primarily concerned with CO,, although much of the discussion applies equally to many other SCFs. For over a century, the unique properties of the supercritical state have enticed scientists and engineers to explain and harness them. Yet even today, the science and technology of SCFs remains far from mature and opporhmities abound for spectacularreturns on research investments. In the final analysis, our book will have met its objective if the reader is persuaded to join the quest for a deeper understanding of this intriguing field. El Segundo, CA September, 1997

John McHardy

Lowell, MA September, 1997

Samuel P. Sawan

Contents

1 The Supercritical State

...........................................

Gurusamy Manivannan andSamuel I! Sawan 1.0 INTRODUCTION ................................................................... 2.0 PHYSICAL TRANSFORMATIONS OF PURE SUBSTANCES ............................................................ 3.0 CRITICAL CONSTANTS ......................................................

1 1

3 5 CRITICAL AND BOILING POINTS ...................................... 7 SUPERCRITICAL STATE ..................................................... 9 SOLVENTS AND ANTISOLVENTS ..................................... 9 SOLUBILITY ISSUES OF SUPERCRITICAL FLUIDS ....... 10

4.0 5.0 6.0 7.0 8.0 HILDEBRAND SOLUBILITY PARAMETER ..................... 12 9.0 CARBON DIOXIDE ............................................................. 13 10.0 CANDIDATES FOR SUPERCRITICAL FLUIDS ................ 16 11.0 CONCLUSIONS ................................................................... 19 REFERENCES ..................................................................... 20

............ 22

2 Solubility in Supercritical Fluid Cleaning Kenneth E.Laintz, L Dale Sivils. and U! Dale Spa11

ABSTRACT ..................................................................... 1.0 INTRODUCTION .................................................................

xi

22 23

xii

Contents

2.0 CONTAMINANT SOLUBILITY .......................................... 3.0 MODELS FOR CLEANING ................................................. 4.0 CONCLUSION ..................................................................... ..................................................................... REFERENCES

3 Design for Cleanability

.........................................

25 28 36 37

38

Ramamurthy Nagarajan 1.0 INTRODUCTION ................................................................. 2.0 PROCESS DESIGN GUIDELINES ...................................... 2.1 Use of Water-Soluble Cutting Fluids ................................

38 39 40 2.2 Minimization of WIP. Implementation of CFM ................ 43 2.3 Rinsing after Machining ................................................... 44 2.4 Parts Handling after Final Clean ...................................... 45 3.0 CONCLUSIONS ................................................................... 46 APPENDIX 1: GUIDELINES FOR DESIGN OF PARTS FOR CLEANABILITY .................................................................. 47 A1.1 Introduction .................................................................. 47 Al.2 DFM/DFC Compatibility ........................ Al.3 Guidelines for Design for Cleanability ..... References (for Appendix 1) ................................................... 55 APPENDIX 2: VENDOR PROCESS CONTAMINATION CHECKLIST ..................................................................... 55 A2.1 Receivinglhpection (Wr)at the Vendor ....................... 55 A2.2 Manufacturing ................................... A2.3 DegreasinglCleaningDrying .............. A2.4 Final Inspection .................. ...................................... 63 65 A2.5 Storage ..................................................................... A2.6 Shipping ...... A2.7 Employees ..... A2.8 Management ................................................................. 68

4 Dynamics Of Particle Removal By Supercritical 70 Carbon Dioxide Shirish M. Chitanvis. Chris FK Patterson. W Dale Spall.

.....................................................

and Kenneth E. Laintz

ABSTRACT ..................................................................... 1.0 INTRODUCTION ................................................................. 2.0 THE BOUNDARY LAYER IN TURBULENT FLOW OVER A PLATE ................................................................... 3.0 ADHESIVE FORCES ...........................................................

70 70 73 77

Contents

xiii

4.0 SUPERCRITICAL CARBON DIOXIDE .............................. 5.0 CONCLUSIONS ................................................................... REFERENCES .....................................................................

80

5 Surfactants and Microemulsions in Supercritical Fluids Kevin Jackson and John L Fulton

...............................................

85 86

87

ABSTRACT ..................................................................... 87 1.0 MTRODUCTION ................................................................. 88 2 .0 MICROEMULSION STRUCTURE ...................................... 90 3 .0 MICROEMULSIONS IN SUPERCRITICAL FLUIDS ......... 94 4.0 MICROEMULSIONS IN C02............................................. 104 5.0 SUPERCRITICAL MICROEMULSION CLEANING PROCESSES ................................................................... 107 112 6.0 KINETICS OF CLEANING ................................................ 7 .0 RECENT DEVELOPMENTS ............................................. 115 8.0 CONCLUSIONS ................................................................. 116 ACKNOWLEDGMENT ............................................................. 117 REFERENCES ................................................................... 117

6 Evaluation of Supercritical Fluid Interactions with 121 Polymeric Materials Samuel I? Sawan. Yeong-Tarng Shieh. Jan-Hon Su.

...........................................

Gurusamy Manivannan and K Dale Spa11 1.0 BACKGROUND ................................................................. 121 2.0 SORPTION. SWELLING AND DISSOLUTION OF CARBON DIOXIDE IN POLYMERS AT ELEVATED PRESSURE ..... 124 124 2.1 Introduction ................................................................... 2.2 Experimental Procedure ................................................. 126 2.3 Supercritical Carbon Dioxide Treatment and Testing 126 Procedures ................................................................... 128 2.4 Results and Discussion .................................................. 2.5 Conclusions ................................................................... 140 141 3.0 THERMAL PROPERTIES ................................................. 3. 1 Introduction ................................................................... 141 3.2 Experimental Procedure .................................... 3.3 Results and Discussion .. 3.4 Conclusions ................... 4.0 MECHANICAL PROPERTI .................. 150 4.1 Introduction ................... 4.2 Experimental Procedure ................................................. 151

xiv

Contents 4.3 Results and Discussion .................................................. 5.0 CONCLUSIONS ................................................................. ACKNOWLEDGMENTS ........................................................... REFERENCES ...................................................................

7 A Survey on the Use of Supercritical Carbon Dioxide as a Cleaning Solvent FK Dale Spall and Kenneth E. Laintz

.............................

ABSTRACT ................................................................... 1.0 INTRODUCTION ............................................................... 2.0 EXPERIMENTAL PROCEDURE ....................................... 3.0 RESULTS AND DISCUSSION .......................................... 4.0 TABLES ................................................................... 5.0 CONCLUSION ................................................................... ACKNOWLEDGMENTS ........................................................... REFERENCES ...................................................................

151 158 159 160

162 162 163 164 166 173 193 193 194

8 Precision Cleaning With Supercritical Fluid: A Case Study 195 John E. Giles. Jr., and Robert G. Woodwell. Jr.

...........................................................

INTRODUCTION ............................................................... 195 PRECISION INERTIAL INSTRUMENTS ......................... 197 ADVANTAGES OF SUPERCRITICAL FLUIDS ............... 198 SYSTEM ACQUISITION .............................. 5.0 PROCESS DEVELOPMENT ......................... 5.1 Initial Testing ................................................................ 205 5.2 Results of Initial Testing ................................................ 206 5.3 Cleaning of Cracks and Crevices .................................... 207 6.0 COMPATIBILITY TESTING ............................................. 208 6.1 General ................................................................... 208 ................................................................... 209 6.2 Epoxies 209 6.3 Thermoplastics .............................................................. 6.4 Damping Fluid ............................................................... 210 6.5 Hardware Compatibility................................................. 211 7.0 SYSTEM MODIFICATIONS .............................. 8.0 PRODUCTION IMPLEMENTATION ................................ 213 ........................................ 213 8.1 Bulk Fluid Removal .. 8.2 End Housing Cleani ........................................ 215 9.0 LESSONS LEARNED ........................................................ 217 ACKNOWLEDGMENTS ........................................................... 219 REFERENCES ................................................................... 219

1.0 2.0 3.0 4.0

Contents

9 Scaleup Considerations

.......................................

m

220

.

Max R Phelps. Laura J. Silva. and Michael 0 Hogan 1.0 INTRODUCTION ............................................................... 220 1.1 Pre-Selection Activities .................................................. 221 1.2 Scaleup Approach ......................................................... 221 2.0 DESIGN CONSIDERATIONS ........................................... 222 3.0 DIMENSIONLESS VARIABLES ....................................... 224 4.0 PRINCIPLE OF SIMILARITY ........................................... 227 4.1 Geometrical Similarity ................................................... 227 4.2 Mechanical Similarity .................................................... 228 4.3 Thermal Similarity......................................................... 229 4.4 Chemical Similarity ....................................................... 230 5.0 SUPERCRITICAL CARBON DIOXIDE SYSTEM SCALEUP ................................................................... 231 5.1 Experimental Approach ................................................. 231 5.2 Equipment ................................................................... 232 5.3 Coupon Contamination Procedures ................................ 234 5.4 Coupon Cleaning with Gravimetric Analysis ..................234 5.5 Results ................... 5.6 Discussion of PNL Scale 6.0 CONCLUSIONS ................................................................. 243 ACKNOWLEDGMENTS ........................................................... 244 ................................................................... 244 REFERENCES

10 Equipment Cost Considerations and Financial Analysis of Supercritical Fluid Processing 245

........

Charles K Smith and Glen Huse 1.0 COST DRIVERS ................................................................ 1.1 Process Parameters ........................................... 1.2 Equipment Considerations................................. 1.3 Operating Costs ............................................................. 2.0 COST OF OWNERSHIP FOR SCF CLEANING ................................................................... SYSTEMS 2.1 Operating Costs ............................................................. 2.2 Net Cash Flow ............................................................... 2.3 Cumulative Net Cash Flow ............................................ 3.0 RELATIVE COST OF OWNERSHIP FOR SCF VERSUS OTHER CLEANING SYSTEM .......................................... ............................. 4.0 CONCLUSIONS .......................... 5.0 ACKNOWLEDGMENT ......................................................

246 259 260 261 262 262 263 265 265

mi

Contents

11 A Practical Guide to Supercritical Fluid Cleaning

...............................................................

267

Robert . I Purtell 1.0 INTRODUCTION ............................................................... 267 2.0 SOLUBILITY:THEORETICAL AND EMPIRICAL .......... 269 3.0 TRANSPORT ................................................................... 272 4.0 REACTOR DESIGN ........................................................... 275 5.0 PARTS DRYING ................................................................ 277 278 6.0 SUMMARY ................................................................... ACKNOWLEDGMENTS ........................................................... 279 REFERENCES ................................................................... 279

Index

.........................................................................

281

The Supercritical State Gurusamy Manivannan and Samuel l? Sawan

1.0

INTRODUCTION

A hundred and twenty-five years ago, the behavior of highly compressed gases was a major research field that interested many famous chemists of the day, including even Mendeleev.['] In 1869, the existence of an additional state of matter became known when Andrews discovered the critical phenomenon and supercritical state. The values reported by him for the critical point of carbon dioxide are in close agreement to the presently accepted values. Later in 1879, the phenomenon of supercritical fluid (SCF) solubility was studied by Hannay and Hogarth.I21 They found that gases could be good solvents under supercritical conditions and the dissolving power of a SCF is highly pressure-dependent. There was substantial controversy about the initial finding of the pressure-dependence of solubility in a supercritical fluid after it was first reported at the Royal Society Meeting by Hannay and Hogarth. Small changes in pressure continuously alter the density of these fluids from gas-like to liquid-like, thereby allowing their solubility power to be adjusted over wide ranges. I

2

Supercritical Fluid Cleaning

Supercritical fluid technology has been widely used in extraction and purification processes in the food and pharmaceuticals indu~try[~]-[~] and for techniques such as supercritical fluid chromatography.[6]-[8] Recently, there has been a significant increase in interest of the use of sub- as well as supercritical (SC) carbon dioxide as a substitute for chlorofluorocarbons (CFCs) for a variety of specific and specialized applications[g]-[' in which the choices of environmentally acceptable alternatives are quite limited. Today most chemists are familiar with the critical point of a substance as defined by the critical temperature and pressure, T, and P, respectively, but often their knowledge goes no further. The behavior of gases close to the critical point is so far removed from ideal that, the topic is usually ignored. Hence, many scientists are unaware that near-critical fluids display unusual and intriguing properties which can lead to new chemistry with exciting applications. A substance is said to be in the gaseous state when heated to temperatures beyond its critical point. However, the physical properties of a substance near the critical point are intermediate between those of normal gases and liquids, and it is appropriate to consider such supercritical fluid as a fourth state of matter. For applications such as cleaning, extraction and chromatographic purposes, supercritical fluid often has more desirable transport properties than a liquid and orders of magnitude better solvent properties than a gas. Typical physical properties of a gas, a liquid, and a supercritical fluid are compared in Table 1. The data show the order of magnitude and one can note that the viscosity of a supercritical fluid is generally comparable to that of a gas while its diffisivity lies between that of a gas and a liquid. Table 1. Physical Properties of a Gas, Liquid, and Supercritical Fluid ~~~~~~~~~~

Density g/ml Gas Liquid Supercritical Fluid

1 x 10-3 1.o 3 x 10-I

~

Diffisivity cm2/sec 1 x 10-1 5 x lo4 1 x 10-3

Dynamic Viscosity glcm sec 1 x 10-4 1 x 10-2 1 x 10-4

The Supercritical State

3

2.0 PHYSICAL TRANSFORMATIONS OF PURE SUBSTANCES

The simplest applications of thermodynamics to chemically significant systems involve the phase transitions that pure substances undergo. Thephase of a substance is a form of matter that is uniform throughout in chemical composition and physical state. The word p h e comes fiom the Greek word for appearance. Thus, we speak of the solid, liquid, and gas phases of a substance, and of different solid phases distinguished by their crystal structures (such as white and black phosphorus). A p h e transition, spontaneous conversion of one phase to another, occurs at a characteristic temperature for a given pressure. Thus, at 1 atm, ice is the stable phase of water below OOC, but above O°C the liquid is more stable. The difference indicates that, below OOC, the chemical potential of ice is lower than that of liquid water, p(so1id) < p(1iquid) (Fig. l), and that above OOC, p(1iquid) < p(so1id). The transition temperature is the temperature at which the chemical potentials coincide and p(so1id) = p(1iquid).

Gas

0

Stable Solid State

1 staMB I

Liquid state

Temperature

I I

Stable & e o i State

I I ,

Figure 1. The schematic temperature dependence of the chemical potential of the solid, liquid, and gas phases. The phases with the lowest chemical potential are most stable at that temperature (shaded area).


4

Figure 2 is a phase diagram or temperature-pressure chart showing the phase transitions for a substance such as carbon dioxide. It is a graphical way to summarize the conditions under which the different states of that substance are stable. It maps the ranges or regions of pressure and temperature over which the various phases are thermodynamically stable (i.e., with lowest Gibbs energy, AG = AH - TU). The phase boundaries separating different regions delineate the values of P and T over which two phases coexist in equilibrium. As one can see, the solid-liquid phase boundary is a plot of the freezing point at various pressures, the liquid-vapor boundary is a plot of vapor pressure of the liquid against temperature, and the solid-vapor boundary is a plot of the sublimation vapor pressure against temperature. Solid, liquid and gas phases are all in the equilibrium at the intersection of these three boundaries, known as the triple point.

1

Supercritical Fluid

i Crlt\c.l Pressur8

Liquid

I

cr1tic.i Point

Pressure

I

Temperature

CritlcaI ~emperature

Figure 2. A phase diagram showing the Merent phases and supercritical region. The critical point, critical pressure, and critical temperature are indicated.


5

3.0 CRITICAL CONSTANTS Figure 3 shows a series of isotherms (constant temperature lines) for a substance close to its critical point. Consider what happens when the volume of a sample of gas in a closed cylinder (initially in the state marked A in Fig. 3) is compressed at constant temperature by applying pressure with a piston. Near the point A, pressure of the gas rises in approximate agreement with Boyle’s law (the pressure and volume of a fixed amount of gas at constant temperature are related by Boyle’s law: P Y = constant). When the volume has been reduced to B, serious deviations from that law begin to appear. At a point C (-60 atm for COz), all similarity to ideal gas behavior is lost and suddenly the piston slides in without any further rise in pressure; this state is represented by the horizontal dashed line CDE. Examination of the vessel reveals that just to the left of C a liquid appears, and there are two phases separated by a sharply defined surface. As the volume is decreased from C to E through D, the amount of liquid increases and the gas condenses without any further resistance. The pressure corresponding to the line CDE, when both liquid and vapor are present in equilibrium, is called the vaporpressure of liquid at the experimental temperature. At point E, the sample is entirely liquid and any fbrther reduction of volume requires the exertion of considerable pressure, as is indicated by the sharply rising line to the left of E. Note that even a small reduction of volume from E to F requires a great increase in pressure. The isotherm (plot of pressure versus volume) at the temperature T, plays a special role in the theory of the states of matter. An isotherm behaves in accordance with the gas laws slightly below T,. At certain pressure, a liquid condenses from gaseous state and is distinguishable from it by the presence of a visible interface. If, however, the compression takes place at T, a surface separating two phases does not appear and the volumes at each end of the horizontal part of the isotherm have merged to a single point, critical point of the gas. The temperature, pressure, and molar volume at the critical point are called the critical temperature To criticalpressure P, and critical molar volume V, of the substance respectively. Collectively, P,,V, and T, are the critical constants of a substance.

6


2o ; 0

t 0

/

QC

0.2

0.4

V,AL

0.6

mor’)

Figure 3. Experimental isotherms of carbon dioxide at several temperatures. The critical isotherm is the isotherm at the critical temperature (3 1.04OC).

At and above T,, the sample has a single phase that occupies the entire volume of the container. Thermodynamically such a phase is, by definition, a gas. However, the unusual physical properties of a substance close to its critical point warrant special terminology. The single phase that fills the entire volume at T > T, may be much denser than one normally considers typical of gases, and the name supercritical fluid is preferred and the area above and to the right of the critical point in Fig. 2 is known as the supercritical region. A fluid in this region is neither a liquid nor gas-but possesses some of the properties of each. At this stage, no matter how much pressure is applied, a fluid in this region will not condense and no matter how much the temperature is increased, it will not boil. Since these materials are neither liquid or gas, we call themfluids.

The Supercritical State 4.0

7

CRITICAL AND BOILING POINTS

The significance of the supercritical state may also be appreciated by comparing liquid-vapor transitions at constant pressure (open vessel) and constant volume (closed vessel). The behavior of a liquid heated in an open vessel differs significantly fiom that of a liquid in a sealed vessel (Fig. 4). In an open vessel, the liquid vaporizes from its surface as it is heated. The vaporization can occur throughout the bulk of the liquid at the temperature at which its vapor pressure would be equal to the external pressure, and the vapor can expand freely into the surroundings. The condition of fiee vaporization throughout the liquid is called boiling. Temperature at which the vapor pressure of a liquid is equal to the external pressure is called the boiling temperature at that pressure. Note that a liquid does not suddenly start to form a vapor at its boiling temperature; even at lower temperatures there is an equilibrium between the liquid and its vapor. At the boiling point, the vapor pressure is great enough to overcome atmospheric pressure and vaporization can occur freely. For the special case of an external pressure of 1 atm, the boiling temperature is called the normal boiling point.

A

B

C

Figure 4. (A/ A liquid in equilibrium with its vapor. @) When a liquid is heated in a sealed container, the density of the vapor pressure increases and that of the liquid decreases slightly. Due to vaporization, the quantity of liquid decreases. (C) At critical temperature, interface between the fluids disappear, at which condition the densities of liquid and vapor are equal.

8


When a liquid is heated in a sealed vessel, boiling does not occur. Instead, the temperature, vapor pressure, and density of the vapor rise continuously. At the same time, density of the liquid also decreases as a result of its expansion. There comes a stage at which density of the vapor is equal to that of remaining liquid and the surface between two phases disappears. The temperature at which the surface disappears is the critical temperature, T,, and the corresponding vapor pressure is the critical pressure, P,. At and above this temperature, a single uniform phase fills the container and an interface no longer exists. That is, above the critical temperature the liquid phase of the substance does not exist. For a reduced temperature (TR= T/TJ in the range 0.9-1 2, the reduced fluid density (pR=p/p,) can increase from gas-like values of 0.1 to liquid-like values of 2.5 as the reduced pressure (PR = PP,) is increased to values greater than 1.O. But as TR is increased to 1.55, the supercritical fluid becomes more expanded and reduced pressures greater than ten are needed to obtain liquid-like densities. By operating in the critical region, the pressure and temperature can be used to regulate density, which in turn regulates the solvent power of a supercritical fluid. When utilizing additives in conjunction with a single solvent, binary phase diagrams become usehl. Actually single component PT phase diagrams are of limited value for operations such as supercritical chromatography or extraction purposes, because any analyte or solvent present generates a binary phase system. When injecting a sample dissolved in a liquid, ternary or even higher phase diagrams must be considered. Fortunately, in many cases one can look at an ideal diluted solution without pondering over state parameters too much, so it is appropriate to look at the properties of binary mixtures which can differ profoundly from single components. For example, it is not possible to interpolate the critical point for carbon dioxide/ methanol mixtures from the critical point of each component. Depending on the mole fraction and temperature, there exists a distinct pressure maximum, while the critical temperatures lie between those of the pure compounds.[']


9

5.0 SUPERCRITICAL STATE

Supercritical fluids make ideal solvents because their density is only about 30% that of a normal fluid, a factor sufficient to provide for good solvent capability, but low enough for high difksivity and rapid mass transfer. A sample of gas is supercritical whenever its temperature and pressure are above their critical values, but in practice the operating temperatures are not far above T,. A simple way to predict the solvent characteristics of a low-boiling substance in its supercritical state is to compare the boiling point (Tb) and critical temperature (T,) of substance. The Guldberg’s rule:

can be applied[12]for low-boiling point liquids. As discussed earlier, supercritical fluids posses properties which resemble gases and liquids. The diffision coefficient of a SCF is somewhere in the middle between those for gases and liquids and, the viscosity is similar to that of gases while the density is close to that of liquids. If only density is the requirement, one could operate near the critical point. However, the density changes with a maximum rate near the critical point and small pressure differences have a marked effect on density. While raising pressure increases the density, raising temperature will always decrease the density, but will only decrease the solubility at pressures below 350 bar. Solvent properties improve with fluid density and hence can be strongly influenced by changes in pressure. The effects of temperature are more complex.

6.0

SOLVENTS AND ANTISOLVENTS

Supercritical fluids have the ability to dissolve nonpolar solids, and this is what makes them usehl for various applications, especially cleaning, though the fact that solids can dissolve in gases remains counterintuitiveto most scientists. A number of attempts have been

10


made to quanti@ the solvent strength of supercritical fluids, usually through solvatochromic shifts in the W-visible absorption bands of organic The precise positions of these bands change as the density of the fluid increases; the solvent strength of the fluid at a particular density is determined by finding a conventional solvent of known strength in which the dye displays an absorption band at the same wavelength as in the supercritical fluid. Although supercritical fluids can rarely match the solvent properties of conventional solvents such as hydrocarbons, they can be good solvents for highly fluorinated compounds. For example, fluorocarbons are notoriously insoluble in conventional solvents but are very soluble in supercritical carbon dioxide. This exceptional solubility has been exploited by DeSimone and colleagues[14]to replace CFCs in the synthesis of fluoropolymers. The solubility of solids depends on the density of the fluid and hence on the applied pressure; increasing the pressure increases solubility, while reducing the pressure precipitates the solid. This tuning of solubility can be exploited to control the composition and morphology of polymers and similar materials.

7.0

SOLUBILITY ISSUES OF SUPERCRITICAL FLUIDS

For supercritical fluids, the solvent strength of a given fluid is primarily dependent upon its density and pressure. It is much easier to directly measure and control the pressure of a supercritical fluid in a given operation than to measure and control the fluid density. It is possible, however, to determine a relationship exists between the fluid pressure and its density, thus allowing indirect measurement and control of the density. For an ideal gas, the relationship is simply, PV/RT= 1, where Vis the molar volume (reciprocal of the the mass molar density). From the molecular weight of the gas density, p , can be calculated as MN. The simple equation breaks down at the high densities characteristic of supercritical fluids, but, the work ofPitzer et al.[15][16] allows the ideal gas law to be extended by introducing a term called the compressibility factor, z. It is a

w),


11

fbnction of the pressure, temperature, and molecular identity of the fluid. The gas law then becomesPV/RT=z. Pitzer was able to reduce the molecular identity terms ofz to a single number called the acentric factor, ar. This factor attempts to account for both molecular size and shape. The value of the acentric factor was taken to be the ratio of the vapor pressure of the substance at 70% of its critical temperature to that at its critical temperature. References 15 and 16 detail the reasons for this choice. Thus, at a given temperature, pressure, and known acentric factor, z can be determined from the tabulated values which are given in Ref 17. This allows V to be determined from the gas law (y = zRTP) and thus the mass density to be determined if the molecular weight of the fluid is known. Further, the solvent power (as measured in milligrams of solute per gram of solvent) increases exponentially with the increase in pressure above the critical pressure (greater than 350 bar for carbon dioxide), with relatively little increase in solvent density. At these high pressures (350-700), a rise in temperature causes an increase in the solvent density, because a concurrent increase in vapor pressure of the solute overcomes the decrease in solvent density and leads to a significantly greater solvent power. The solvent power of the supercritical fluid increases exponentially with relatively small increases in temperature or pressure in this high pressure region of 350-700 bar and 50-1 10°C. During the early 1980s, John Freidrich and Egon Stahl discovered that at comparatively higher pressures and temperature (above 350 bar and SOOC), the solvent power of supercritical fluids is much greater than one would expect from ideal solubility calculations. The actual solubility deviates from the ideal solubility (which can be calculated fiom the vapor pressure of the condensed phase) by several orders of magnitude because of intermolecular interactions that assist in the solvation of the solute. This surprising but interesting apparent increase in the vapor pressure of a poorly soluble substance can be most easily understood as an “evaporation,” because the solute becomes part of the gas phase. The ratio of the much greater actual solubility to the ideal solubility is called the enhancementfactor.

12


8.0 HILDEBRAND SOLUBILITY PARAMETER

The Hildebrand solubility parameter, 6, is a semi-quantitative entity related to the thermodynamic properties of dense gases (supercritical fluids) and solutions.[’*] The solubility parameter in calories per cubic centimeter is calculated from the equation:

where P, is critical pressure in atmosphere, p, is the reduced density and pxliq) is the reduced density of liquid.[19] Reduced density is the ratio of supercritical density at a given pressure and temperature to critical density, p, whereas the critical density is the density of supercritical fluid at its critical pressure and critical temperature.

Based on the semi-empirical calculations as shown in Ref. 18, the rdliq) is given by: pdliq) = 2.66/pC From Eqs. 1-3, the following relationship can be derived to relate the Hildebrand solubility parameter with the density of supercritical fluid. Eq. (4)

6 = 1.25PC’”p/2.66= 0.47 P,’”p

Equation 4 was used to calculate the Hildebrand solubility parameters for various supercriticalfluids. Here p is the density ofthe supercritical fluid which is related to the pressure and temperature as described earlier. The enhancement of solvent power obtained by compressing a gas into its critical region can be demonstrated dramatically. An estimate of the solubility of a solid in an SCF solvent can be made using the following expression:[7]


13

where the subscript 2 refers to the solid component, PFb is the sublimation pressure of the solid, v $ is the molar volume of the solid, and 42 is thefugacity coeflcient (a factor that accounts for specific repulsive and attractive interactions that occur between solvent and solid component molecules). On comparing the experimentally observed solubility of naphthalene b2)at 100 bar and 12OC with the solubility calculated of the SCF phase (assumed to behave like ideal gas), it can be found that: y2 observed ly2 calculated = 16,156.

9.0

CARBON DIOXIDE

Gaseous carbon dioxide is very inefficient as a solvent for liquid and solids, but as its pressure is increased, accompanied by a concurrent increase in density, its extractive power improves. The solubility of organic compounds in carbon dioxide increases with higher density and temperature. Liquid carbon dioxide at 2OoC and equilibrium pressure of 56 atm is a usehl solvent because its density is relatively high at 0.77 g/ml. In 1954, Francis reported the mutual solubilities of liquid carbon dioxide with each of 261 other substances and presented triangular graphs of 464 ternary systems in a single paper.[*'] The solvent properties are retained as the temperature is increased towards 3 l0C, the critical temperature for carbon dioxide, there are no differences in any of the properties, and in fact two phases no longer exist. Above 3 l0C, there exists only one phase, a supercritical fluid, as illustrated in Ref 7 and in Fig. 5 . The phase diagram for carbon dioxide is shown in Fig. 5 . The features to notice include the positive slope of the solid-liquid boundary (the direction of this line is characteristic of most substances), which indicates that the melting temperature of solid carbon dioxide rises as the pressure is increased. Note also that, as the triple point

14


(where the three phase boundaries coincide) lies above 1 atm, the liquid cannot exist at normal atmospheric pressures whatever the temperature, and the solid sublimes when left in the open (hence the name d y ice). To obtain the liquid, it is necessary to exert a pressure of at least 5.1 1 atm. Cylinders of carbon dioxide generally contain liquid and compressed gas; at a temperature of 25OC that implies a vapor pressure of 67 atm. When the gas squirts through a restrictive orifice, it cools by the Joule-Thomson effect, so that when it emerges into a region where the pressure is only 1 atm it condenses into a finely divided snow-like solid. Table 2 shows the relationship of a complete range of pressure-temperature-density for SC COz.

0.928

F

0.464

6 c E 0

7.38

73.8 Pressure in bar

738

Figure 5. Phase diagram of supercritical and near critical carbon dioxide, showing the density and pressure relationship. The shaded area represents the supercritical state.

The Supercritical State I5

16


10.0 CANDIDATES FOR SUPERCRITICAL FLUIDS

Table 3 (constructed from Refs. 21-24) lists various physical properties for a variety of candidate SCF solvents. A cursory inspection of the entries in this table shows that many hydrocarbons have a critical pressure close to 45 bar and the critical temperature of SCF solvents increases as the molecular weight of the solvent increases or as the polarity or intramolecular hydrogen bonding of the solvent increases. This means that the solvent’s vapor pressure curve extends to very high temperature. Water is also included in the table to make one point-the solvent that we are all most familiar with is a poor candidate from both engineering and safety standpoint. The critical temperature and pressure are among the highest for common solvents. Ammonia is very unpleasant to work with since a fume hood or other venting precautions are needed to keep it out of the laboratory atmosphere. One of the alternative fluids of potential interest is nitrous oxide. It is attractive since it has molecular weight and critical parameters similar to carbon dioxide, yet has a permanent dipole moment and is a better solvent than carbon dioxide for many solutes. There are evidences of violent explosive reactions of nitrous oxide in contact with oils and fats. For this reason, nitrous oxide should be used with great care and is not suitable as a general purpose extraction fluid.[*’] As Table 3 shows, the critical temperatures of gases and liquids can differ by hundreds of degrees; this difference suggests the use of specific supercritical fluids for specific applications. For example, because the critical temperatures of carbon dioxide, ethane, and ethylene are near ambient, they are attractive solvents for processing heat-sensitive flavors, pharmaceuticals, labile lipids, and reactive monomers. Substances that are less temperature-sensitive, such as most industrial chemicals and polymers, are readily treated with C3 and C4 hydrocarbonswith critical temperatures in the range 100-1 5OOC; C3 and C4 hydrocarbons are generally better solvents for polymers than C2 hydrocarbons. Still higher molecular weight hydrocarbons, such as cyclohexane and benzene, with high critical temperatures of 250-3OO0C,are used to process nonvolatile substances such as coal and high molecular weight petroleum fractions.[7]

The Supercritical State 17



19

11.0 CONCLUSIONS

After many years of academic interest, supercritical fluids have emerged as revolutionary industrial solvents. In addition to its unique solubility characteristics, a SCF possesses certain other physiochemical properties that add to its attractiveness. For example, even though it possesses a liquid-like density over much of the range of interest, it exhibits gas-like transport properties of difisivity and viscosity. Additionally, the very low surface tension of SCFs allows facile penetration into microporous materials. For cleaning issues, this property plays a significant role since SCF can penetrate into and in between the yarns and clothes. The properties of gas-like difisivity and viscosity, zero surface tension, and liquid-like density combined with the pressure dependent solvent power of a supercritical fluid have provided the impetus for applying SCF technology to a gamut of separations problems experienced in many segments of the industry. The practical as well as fbndamental principles governing SCF processing are considered in detail in later chapters with particular reference to the various aspects of cleaning. From the earlier reported literature, one can conclude that a SCF treatment system can be tailored to realize most desired applications by adjusting pressure andor temperature and other influencing parameters. With the more detailed understanding of various issues, it can be realized that they are ideal environmental friendly candidates to replace ozone depleting chlorofluorocarbons. The following chapters will reinforce this theme based on the principles and practices of exploiting the unique and unusual physical properties of supercritical fluids explained in this chapter.

20


REFERENCES 1. Mendelejeff, D., Pogg. Annalen Physik Chemie, 141:618 (1870) 2. Hannay, J. B. and Hogarth, J., On the solubility of Solids in Gases, Proc. Roy. SOC.London, 29:324 (1879); 30:178 (1880) 3. King, J. W., Johnson, J. H., and Friedrich, J. P., Extraction of fat tissue from meat products with supercritical carbon dioxide, J. Agric. Food Chem., 37:951 (1989) 4. Wehling, R. L., Froning, G. W., Cuppet, S. L., andNiemann,L., Extraction of cholesterol and other lipids from dehydrated beef using supercritical carbon dioxide, J. Agric. Food Chem., 40:1204 (1992) 5. Bruno, T. J. and Ely, J. F., Supercritical Fluid Technology; Review in Modern Theory and Applications, CRC Press (1991) 6. Williams, D. F., Extraction with supercritical gases, Chemical Eng. Sci., 36:1769 (1981) 7. McHugh, M. A. and Krukonis, V. J., Supercritical Fluid Extraction: Principles and Practice, 2nd Edition, Butterworth-Heinemann ( 1994) 8. McHugh, M. A., Krukonis V. J., and Pratt, J. A., Supercritical fractionation of polymers and copolymers, Trends. Polym. Sci., 2:301 (1994) 9. Pirrotta, R. and Pava, T., Supercritical carbon dioxide-performance with potential, Precision Cleaning, II:37 (1994) 10. Agopovich, J. W., Fluorocarbons and supercritical carbon dioxide service niche needs, Precision Cleaning, 111:15 (1995) 11. Spall, W. D., Supercritical carbon dioxide precision cleaning for solvent and waste reduction, Int. J, Environmentally Conscious Design and Manufacture, 231 (1993) 12. Wenclawiak, B., Analysis with Supercritical fluids: Extraction and Chromatography,Springer Verlag, Heidelberg (1992) 13. (a) Yonker, C. R., Frye, S. F., Kalkwarf D. R., and Smith, R. D. Characterizationof supercritical fluid solventsusing solvatochromicshifts, J. Phys. Chem., 90:3022(1986); (b) O’Neil, M. P., Kruus, P., and Burk, R. C., Sohatochromicparametersand solubilitiesin supercriticalfluid systems, Can. J. Chem., 71:1834 (1993) 14. DeSimone, J. M., Guan Z., and Elbernd, C. S., Science, 257:945 (1992); Chem. Br., 29:lO (1993) 15. Pitzer, K. S., J. Am. Chem. SOC.,77:3427 (1955) 16. Pitzer, K. S., Lippmann, D. Z., Curl, R. F., Jr., Huggins, C. M., and Petersen, D. E., J. Am. Chem. Soc., 77:3433 (1955) 17. Fjeldsted, J. C., Jackson, W. P., Peaden, P. A., and Lee, M. L., J. Chrom. Sci., 21:222 (1983)


21

18. Hildebrand, J. H. and Scott, R. L. The Solubilities of Nonelectrolytes, 3rd Edition, Reinhold, New York (1950) 19. Giddins, J. C., Myers, M. N., McLaren, L., and Keller, R A., High pressure Gas Chromatography of Nonvolatile Species, Science, 162:67 (1968) 20. Francis, A. W., Ternary Systems of Liquid Carbon Dioxide, J. Phys. Chem., 58: 1099 (1954)

C.,and Laby, T. H. (Eds), Tables ofphysical and Chemical Constants, Longman, London, ( 1973) Gray, D. E., (Ed), American Institute ofphysics Handbook, McGraw Hill, New York (1972) Atkins, P., Physical Chemistry, Fifth Ed., W. H. Freeman & Co., New York (1994) Lide, D. R, (Ed)., Handbook of Chemistry and Physics, CRC Press, Boca Raton (1993) Raynie, D. E., Warning concerningthe use of nitrous oxide in supercritical fluid extractions,Anal. Chem., 65:3127 (1993)

21 Kaye, G. W. 22 23. 24. 25.

Solubility in Supercritical Fluid Cleaning Kenneth E. Laintz, L. Dale Sivils, and FK Dale Spa11

ABSTRACT With the ban of chlorofluorocarbons imminent, replacement solvents for many industrial processes must be implemented. The use of supercritical carbon dioxide as a cleaning solvent offers many advantages for the cleaning of selected materials. For supercritical CO, to be effective as a replacement solvent for cleaning, the contaminants normally removed using organic solvents must also exhibit solubility in supercritical CO,. For this reason, solubility as it relates to cleaning is discussed in this chapter. A basic model for contaminant removal from a surface is presented. The implications of this model as it relates to cleaning and contaminant solubility are also discussed.

22

Solubility in Supercritical Fluid Cleaning

23

1.0 INTRODUCTION

When consideringthe cleaning of an item, the type of soil that is to be removed must also be considered. In general, the type of soil and the degree of final cleanliness determine the cleaning method. Cleaning processes remove contaminants by overcoming the forces that bind the contaminant on the surface. These forces are varied and can range from electrostatic to molecular attraction forces. Commonly employed cleaning methods utilize mechanical abrasives, detergents, chemical reactants, and solvents. Cleaning with supercritical fluids (SCFs) generally falls into the category of cleaning with solvents. Solvent cleaners work by dissolving contaminants found on the substrate being cleaned. This dissolution process is simply illustrated in Fig. 1. In the figure, the contaminant is illustrated by the globular shapes sitting on the surface to be cleaned. The dissolution of the contaminants occurs as the solvent surroundsthe contaminants and lifts them from the surface, thus leaving it clean. Of course, this is a very simplistic representation, and many factors are involved in the dissolution of a contaminant. In order for a solvent to effectively dissolve a contaminant in this manner, the contaminant must be soluble in the solvent. In basic terms, solubility is the maximum amount of a substance, or solute, that can be dissolved in a given quantity of solvent. In general, for the dissolution of solutes or contaminants, the concept of like-dissolves-like is often used. This basically means that an inorganic or polar substance such as salt generally dissolves in a polar solvent such as water. On the other hand, a nonpolar compound such as an aliphatic hydrocarbon dissolves in a nonpolar solvent such as hexane. Besides aqueous based cleaning techniques, most common solvent cleaning methods use nonpolar solvents such as 1, 1, 2trichlorotrifluoroethane, or Freon- 113, for the removal of hydrocarbon-based soils such as oils and greases. Since chlorofluorocarbon solvents like Freon-1 13 have been banned by the Montreal Protocol, alternative solvents with similar dissolution capabilities are needed. One replacement solvent for solvent-based cleaning methods is supercritical carbon dioxide.

24

Supercritical Fluid Cleaning Contaminant on Surface

Contaminant Dissolved Off of Surface

Figure 1. Simplified schematic of a contaminant dissolving off a surface.

In general, knowledge of the solubility of a particular compound in an SCF is important when considering SCF processing of that particular compound. For this reason, many theoretical calculations and experimental measurements have been made to determine the solubilities of a wide variety of compounds in various supercritical fluids as a hnction of temperature and pressure. Most of these measurements have been made on pure compounds, and in an industrial setting, the desired compound or contaminant often exists as a mixture of several compounds. In such cases, pure component solubilities can only suggest at extractability or cleanability. This knowledge does, however, provide a basis for designing an SCF processing or cleaning scheme. Many articles have appeared in the literature that list solubilities of a wide variety of pure components under a variety of temperature and pressure conditions. A recent


25

review article by Bartle, et al., compiles solubility data on a wide variety of solids and liquids in supercritical CO, and is an excellent source for initial compound solubility data.['] When using supercritical CO, as a replacement solvent for cleaning methods, the solvent strength of the SCF must be considered. The solvent strength of supercritical CO, is dependent upon temperature and pressure, but generally, it is considered a nonpolar solvent. Therefore, using a Zikedissolves-like scenario, supercritical CO, is best suited for the removal of nonpolar hydrocarbon-based soils like various machining and lubricating fluids. These compounds are all mixtures of many single components, and accurate solubility determinations of such mixtures have yet to be made. For this reason, the intent of this chapter is to present a basic model for contaminant removal from a surface. The implications of this model relating to cleaning and contaminant solubility will also be discussed.

2.0

CONTAMINANT SOLUBILITY

Solubilities are defined either as the mole fraction of solute in solution, x, or by the amount of solute per unit volume, or concentration, C, at saturation. The concentration in moles per unit volume is given by:

where Vis the molar volume of the pure fluid. Contaminant volatility and SCF solvating power are two major factors that determine the solubility of a contaminant in an SCF. Supercritical fluid solvating power is primarily a hnction of density. In general, higher densities result in higher solvating power. These higher densities are most readily achieved through increasing pressure. However, temperature is an important variable in contaminant solubility since the volatility of a compound is directly related to temperature. Depending on equipment setup, increasing the temperature in a cleaning system can

26


reduce solvent density, but the increase in contaminant volatility generally overcomes the decrease in density and enhances contaminant removal. The solubility of a solid or semisolid in a liquid is related to the heat of hsion of the solid and the temperature of the solution by:

where x is the mole fraction of the dissolved solute, Myis the heat of hsion, T is the melting point of the solute and Tis the temperature of f the solution. Equation 2 is valid if the change in the heat capacity and volume at different T’s is ignored. Ideal solution theory also requires that there is only dispersion force interaction between the solute and solvent. It can be seen from Eq. 2 that an increase in solution temperature will increase the mole fraction of the dissolved solute. In an SCF system at constant pressure, the solubility of a solute will fall initially with increasing temperature because the SCF density decreases, and hence its solvating power decreases, with increasing temperature. However, as the temperature rises, the volatility also rises, and eventually this effect exceeds the effect of the falling solvation and the solubility rises. Section A-B in Fig. 2 shows that the solubility falls as the contaminant is diluted by the fluid. The rapid rise in solubility in Sec. B-C occurs at pressures quite higher than the critical pressure because of the rapid rise in density, and therefore solvating power, of the SCF at around this pressure. This region has been defined by King as the thresholdpressure which is the pressure at which the solute begins to dissolve in the SCF.f2]Obviously, this pressure is technique dependent and varies with the analytical method sensitivity used to measure the solute concentration in the SCF. A decrease in solubility, as shown in region C-D, may occur at higher pressures due to repulsive forces that may squeeze the solute out of solution. For moderately volatile solutes, a rise in solubility, as shown in section D-E, can occur if there is a critical line in the mixture phase diagram at higher pressures.

Solubility in SupercriticaJ Fluid Cleaning

27

f\

Pressure Figure 2. Schematic representation of compound solubility in a supercritical fluid as a function of pressure.

Solubility can also be enhanced by the presence of other compounds. This phenomenon is caused by one or more compounds acting as solubility enhancers for other compounds present on a surface. This phenomenon is sometimes called the local cosolvent effect. A typical method of enhancing contaminant solubility is through the addition of a small amount of secondary solvent to the SCF cleaning system. Alcohols are commonly used in this manner to increase solubilities of more polar contaminants. However, more subtle local cosolvent effects have been observed. Perhaps a classic example was first reported by Kurnik and Reid.[3] In their study, they observed that the solubilities of both naphthalene and benzoic acid in supercritical CO, were enhanced by 107% and 280%, respectively, when both species were present. It has also been shown that there needs to be enough of a secondary component present in solution about the local contaminant environment to enhance the solubility of another compound.[4] This example demonstrated that an excess of phenanthrene promoted the solubility of anthracene in supercritical CO,, but since anthracenewas only present in very small quantities, it did not help to enhance the overall solubility of phenanthrene. A

28


similar result was seen in a cleaning study on the removal of TlUM?SOL, which is a water miscible cutting fluid, from stainless steel, copper, aluminum, and brass coupons.[5] In this example, the metal substrates were cleaned using a SUPER SCRUB^ CO, cleaning system, manufactured by EnviroPro Technologies equipped with a 60 liter cleaning vessel. TRlM?SOL was shown to have an average removal rate of 44.5% from the metal surfaces when it was applied at a contamination level of 10 pg/cm2. In contrast, the removal efficiency increased 167% to a removal rate of 74.5% when the contamination level was increased to 50 pg/cm2. This phenomenon is caused by one or more components acting as solubility enhancers for other components. With more contaminant present at 50 pg/cm2, it is possible that the overall solubilities of the contaminants were increased to yield the higher extraction efficiency. More detailed theoretical discussions of solubility and solute-solvent interactions and these interactions as applied to SCFs can be found else~here.[~][~I

3.0

MODELS FOR CLEANING

In order to present a cleaning model, there are several factors that must be considered. The first factor under consideration is the partition coefficient. This coefficient is a ratio ofthe concentration of a contaminant on the surface to its concentration in the solvent and can be represented as:

Eq. (3)

P = [c,] [CFI

where P is the partition coefficient, [GI is the concentration of the contaminant on a surface, and [C,] is the concentration of the contaminant in the SCF solvent. In the case of cleaning, if [C,] is plotted as a function of [C,] as shown in Fig. 3, the distribution behavior of a contaminant between the surface and the SCF solvent can be illustrated. The curves shown in the figure relate the amount of


29

contaminant on the surface to the amount in solution in the cleaning solvent. Initially, [q] is high, and as the cleaning cycle progress, [GI decreases as [C.’] increases. In a cleaning process, line C of Fig. 3 represents a case where the partition coefficient is constant over a wide range of contaminant concentrations. Typically, however, a relationship similar to lines A or B is usually observed. Curve B is indicative of a contaminant with limited solubility in the cleaning solvent or where surface interactions limit contaminant dissolution. CurveA can be thought of as an adsorption isotherm. In this case, the concentration of contaminant in solution is related to the amount adsorbed onto a surface. As the curve approaches some asymptotic value, contaminant adsorption on the surface becomes the controlling factor in the dissolution of contaminant from the surface regardless of solubility. In general, such a curve results for a contaminant that is highly soluble in an SCF thus indicating the importance of the adsorption isotherm to SCF processing.

Figure 3. Plot of contaminant partition coefficient as a function of surface, [CsJ and fluid, [CF],concentrations.

30


Overall, three different partitioning phenomena must be considered when cleaning a contaminant fiom as surface. These three different partition coefficients, represented as P,are illustrated in Fig. 4. The contaminant can partition between the surface and the solvent directly which is represented as pSF, where the subscript S corresponds to the surface, L to the liquid contaminant, and Fcorresponds to the SCF. The term pSF corresponds to the adsorption isotherm is which was previously discussed (curve A in Fig. 3). The term the partitioning of the contaminant between the surface and the bulk contaminant phase, liquid, in this case, which can be thought of as an oil, for example. It should be noted that the term is equally applicable to a solid contaminant. Finally, the contaminant partitions between the bulk liquid and the cleaning solvent, represented as P 4F' The partitioning processes are usually expressed in terms of equilibrium concentrations. This is because contaminant partitioning occurs back and forth between two phases. For example, in the case of PW, the contaminant can partition into the cleaning fluid and back into the liquid adsorbed on the substrate surface. If the contaminant is not very soluble in the cleaningfluid or ifthere is strong surface adsorbtion, the partitioning into the cleaning solvent is going to be small. However, this particular partition coefficient, PW, is directly related to solubility. Since solubility is related to contaminant vapor pressure and temperature, increasing the temperature of the cleaning cycle increases partitioning into the cleaning fluid, thus promoting overall cleaning of the surface. One can ideally visualize the partitioning process during a cleaning cycle as the contaminant partitioning initially into the solvent with subsequent partitioning fiom the surface into the liquid. Finally, when the contaminant concentration on the surface reaches a point where essentially no liquid is left, the contaminant partitions directly into the solvent. The partition coefficients are not the only parameters that must be considered in the cleaning process. The d f i s i o n coefficient, D, encompasses molecular motion and the tendency for a molecule to move from a concentrated region, the contamination site, to more dilute regions, the bulk solvent. Again, the subscript Scorresponds to the surface, L to the liquid (or solid) contaminant, and F to the SCF. In order for contaminant removal and subsequent cleaning to occur,

GL


31

Bulk Flow

DF Boundary

Contaminant

Surface

Figure 4. Simplified model of contaminantremoval in a supercriticalfluid. This model is applicable to both solid and liquid contaminants.

the contaminant must enter into the bulk flow of the supercritical solvent. Several processes need to occur for this to happen. There is partitioning from the substrate surface into the bulk liquid of the contaminant. The contaminant then diffises to the surface in contact with the cleaning solvent, represented as DLin Fig. 4, where it can partition into the fluid phase of the boundary layer. This boundary layer can simply be thought of a viscous solvent layer where there is essentially only static, nonflowing solvent in contact with the surface of the item being cleaned and the contaminant itself Finally, the contaminant must diffise from the surface through the boundary layer and into the bulk flow, D in order to be removed from the F' surface. Since SCFs have gas-like difhsivities which are orders of magnitudes higher than liquids, it is often claimed that SCF extraction processes do not experience the mass transfer limitations of a liquid extraction process, thus yielding more efficient extraction or cleaning. This would be the case if the rate-limiting step of an extraction process is the actual transfer of a contaminant from the surface of a substrate to the extraction medium, denoted by the adsorption iso-

32 Supercritical Fluid Cleaning therm or PSF In reality, resistance to difision in the liquid phase, DL, is more likely the rate-determining step since extraction from the liquid phase into the SCF phase is probably more typical in an industrial cleaning situation. In this case, the gas-like diffision characteristicsof the SCF solvent will not have an enhancing effect on the overall mass transfer rate. Now, as it turns out, the thickness of the boundary layer is quite important in terms of contaminant removal. If the boundary layer is thick, the difision of the contaminant into the bulk solvent layer is slow which would contribute to a very slow dissolution process. An example of such a scenario might be a static dissolution or extraction of a contaminant where no flow is occurring within the cleaning or extraction chamber. In this case, the boundary layer is essentially the bulk solvent, and once this solvent becomes saturated with the contaminant, dissolution or cleaning ceases. On the other hand, if the boundary layer is quite thin, as in a dynamic or flowing process, contaminant difision across the boundary layer into the bulk fluid phase occurs rapidly thus facilitating cleaning. This then suggests that, for maximum cleaning potential, the flow of the solvent across the surface being cleaned needs to be maximized. Overall, the thickness of the boundary layer and contaminant solubility affect the time it takes for contaminant removal or extraction. This is illustrated in Fig. 5. Curve A of the figure represents several cleaning or extraction processes. Such a rapid removal rate is generally observed for a very soluble contaminant and for a soluble contaminant in a cleaning process that incorporates an efficient mixing system in order to generate a thin boundary layer. The extraction profile represented by curve A approaches an asymptotic value of an adsorption isotherm where surface interactions become dissolution limiting. Line C of Fig. 5 represents a case where the partition coefficient is constant over a wide range of contaminant concentrations. This can be due to a thick boundary layer as in a static process or due to a solubility-limited contaminant. Curve B of the figure can be thought of as process intermediate to those of curves A and C where the boundary layer difision is effected by contaminant solubility or surface interactions. Of course, all contaminant dissolution processes depend on contaminant solubility in the cleaning


33

solvent. If the contaminant is insoluble in the cleaning solvent, an increased flow would, at best, simply move the contaminant along the surface by shear velocity forces and not dissolve it.

C Mass

Extraction Time Figure 5. Plots of total contaminant extracted as a function of time.

In addition to contaminant partitioning across the boundary layer, there is also partitioning directly from the surface into the fluid phase. The partition coefficient associated with Pw. in Fig. 4 is dependent on solubility which in turn is dependent on temperature. The partition coefficients associated with PsL and PsF are hnctions of surface interactions. These interactions consist primarily of sorption phenomena such as chemisorption or physisorption. In the case of chemisorption, the contaminant is chemically sorbed onto the surface through hydrogen bonding or other attractive chemical forces. In the case of physisorption, the contaminant can be thought of as physically trapped on the surface. For example, one would expect physisorption to be small on a highly polished surface. On the other hand, a highly porous surface would be expected to provide physical barriers to contaminant removal from sites within the substrate matrix. In any case, both chemisorption and physisorption are related to the adsorp-

34


tion isotherm. If there is minimal surface interaction, P and P are SL. .SF large, thus indicating the partition into the SCF phase is umdirectional. Conversely, if these coefficients are small (meaning a high degree of surface interaction), it will be difficult to dissolve the surface layers of the contaminant. The SCF also partitions into the liquid or solid contaminant, and depending on the rate of this diffision, can either increase or decrease the kinetic rate of removal into the bulk solvent layer. Figure 6 shows three possibilities that can result when cleaning a contaminant from a smooth substrate surface. The top model shows the desired result of cleaning which occurs when the solubility of a contaminant is high enough to overcome any chemical or physical sorption mechanisms which bind it to the surface. Possibility I can occur when the contaminant is not sufficiently soluble in the SCF and has high intramolecular attraction which results in high surface tension. In other words, the contaminant likes itself better than the SCF or the substrate surface and subsequently forms beads on the surface. Possibility I1 can happen when PsL> pLF. The contaminant is partially solvated by the SCF but sorption to the substrate surface prevents complete contaminant removal. Bulk Flow

Desired Surface

Surface Bulk Flow

+

Possiblity I

Contaminant

Surface

Surface

Bulk Flow

> Possibility 11

---+

Surface

Surface

Figure 6. Simplified model of contaminant removal from a smooth surface.


35

Removing contaminants from rough surfaces can be a challenging proposition. Physical sorption plays a much larger role than chemical sorption when trying to dislodge contaminants. Supercritical fluids enjoy a distinct advantage over conventional liquid solvents when cleaning surfaces that have contaminants trapped in crevices or other physical barriers. This is because supercritical fluids have much lower viscosities and surface tensions than conventional liquid solvents. This enables SCFs to have much more penetration into microcrevices or cracks to dissolve trapped contaminants. Figure 7 shows what can happen when cleaning contaminants from rough surfaces. Obviously, the desired result is the complete removal of the contaminant from the crevice. This will occur if the contaminant is sufficiently soluble and it partitions to the bulk SCF flow. Less soluble contaminants can be removed if enough turbulence is applied to the system and the partially soluble contaminants are physically flushed from the physical barriers. Partial contaminant removal can occur if the contaminant’s solubility is low and the bulk flow turbulence is insufficient to flush out the trapped contaminants. Nevertheless, if a trapped contaminant has a low solubility and a high affinity to the substrate surface it will be very difficult to remove regardless of the solvent system used. Bulk Flow

---

Desired

4

v Surface

Surface

Bulk Flow

Possiblity

---+ 7 9

f

-

Surface

f

Contminmt

Surface

Figure 7. Simplified model of contaminant removal from a rough surface.

36


A final place in the overall cleaning process where contaminant solubility is an issue is in the separation process where the contaminant is removed for the solvent. In the separator, the spent cleaning solvent is passed from the cleaning chamber into a vessel where it expanded into a gas. The extracted compounds are collected in the separator, and the gaseous CO, is passed back into the flow stream to be used again in the cleaning process. If the contaminant is miscible with the fluid and has a high vapor pressure, the possibility of contaminant carry-over into the solvent reservoir is possible because such types of compounds are difficult to completely remove from the CO,. While feasible from a solubility standpoint, the removal of volatile contaminants in CO, is difficult from an operational standpoint. Current studies indicate that the concentration of low vapor pressure compounds must be kept below 1 part per million to achieve cleanliness levels at or below 1 pg/cm[21[8]Efficient separator designs are engineering issues, but these are indeed commercially available.

4.0

CONCLUSION

In general, the higher the solubility of the contaminant, the better the cleaning or removal efficiency. Solubilities can be increased by increasing the SCF density. This is achieved through increasing pressure. However, an SCF cleaning method should be optimized to run at the lowest pressure possible to reduce overall equipment costs. Solubility can also be increased through increasing temperature. For this reason, the cleaning method needs to optimized to run at as high a temperature as reasonably possible. In most cases, temperature and pressure must be optimized simultaneously in order to give the most efficient contaminant solubility. Contaminant solubility can also be increased by minimizing the solvent boundary layer by operating in a turbulently flowing system. Finally, contaminant solubility can be increased through the addition of a secondary compound such as a cosolvent. In summation, all variables must be optimized together to achieve the highest contaminant solubility and most efficient cleaning process regardless of the solvent choice


37

REFERENCES 1. Bartle, K. D., Clifford, A. A., Jafar, S.A., and Shilstone, G. F., Solubilities of Solidsand Liquids of Low Volatility in SupercriticalCarbon Dioxide, J. Phys. Chem. ReJ Data 20(4):713-756 (1991) 2. King, J. W., Fundamentals and Applications of Supercritical Fluid Extraction in Chromatographic Science, J. Chromatogr. Sci., (27):355364 (1989) 3. Kurnik, R. T., and Reid, R. C., Solubilityof Solid Mixtures in Supercritical Fluids, Fluid Phase Equifib.,8:93-105 (1982) 4. Kwiatkowski, J., Lisicki, Z., and Majewski, W., An Experimental Method for Measuring Solubilitiesof Solidsin SupercriticalFluids, Ber. Bunsenges. Phys. Chem. 88:865-869 (1984) 5. Williams, S. B., Laintz, K. E., Barton, J. C., and Spall, W. D., Elimination of Solvents and Waste by Supercritical Carbon Dioxide in Precision Cleaning,LosAlamos National Laboratory Report, LAUR-94-33 13 (1994) 6. Reid, R C., Prausnitz, J. M., and Poling, B. E., The Properties of Gases & Liquids, 4th Ed., McGraw Hill, Inc., New York (1987) 7. McHugh, M. and W o n i s , V., Supercritical Fluid Extraction, 2nd Ed., Butterworth-Heinemann, Boston ( 1994) 8. Spall, W. D., SupercriticalCarbon Dioxide Precision Cleaning for Solvent and Waste Reduction, Int’l. J. Environ. Conscious Design & ManuJ, l(2-4): 1-10 (1992)

Design for Cleanability Rammarthy Nagarajan

1.0 INTRODUCTION

In the rush to replace chlorinated solvents in cleaning applications with water or other inert solvents, the hndamental problem, i.e., how the contaminants deposit on and adhere to the substrate, has often been overlooked. In order for the solvency of the cleaning agent to become less important, the parts and their manufacturing processes have to be designed in such a way that capillary forces of adhesion do not become predominant, or such that the capillary medium is easy to dissolve. Design of microelectronic components and subassemblies for nonsolvent cleanability is crucial. Rules for design for cleanability are included here as Appendix 1. Some of these rules are self-explanatory, but others are discussed in more detail. Designing parts to be water-cleanable requires, for instance, that the print specifL the use of water-soluble or no-residue fluxes. Porous castings tend to retain moisture; low-carbon steels tend to rust easily. Certain elastomers tend to absorb moisture, swell, and later emit water vapors inside a product. If the part has features, 38

Designfor Cleanability

39

(e.g., high depth-to-diameter ratio blind holes) that are inherently difficult to dry, corrosion, pitting, etc., become a real possibility. Parts should be designed so that they do not provide locations for pools of the cleaning fluid to accumulate. Design of the part should not incorporate such tight tolerances on mechanical, geometrical, and material properties that machining it requires the use of a heavy oil as cutting fluid. Vendors typically have extensive experience in dealing with these issues, and their early input can significantly impact the producability of the part without recourse to oil-based vehicles. Of course, design and development engineers are not, nor are they expected to be, experts in cleaning or contamination control. But they do need to work closely with the cleaning engineers to achieve a design that allows a painless transition for the supplier from the solvent-cleaning practice of yore to more environmentally-conscious cleaning.

2.0

PROCESS DESIGN GUIDELINES

Once component and subassembly designs have been optimized for solvent-free cleaning, the responsibility resides with the vendor who is manufacturing the part to design his machining processes so that the end-product is suitable for the same purpose. Most vendors do not possess sufficient expertise in cleaning to make all the right decisions; then, it becomes imperative that they be assisted in achieving this desirable result. The following rules have been derived to serve as guidelines for part suppliers to design machining process such that the part is ultimately amenable to nonsolvent cleaning. The guidelines are stated here, and expanded upon in subsequent sections: 1. Use of water-soluble cutting fluids throughout the machining process

2. Minimizationof work-in-progress (WIP); implementation of continuous flow manufacturing (CFM) 3. Use of rinse tanks between successive machining operations

40

Supercritical Fluid Cleaning 4. Handling of parts

with gloves after final cleaning and during final inspection 5 . Parts coverage during storage 6. Shipping of parts in appropriate, clean, nonshedding bags or reusable containers 7. Use of water-soluble solder pastes and fluxes, in applications where soldering is involved These are discussed in order here. 2.1

Use of Water-Soluble Cutting Fluids

The importance of this item cannot be overstated, and is perhaps best illustrated with some production examples. There is some evidence that large (> 25 pm) metallic particles inside the headdisk assembly (HDA) of a direct access storage device (DASD) damages the disk substrate by means of high-velocity impingement. The large machined castings that enclose the HDA are obvious suspects because of their large exposed, machined surface area within the disk enclosure. When, in a particular disk drive, these suspicions were confirmed by particle counts on these surfaces, a systematic effort was undertaken to clean up the process at the supplier of the housings. As it turned out, two of the cutting fluids used by the supplier were water-soluble, but the third was not. The vendor was encouraged to contact the manufacturer of the coolant regarding the availability of a water-soluble substitute for the lone oil-based vehicle. The vendor was then informed about a water-soluble, synthetic grinding-fluid concentrate, which was essentially an aqueous solution of organics. The fluid was immediately incorporated into the process. The parts also underwent lapping, a notoriously contaminating operation. In the manufacture of the cast aluminum housings, the top and bottom halves were mated along several polished (lapped) surfaces. Lapping involved positioning the housings on a large lapping table where a continuous supply of lapping compound was fed in with a carrier. The lapping compound used by the vendor was alumina, and the carrier was oil-based. This necessitated that, after lapping, the polished surfaces had to be cleaned by immersing in a methyl

Design for Cleanability

41

chloroform or 1,1,l-trichloroethane(TCA) tank. It became apparent that, apart from dictating the use of a chlorinated solvent, the oilbased carrier was contributing significantly to the cleaning difficulties encountered later down the line. It was very difficult to remove from the surface, and probably even penetrated into the lapped surface to some extent. It was recommended to the supplier that the petroleum based vehicle be replaced by a water-soluble synthetic vehicle. This was adopted, tested, and proved to be viable. This substitution also eliminated the need for the 1,1,l-trichloroethane solvent, which, apart from its hazardous aspects and environmental impact, also tended to form chlorinated deposits on the surface of the part. Water-soluble coolants typically do not form hard-to-remove residues on part surfaces; they contain several buffers and stabilizers that prevent this from happening. In addition, in a wet state, they are easily rinsed off the surfaces. Even when (say, over a long weekend), residues do form on the surface, these can be removed by the use of a detergent. The soap may be chosen primarily based on industrial hygiene, waste disposal, and rinsability concerns. Enhancements in cleaning efficiency that could be attained by the use of more expensive chemicals are not likely to be more than marginal, provided machining chips and other debris are not glued onto the surface by oily substances. Use of water-soluble cutting fluids also puts less of a premium on just-in-time (JIT) manufacturing by minimizing the contamination consequences of allowing parts to dry during staging between operations. However, JIT (or CFM) will still help, as discussed later (Sec. 2.2). Extreme-pressure lubricants for machining on dead centers are perhaps the most difficult to convert fiom oil-based to waterbased. Center tapping operations typically require heavy lubricity in the fluid. However, one supplier recently made the switch with great success. With proper additives, a waterbased biostatic (controlled bacterial growth) coolant was found to have the necessary load carrying characteristics to handle high-severity tapping applications. This change, implemented in Oct. 1989, made possible the elimination of a TCA-dip rinse following the center tap, as well as the replacement of a Freon-ultrasonic cleaner at the end by an aqueous ultrasonic system. The resulting improvement in cleanliness (see


42

Fig. 1) was quite startling. Cleanliness data presented here were obtained by ultrasonically extracting residual surface contaminates from the part in a water-surfactant mixture for one minute, then sampling the water using a liquid particle counter. The counter can provide detailed particle size versus concentration information; for the sake of clarity, only the cumulative number of particles larger than 5 micrometers in diameter is shown here. Although the data presented pertain only to particulate contamination, the reduction in the quantity of residual organics on the surface was even more significant. Efforts are currently underway to implement waterbased coolants in grinding applications, traditionally a stronghold of oil-based carriers. The initial development costs in such an undertaking are easily offset by the anticipated paybacks in part cleanability. In many instances, current solvent rinse stages can be eliminated altogether, and dependency on solvent-cleaningis greatly reduced as well. There are still several challenges to overcome in this arena, particularly in the context of machining light metals like magnesium.

11

0

4

i

7 6

Figure 1. Improvement in part cleanliness after switching from an oil-base to a water-base coolant in a center-tapping application.


43

2.2 Minimization of WIP, Implementation of CFM Adaptation of a pull system at the part supplier, wherein the product is pulled through the manufacturing process at the rate of customer demand, is deemed preferable to the push system, in which the product is pushed through the process in batch quantities, from a waste-elimination-quality-improvement perspective. But the JIT manufacturing concept has excellent contamination control implications as well. Capillary adhesion of particles to surfaces is a timedependent phenomenon; with time, the liquid bridges that glue particles onto surfaces become stronger, and more difficult to break. In production situations, it has been observed that cleaning problems often arise when parts are left as WIP to dry over a long period. Wetting of the particlehubstrate interface by the cleaning medium becomes very difficult in such cases. If, in addition, the capillary liquid is oily or greasy, the solvent characteristics of the cleaning agent become of paramount importance, and chlorinated solvents, such as chlorofluorocarbons, become indispensable where inventory is stockpiled. Even under conditions where the residue is an agglomeration of particles glued to the surface by water, conventional cleaning methods sometimes prove inadequate; specialized cleaning, such as high pressure spraying or the use of strong chemicals, must then be resorted to. Parts of older vintage are always more difficult to c1ean;just in time cleaning, which minimizes the time delay between cleaning and processing, is the way to reduce contamination effectively. This philosophy is applicable not only to machined metallic components, but equally to advanced technologies, such as circuits manufacture, including lamination, photography, resists, screen printing, multilayer build, and automatic optical inspection equipment. While dried-on contamination could conceivably be loosened by presoaking in a solvent or a soap-water mixture, followed by an energetic dislodgment procedure, the added cycle time is often unacceptable. The problem becomes exacerbated when the end-user does not practice inventory control based on Jirst-in, Jirst-out principles. This double jeopardy usually results in a highly contaminated product. Thus, to get clean parts in, and to assemble clean

44


components into the final product, it is imperative that CFM be practiced down the line. 2.3 Rinsing after Machining

Immersion-type or spray-type rinsing after every machining operation can contribute significantly to producing an easily cleanable part. Although immersion tanks are currently more common among component suppliers, the frequency at which the rinse water gets dumped and refilled leaves something to be desired. Obviously, when the liquid in the rinse-tank is of the same color and consistency as the coolant or lubricant, it’s time to recharge! But the control over the process should be much tighter than that. Discharging of the rinsewater should be related to the number of parts processed, and this relationship should be established rationally before production rampup begins. If a spray-rinse is used, the water, which usually needs to be recycled, should be filtered for particles as well as chemicals. In the case of the cast aluminum housings discussed earlier, the solventimmersion rinse tanks, in use when oil-based cutting fluids were used, were replaced by low-pressure spray washers, which were recirculated and filtered to 5 pm. There should be such a separate precleaner to remove the cutting fluids and gross metal chips and shavings; doing this in the final cleaner, however attractive it may look from a logistics and cost point of view, is guaranteed to defeat the purpose by exceeding the filtration capability of the final-clean system, and by redepositing gross contamination on partially-cleaned parts. These interim cleaners also support the philosophy of just-intime cleaning. Chlorinated solvents and alcohols find perhaps their widest application as quick-dip-rinse media. When water-soluble coolants are installed, and solvent tanks become obsolete, they must be replaced by a waterbased rinse system. Total reliance cannot, and should not, be placed on the final cleaning system in view of the fact that specialized solvents cannot be used anymore. Water, while an excellent departiculating agent, is not as effective in removing nonpolar contaminants from surfaces. With chemical contamination emerg-


45

ing as the biggest threat in many industries, residues of lubricants, coolants, surfactant, mold releases, and other extraneous fluids will not be tolerated on parts to be assembled. This will place increased emphasis on removing such chemicals from surfaces as soon as possible after they are placed there. 2.4 Parts Handling after Final Clean

Once an appropriate cleaner has been installed for the particular part to be cleaned, and it has been verified that the system delivers the required cleanliness and dryness, a discipline has to be put in place for how the parts are subsequently to be handled, stored, inspected, and shipped. Unless great care is exerted in dealing with these aspects, parts cleaning at the end-user using an aqueous system may turn out to be inadequate for functional purposes. Unless parts are handled with gloves after final-cleaning, fingerprints, skin flakes, and oils can transfer to the parts by contact; these are very hard to remove afterwards. Gloves should be worn during inspection as well. Parts should be covered during storage after inspection to prevent shopdebris from accumulating on their surfaces. Polyethylene sheets draped over the parts provide good protection, and do not shed particles themselves. The air quality of the storage area should be monitored using an air particle counter, and any deteriorations should be promptly addressed. Proper packaging materials should be used, as specified by the customer. The shipping bags should have a minimum cleanliness requirement on them. If the parts are not to receive a final wash at the assembly facility prior to assembly, then the bags should be of clean room quality and consistency; double-bagging, with the clean bag as the inner bag, is recommended in this case. Reusable containers are gaining in acceptance as packaging costs escalate along with product cleanliness requirements. The shipping container should not be allowed to transfer hard-to-extract contaminants (both particulate and film-type) to the part surface. If the part is designed and manufactured to be water-cleanable at the vendor, and no additional contamination (except loosely-adhering environmental debris) is allowed to accumulate on the part prior to its receipt at

46

Superctitical Fluid Cleaning

the assembly plant, it should, by definition, also be water-cleanable prior to assembly. This systems approach appears to be lacking in many short- sighted and hurried CFC-replacement efforts completed recently. 3.0 CONCLUSIONS

In this chapter, several guidelines have been presented for designing machining processes to manufacture parts that can be cleaned without recourse to chlorinated solvents. These are meant to supplement rules for designing parts for the same purpose. These rules are not meant to be all-inclusive; they are rather aimed at creating a mindset among product design, process design, manufacturing, and contamination control engineers that they need to collaborate hlly and early with one another to produce parts that are easily cleanable. The underlying premise of this paper is that if an effort to replace chlorinated solvents in metal cleaning applications is to be successful, the first order of business is to render the manufacturing process compatible with that subsequent effort. The allocations of investigative resources must be at least equally divided between improving the manufacturing process and developing alternative cleaning processes. For this purpose, a Vendor Process Contamination Checklist (Appendix 2) has been developed at IBM San Jose, and used extensively to identi@potential contamination exposures in the supplier process. Judicious use of this checklist was indeed primarily responsible for early success in replacing chlorinated solvents at several IBM San Jose vendors. The checklist emphasizes a systems approach, a comprehensive outlook that ignores nothing, from raw materials receivinghnspection at the vendor to final shipping to IBM, in the manufacturing process. It is based on the principle that solvent replacement in cleaning is not a stand-alone effort, but rather the culmination of a series of efforts to tailor part and process design to make this happen.


47

APPENDIX 1: GUIDELINES FOR DESIGN OF PARTS FOR CLEANABILITY

Al. 1 Introduction Design for X, where X equals manufacturability, assembly, automation, etc., has become a widely accepted practice in the U.S. industry. Here, a novel design concept-design for cleanability-of particular relevance to the microelectronics industry is introduced, and appropriate rules are formulated. In this era of increasingly tight contamination control requirements for microelectronic devices, increasingly severe environmental restrictions make it imperative that the thinking be done up-front in the design stage about fbture cleaning needs for the component parts. The guidelines stated here, it is hoped, will help create such a timely awareness among the design and development community. Design for manufacturability (DFM) considerations are heling early manufacturing involvement in the design and development processes. DFM is intended to produce innovative products that are, at the same time, reliable. In the case of the high-end DASDs (direct access storage devices), for example, HDA (head disk assembly) contamination is among the key factors determining the reliability of the storage device. While a significant part of the (particulate and chemical) contamination is indeed contributed by the in-house assembly process itself and subsequent HDA operation, initial component cleanliness doubtless has ramifications for product performance. Typically, the read-write heads and the particulatehhin-film disks are manufactured in-house in clean room (Class 100) environments; however, many of the mechanical components of the assembly, such as the drive motor and cast housings, are vended out to machine shops, whose knowledge of contamination control concepts is rudimentary, at times. The high premium placed on building only clean parts into the HDA gives rise to a new conceptDesign for Cleanability (DFC). Parts cleaning is becoming an increasingly complex problem as solvent cleaning is phased out at many vendors. Cleaning processes

48

Supercn'tical Fluid Cleaning

are becoming very part-specific in the absence of one catchall technique. At the same time, there is reduced tolerance for particulate as well as organic contamination in the many new semiconductor products. This trend will only be reinforced in future. All this is placing increased responsibility on the supplier to tailor his cleaning procedure to the part. If the part, by the nature of its design, is not easily cleanable, getting it acceptably clean could prove to be an expensive, time consuming and, sometimes, impossible undertaking. This is true even if the parts were indeed designed for manufacturability and assembly. Thus, the need is clear to design parts and assemblies by taking into account their cleanability as well.

A1.2 DFM/DFC Compatibility

DFM rules are well documented.['] These include: 1. Design for uniaxis assembly (stacking of parts)

2. Reduction in number of different parts and part numbers 3. Use of standard parts 4. Elimination of threaded fasteners and springs 5. Design for robustness (compliance) 6. Elimination of adjustments

7.Design for ease of orientation It is interesting to note that several of these rules, if followed rigorously, would also result in easily cleanable parts. For example, the uniaxis assembly rule states that all parts will be inserted onto the base from the same direction. This simplifies the manufacturing process because fewer operations and less movement are required. But these same improvements also result in less assembly debris being generated. Reducing the number of parts and part types implies that there are fewer variables in the manufacturing process; the same holds true for the cleaning process as well. The decrease in the number of variables allows full optimization of existing cleaning resources. Having standard parts, Le., incorporating the same parts in various products, will offer the same advantage, but, of course,


49

rapid technological advances may render this impractical. Elimination of screws and springs is discussed in greater detail later on, but they represent a severe contamination exposure as well as a manufacturing nightmare. Adequate tolerancing in design is necessary for a robust manufacturing process, and, often, for a robust cleaning process; etching and erosion associated with several chemical-based and sonic cleaning procedures cause variations in dimensions and other physical properties of the part, and if the tolerances are unreasonably tight, sufficient cleaning cannot be accomplished. Compensatory adjustments entail nonrepeatable operator intervention, and, very often in a clean room environment, a contamination incident; it is oRen said that people are the biggest source of contamination in a clean room. Tighter tolerances may eliminate this occurrence, as long as they are not excessively tight. While ease of orientation is clearly a manufacturability and assembly issue, it is just as much a cleanability and dryability issue; this will be expanded upon later. Apart from these common rules for DFM and DFC, there are issues peculiar to parts cleaning that need to be borne in mind during the design stage as well. In what follows, additional guidelines are formulated to aid the design and development engineers in designing parts specifically for cleanability (in cases where this can be accomplished without compromising functionality).

A1.3 Guidelines for Design for Cleanability

The additional DFC rules may be briefly stated as follows: 1. Use of easily cleanable materials 2. Use of surface coatings/polishes/treatmentto promote cleanability 3. Compatibility of all components with water cleaning 4. Elimination of blind holes 5 . Use of adhesives and other materials that do not outgas or emit corrosive vapors 6. Effective sealing ofthe sensitive components from external particulate/chemical contamination

50

Supercritical Fluid Cleaning 7. Avoidance of materials that erode during ultrasonic

cleaning 8. Design for ease of drying after aqueous-clean 9. Design for ease of packaging 10. Early supplier involvement in design 11. Feedback of cleaning problems to part designer 12. Implementation of sign-off by cleaning engineer prior to print release for vended parts. 13. Close collaboration between the desigddevelopment and contamination control functions. Some of these guidelines are covered in greater detail in the remainder of this section. 1. Materials. The material@)the part is made of is of fundamental importance in determining its cleanability. The force of adhesion between the contaminant particle and the part surface depends on bulk properties of the material, such as its hardness, the surface properties of the material, such as its surface energy, and (surface/cleaning mediudparticle) interfacial properties, such as Van der Waals forces of Various materials (steel, stainlesssteel, aluminum, plastics, rubber, etc.) and various alloy compositions (300-series versus 400-series stainless steels, etc.) may be ranked with respect to their surface adhesion characteristics. See Table A1 for a partial list, in which a rank of 1 corresponds to a highly-cleanable substrate, and higher numbers represent surfaces that are progressively more difficult to clean. This ranking is based strictly upon particle/substrate adhesion characteristics and is independent of the cleaning method. Thus, although plastics and rubbers typically exhibit loosely-bound contamination, devising a procedure to clean them may be a more demanding task than devising a cleaning procedure for stainless steel parts, for example. The materials and surface treatmentdcoatings listed in this table are those typically found in DASD HDAs. Of course there is a cost impact here; typically, easyto-clean materials are more expensive than those that are not easily cleanable. A compromise, based on cost and functionality, may be necessary to achieve optimal price-performance characteristics.


51

Table Al. Ranking of Materials with respect to Surface Particle Removability 1. Plastics

2.

3.

4.

5.

6.

7.

8.

9.

a. Polycarbonate without carbon fill b. Polysulfone with carbon-fill c. Polycarbonate with carbon-fill d. Polysulfonewithout carbon-fill Rubbers a. Paralene-coated b. Liquid injection molded c. Uncoated polyethylene d. Uncoated silicone E-Coated Surfaces a. Stainless-steel b. Nd/FeJB c. Copper d. Aluminum e. Magnesium Electroless Nickel Plated Surfaces a. Steel b. Iron c. Aluminum Stainless Steel a. 400 series (Martensitic) b. 300 series (Austentic) c. PH (h.ecipitation-Hardened) NutsMrashers (stainless steel) a. Chempolished + Electropolished b. Electropolished c. Passivated Steel a. No surface treatment Aluminum a. Wroughthot machined b. Anodized c. Wroughtlmachined d. Stamped e. Forged f. cast Screwdsprings (stainless steel) a. Chempolished + Electropolished b. Electropolished c. Passivated

52


2. Surface Finishes. Of almost equal importance are surface finishes and surface coatings. For instance, E-coating (or electrophoretic painting) is an effective contaminant-containment for many machined metal surfaces such as aluminum and magnesium surfaces, and for magnets. Other surfacetreatments applied in past and current IBM disk drives are listed in Table A1 . 3. Aqueous Cleanable. All parts should preferably be cleanable in water or a “safe” solvent, unless they come out of the production process clean enough to be double-bagged and shipped directly to the clean-room assembly area. When soldering is involved, for example, the specification should call out for the use of water-soluble or noclean fluxes. Parts and assembly design should allow suppliers to quickly switch to a non-CFC solvent. 4. Elimination of Blind Holes. Design should minimize blindholes which typically amass excessive particulate contamination; through-holes should be avoided as well, although they do not represent as high a contamination exposure. This is stated from a part cleaning perspective; however, this should be weighed against the fact that blind-holes act as effective getters for debris generated during rework and assembly processes. 5. Fastenings. Screws should, in general, be eliminated to the extent possible. Design should incorporate as many snap-on and snap-together features as possible, although it is admittedly simplistic to say “replace a screw with a clip;” riveting and gluing surfaces together are other alternatives. Threaded fasteners and springs are disavowed on the basis of DFM principles as well. It is difficult to get the thread started, and cross-threading may be a frequent defect. Springs are difficult to handle and to package. Screws and springs are difficult problems fi-om a cleaning standpoint[*]due to the nonplanar nature of their surfaces. The crevices between screw threads experience highly-attenuated ultrasonic energetics, thus resulting in practically no ultrasonic cleaning action. Spraying them one by one is costly and unacceptable in terms of cycle time, and spraying them in batches reduces the cleaning efficiency considerably. Many of these considerations carry over to springs as well. Thus the ideal design would be one in which all of the functions of fasteners and springs would be accomplished without them.


53

6. Adhesives and Sealants. The use of adhesives, while generally nonthreatening from a particle contamination viewpoint, could be a potential source of chemical (organic) contamination. Prior to their incorporation in the design, adhesives should be cleared with materials scientists with respect to their outgassing characteristics, contact transferability, etc. Actually, all componentdmaterials should pass outgassing tests, with greases and other residual fluids requiring special attention. In addition to organic outgassing, some materials can emit corrosive vapors as well, e.g., chloride contamination originating from chromated aluminum parts. This cannot typically be allowed from a corrosion viewpoint. The product should also be protected from external chemical contamination by designing enclosure materials to provide good sealing. 7. Ultrasonic Cleaning. Potential damage of parts during cleaning should be borne in mind during the design stage. For example, bare aluminum and other soft metals erode during ultrasonic cleaning; plastics with carbon filling tend to shed carbon fibers during sonic extraction. It is not feasible to alter the cleaning procedure on a part-by-part basis, especially when multiple parts are processed through the same cleaner. Thus, given a single in-house cleaning system, materials should be chosen not only to optimize cleanability, but also to minimize surface deterioration during cleaning. 8. Drying. Drying after aqueous cleaning is sometimes a difficult problem. If parts are designed such that they provide locations for pools of the cleaning fluid to accumulate, requirements for complete removal of water and suspended particulates from such locations may place an undue burden on the drying system. Conventional convection-dryers may need to be replaced by highly-directed automated chase-offtype air-knife dryers which are likely to be many times as expensive. Determining the correct orientation of such parts in cleaning racks to obtain maximum drainage becomes a timeconsuming step in optimizing drying and cleaning efficiencies. Incomplete fluid removal may also result in rusting, corrosion, pitting, etc. From this viewpoint, an easily-drained surface is a clean surface; the design of the part should reflect this truism. Materials that are porous or permeable to moisture would also be very difficult to dry after waterbased cleaning.

54

Supercritical Fluid Cleaning 9. Packaging. As organic contamination gains in importance,

the contribution of the packaging material to the total contamination level of the delivered part becomes increasingly significant. If the package has to be specifically designed for the part, it becomes difficult to enforce clean packaging guidelines to the letter. Thus, ease of packaging of the part determines to a large extent how tightly the cleanliness of the container can be specified and controlled. IO. Supplier Involvement. Industries in the United States are beginning to commit to the strategy of developing a limited supplier base, and working with them closely in order to effectively implement just-in-time manufacturing. In order to optimally institute design-forcleaning, it is recommended that the design group work in close early involvement with the intended supplier for each part. The supplier can bring the benefit of extensive experience of having manufactured and cleaned similar parts in the past. Suppliers are usually eager to eliminate solvent cleaning, but solvent use is often dictated by the fact that the machining process requires the use of an oil-based vehicle (e.g., in tapping, grinding and lapping operations), which in turn is dictated by the design of the part, the tolerances on its mechanical properties and geometrical tolerances, properties of the material (such as strength and heat capacity), etc. It becomes imperative, then, for the supplier to have a say regarding the safe and environmentally-sound manufacturability of the part. This will be a litmus test on whether the microelectronics industry is willing to accept parts vendors as “business partners” and incorporate their concerns into its decision-making. 11. Feedback. A feedback loop needs to be established that conveys information regarding the performance of, and problems associated with various parts back to their original designers. For example, every time a part fails to meet cleanliness requirements, and it is determined that nothing has gone haywire in either the vendor or the in-house cleaning systems, the responsible desigddevelopment engineers need to be made aware of the incident, since they must share (to varying extents) in the responsibility for the failure. Investigations of line-rejects and line-down situations involving contamination-related causes should be conducted by teams in which the partdesigner is integral. The design group should have access to, and be


55

familiar with, parts cleaning data, problems and history during the life of every precision assembly product. This list is by no means exhaustive, nor is it meant to be. Rather, it is designed to create a mindset among desigddevelopment engineers that cleanability, just like manufacturability and ease of assembly, is a key price-performance issue that needs to be factored in at the outset. In the highly competitive microelectronics marketplace, effective contamination control, when properly integrated with product development, may turn out to be a big advantage. This concept has more recently been discussed in more detail in Ref. 3. References (for Appendix 1) 1. Bancroft, C. E., Introduction to Design for Manufacturability, Internal Report, CIMTechnology Center, IBM,Boca Raton, FL,(March 13, 1989) 2. Nagarajan, R, SurfacePropertiesand Cleanability:A Rational Correlation, J.IES, 33(1):26-33 (Jan./Feb. 1990) 3. Sipitkowski, J. M., Design for Cleanability: Part Geometry and Surface Finish, Precision Cleaning, 1( 1):55-57 (Nov./Dec. 1993)

APPENDIX 2: VENDOR PROCESS CONTAMINATION CHECKLIST A2.1 Receivinghspection (R/I) at the Vendor o Are all incoming materials raw (stock) materials? ESNOo Are some of the incoming materials parts supplied by a secondary vendor that will be assembled at the primary vendor?

ESNOo Do stock materials receive at least a cursory visual inspection for gross contamination?

YES-.

NO-

56

Supercritical Fluid Cleaning o How is the quality of incoming stock controlled? o If some parts are received for assembly, do the parts have customer cleanliness specs on them?

=s-

N O - . NIA-

o If they do, are they being final inspected at the secondary vendor?

=s-

NO-

NIA-

o Ifthey do, are they being IUI’d at the primary vendor for cleanliness? =sNONIAo Does the secondaryvendor have a “black box’’ operation in his process?

=s-

NO-

N/A-

o Is inventory managed properly, on a first-in first-out (FIFO) basis?

ES-

NO-

A2.2 Manufacturing

o Is the manufacturing process hlly documented with the customer?

=s-

NOo Briefly outline/sketch the steps in the manufacturing process. o How many parts are produced per day? o Is continuous flow manufacturing (CFM) being practiced?

=s--

NO-

oOn average, how long does the part wait between successive operations? MINUTESHOURSDAYS-

Designfor Cleanability o Are products for other customers being manufactured in the same area? ESNOo If so, is there any chance of cross contamination? ESNONIA01s machining debris being effectively (across all size ranges) removed between stages? ESNONIAIf so, state how:

o Are the intermediate cleaners diplrinse tanks or spray units? SPRAY CLEANERSN/ASTATIC TANKSo If dip type, how often is the cleaning fluid replaced? NIAONCE A DAYTWICE A DAYOTHER (Explain) o Has the nature of the cutting tool (hardsoft, sharp/dull) been correlated to the amount of machining chipdfines generated? ESNON/A oHave tool-bit changes been looked at as a means to reduce machining debris generation? YESNONIAo During machining, are the parts spray cooled or flood cooled? FLOODNIASPRAYo Are all the coolants being used compatible with the downstream cleaning process? In other words, if the cleaner is water based, are all coolants and other process fluids water soluble? ESNON/A-

57

58

Supercritical Fluid Cleaning o If oil-base coolants or lubricants are in use, has the vendor explored the use of water-soluble counterparts? ESNONIAo Are Manufacturer’s Safety Data Sheets (MSDS) for all the chemicals used in the manufacturing process documented with the customer? YESNOIf not, provide MSDS for all chemicals used in manufacturing. o If the process fluids dry on the part surface, do they appear to form a lacquer by means of polymerization reactions? ESNONIAo Are air particle counters (APCs) used to monitor the quality of the manufacturing environment?

ESNOo If so, at what intervals? MONTHLYDAILY- WEEKLYExplain how the APC data is used.

o Is deburring done? ESNOIf so, explain how.

o Is deflashing done? If so, explain how.

ES-

NO-

o Is hand wiping of the surface being done? ES-

NO-

NIA-


o Is oil being added (on bolt threads, and so on)? YESNOIf so, explain how it is being removed.

o Are there any “black box” operations at the vendor? YES__ NOo Ifparts undergo a surface treatment such as passivation, polishing, epoxy coating, and so on, are these processes vended out to a secondary supplier? ALL- SOMENONEo Are platinglcoating processes at the primary vendor certified and qualified by the customer? YESNON/Ao Are platinglcoating processes at a secondary vendor certified and qualified by the customer and by the primary vendor? ywNON/Ao If surface platinglcoating is done, are inspection procedures in place for porosity, adhesion, and so on? YESNONIAIf so, according to what specification? o If soldering is involved, are water-soluble fluxes used? YESNO-. N/AoIf soldering is done, are water-soluble solder pastes used? YESNON/Ao If soldering is done, are no-residue or low-residue fluxes used? =sNONIA-

59

60


A2.3 DegreasingKleaningDrying oHas the cleaning process been documented with the customer? ESNOBriefly outlindsketch the cleaning, rinsing, and drying processes.

o Is the wash fluid or rinse fluid being recycled?

YES-

NOIf so, describe the filtration arrangement.

o Is online monitoring of fluid cleanliness being done?

ym-

NOo Are chlorofluorocarbons (CFCs) or other chlorinated solvents used in cleaning and/or drying?

ESNOo If so, what is the estimated consumption of the cleaning (and/or drying) medium per year (in gallons)? o If so, what is the estimated recovery/recycling of the cleaning (and/or drying) medium (as percentage of consumption)? o Is hot air being used for the drying of parts? ESNOo If so, is the air filtered? asNONIAIf yes, describe the air filtration arrangement

o Is the air oil free? YESNO-

NIA-


o Is there a possibility of overloading the cleaner? For example, are grossly contaminated parts being prewashed in the same system where final cleaning is done? ESNO01s there a separate prewash unit to remove gross contamination?

=s-

NOo Have cleaner operating parameters (temperature, pressure, flow rate, and so on) been optimized?

=s-

NO-

o Is the cleaning system water based?

YES-

NOo If so, is deionized @I) water being used? =sNONIAo If so, is the DI water quality monitored on a continuous basis? ESNONIAoDescribe how cleaner efficiency is monitored and maintained.

o How consistent is the cleaning process-is significant part-to-part or lot-to-lot variation?

there

ESNOo Is statistical process control (SPC) data on cleanliness being kept? ESNOo If so, do the records extend back A WEEK?- A MONTH?A YEAR?o Explain briefly how the SPC data is used.

NIA-

61

62


o Is MSDS information on all the chemicals used in the cleaning process documented with the customer? ESNOIf not, provide MSDS for all chemicals used in the cleaning stage. o Are the surfactantsused in the cleaning stage compatible with the process fluids used in the manufacturing stage? In other words, are they completely mutually soluble? ESNOo I f process fluids used in machining tend to form a lacquer on the surface of the parts, has the surfactant been chosen so that it can dissolve the lacquer? ESNONIAo If the detergent contains fatty acids, has the system been designed so that they can be rinsed off, ESNONIAo Do wash baths contain surfactant concentrations high enough to leave a lubricating residue? ESNOo Are any toxic solvents used? ESNOo If so, do they leave any residues on the part? ESNON/Ao Are all the detergentdsurfactants used in the washing section being completely removed in the rinsing section?

yw-

NOo Describe how part cleanliness before and after the cleaner is being monitored.

oDescribe how the particle concentration level of the washer/rinse outlet liquid streams is being monitored.

Designfor Cleanability oHow often is the liquid replaced, and based on what criterion? OHOWflexible is the cleaning system? If major design changes are made on the part, can the unit still be used without major modification? ESNOoIf ultrasonic cleaning is done, briefly describe the sonicating procedure.

o If sonic cleaning is presently done, has there been any evidence of erosion of parts due to over sonication? ESNONIAo Is there final manual cleaning of any leftover visible contaminants? ESNOIf so, describe the procedure:

A2.4 Final Inspection o Does the vendor have the customer’s cleanlinesddryness specification documents? ESNOoDoes the vendor have the specified cleanliness (or dryness) levels for the part(s)?

ES-

NOIf so, state the cleanliness (and/or dryness) specifications.

oDoes the vendor have someone in charge of contaminatiodquality control? ES-

NO-

63

64


o What analyses (particulate/chemical/outgassing/ nonvolatile residue) are being done on the parts to ensure conformance to the customer’s cleanliness specifications? o What percentage of parts are being tested? -Yo If less than loo%, explain on what basis. o Is final inspection data being reported to the customer?

YES-

NOIf yes, state in what form, and at what frequency? o Are the contamination monitoring equipment at the vendor properly correlated with the customer’sin-house test instruments?

YESNOoHas someone been sent by the customer to install the vendor equipment, and instruct operators on its use?

YESNOo Is periodic calibration and/or equipment servicing being done? YESNOo If so, at what frequency? TWICE A YEARONCE A YEAROTHER (Explain) o Has it ever happened that readings taken at the vendor and the customer FUI readings on the same batch differ considerably? YES-

NO-


oIf so, can transportation debris hlly account for the discrepancy? ESNON/Ao Is there any disassembly/reassemblyof parts during final inspection? ESNOo Are nonshedding gloves worn during inspection when handling final-cleaned parts? ESNO01s the environmental quality in the inspection area monitored?

Y-ESIf so, how?

NO-

o How good is traceability? For example, if one part in a lot is bad, can the vendor go back through the process and find the reason?

ES-

NO-

A2.5 Storage o How long are parts stored after final inspection before shipping? HOURSDAYSWEEKSMONTHSo Is the air quality of the storage area monitored? ESNOo If so, at what frequency? DAILYWEEIUYOTHER (Explain)

MONTHLY-

65

66

Supercritical Fluid Cleaning o Are finished parts covered with nonshedding sheets during the time of storage?

ES-

NOo Are they ever exposed to shop conditions during storage after final clean? ESNOo Are any parts left to dry overnight, or left uncovered overnight for any other reason? ES-

NO-

A2.6 Shipping

o Does the vendor have the customer’s packaging specification documents?

ES-

NO01s the vendor aware of who the customer-approved packaging material suppliers are?

ESNOo Is the vendor purchasing bags from customer-approved suppliers?

ES-

NOo Are the bags of the type recommended by the customer for the particular part? ESNONIAo Is the packaging being done in a clean environment? ESNOIf yes, state Federal Standard 209D class of the clean room. o Describe the package sealing procedure.


67

o Are parts being shipped from the same facility where they were manufactured? oIf antistat bags are used, do they meet customer cleanliness criteria? oIf rigid packaging is used, does it meet customer cleanliness criteria?

o Is the number of parts packed in each bag consistent with customer recommendations?

ES-

NO-

o Is the required humidity being maintained within the package? o Are desiccants used for this purpose?

Apart from these process-specific questions, there are some that pertain to operator training and management attitudes.

A2.7 Employees oHave operators been made aware of the need for (particulate and organic) contamination control (through video presentations, and so on)?

ESNOo Are operators familiar with requirements for working in a contamination controlled environment (change procedures, good manufacturing practices, contraband substances, and so on)? YES-

NO-

68

Supercritical Fluid Cleaning o Are they aware of simple, quick contamination monitoring procedures, such as tape tests? ESNOo Are employees encouraged to come up with their own suggestions for improving product quality? ESNOo Are employees allowed to take responsibility for the parts they produce, and given leeway (subject to management approval) to make improvementsthat they deem necessary? ESNOo If the process involves clean-room work, are operators properly trained for it? ESNONIAo Do they wear proper gloves and other apparel suitable to their work station?

ES-

NO-

A2.8 Management o Are they responsive to the customer’s cleanliness requirements? ESNOo Are they willing to work with the customer when necessary to improve parts cleanliness? ESNOo Are they concerned with product quality? ESNOo Are they responsive to the customer’s suggestions and recommendations? ES-

NO-


o How many years of association does the vendor have with the customer? o Is the vendor willing to let the customer’s representatives inspect their product manufacturing operations?

ESNOo Is the vendor willing to let the customer’s representatives inspect their product cleaning facilities? YES-

NOo Is the vendor willing to let the customer’s representatives inspect their product packaging procedure?

ES-

NO-

o Is there effective two-way communication between the customer and the vendor?

YES-

NO-

69

~~

Dynamics Of Particle Removal By Supercritical Carbon Dioxide Shirish M. Chitanvis, Chris W Patterson, M? Dale Spall, and Kenneth E. Laintz

ABSTRACT

We have investigated the possibility of cleaning semiconductor wafers contaminated by particles by flowing supercritical carbon dioxide over them. The hydrodynamics of turbulent flow of supercritical carbon dioxide over a flat plate was calculated. Through these calculations it is proposed that the drag on a particle sitting on a flat plate in the viscous sublayer is capable of rolling submicron-sized spheres off the wafer at sufficiently high velocities. 1.0 INTRODUCTION

Contamination of semiconductor wafers during manufacturing is a common occurrence. Particle adhesion cannot always be prevented in the manufacturing stage, making it necessary to clean the 70

Dynamics of Particle Removal

71

wafer in a subsequent stage. As feature sizes on integrated circuits become smaller, the need for ensuring particle removal increases. In the next generation of semiconductor circuits, it is expected that feature sizes of less than a quarter-micron will be attained. It will therefore become imperative that all particles greater than about a tenth of a micron be removed from semiconductor wafers. Several schemes exist for cleaning wafers, such as an ultrasonic method,[’] high pressure jet scrubbing[2]and carbon dioxide snowjet,[3] among others. The use of air to roll away particles has been investigated previously by Bhattacharya and Mittal14]and by Zimon.[’] The dynamics of liquid jets for cleaning has been studied recently by Gim et a1.I6I Here we investigate the possibility of flowing supercritical carbon dioxide over a dirty wafer in order to roll away spherical particles. There are two advantages of using supercritical carbon dioxide over air. The first advantage is that the density of supercritical carbon dioxide is much larger than that of air, thereby dramatically increasing the drag on the particle. The second is the ability of supercritical carbon dioxide to dissolve organic material. Since some of the contaminant particles are indeed organic in nature, supercritical carbon dioxide will aid in loosening the adhesion between such particles and the surface on which they are located. Furthermore, a liquid such as water is not an option for cleaning semiconductor wafers, as it would tend to hydroxylate the surface of the wafer and damage it in other ways as well. The basic idea behind using flow to remove particles from surfaces is the formation of a boundary layer near the surface. Within this boundary layer, there is a shear in the velocity field, leading to a stress which rolls particles away from any given position. As one might guess intuitively, the wall shear in the case of turbulent flow is much greater than in laminar flow. In laminar flow, the boundary layer is relatively thicker, allowing the velocity to change gradually to its stream value. In the case of turbulent flow, the viscous sublayer which develops right next to the wall is much thinner, causing a much more abrupt change in the velocity field, thereby setting up a larger wall shear. The velocity profiles for turbulent and laminar flow are compared in Fig. 1 as a hnction of distance from the wall. It can be seen fiom the figure that laminar flow shows a gradual velocity

72


change while that change is abrupt for turbulent flow. Therefore, turbulent shear is much larger than that present in laminar flow. We must note however, that the viscous sublayer displayed for the case of turbulent flow in Fig. 1 must not be interpreted too literally as a laminar sublayer. While this layer may be relatively quiescent compared to the turbulent core of the flow, there is still turbulent motion within this sublayer.

Velocity ( a d s ) 300

12

.

lo

'

bars

/

/

-

I

8 .

0

6 .

0 /

4 '

/

0.1

0.2

0.3

0.4

.

0.5

Distance from w a l l loa)

Figure 1. Velocity past a flat plate as a function of distance from the surface of the plate under turbulent and laminar flow conditions using a flow (stream) velocity of 1 1 c d s . Notice the gradual change in the laminar flow (dashed line), compared to the turbulent case (solid line). It follows that the shear in the turbulent case,near the plate is much larger than in the laminar case.

In our treatment of particle removal we shall make a number of simplifjling assumptions. We shall assume that the force of adhesion is only due to Van der Waals force of attraction. We shall not consider other possible forces such as electrostatic interactions or gravity. We shall also assume that the surface is smooth. In other words, the roughness of the surface is due to randomness at the atomic level. This means that over the width of the particle, the surface height varies by a few angstroms at most. This assumption allows us to readily determine the force of adhesion of glass particles


73

on a silicon wafer. In this chapter, we have not investigated how the particles came to be attached to the wafers.['] We utilize the physics of rolling particles on a surface as developed by Bhattacharya and Mittal.i4] Our treatment differs fkom that of Bhattacharya and Mittal[41in that we provide a more detailed description of the turbulent boundary layer which is formed in the steady state when a fluid flows over a flat ~late.1~1 The drag force we use is consistent with the treatment of Gim et a1.16] In Sec. 2 we discuss the properties of turbulent flow and, in particular, the flow velocity as a hnction of distance from the wall in the turbulent regim-the so-called law of the wall. In this respect, we believe our treatment of particle removal is unique. We discuss the magnitude of adhesive forces in Sec. 3. In Sec. 4 we apply our treatment of boundary layer flow to supercritical carbon dioxide and we calculate the requirements for spherical glass particles to be removed from a smooth silicon surface using supercritical carbon dioxide.

2.0

THE BOUNDARY LAYER IN TURBULENT FLOW OVER A PLATE

Turbulent flow over a flat plate is characterized by three regions:r8] (a) a viscous sublayer often called the laminar sublayer, which exists right next to the plate, (b) an adjacent turbulent boundary layer, and (c) the turbulent core. Viscous forces dominate inertial forces in the viscous sublayer, which is relatively quiescent compared to the other regions and is therefore also called the laminar sublayer. This is a bit of a misnomer, since it is not really laminar. It is in this viscous sublayer that the velocity changes are the greatest, so that the shear is largest. Viscous forces become less dominant in the turbulent boundary layer. These forces are not controlling factors in the turbulent core. In the supercritical region, the kinematic viscosity of carbon cm2 s-~,[~] so that even for a low flow speed u of dioxide is D 100 c d s , the Reynolds number Re = u slD is -lo5, s being the typical length scale chosen here to be -1 cm. At such Reynolds numbers we

-

74


expect the flow of supercritical carbon dioxide over a flat plate to be turbulent at least in the vicinity of the plate. We use standard empirical laws of turbulent flow to get the velocity distribution both near and far from the wall. The velocity in the laminar boundary layer grows linearly and connects in a continuous fashion with a logarithmic velocity profile in the turbulent boundary layer:[”]

Z = ~ U * ~ K;

=0.4437; c=0.15; B = 5 . 5

where z is the wall shear, p is the fluid density at the wall, D is diffusionconstant, u* is a velocity called the frictional velocity, and K, c and B are empirically determined dimensionless constants. The transition between the laminar region and the turbulent region is made by demanding continuity, which yields a self-consistent equation to be solved:

whereR,=y,u/D. The solution isR,= 114. We can use this value to estimate the stream velocity required to develop a boundary layer of thickness y,= 6. For 6 = 100 microns, we get u, 5 cm/s, using D 5.4 lo4 cm2/s at T = 325 K. From Eq. (l), u* = u,/(R,)” 0.5 cm/s. Now, there exists a relation between the frictional velocity and the stream velocity U

-

-

Eq. (3)

u=

0.0791

-


75

-

wherefis Fanning’s friction factor 0.04 which finally yields U - 13 c d s . When we examined these values with detailed numerical calculations, it turns out that the estimates of the velocity required to achieve a certain thickness of the viscous sublayer is quite accurate as long as realistic estimates of the density and difision constant are used. We notice from Eq. (1) that the wall shear stress z is constant within the laminar boundary layer, and for this case is 0.25 dyne/ cm2. As far as calculation of the wall shear stress for a known stream velocity is concerned, we point out that the use of Fanning’s factor alone suffices to determine its value, and we do not need all the formulas stated above. In order to get the entire velocity profile, we need to use realistic values of the density and difision constant as fbnctions of pressure and temperature, and then solve the above equations numerically for the velocity in the two different regions and match them at the edge of the boundary layer. We have given an example of such a calculation in Figs 2 and 3. These two figures are calculations done at 300 bars and 400 bars respectively. These figures show basically that there is a greater sensitivity to the stream velocity rather than temperature. We also learn from these figures that for velocities on the order of a few cm/s, the thickness of the boundary layer is about 100 microns. One might be tempted to conjecture that this layer being turbulent (albeit less turbulent than the regions fbrther away from the plate) will have an eddy swooping down once in a while to somehow “pick” a particle off the surface. We believe this is not very likely, since the sublayer is on the order of a few microns for the velocities under consideration, while the particles we would most like to remove are less than a micron in radius. In fact, one may use standard estimates for turbulence variables such as the dissipation (loss of kinetic energy per unit time), E u2/6 (where u, is a measure of the typical velocity in that region and 6 is the thickness of the viscous sublayer), and the turbulent kinetic energy (mean square velocity fluctuations), k ( ~ 6 ) ~to’ ~obtain a kinetic “pressure” due to turbulence pk dynes/cm2, p being the characteristic density. Multiplying this pressure by the cross-sectional area of a micron-sized particle yields a

-

-

-

-

76


-

force of 3.14 x dynes. This force due to turbulence may be compared to the force of adhesion (see Eq. 5 in the next section), which is 0.0052 dynes for a micron sized particle. It is therefore clear that turbulence cannot scoop up a particle and move it along the surface.

-

V e l o c i t y (an/s) 300

bars

- e - -

_-----Distance from wall

Figure 2. Velocity profiles for supercritical CO, calculated at 300 bars of pressure. The solid line corresponds to 325 K and a viscous sublayer thickness of 100 microns, the dashed line corresponds to 375 K and a viscous sublayer thickness of 100 microns, and finally the dot-dashed line corresponds to 325 K, but a viscous sublayer thickness 200 microns.

V e l o c i t y (cm/s) 400

10

bars

1 from wall fan)

Figure 3. This figure is similar to Fig. 2, except that the pressure used is 400 bars. Notice the correspondingly higher velocities required to attain the same viscous sublayer thicknesses (100 microns and 200 microns).


77

On the other hand, it is in the viscous sublayer that the velocity changes are largest, so that the shear is greatest here. It may be possible for the shear stress in this region to move a particle. We develop this idea in the next section.

3.0 ADHESIVE FORCES As mentioned in the introduction, in addition to assuming a spherical particle and a smooth surface, we shall assume that electric charges are not present in the problem, and the effects of gravity will be ignored as well. The adhesive force between a neutral particulate contaminant and the wafer is expected to be due to the attractive Van der Waal's interaction between molecules.[''1 This is a macroscopic force found by averaging over the force between all the molecules of a particle and the neighboring surface. For a spherical particle sitting on a flat wafer, it is known that surface roughness will cause the mean distance of separation between the particle and the wafer to be nonzero. The attractive force between these two entities acts along the normal between the sphere and the wafer, and is given by:

where A is Hamaker's constant (J), r is the radius of the spherical particle, and z is the distance of separation. Experimentally, it is found that h 10 eV. In our theoretical study in this paper, A and h need to be considered as empirical parameters which will change depending on the nature of the particle and the surface upon which it is placed. The dielectric properties of the intervening fluid, eg., supercritical carbon dioxide, will change the strength of the interaction between the particle and the surface. Empirically, the separation distance for a smooth surface is found to be 0.4 run, because at distances smaller than this the interaction becomes repulsive. Using these numbers, we get:

-

-

78


where the adhesive force Fuis in dynes, I is in cm, and the negative sign indicates an attractive force. For I = 1 micron, the adhesive force is 0.00521 dynes. The force due to the shear in the viscous sublayer on a particle lying on a surface, is analogous to the formula for Stoke’s drag:[12]

-

One sees immediately fiom this equation that when U 1000 c d s , the shear force will exceed the adhesive force for I = 1 micron. This puts movement of the particle due to the shear force in the realm of reality. In general, for the particles to move, F, must be greater than the force of friction Fr which is given by the following equation:

where p is the coefficient of friction. The shear force is proportional to the square of the radius of the particle while the force of friction grows only linearly with the radius. Thus, there will be a radius below which the shear force can no longer overcome the force of friction. The radius of the smallest particle which can be moved is therefore given by:

The idea of utilizing a coefficient of friction is not new, having been proposed and tested for the cleaning of flat plates with air at least two decades ago by Zim~n.*~] Similar ideas have been proposed for the cleaning of waferddiscs by spinning them.[13] If the particles that


79

need to be removed are charged, the force of adhesion may be electrostatic in There appears to be empirical evidence that the approach outlined above is applicable to the sliding of particles.[5]For cases of interest to us, the particles appear to roll, rather than slide off the wafer.[6] As far as we are aware, well-characterized experiments have not been performed for the case of supercritical carbon dioxide. We now turn to the theory of rolling particles off surfaces. We use the treatment of Bhattacharya and Mittalr41to compute the radius of the smallest particle that can be moved due to the torque exerted on the particle by the wall shear. As the particles sit on the surface, there will be a very small deformation xo caused by the adhesive force between the particle and the surface. This deformation is given by the following equation:

-

where E is the modulus of deformation 8.0 10” dynes/cm2 for silica. When rolling occurs, we may equate the torques acting on the sphere due to the vertical force of adhesion and the horizontal force of drag to find

Using Eq. (4), Eq. (6) and Eq. (9), we obtain:

Thus rolling yields a V‘” dependence for r ~ nThis . may be contrasted with the inverse quadratic dependence obtained from Eq. (8), for sliding particles. In fact, we suggest that experimental data on r,, could be used to differentiate between the two different mechanisms discussed here, and indeed to veri& whether such predictions are correct.

80


In the next section, we specialize these formulas to the case of supercritical carbon dioxide. 4.0 SUPERCRITICAL CARBON DIOXIDE We pointed out in the introduction that using supercritical CO, as a removal agent would be more advantageous than air, and this is borne out by Eq. (1 1). This equation shows that the radius of the , a p-22ndependence, smallest particle that will be removed, r ~ , ,has implying that for a given flow velocity, supercritical CO, with a density of 1 g/cc will remove much smaller particles than air. Of course, one could use air and increase its flow velocity appropriately in order to overcome the deficit caused by its low density. One would have to make certain that the air was free of contaminants which could damage the wafer being cleaned. On the other hand, supercritical CO, has the inherent advantage of being able to dissolve organic compounds, which should help loosen organic particle contaminants on the wafer surface. Thus, besides flow characteristics, one needs to consider that chemistry could make a difference in this process. The Hamaker constant A which appears in the adhesive force defined by Eq. (4) is expected to be smaller for the case of supercritical CO, flowing over particles containing organic material and placed on a wafer. Supercritical CO, is also a drying agent, and this may be of some use in the cleaning process as well. The critical temperature of CO, is Tc = 3 1"C,and the critical pressure Pc = 73.8 bars. In Fig. 4, we plot the viscosity of supercritical CO, as a function of temperature,['51 varying pressure parametrically. In Fig. 5 , we plot the density of carbon dioxide over the same range of temperature and pressure.[9]In order to maximize the force on particles, it is desirable to increase the density (see Eq. 10). There is only a weak dependence (see Eq. 10) of the particle removal efficiency on the viscosity, through the Fanning factor. The Fanning factor changes very little as the Reynolds number is changed. In any event, we can optimize the process by choosing to work in a regime where the viscosity is relatively low. Accordingly, we choose a temperature of 325 K, and a pressure of 300 bars based on Figs. 4 and 5 .

-


81

Lowering the pressure would not only decrease the density but bring the operating point closer to the critical pressure, which is not desirable. Increasing the pressure would increase the viscosity. Raising the temperature decreases the density, while lowering it would not only increase the viscosity, but again bring us uncomfortably close to the critical temperature. This is due to dramatic density changes as a function of temperature and pressure near the critical point. We recognize in this manner that there is not a truly optimal operating point. We obtain a range of parameters defining an operating regime whose boundaries are defined by the proximity to Tc and P,,and within which one can perform cleaning experiments.

n W

v,

20 [L:

0

I

v

c v)

0 0

v, >

TEMPERATURE (KELVIN) Figure 4. Viscosity as a function of temperature for supercritical C02for three different pressures.

82

SupercMcal Fluid Cleaning

W

L3

TEMPERATURE (KELVIN) Figure 5. Density as a function of temperature for supercritical CO, for three Werent pressures.

In Fig. 6, we show the thickness of the boundary layer as a fbnction of the stream velocity of supercritical COz for temperatures of 325 K (solid line) and 375 K (dashed line) at a fixed pressure of 300 bars. In this regime, the density of COz is 0.94 g/cc and its viscosity is 7.4 x lo4 gkm-s. The thickness ofthe laminar sublayer is seen to depend inversely on the stream velocity. This figure indicates that for a stream velocity of 200 cm/s, the boundary layer thickness is about 2 microns.

-

-

Dynamics of Partkle Removal

83

Stream Velocity (cm/s) 300

bars

250 200

.

.

I 0.0002

0.0004

0.0006

0.0008

Boundary Layer (an)

0.001

Figure 6. The relation between the stream velocity and the thickness of the viscous sublayer for 325 K (solid line) and 375 K (dashed line) at a pressure of 300 bars. The thickness ranges from about 1 micron for 350 cm/s to 10 microns for 50 cm/s.

In Fig. 7, we plot the radius of the smallest particle that can be rolled as a fbnction of the stream velocity for temperatures of 325 K (solid line) and 375 K (dashed line) at a fixed pressure of 300 bars. We notice first of all from this plot that a difference of 25 K in the operating temperature causes a change by almost a factor of two in the radius of the smallest particle which can be moved. It is clear that it is more efficient to work at the lower of the two temperatures that are displayed. Such plots allow us to design experiments with a goal of removing particles larger than some critical value. As pointed out in the introduction, the critical size of particle contaminants is expected to be of the order of 0.1 micron for the next generation of semiconductor wafers. Figure 7 shows that the stream velocity of supercritical C02 required to remove particles of this size is about 200 c d s at 375 K, and the required stream velocity is about 100 c d s at 325 K. Such velocities are attainable with currently available pumps.

84


Minimum Radius (microns) 300

0.08

’

0.04

.

0.02

bars

\

’

\2,

Figure 7. The theory of rolling developed for turbulent flow conditions can used to calculate (at 300 bars) the radius of the smallest particle that can be moved as a function of the stream velocity for 325 K (solid line) and 375 K (dashed line). Notice the sensitivity to temperature. Velocities of 100 c d s to 200 c d s are needed to move particles with a radius of about a tenth of a micron.

To compare the cleaning characteristics of supercritical carbon dioxide with that of air, we have plotted in Fig. 8 the radius of the smallest particle (in the range of a few microns) that can be removed with air at 1 atmosphere and a temperature of 300 K. It shows that even for a relatively high velocity of 1000 c d s , the minimum radius is 0.2 microns. This may be compared to 0.005 microns for supercritical C 0 2 . Requirements of such fast flow for air will probably force the use of jets. This raises an additional complication of having to install a mechanism to “sweep” the wafer clean. The sweeping motion will most certainly lead to additional contamination fi-om lubricated joints, etc. Such complications do not arise for the case of supercritical C 0 2 , given the relatively low velocities needed.

-


M i n h u m Radius (microns)

85

Air F l o w

Figure 8. In order to compare the cleaning efficiencies of air and supercritical CO,, we have displayed in this figure the radius of the smallest particle that can be moved with air at one atmosphere and 300 K, for velocities comparable to those used in the previous figure. Notice that the velocities required to move the particles with air are much higher than in the case of supercritical CO,.

5.0 CONCLUSIONS

We have shown that particles can be theoretically rolled off semiconductor wafers using a turbulent flow of supercritical C 0 2 over them. Flow speeds of a few hundred c d s will be required to remove particles less than a tenth of a micron in radius. The relative merits of using supercritical carbon dioxide over air include the use of lower flow velocities with supercritical CO,, it is pointed as well as an advantage in terms of chemistry over air. Supercritical C 0 2 dissolves organic molecules, implying that it can loosen the adhesion between organic contaminants (particles) and the semiconductor wafer.

86


REFERENCES 1. Mayer, A. and Shwartzman, S., J. ElectronicMater., 8:855 (1979) 2. Musselman, R. P. and Yarbrough, T. W., J. Environmental Sci., Vol. 51 (1987) 3. Archibald, L. C. and Lloyd, D., Particles on Surfaces, Vol. 3, (K. L. Mittal, ed.),Plenum Press, New York (1988) 4. Bhattacha~ya,S.andMittal, K. L., Surfiace Technology, 7:413 (1978) 5. Zimon, A., Adhesion o f h s t andpowder, p 252, Plenum Press, New York (1966) 6. Gim, R, Lesniewski, T., and Middleman, S., Particles on Surfaces, Detection, Adhesion and Removal, Vol. 3, (K. L. Mittal,ed.),Plenum Press, New York (1988) 7. Vatistas, N., Particle Adhesion to surface under turbulent flow conditions, in Particles on Surfaces, Vol. 3, (K. L. Mittal, ed.),Plenum Press, New York (1988) 8. Stanisic, M. M., TheMathematical Theory of Turbulence, p. 63, SpringerVerlag, New York (1988) 9. Ely, J. F., NIST COZPAC, Proceedings of the 65th Annual Convention of the Gas Processors Association, San Antonio, TX (1986) 10. Amsden, A. A., O’Rourke, P. J. and Butler, T. D., KIVA-11: A program for chemically reactive flows with sprays, Los Alamos National Laboratory Report LA-11560-MS, Appendix B (1989) 11. Israelichvili, J. N., Intermolecular and Surface Forces, p. 128 Academic Press (1991) 12. Goldman, A. J., Cox, R G. and Brenner, H., Chem. Eng. Sci., 22:637 (1967) 13. Reed, J., The Adhesion of small particles to a surface, Particles on Surfaces, (K. L. Mittal, ed.),2:3, Plenum Press, New York (1988) 14. Bowling, A., A Theoretical Review of Particle Adhesion, Particles on Surfaces, 1:129 (K. L. Mittal,ed.),Plenum Press, New York (1988) 15. Reid, R. C., Prausnitz, 1. M. and Shemood, T. K.,Properties of Gases and Liquids, Third Edition, McGraw-Hill Book Co., New York (1977)

Surfactants and Microemulsions in Supercritical Fluids Kevin Jackson and John L. Fulton

ABSTRACT A new class of cleaning solvents is described in which the unique solvent power of a supercritical fluid is united with the broad solvation capabilities of a reverse microemulsion. The application of these systems to cleaning processes is discussed. A reverse microemulsion has a structure similar to that of a conventional waterbased surfactant system with the exception that the nanometer-sized structures are inverted-the core of the micelle is an ultra-small droplet of water and the exterior is the oil or supercritical fluid phase. The advantages of supercritical microemulsions over conventional liquid or aqueous based systems are (0 energy savings (no drying is required), (ii) environmental benefits since the fluids may be benign and the contaminants or surfactants can more easily be recovered, (iii) microemulsions have a high capacity for oils or other lipophilic 87

88


materials, (iv) the selectivity for the contaminants may be adjusted by density manipulation, and (v) improved cleaning or extraction efficiency because of the high diffisivities and low viscosities inherent in supercritical fluids. A potent, new solvent is evolving-a supercritical carbon dioxide microemulsion. A CO,-based microemulsion is especially attractive since CO, is very abundant, relatively inexpensive, and environmentally benign at this scale of use. Applications of this system to cleaning processes appear very promising. The range of industrial applications that supercritical microemulsions aspire to cover is extensive and includes areas such as separations, reactions, and spray coatings (paints, lacquers, enamels and varnishes). Applications to cleaning could include dry cleaning, the separation of dyestuffs, the extraction of contaminants from soils, the regeneration of activated carbodcatalysts, and the removal of strongly polar or ionic species from printed circuit boards, polymers, foams, aerogels, porous ceramics, and laser optics. The combination of colloidal technology with the technology of supercritical fluids greatly enhances potential applications in cleaning processes.

1.0

INTRODUCTION

Supercritical fluids are attractive cleaning solvents because of their excellent mass transport characteristics and their unique physical properties that allow extracts to be isolated with relative ease. Carbon dioxide is especially attractive since it is inexpensive and environmentally benign in comparison to liquid solvents that it might replace. It also has a relatively low critical temperature which makes it an ideal solvent for extracting thermally labile substances. There is however a limitation of supercritical fluids having moderate critical temperatures in that they are poor solvents for polar or ionic species that are typically one of the target components in a cleaning operation. One strategy to overcome this basic limitation of supercritical fluids is the incorporation of a microemulsion phase to greatly enhance the solvent power of the solution. The combination of the unique properties of supercritical fluids (e.g., low viscosities, high diffusion

Sujbctants and Microemulsions

89

rates, solvent properties, etc.) with those of a dispersed microemulsion (or reverse micelle) phase creates a whole new class of solvents. The microemulsion phase adds to the properties of the supercritical fluid what amounts to a second solvent environment, which is highly polar and which may be manipulated by changing the pressure. Polar or ionic species partition into the aqueous microphase of the microemulsion. As a result, these microemulsions greatly expand the range of applications of supercritical fluids for cleaning processes. Although it is not quite a decade since the discovery of microemulsion formation in supercritical fluids['] there has been a in order to great deal of interest in understandingtheir propertie~[~I-[~1 efficiently apply this new technology. Microemulsions formed in conventional liquid solvents such as alkanes have been known for a much longer time and a variety of applications have been explored including use for general separations, enhanced oil recovery,[8] as reaction media for catalyti~,[~l['~] or enzymatic reactions,[' for general catalytic and semiconductor application^,['^]-[' 51 as a mobile phase in chromatographic separations,['6]and for the separation of proteins and biomolecules from aqueous ~olution.['~1['~] Many similar applications of supercritical microemulsions have been explored.[' 91[201 The technique may be viewed as an alternative to the addition of cosolvents or modifiers (sometimes termed entrainers) that are commonly used in supercritical fluid technology to enhance the polarity of the fluid. For cleaning processes, however, these cosolvents may be toxic or detrimental in various ways to the substrate. In addition, these modifiers are usually more difficult to separate downstream from the process due to their high volatility. In contrast, surfactants typically have very low volatility and thus interact to a much lesser degree with the substrate. Furthermore, they often dramatically improve the solubility of polar species, well beyond that of simple modifiers. Because of the obvious importance of a C0,-based system, there has been much work in the area of developing this type of supercritical fluid mi~roemulsion.[~1[~~]-[~~] On the surface, the strategy is simple and interesting parallels can be made to the first soap or surfactants used in water thousands of years ago. For instance, with

90


water, we start with an abundant, benign solvent and add surfactants that will aid in the solubilization of oils in order to improve the cleaning efficacy. For a CO, microemulsion we reverse the thinking. We start with the least expensive nonpolar solvent and add surfactants to increase the solubility of polar species-again dramatically improving the cleaning properties. As discussed herein, there are many applications where there are distinct advantages in using a C0,based microemulsion over aqueous- or liquid-based systems. The application of supercritical microemulsions to cleaning is a new area and there have been no published results to date applying these systems to cleaning operations but the combination of surfactants with fluids would be a natural choice for many systems. The information in this chapter is to be used as a guide to the development of cleaning applications based upon supercritical fluid-based microemulsions. 2.0 MICROEMULSION STRUCTURE

It is helphl to begin by describing and defining the basic properties of microemulsion systems. Microemulsions are clear, thermaj.tnamically stable solutions that generally contain water, a surfactant, and an oil. An example of the structure of a microemulsion phase is shown in Fig. 1. The “oil” in this case can be a nonpolar or low polarity liquid solvent or a supercritical fluid phase. The sufiactant molecules form organized molecular aggregates and preferentially occupy the interface between the oil and water phases. Although microemulsions are thermodynamically stable, they are not static and are continuously breaking and reforming. The oil and water microdomains have characteristic structural dimensions between 5 and 100 nm. Aggregates of this size are poor scatterers of visible light and hence these solutions are optically clear. Water-in-oil (w/o) microemulsions can have a multitude of different microscopic structures including sphere, rod, or disc shaped aggregates. Micelles are the smallest aggregate entity and are generally defined as binary systems containing a surfactant that forms organized clusters in either water (normal micelles) or in oil (reverse or inverted micelles). When

Surjktants and Microemulsions

91

small amounts of water are added to a reverse micelle system, the water preferentially partitions into the hydrophilic core, causing it to swell and yielding (at some arbitrary water content) a water-in-oil microemulsion. At low water concentrations, such inverted microemulsions typically contain spherical, nanometer-sized droplets of water. These microdomains of water are dispersed into a nonpolar solvent that is sometimesreferred to as the continuousphase solvent.

Figure 1. Structure of a water-insil microemulsion droplet showing partitioning of polar and nonpolar solutes into the water and oil microphases.

The shape, size, and structure of these dispersed droplets depend upon a multitude ofvariables including the surfactanttype, ionic strength, the presence of cosurfactant(s), and the amount of water. Commonly used surfactants are of the five general categories; anionic, cationic, nonionic, amphoteric and zwitterionic. The nature of amphoteric surfactants, i.e., whether or not they behave as anionic or cationic species, is dependent on the pH or ionic strength of the aqueous phase.

92


The amphiphilic nature of surfactant molecules having both highly polar or charged hydrophilic head groups and nonpolar, lipophilic tails, promotes aggregation. Their selective absorption into the interface reduces the surface tension of the medium. Hence, the origin of the name surfactant is derived from the expression, surfaceactive agent. Cosurfactants are small amounts of somewhat amphiphilic substances, typically low molecular weight alcohols, having some solubility in both oil and water, that greatly enhance the water- or oil-solubilizing capacity of the microemulsion. The shape of the surfactant molecule is a very important factor in predicting the types of aggregates formed. For reverse micelles and water-in-oil microemulsions, the surfactant must have a “wedge” shape where the hydrocarbon tails have a much higher volume than that of the polar head groups. This allows for efficient packing to form a stable inverted structure. Typical reverse micelle surfactants have twin alkyl tails in order to satis@ this packing requirement. Another option is to use cosurfactants such as low molecular weight alcohols in combination with single-tail surfactants to obtain the proper surface curvature required for reverse or inverted structures. Microemulsions have the ability to partition polar species into the aqueous core or nonpolar solutes into the continuous phase (See Fig. 1). They can therefore greatly increase the solvation of polar species in essentially a nonpolar medium. The surfactant interfacial region provides a dramatic transition from the highly polar aqueous core to the nonpolar continuous-phase solvent. This region represents a third type of solvent environment where amphiphilic solutes can reside. Such amphiphilic species will be strongly oriented in the interfacial film so that their polar ends are in the core of the microemulsion droplet and the nonpolar end is pointed towards or dissolved in the continuous phase solvent. When the continuous phase is a near-critical liquid (TR= T/T, > 0.75) or supercritical fluid, additional parameters such as transport properties, and pressure (or density) manipulation become important aids in applying this technology to chemical processes. The amount ofwater added to a water-in-oil (w/o) microemulsion is defined by the molar water-to-surfactant ratio, W; Generally, the greater the W value, the larger the size of the nanometer-sized

Sujirctants and Microemulsions

93

droplets in the microemulsion phase. When the volume of the dispersed water phase exceeds approximately 26% of the total microemulsion volume, a transition may occur from an inverted structure to what is postulated as a bicontinuous structure[25]and finally to normal structures at water volumes greater than 74%.[261 The ability to dissolve highly polar or ionic species in the polar microdomains of the microemulsion is, to a large part, dependent upon the amount of water incorporated into the microemulsion. The molar water-to-surfactant ratio ( W ) is a convenient measure of the polar solvent strength of the microemulsion. For ionic surfactants, the microemulsion will have a solvent power for polar species approaching that ofbulk water when Wis above 10. At lower Wvalues, the water acts to hydrate the head groups and counter-ions of the surfactant yielding a microsolvent environment that has a dielectric constant appreciably lower than that ofbulk water and, hence, a much reduced capacity for the dissolution of ionic or highly polar species. Manipulation of the Wvalue of the microemulsion can be viewed as a means to adjust the polarity of the microemulsion for cleaning operations. For supercritical microemulsions, the JV value can be adjusted through controlled addition of water to the solution or through the manipulation of the pressure or temperature of the solution. It should be noted that high concentrations of ionic species can alter the phase stability of microemulsions based upon ionic surfactant systems. Nonionic surfactant systems are much less susceptible to this effect. The curvature ofthe interfacial film ofthe microemulsion droplet is determined by a balance between the electrostatic interactions of the head groups and repulsive interactions of the surfactant tail group. Addition of ionic solutes can upset this delicate balance and induce phase separation. By changing the structure of the surfactant or through the addition of cosurfactants one can restore this balance and thus allow the dissolution of high concentrations of ionic species. A universal property of surfactant solutions is the existence of a critical micelle concentration (CMC) representing the minimum amount of surfactant required to form aggregates. The CMC also represents the solubility of the surfactant unimer in the oil or continuous phase solvent. At surfactant concentrations above the CMC,

94


surfactant unimers exist in equilibrium with the surfactant making up the micellar aggregates. The CMC in reverse micelle systems formed with ionic surfactants is typically lo4 to 10” molar. For a reverse micelle system in a supercritical fluid, the CMC is dependent upon the type or density of the fluid, the amount of water, and the temperature and pressure of the system. In order to obtain enhanced solvent powers of a microemulsion for a cleaning operation, one must operate well above the CMC of the particular surfactant. Also the capacity of the microemulsion to solvate polar or ionic species is naturally dependent upon the concentration of reverse micelles in the solution. Finally, in the discussion of reverse microemulsion systems, mention should be made of one of the most widely studied systems. The surfactant, sodium bis(2-ethylhexyl) sulfosuccinate or AerosolOT (AOT), is one of the most thoroughly studied reverse micelleforming surfactants since it readily forms reverse micelle and microemulsion phases in a multitude of different solvents without the addition of cosurfactants or other solvent modifiers. The phase behavior of AOT in liquid alkane/water systems is already well documented. Indeed, the first report of the existence of the formation of microemulsions in a supercritical fluid involved an AOT/alkane/ water system.[’] The spherical structure of an AOT/nonpolar-fluid/ water microemulsion droplet is shown in Fig. 1. In the now wellknown structure, it can be seen that the two hydrocarbon tails of each AOT molecule point outward into the nonpolar phase (e.g., supercritical fluid). These tails are lipophilic and are solvated by the nonpolar continuous phase solvent whereas the hydrophilic head groups are always positioned in the aqueous core.

3.0

MICROEMULSIONS IN SUPERCRITICAL FLUIDS

Most of the early work involving microemulsions in supercritical fluids utilized the supercritical alkanes, ethane and propane, with the surfactant AOT. Table 1 gives a summary of the surfactant systems that have been studied in supercritical hydrocarbon solvents. More recently, there has been some success with the formation of

Surfactants and Microemulsions

95

microemulsions in supercritical CO,, and these systems are summarized in Table 2. Table 1. Supercritical Fluid Surfactants and Solution Conditions GaS

Surfactant

Methane AOT (Tc432'C) (Pc=46 bar) Ethane AOT (Tc=32'C) (Pc=49 bar)

Temperature ("C)

Pressure (bar)

25

1650

37

250-850

'

250 220-350 250-350 Up to 1500

"

66

"

"

66

"

"

'L

"

'C

6'

"

66

"

" 66

" " " " " '6

CI2EOl C,2E02 Dodecanol CIZEOO CIZEO, C,EO5 C,E% AOT

Propane (Tc=97OC) (Pc=43 bar) 6'

LC

" 'L

"

'1

"

"

'I

'

37-100 37 37-50 40

380-500 100-300 Up to 400

"

"

"

"

'

90-450

"

"

"

'

"

"

103

100-300

103-110 110 up to 150 103 100

100-350 200 10-200 Up to 350 70-600 120-500 350 200-500

"

DDAB

(6

6

"

Lecithin

35

Propylene

"

Reference

96


Table 2. Supercritical Carbon Dioxide Surfactants and Solution Conditions Surfactant

Temperature Pressure

("0 Perfluorooctanoic acid 50 " Polyoxyethylene 4-lauryl ether Polyoxyethylene 3(Cll~14-iso-C13 rich) alkyl ether " " Polyoxyethylene 2-cetyl ether 'L Polyoxyethylene 4-nonyl phenol ' Polyoxyethylene 6-nonyl phenol " 2-Tertdodecylmercaptoethanol ' Glyceryl monostearate Moxpholinium peffluorooctanate ' ' Bis(2-ethylhexyl) amine ' n-Hexadecyl morpholine " F(CF2)7.5S02N(C2H5) (c2H40H) " 8' F17 S0zN(C2H5)H Ami nopolyfluorosulfonate 'gF1 7S02N(C~H~)(C~C00K) Trimethyl fluoroalkyl ammonium iodide 40 Silicone based surfactants

W

(bar) lHpl/[surf.l 97 388

(47) (47)

257 304-432 388 388 169-377 388 133 213-278 202 169 109 136-160 300-500

6-8

160-500 300

-

-

12

Bis(poly(hexafluoropropy1ene " 0xide))hydroxy aluminum Sodium poly(hexafluoropropy1ene " oxide) carboxylate Ammonium poly(hexafluoropropy1ene 0.8 ACFM Pump pressure could not be monitored C 0 2 temperature in cleaning vessel was not consistent and was hotter at the top than at the bottom

SOLUTION Replaced with 6000 psi disk Installed particle Wter before flow meter Installed separate heater controls Installed gauges Changed to micrometering valve Reconfigured line and added line heaters Installed heater and

I

ltemperature controller IHeater control for inlet was llnstalled controller and poorly designed thermocouples

No insulation on internal tubing C 0 2 was entering pump as gas instead of liquid No pump pressure gauges Insufficient heatingof C 0 2 entering vessel. Poor positioning of extractor control thermocouple

Installed insulation Installed tubing insulation, heat exhanger and chiller Installed pressure wges Designed, fabricated and installed temperature controlled water bath with heat exhanger and repositioned thermocouples

Precision Cleaning: A Case Study 203

204 Supercritical Fluid Cleaning The most important problems were associated with the lack of sufficient heat being supplied to the separator where dissolved materials were collected, and the inability of the pump to maintain high flow rates. The carbon dioxide was supplied to the system through a manifold of standard pressurized gas cylinders with siphon tubes to supply liquid carbon dioxide to the pump. At flow rates greater than 0.8 Actual Cubic Feet per Minute (ACFM), the liquid would begin to flash to gas in the transfer tubing and within the pump heads and therefore the high pressure pump could not maintain flow rates above 0.8 ACFM. The solution was to install a heat exchanger and chiller to cool the incoming carbon dioxide to approximately -20°F to ensure that it remained liquid all the way to the pump inlet and in the pump head. The flow rate could be maintained at 6.5 ACFM with a constant supply of liquid carbon dioxide. The addition of the chiller system for the pump added to the problem for the as-supplied heating system for the pressurized carbon dioxide. The system was not capable of maintaining the pressurized carbon dioxide at the required process temperatures of approximately 170-185°F. The problem was magnified by the position of the cleaning vessel temperature control thermocouple. The system was originally designed with the inlet on the chamber bottom and outlet on top with the control thermocouple near the inlet port. When the thermocouple sensed low temperature carbon dioxide entering the chamber it would turn on full heat to the vessel. This resulted in the carbon dioxide exiting the vessel well above the set process temperature. The problem was resolved by designing and installing a temperature-controlled water bath that the pressurized carbon dioxide would pass through prior to entry into the extraction vessel. The water bath would raise the temperature of the carbon dioxide to the required process temperatures at both high and low flow rates. Freezing of the separator during operation at flow rates above 0.5 ACFM was another major problem with the original equipment. Originally, the system was designed as an open loop, with the gaseous carbon dioxide being exhausted to the building vent system after passing through the cleaning chamber and separator. The separator was not designed for the amount of expansional cooling encountered at the desired flow rates and would freeze due to the lack of sufficient

Precision Cleaning: A Case Study

205

heat input. This was corrected by adding additionaljacket heaters to the separator and using liberal amounts of heat tape and insulation on equipment plumbing. Addition and control of heating elements had to be carehlly considered due to the heat sensitive nature of the damping fluid being extracted. This fluid could not withstand temperatures greater than ZOOOF without degradation, and therefore the heating elements had to be designed and controlled so as to not exceed these temperature limits while still supplying the required heat input. In summary, the problems associated with this system were related to the fact that supercritical fluids had been used previously for batch extraction processes and there had been no significant development to this time of supercritical fluid extractors for the specific application of precision cleaning. These problems were exacerbated by an unwillingness of the vendors to discuss many of the problems they had already encountered and solved, based on “proprietary” process concerns.

5.0

PROCESS DEVELOPMENT

5.1

Initial Testing

M e r the SCF system was operational, it was necessary to determine process parameters for removal of the common contaminants. Temperature, pressure, flow rate and time were identified as critical parameters and a number of experiments were designed using Taguchi methods to minimize the number of experiments necessary to accurately determine the effects of these parameters on the cleaning effectiveness. Damping fluid, Krytox 143AZTMlubricant, and hydrocarbon vacuum pump oil were selected as the contaminants for removal. In each experiment a known weight of each material was put in a cleaned and tared Petri dish. The system was brought up to pressure and temperature before flow was initiated at a specified rate for a specific time. The chamber was then slowly vented to atmosphere and the Petri dish removed and reweighed to determine percent removal by weight.

206 Supercritical Fluid Cleaning 5.2 Results of Initial Testing The results of the these experiments showed that the damping fluid, KrytoxTM,and hydrocarbon oils all could be successhlly removed using supercritical carbon dioxide at pressures in the range of 1200-3000 psi and temperatures in the range of 90 to 185°F. The KrytoxTMlubricant was easily removed at the lower extremes, while the hydrocarbon oil was the most difficult to remove. Determining the optimum parameters was a long process which required many hours of engineering and technician time and spanned a time period of over one year. In addition to these initial multivariant experiments, a large number of independent experiments were conducted to veri@ the predicted results. Figure 2 is typical of the solubility curves generated during these additional independent tests, These curves, while specific to the tests being performed, indicate the relationship between temperature, pressure and solubility of the soils and fluids being removed. These curves indicated that within the temperature and pressure parameters tested, the solubility of damping fluids increased with both temperature and pressure. It was our practice, therefore, to select the highest pressure and temperature conditions to effect the best solubility of the specific contaminant when implementing a cleaning process. This decision was then evaluated through additional testing to address the compatibility ofthe parts being cleaned with the selected temperatures and pressures. Temperatures greater than 185°F were not evaluated because of the compatibility problems associated with halogenated damping fluids as discussed earlier.

-

e160 'F 140 *F

1500

2000

2500

3000

PRESSURE (psi)

Figure 2.

Solubility of damping fluid in supercritical CO,.

3500


207

5.3 Cleaning of Cracks and Crevices Since supercritical fluids were chosen for their ability to penetrate small cracks and crevices, additional tests were performed to evaluate this characteristic. A test cube, modeled after a similar fixture fabricated by Fenanti Aerospace, was developed and manufactured to aid in this study. The cube had a number of blind holes, tapped holes, channels and crevices to simulate actual hardware. Beryllium, 300 Series stainless steel and aluminum cubes were constructed to simulate the common metals found in the instrument. In addition, the sides of the cube were removable to facilitate deposition of the contaminants into these blind holes and crevices and later analysis of cleaning effectiveness. The base of the cube was equipped with a scanning electron microscope (SEW mount so that the cube could be examined directly in the SEM. Figure 3 is a photograph of a test cube. Extensive evaluations with these test cubes indicated that supercritical fluids were indeed effective at removing contaminants from cracks and crevices.

Figure 3.

Photograph of Honeywell test cube.

208 Supercritical Fluid Cleaning

One of the other factors which was investigated in the process development stage was the use of multiple pressure changes during the cleaning process. The idea was that decompression would cause a density change in the supercritical fluid. The resultant mass transfer of carbon dioxide would then carry trapped contaminants out of tight spaces more effectively than simply flowing supercritical fluid over the parts. The Honeywell test cube was cleaned, weighed and contaminated with a known amount of damping fluid. The cube was then cleaned using both variable pressure and constant pressure processes and the removal effectiveness was compared for the two processes. The variable pressure process, which we called the pumping process, had an average efficiency was 95% whereas the constant pressure method averaged greater than 98% efficiency. Based on these results, it was determined that the pumping method was no more effective than constant pressure and was much more labor intensive on a manual system.

6.0

COMPATIBILITY TESTING

6.1 General

M e r determining the optimum parameters for removal of contaminants, it was necessary to determine if these conditions were compatible with the wide variety of materials and component assemblies which are present in the instrument. It was assumed, and later confirmed, that metallic components would not be effected by the pressures, but there were definite concerns with polymer components such as epoxies, thermoplastics such as DelrinTM,and various organic and inorganic coatings. Samples of the materials were prepared and tests were performed to address issues such as weight gains, dimensional changes and material degradation. A humorous example of compatibility testing occurred during the initial development of the SCF system. A group of executives were touring the facility and were given a demonstration of the SCF system as a possible replacement cleaning process for other product lines. There was a great deal of interest in this process and one

Precision Cleaning:A Case Study

209

executive asked if we could clean the residual oils from his plastic handled pocket knife. We assured him that this was a simple task and placed the knife in the cleaning chamber for a 10 minute cleaning cycle. At the end of the cycle we opened the chamber and removed the pocket knife, minus the plastic handles. The handles were deformed and laying in the bottom of the chamber. It appeared that the plastic could not withstand the heat or exposure to supercritical carbon dioxide. Needless to say, numerous apologies were made and we offered to replace the knife. This serves to illustrate the need for extensive compatibility testing prior to performing demonstrations for executives. 6.2

Epoxies

Epoxies, both filled and unfilled, are used throughout precision instruments as potting materials and sealing adhesives. Degradation or dimensional changes of these epoxies would cause serious problems and many tests were conducted to determine the effects of supercritical carbon dioxide on the various epoxies. The specific epoxies were filled and unfilled mixtures of a proprietary amine cured epoxy compound. Results of these coupon tests indicated that the supercritical carbon dioxide would cause a weight gain of approximately 0.1 to 2.6%. This weight gain appeared to be due to absorption of carbon dioxide by the epoxy and was reversible by using a 2830" Hg vacuum bake at 180-200°F. Slight dimensional changes in the epoxy were also detected, but were reversible with the use of a vacuum bake. No significant changes were detected in mechanical or thermal properties of the epoxies following exposure to supercritical carbon dioxide. 6.3 Thermoplastics DelrinTMis used as a bearing retainer (cage) material and was evaluated since removal of KrytoxTMlubricant from bearing assemblies was an area targeted for implementation of supercritical fluid cleaning. This material also exhibited slight weight gains due to absorption of carbon dioxide. The weight gains of retainer samples

21 0 Supercritical Fluid Cleaning were approximately 1% after exposure to supercritical carbon dioxide. Swelling on the order of 0.1 to 0.7 % was measured after exposure. Both the weight and the dimensions returned original values after approximately 20-24 hours at 129°F. TeflonTMis another thermoplastic used extensively in fixturing which would be exposed to supercritical fluid cleaning. It showed weight gains and dimensional changes which were greater than those seen in Delrinm, but after vacuum baking and exposure to room temperature these changes were similar to those seen with exposure to CFC-113. The testing of these specific polymer materials indicated that the absorbed carbon dioxide will outgas over time and that the rate of outgassing can be increased by increases in temperature and/or vacuum baking. These materials were considered compatible with supercritical cleaning for our applications. 6.4

Damping Fluid

Compatibility of a halogenated damping fluid was also evaluated to ensure that the fluid was not being degraded by exposure to supercritical carbon dioxide. This was important for cleaning of tooling and fixturing since these components contained relatively large amounts (40-65 grams) of this expensive fluid. As discussed previously, the supercritical process was selected for the ability to reclaim the damping fluid from fixturing less expensively by avoiding solvent contamination. To test compatibility, the fluid was exposed to supercritical carbon dioxide for 30 minutes at 4000 psi and 185°F. Following exposure, the remaining fluid was analyzed for viscosity, density, corrosion potential, chemical saturation and purity (by high pressure liquid chromatography-HPLC) and compared to a sample fi-omthe same lot of fluid which had not been exposed to supercritical fluid. All values were within specification requirements, with no significant changes, indicating that supercritical fluid did not affect the damping fluid. The HPLC chromatogram of the damping fluid suggested that there may have been some hydrocarbon contamination by the carbon dioxide, but it was not significant enough to fail the purity requirement. Also present were several other compounds attributed to contamination picked up in the SCF cleaning system.


211

This indicated the necessity to completely clean the system when changing between different parts to be cleaned. A typical cleaning process involves performing a 1 hour SCF run with an empty chamber. Cross-contamination can be a problem when using one piece of equipment to clean multiple contaminants, especially when one of those materials is being collected for reuse. 6.5 Hardware Compatibility

In addition to material compatibility testing, there were a significant number of hardware compatibility tests performed. Unlike the materials testing, this testing detected problems with certain pieces of hardware. End housings, as described earlier, were put through a variety of tests including electrical, magnetic, dimensional and microscopic examination to determine if any changes were detected. No impact of the SCF process was detected by these tests other than small weight increases and dimensional changes which were reversible by vacuum baking. Hermetically sealed assemblies were not as compatible and, as expected, did not tolerate the pressure of the cleaning process. A motor stater, composed of wire windings potted in place with an unfilled epoxy was also found to be incompatible with the cleaning process. In this case, the carbon dioxide would fill voids in the epoxy and then during depressurization cause fiacturing at these voids. Later this problem was alleviated by cleaning at a lower pressure and decreasing the decompression rate. The issues raised during testing of these different assemblies illustrate the need for effective and complete compatibility analysis prior to implementation of a cleaning process. Materials analysis alone is insufficient when determining if supercritical fluid cleaning can be used in a proposed application.

7.0

SYSTEM MODIFICATIONS

As development of the SCF cleaning process continued, it became apparent that certain changes were required to the cleaning

212 Supercritical Fluid Cleaning system. These were needed to improve cleaning efficiency and enhance the capabilities of the system. The first modification was the acquisition of an additional cleaning vessel. This vessel had interior dimensions of 3 inches diameter and 8 inches deep (0.9 liter) and was equipped with a magnetically driven stirring propeller, The smaller chamber would allow a lower cleaning volume for experimentation purposes and the lower thermal mass would allow more precise control of the extractor temperature. The system was connected to the pump station through a solenoid valve so that either chamber could be selected for use. The stirring capability was added so that the effects of stirring on cleaning effectiveness could be evaluated. It was believed that stirring would improve the cleaning efficiency for bulk removal of damping fluids, but would not greatly improve efficiency during removal of small amounts of contaminants trapped in deep cracks and crevices. During execution of the large number of test plans, it became apparent that monitoring and adjusting the temperature, pressure and flow rate were difficult to perform in a completely manual mode. A PC-based monitoring system was added to the SCF system to provide monitoring of the various temperatures and pressures within the system. The PC was set up to control and monitor the SCF system, but due to time and budget constraints only the monitoring hardware was installed. With this system, monitoring was performed of chamber inlet and outlet temperatures, separator temperatures, pump and chamber pressures, water bath temperatures and flow rates. To make this system more readable and meaninghl to the operator, an animated display system was used to display real time data with updates every second. Another change was made when the electrically powered high pressure pump that was supplied with the original system was replaced with an air-operated high pressure pump with higher capacity. The original pump took approximately 20-30 minutes to reach operating pressures with the existing cleaning chambers and would have taken an excessively long time to obtain operating conditions in the larger chamber which was planned for later implementation. The airdriven pump had been used in supercritical cleaning systems for similar applications and had shown excellent performance. The pump


213

operates using an in-plant air supply and would reach operating pressure within a few minutes with the existing extraction chambers. The pump was installed and has worked extremely well. By ducting the exhaust from the pump into the plant vent system, the noise level from this pump was similar to that of the electric pump. The airoperated pump also added another degree of safety to the system. The maximum liquid delivery pressure is directly related to the incoming air pressure. Tightly controlling the inlet air pressure lowers the probability of system overpressure. A third cleaning chamber was purchased and installed. This chamber was much larger (8.5 inches diam. x 16 inches deep, 14.9 liter capacity) than either of the other two chambers. It was equipped with an pneumatically operated hydraulic closure and locking mechanism and was equipped with a variable speed magnetic drive. The larger size was necessary so that all components of a system to fill guidance instruments could be cleaned in one SCF cleaning run.

8.0 PRODUCTION IMPLEMENTATION

With over three years invested in the development and qualification of the supercritical cleaning process, the next phase of the implementation plan was cleaning of production hardware. The two key operations which had been identified early on as potential applications were the cleaning of end housings and the removal of bulk amounts of damping fluids from fill system fixturing. 8.1

Bulk Fluid Removal

The ability of SCF to remove contaminants is directly related to the number of volume exchanges of SCF within the cleaning chamber and the amount of contaminant. Previous experiments in removal of damping fluid indicated that it would take approximately 3-4 hours at 3500 psi and 185°F with a flow of 1.0 Actual Cubic Feet per Minute (ACFM) to clean fill stand hardware. This was performed in the small (2.4 liter) chamber. Since the volume of the new chamber (14.9 liter) was much greater than the previous chamber, it was anticipated that

21 4 Supercritical Fluid Cleaning the flow rates and cleaning times would need to increase significantly to obtain the same number of volume exchanges. Additionally, large plates which were too big to fit in the 2.5 liter chamber were added to parts cleaned in each batch. These plates hold significant amounts of fluid, thus increasing the amount of damping fluid which needed to be removed and reclaimed fiom the fixturing. The flow rate for use with the large chamber was increased to 2-3 ACFM and the time at pressure (soak time) was increased to 5 hours in order to address the issues discussed above. Visual examination of the hardware following this cleaning cycle indicated that the parts were not completely clean and another 3 hour run was necessary to completely remove the damping fluid. Another process problem which became apparent was the time necessary to vent the large chamber at the end of a cleaning cycle. The system was limited by the separator design and cannot vent faster than approximately 3 ACFM. The large chamber required approximately 1.5 hours to return to ambient conditions. To achieve the most efficient process times, it was desirable to perform the cleaning in one cycle rather than multiple shorter cycles. Due to the difficulties listed above, it was decided to implement mechanical stirring into the large chamber to improve the cleaning efficiency. A four-blade aluminum fan was procured and electroless nickel plated to reduce particle generation. An adapter was manufactured to allow mounting of the fan to the existing magnetic drive at the bottom of the chamber. A removable framework and stainless steel screen were installed to prevent the fill fixturing from contacting the fan blades. Additional cleaning runs were successfblly performed using the stirred extractor and a six-hour soak time. This process allowed a fill stand to be cleaned in one eight-hour shift, including the pressurization and depressurization cycles. One final problem encountered with stimng was related to the strength of the aluminum fan. The adapter on the aluminum fan was not sufficient to withstand the reaction torque generated by the density of the SCF and subsequently allowed the fan to spin on the adapter and deform the fan blades. The aluminum fan was removed and a three-blade polished stainless steel mixing propeller was procured. The magnetic drive assembly was removed and the shaft


215

and propeller threaded with a left-hand thread to prevent slippage. With these modifications the implementation of SCF for bulk fluid removal was complete. 8.2 End Housing Cleaning

End housing cleaning has always been one of the most time consuming cleaning procedures. Small cracks and crevices present at the intersection of the laminated pole faces and epoxy potting can trap small quantities of damping fluid. Additionally, these openings will allow damping fluid to travel behind the pole faces to small areas of porosity in the potting epoxy. When the fluid becomes trapped in these voids, it is very difficult to remove since it is not directly exposed to the solvents. As discussed previously, vacuum pressure cleaning (VPC)with CFC-113 has been the historical process for end housing cleaning. Evaluation of cleaning effectivenessis performed by visual inspection after vacuum Lake. If fluid is seen, the end housing is processed through ano-iherVPC cycle. At the end of the day, the end housing is placed in a vacuum oven at 200°F overnight to try and draw out additional fluid. The end housing is inspected again the next morning for evidence of additional fluid. Supercritical fluid has long been proposed as an alternative to VPC for end housing cleaning due to its high diasivity and solvent characteristics. During the compatibility testing, epoxy coupons as well as scrap end housings were used to evaluate the compatibility (both material and performance properties) of SCF with end housings. Results indicated that the end housings were compatible with SCF. Although compatibility testing indicated that SCF was an acceptable process, implementation of this process was much more difficultthan it originally seemed. In conjunction with the original evaluations, end housings from different inertial instruments were evaluated to determine if SCF cleaning could beimplemented. Several SCF processes were evaluated, including the pumping method (see Sec. 5.3) where the pressure was rapidly varied between 2000 and 3500 psi. Following this cleaning process, the end housings exhibited numerous cracks (sev-

21 6 Supercritical Fluid Cleaning

era1 mils in length) in the epoxy potting at the corners of magnetic pole faces after one cycle and this pumping process was abandoned. Since it appeared that multiple cycles of SCF cleaning would be necessary to clean end housings, and that this process may cause cracking of the epoxy, additional studies on the effects of multiple cycles were initiated. At that time, the cleaning process consisted of a 500 psi per minute ramp rate to 3500 psi, with a 40 minute cleaning cycle. After eight cycles, small cracks (3-5 mils) began to appear in the epoxy at the corners of the pole faces on some of the end housings. All of the end housings exhibited cracking after 8 to 19 cycles of SCF cleaning. Honeywell decided to limit the exposure of end housings to a maximum of 4 cycles of SCF until the reasons for the end housing cracking were defined. Also at this time, the process parameters for SCF cleaning of end housings was changed. This change slightly lowered temperatures and pressures, without significantly affecting the solubility of the damping fluid in SCF. Using this new procedure, a large number of end housings were cleaned with SCF in support of a project to build instruments without the use of ozone-depleting chemicals. The process was limited to four cycles of SCF followed by vacuum pressure cleaning. The majority of the end housings were clean after 4 SCF cycles followed by 2-5 VPC cycles in a perfluorocarbon solvent. All end housings passed the acceptance testing required prior to assembly into instruments. This was considered a major advancement since the number of W C cycles, and thus the process time, were decreased dramatically. Although this was a usable process, it was desirable to perform end housing cleaning using SCF as the sole cleaning process. In order to accomplish this task, an extensive study was undertaken to determine why cracking occurred with multiple SCF cycles and develop a process which would not produce cracking of the epoxy. If this were possible, then no limits would be needed on the maximum number of SCF cleaning cycles. A test plan was developed to expose a number of scrap end housings to 20 cycles of SCF and evaluate the effects of these multiple cycles through both visual (for cracks) and performance


21 7

testing. At the same time, a number of compatibility studies were performed on components of the end housings to determine the reasons for the cracking phenomena. One of the variables for the evaluation were the ramp rates which were used for the pressurization and depressurization cycles. In addition, a vacuum bake after every SCF cleaning cycle was added. These parameters were thought to be important since the most promising theory was that the cracks were created through stresses in the epoxy. This theory was supported by the direction of the cracks (diagonal from pole face corners) and that the epoxy was absorbing carbon dioxide which caused weight gains and slight amounts of swelling in epoxy coupons. The results of the evaluations indicated that lowering the ramp rate from approximately 500 psi/min to 150 psi/min and performing a vacuum bake after every cycle drastically decreased or eliminated the evidence of cracks or discolored (stressed) areas in the epoxy. This indicates that the lower ramp rates decreased the stress on the epoxy and that the vacuum baking acted as a stress relief by removing the carbon dioxide gas which was absorbed in the epoxy. Based on these evaluations, a number of production end housings were cleaned using the SCF process which had been developed. The end housings passed all dimensional and performance tests prior to and after SCF cleaning and were assembled into instruments earmarked for fbrther evaluations to determine if instrument performance met all requirements. In summary, the road between initial development of alternative cleaning processes and implementation of those processes into the production environment is often long and full of potential potholes. This is especially true when implementing those processes for precision cleaning where the parts are typically complex and expensive, and the results of changes are not immediately apparent.

9.0

LESSONS LEARNED

Several important lessons were learned during this project. One of the first and most important was related to the equipment necessary for performing precision cleaning using supercritical fluid. The

21 8 Supercritical Fluid Cleaning majority of supercritical extraction development and equipment manufacturing has been in the field of supercritical extraction for the food industry or chemical processing industry. Equipment requirements for precision cleaning with supercritical fluids are different when one is are trying to remove all of the contaminants rather than selectively extract contaminants. Because of this, vendors were only partly able to design and build equipment suitable for cleaning applications. Compounding this problem was the unwillingness of vendors to k l l y reveal, and to take into consideration, design improvements and equipment requirements that they had already documented. This led to a situation where large amounts time and money were expended to find out what the vendor already knew concerning equipment design and the SCF process. System requirements should be carehlly considered and thoroughly documented so that there will be no misunderstandings between the procuring agency and the vendor. It would also be desirable to include acceptance testing as a condition for purchase. Supercritical fluids are compatible with a wide range of materials, but there are special concerns which need to be addressed with polymeric materials. As discussed in Sec. 6.1, these concerns are especially important when providing demonstrations for executives. Most polymers will absorb the gas phase which causes, at a minimum, swelling and weight gain. This absorption is typically reversible through vacuum baking. Also, by tailoring the process parameters, supercritical fluids can sometimes be used in applications where there were at first compatibility problems associated with a non-optimized process. The process development stage for implementation of supercritical fluids is often frustrating. The industry has made significant progress since the original implementation of supercritical fluid for precision cleaning and there is a much broader knowledge base from which to obtain process information. The Joint Association for the Advancement of Supercritical Technology ( J U S T ) is a valuable asset for accumulating knowledge and membership is recommended for potential users of this technology. Supercritical fluids are a viable alternative for certain high value applications, but as with any alternative there is a learning curve


219

associated with implementation of a new process. Because of the wide variety of contaminants encountered in the field of precision cleaning, the use of any alternative must be considered on a case-bycase basis. This consideration should encompass cleaning ability, compatibility issues, ease of process implementation and environmental acceptability as well as common factors such economics and safety. ACKNOWLEDGMENTS

The work presented here was the result of efforts by a large number of people. Special thanks to Tom Ingram, Ron Boltze and Bob Grab for their support. REFERENCES 1. McHugh, M., and Krukonis, V., SupercriticalFluidExtraction -Principles

and Practices, Butterworth, Boston (1986) 2. Industty Cooperativefor Ozone Layer Protection, Eliminating CFC-I 13 and Methyl Chloroform in Precision Cleaning Operations, United States Environmental Protection Agency, Washington, DC (199 1)

Scaleup Considerations Max: R Phelps, Laura K. Silva, and Michael 0.Hogan

1.0

INTRODUCTION

Successhl scaleup usually combines good teamwork, a strong working knowledge of the process, and some experience in modeling and analyzing experimental results. However, there is no fixed way for achieving scaleup because each situation is different and the methods chosen must apply accordingly. Scaleup considerations for a cleaning system can be viewed from three different levels: overall, process, and mechanistic. Overall scaleup considerations cover the entire cleaning operation, from acquiring and processing raw resources to the geographical distribution of the final product. The process level considers pumps, separators, cleaning vessels, condensers, solvents, and solutes collectively. At the mechanistic level, considerations are specific to each part in the process. This chapter discusses general parameters, processes, and procedures for scaling up supercritical fluid parts cleaning (SFPC) vessels at the mechanistic level. A section is also provided which '1 and includes work at Pacific Northwest Laboratory (PNL),[lOI[l describes the subsequent scaleup approach used in designing a transportable SFPC cleaning unit. 220

Scaleup Considerations

221

1.1 Pre-Selection Activities Before a method for scaleup of a cleaning vessel is selected, the parameters that affect the process need to be understood, as well as how changes to these parameters affect operational results. This knowledge, coupled with an understanding of what is ultimately required of the cleaning process, defines which operating variables must be chosen by the design engineer before the initial vessel design can be established.L2] The operating variables to be considered include temperature, pressure, time, modifier affects, and agitation of the vessel. These variables are typically determined by the physical properties of the cleaning solvent and solute, e.g., solubility of the solute in the solvent, mass transfer rates of the solute into the bulk solvent, and reactivity of the solvent with the solute. These properties are determined in the laboratory from a scaled working model of the process or found in a source reference book. 1.2 Scaleup Approach While the simplest method for scaleup of a chemical reactor is to start with a small working unit that provides the desired results and then progressively build larger units, this method is usually too expensive and time-consuming. Consequently, more semiempirical approaches, such as dimensional similitude, are taken. Other approaches include mathematical modeling or a combined dimensionless similituddmathematical modeling method but are not involved in the scaleup example discussed here. Dimensional similitude is based on principles of similarity and uses dimensionless ratios of the physical and chemical parameters that govern the working model to design the scaled-up prototype. This method is followed when the prototypical unit is expected to be dimensionally similar to the small-scale unit. The degree of success in using dimensional similitude as an approach to scaleup depends mainly on the extent to which the physical parameters necessary to achieve a desired result influence the process.

222 Supercritical Fluid Cleaning Frequently, complete dimensional similitude cannot be reached between two reactors with widely different scales; as a result, this method is usually limited in practice to relatively simple chemical reaction systems. For a cleaning vessel in which the reaction rate is very fast and the process is governed by the physical rates of the process, e.g., mass transfer, heat transfer, etc., the dimensionless groups describing the process consist of fluid mechanic and thermodynamic quantities. The cleaning process can usually be a relatively simple mechanism to describe, making dimensional similitude more easily achievable. The following discussion introduces the considerations applicable to initial design and subsequent scaleup of any supercritical or near-critical fluids parts cleaning vessel, then centers on dimensionless variables and principles for applying these variables to actual scaleup of a parts cleaning vessel. The last part of the discussion focuses on the application ofthese concepts to an agitated, supercritical carbon dioxide, parts cleaning vessel.

2.0

DESIGN CONSIDERATIONS

The nature of the solvent/contaminant interaction is the first consideration for designing the cleaning vessel. If this interaction is governed primarily by physical transport processes, the vessel should be designed to optimize mass transport mechanisms. If, however, it also involves a chemical reaction, then the governing rate of solvation is dependent, to a certain extent, on the reaction between the solvent and contaminant. If the reaction rate is extremely fast, then cleaning is again limited to mass transport mechanisms. If the reaction rate is relatively slow, then cleaning is limited by the chemical reaction. The rate equation is hndamental to the design of any chemical react0r.1~1 This equation describes the rate at which a chemical process takes place with respect to the physical dimensions of the reactor and, therefore, the size of the equipment needed to reach the desired level of cleaning for a given solvent throughput.[4]


223

where 5 0

Fi rj

= rate of chemical speciesj entering the reactor = =

V

=

flj/ dt

=

(moledtime) rate of flow of speciesj out of the system (moledtime) rate of formation of speciesj, a -5 would be the rate of disappearance of speciesj (moles/ time x volume) reaction volume rate of accumulation of speciesj within the system (moledtime).

This equation is true for batch as well as continuous operations. It is also true for semi-batch, stirred-tank, and tubular reactors, although the equation takes on different forms for each reactor. Factors that enter into the rate equation describe the changes that occur in the cleaning process with respect to the physical dimensions of the cleaning vessel, the physical properties of the solvent and solute, operating conditions, and the rate at which cleaning occurs. The dimensional factors are determined by the dependency on the total amount of cleaning accomplished for a given set of dimensions. These factors would be different, for example, for a cleaning process that varied according to the length of a cleaning vessel than one that varied with the volume of the vessel or the surface area of the part being cleaned. Time factors are determined by how the cleaning process changes with time. When the cleaning process proceeds at a constant rate, both the dimensional and time factors are typically incorporated into the mass transfer coefficient. The rate equation includes necessary factors for describing the cleaning vessel for either initial design or scaleup operations.


However, each set of factors entering in to the rate expression is also a potential source of scaleup error. For this, and other reasons, a hndamental requirement when scaling a process is that the model and prototype be similar to each other with respect to reactor type and design. For example, a cleaning process model of a continuousstirred tank reactor (CSTR) cannot be scaled to a prototype with a tubular reactor design. Process conditions such as fluid flow and heat and mass transfer are totally different for the two types of reactors. However, results from rate-of-reaction experiments using a batch reactor can be used to design either a CSTR or a tubular reactor based solely on a hnction of conversion, -rA

3.0

DIMENSIONLESS VARIABLES

Once the vessel type has been selected and the rate expression established for the cleaning process, the design engineer must decide which variables are significant for the scaling process. Mathematical tools in the form of dimensionless variables are usehl for determining these variables. Every chemical or physical process can be defined by a set of dimensionless ratios or variables intrinsic to the process. An example is the Reynolds number, which is defined for a process with internal flow through a tube by solvent properties, flow rate of the solvent, and the diameter of the tube.161

where

NRe = Reynolds number p = density of the solvent p = dynamic viscosity of the solvent V = velocity or average velocity of the solvent flow through the tube D = diameter of the tube


225

The Reynolds number is useful for characterizing flow of a Newtonian fluid through a tube. The variables in dimensional analysis can be arranged to suggest a valid ratio when none exists. Therefore, established dimensionless variables should be used where applicable. A number of these dimensionless variables have been proven to work in scaleup applications. Examples include the Reynolds, Nusselt, Grashof, and Sherwood numbers, all of which are completely described in Perry’s Chemical Engineers Handbook.[’] An important consideration when scaling a process is that resolving all of the dimensionless variables being used simultaneously is not always possible. Sometimes the design engineer must decide which quantities are more important to the overall objectives of the system design. For example, to achieve dynamic similarity in an agitated, unbaffled, cleaning chamber, the following dimensionless quantities must be satisfied:

Impeller Reynolds Number P

(N2DIIp = (N2DIIm

P

=

N

=

DI

=

p

=

p

=

m

Impeller Froude Number

power consumption of the impeller speed of the impeller diameter of the impeller density of the solvent viscosity of the solvent.

226 Supercritical Fluid Cleaning The subscriptsp and m designate the prototype and model, respectively. The impeZZer ReynoZds number is proportional to the standard Reynolds number and is arrived at by assuming that the velocity of the solvent flow through the chamber is proportional to the speed of the impeller and the area through which it acts. The impeZZer Fruude number is arrived at in a similar manner. Now, using these variables, assume a prototype to be designed with an internal volume 125 times that of the model. The solvent density and viscosity are considered fixed. Then for a geometrically similar vessel:

Substituting this value into the equation for dynamic similarity, as based on the power number, gives:

Solving the equation for the Reynolds numbers gives:

However, solving the equation for the Froude numbers gives: 1

N p = -Nm

Js

which conflicts with the requirements for dynamic similarity as based on the Reynolds numbers. Therefore, it is impossible to resolve all three equations if they are equally important. The design engineer


227

must then select a set of criteria that best fits the requirements of the system.[5] For a SFTC system, it is generally more important to satis@the requirements for the Reynolds number.

4.0

PRINCIPLE OF SIMILARITY

When scaling the process, the model and prototype must have Reynolds numbers in the same regime (i.e., laminar or turbulent flow) in order to achieve similar results. Ideally, the two reactors would have nearly identical or identical Reynolds numbers. In order to satisfy this requirement for scaleup, the model and prototype must be similar to each other with respect to reactor design, fluid flow, and physical dimensions. According to the principle of similarity, for every process condition and point in the model, there must be a corresponding condition and point in the pr0totype.1~1This principle is applied to the scaling process by observing similar requirements for several other dimensionless ratios or variables that must be treated in a similar fashion. The process of scaleup can be simplified by dividing the principle of similarity into four separate states or categories: 9

9

Geometrical Mechanical Thermal Chemical

Each of these four states is dependent on the other three. Mechanical similarity is impossible to achieve if the model and prototype are not geometrically similar. The requirements for chemical and thermal similarity are then impossible to achieve without the requirements for mechanical similarity being met. 4.1

Geometrical Similarity

Of the four states of similarity, only the requirements for geometrical similarity are easy to understand and meet. Simply stated, the model and the prototype must be of the same shape and design.

228 Supercritical Fluid Cleaning Ideally, for every point in the model, there should be a corresponding point in the prototype that can be traced out by a constant scale factor.[’] This means that for every point defined by an x, y, and z coordinate in the model there is a corresponding point x ’,y ’ and z ’ in the prototype, which can be defined by the ratios:

This requirement might not hold in actual scaleup practice. The model might be used to test conditionsthrough a single element of the prototype vessel, and the prototype might contain several of these specific elements. For example, the model may be used to explore the effects of solvent flow rates on cleaning efficiency through a specific volume element. The prototype may contain several of these volume elements. For this example, the length of the prototype would not necessarily be scalableby the same factor as the vessel diameter. If all other factors are kept constant, the model and prototype might still meet the requirements for geometrical similarity even though the scale ratio along each axis might be different. This results in a more general expression of corresponding points given by the equation below:

4.2 Mechanical Similarity Mechanical similarity in a SFPC system can only be achieved if both kinematic and dynamic similarity requirements are satisfied. In other words, kinematic similarity is achieved when corresponding particles trace out geometrically similar patterns in corresponding intervals of time. Times are measured from an arbitrary zero for each system, and correpndng times are defined as times such that t ’/t= t is


229

constant; t is the time scale ratio (primed variables are for the pilotscale plant). According to R. E. Johnstone and M. W. Thring: If two geometrically similar fluid systems are kinematically similar, then the flow patterns are geometrically similar, and heat- or mass-transfer rates in the two systems will bear a simple relation to one another. Kinematic similarity in fluids entails both geometrically similar eddy systems and geometrically similar streamline boundary films. Hence, if L is the linear scale ratio, heat- and mass-transfer coefficients in the prototype will be 1/L times those in the model, from which the total quantities for heat or mass transferred can easily be calculated.[4] Dynamic similarity is achieved when the forces retarding or accelerating the movement of mass within the system are correspondingly similar. These are the forces that define the dynamics of fluid systems in motion. In a supercritical fluids cleaning system, the forces of pressure, inertia, viscosity, and interfacial interactions are important for scaleup when it is desirable to predict either pressure drops or power consumption. For most cleaning systems, however, dynamic similarity is usually only of importance as an indirect means of establishing kinematic similarity.141 4.3 Thermal Similarity

Cleaning systems are thermally similar when corresponding temperature differences can be measured by a constant ratio for kinematically similar systems. Corresponding temperature differences are the differences in temperature as measured at corresponding geometrical points at corresponding times. This requirement needs to be established for a cleaning system in which there is a flow of heat through the system. In a supercritical carbon dioxide system, for example, the solvent stream inight be heated as it enters the cleaning vessel. The flow of heat from the solvent to the vessel walls and, consequently, the flow of heat from the entering solvent to the


bulk solvent are clearly dependent on solvent flow rates through the cleaning chamber. If the model and prototype are kinematically similar, and the vessels are constructed of the same material, the flow of heat through the two systems should be similar as well. This is especially important for a carbon dioxide system where solubilities are highly dependent on solvent density, and solvent density is highly dependent on temperature and pressure. 4.4

Chemical Similarity

Cleaning systems are chemically similar when corresponding concentration differences can be measured by a constant ratio for kinematically similar systems. Corresponding concentration differences are defined in the same manner as corresponding temperature differences. The differences are measured at corresponding geometrical points at corresponding times. That is, the total concentration of either the solvent or solvent/solute mixture is not important but, rather, the differences in constituent concentrations at corresponding locations and intervals.

Eq. (12)

Rate of Formation of the Solute / Solvent Mixture Rate of Bulk Flow

and

Eq. (13)

Rate of Formation of the Solvent / Solute Mixture Rate of Molecular Diffusion

In both model and prototype, the formation time for the solvent/ solute mixture will be of the same rate or reaction order, and this requirement fixes the relative velocities in continuous-flow systems. These velocities are incompatible with the velocities necessary for kinematic similarity except at very low or very high velocities.[4]


231

5.0 SUPERCRITICAL CARBON DIOXIDE SYSTEM SCALEUP The principles discussed above can be used to scale up a cleaning process using any supercritical fluid in a cleaning vessel whether or not chemical reactions are taking place between the solvent and solute. This section describes scaleup experiments conducted by PNL on a supercritical carbon dioxide system. To simplifi the process, this example assumes that there are no additional solvents or cosolvents added to, or used with, the carbon dioxide. Carbon dioxide is typically used as a physical solvent in cleaning operations. Solvation occurs by a clustering of the solvent molecules around individual solute molecules, interfering with the weak forces holding the solute together. Mass transport of the solute into and through the solute is governed by the effects of temperature, pressure, and agitation. 5.1 Experimental Approach

A small SFPC test stand (Fig. 1) was assembled to characterize how system dynamics affect cleaning with supercritical carbon dioxide solvent. Several series of experiments were performed to determine the effects of turbulence, solvent temperature, solvent density, and system pressure. The experimental objective was to establish relationships between solubility of the contaminating species and changes in solvent temperature, solvent density, system agitation, and cleaning efficiency. We know that solubility of the solute(s) can be directly correlated to solvent density, as can diffusivity of the solute into and through the solvent.[7] Typically, solubility increases with density, while diffisivity decreases. Density is strictly a fbnction of temperature and pressure. Similarly, mass transfer increases with reduced solvent viscosity and increasing agitation;18]generally, viscosity increases with temperature for a supercritical fluid.[l2I The rate of solvation is dependent on the concentrations of both the solvent and solute and the density and temperature of the solvent. Concentration is fixed by flow rate, solvent volume, and contaminant loading. Thus,

232 Supercritical Fluid Cleaning the entire process can be described by the following process variables: temperature, pressure, flow rate, and agitation of the solvent; contaminant loading and rate of reaction between the solvents and solute; and volume of the cleaning vessel.

Berty ” Autoclave ”

Pressure Regulator

I

Pressure Regulator

Contaminant Trap

Figure 1. Supercritical fluids parts cleaning test stand shown with recycle.

5.2

Equipment

The autoclave used in the experiment (Fig. 2) has a total internal volume of 433 cc and a total inside diameter of 76 mm; the draft tube has a diameter of 50.6 mm. The autoclave was constructed by Autoclave Engineers (a “Berty” autoclave) and was modified at PNL for these tests. Agitation in the autoclave is supplied by an impeller attached through the bottom of the autoclave to a MagnaDrive assembly. The autoclave rests in a ceramic heater assembly capable of heating the autoclave to temperatures greater than 5OOOC. Rated pressure of the autoclave is 400 bar [5800 psi] at 34OOC [944OF]. Power to the heaters is supplied through a Variac, which can be manually adjusted during testing to control temperature. Impeller speed is controlled by a variable speed controller that is part of the MagnaDrive unit. The necessary modifications to the autoclave


233

consisted of changing the MagnaDrive bushings fiom carbon to nylon, performing a complete polishing of all the interior surfaces of the autoclave, and fashioning a coupon holder to position the coupon in the autoclave in the same orientation for every test run.

Coupon Draft Tube Baffles Packing

sT urbine

Figure 2. Autoclave used in scaleup experiments, showing an internal view with the flow profile.

Test contaminants were selected based on two criteria. First, after being cleaned in the SFPC test stand, the contaminants must leave a measurable residue under some of the cleaning conditions within the range of the experiments. Comparison of different system conditionsin terms of the degree of contaminant removal is necessary to establish optimum process conditions. Second, the contaminants must be approximately as difficult to remove as actual contaminants. Two candidates fitting the criteria were selected: Dow Corning@200 Fluid and Castle/Sybron X-448 High-temperature Oil. The Dow Corning Fluid is a poly(dimethy1)siloxane fraction with a viscosity of 100 cSt and an average molecular weight of about 5000. The X-448 Oil is a polybutene/mineral oil mixture. A Mettler AE200 analytical balance was used to perform gravimetric analyses for the various experiments. Reproducibility for this balance is specified at *O. 1 mg.

234 Superentical Fluid Cleaning 5.3 Coupon Contamination Procedures

For each series of experiments, the same stainless steel coupon (4 cm x 3.5 cm x 0.1 cm) was precleaned, coated with the test contaminant, and then placed in the autoclave for testing. For precleaning, the coupon was either washed first in a chloroform bath, rinsed with acetone, and then dipped in an acetone bath or immersed in a 1, 1, 1-trichloroethanebath and wiped with a tissue to remove any residue. M e r precleaning, the coupon was weighed to determine when the mass of the coupon was within 0.1 mg of its original weight. This was set as the standard for perfect cleanness of the coupon. For the first two series of experiments, the coupon was submerged completely in one of the two test contaminants, patted lightly with a lint-fiee tissue to remove excess material, and weighed. Loading was controlled to 0.005 g from run to run to minimize effects of surface concentration differences. In later series, dopants were applied to one side of the coupon which was marked to designate an area so that the contaminated area would remain constant from run to run. The contaminants were applied to this area with a small, shortbristled paint brush. Contaminant mass for each of these series was kept constant from run to run to within f 0.0001 g. 5.4 Coupon Cleaning with Gravimetric Analysis

M e r the contaminant was applied, the coupon was loaded into the preheated autoclave by placing it onto a holder that kept it in a vertical position. The top of the autoclave was then bolted on, and the test started. The autoclave vessel and system were filled with carbon dioxide to the first back-pressure regulator, and residual air was vented from the system through a valve at the top of the autoclave. Once the system was filled, the pump was started at an average feed rate of 57 d m i n of liquid carbon dioxide. For testing at supercritical conditions, timing began when the pressure gauge at the top of the autoclave reached 1100 psig. When the autoclave pressure reached 1100 psig, the impeller was turned on to a predetermined speed. The autoclave pressure was


235

allowed to reach and maintain operating pressure for the remainder of the test run. Approximately 15 sec before the end of the run, the carbon dioxide supply was turned off, and the system vented. When the system pressure dropped to 1100 psig, the impeller was turned off. Pressure letdown also typically took 15 sec. One volume turnover for the autoclave (internal volume = 235 ml with baffles and basket in place) is estimated at 1.3 min, or approximately 3.85volume turnovers for a 5-minute run. 5.5

Results

Figure 3 shows the relationship between the average mass flux of contaminant and system agitation. For equal cleaning intervals, contaminant removal rates increase as the impeller speed increases up to a maximum efficiency, which depends on the contaminant. These data show that higher rates of agitation are needed for effective removal of less soluble species, and that a maximum efficiency can be reached for the more soluble species (silicone oil) at a lower impeller speed than for a less soluble species (high-temperature oil). Increasing system turbulence beyond this point does not improve mass flux, and would be a waste of power. Each data point in Figure 3 represents results based on constant temperature, pressure, dopant loading, and length of cleaning cycle. Although tests were performed showing the relationship between coupon placement and orientation, these data represent results from experiments performed using the same coupon for each cleaning run placed in the autoclave in the same orientation from run to run. Coupon placement and orientation had an insignificant effect on cleaning performance throughout the agitation range. Weight differentials were used to determine mass flux, measured as cleaning efficiency, of dopant from the coupon. Similar experiments were performed at various pressures and temperatures to show the relationships between changes in density at constant pressure or temperature and the effects on cleaning efficiency at various rates of agitation. Figure 4 is a typical plot showing that comparison. As shown in Fig. 4, cleaning was more agitationdependent (Reynoldsnumber) than density-dependent (solubility)in the regions of low flow. Notice the solubility-cleaning-efficiency


plane at Re > 1500. The change in solubility, which is based on solutdsolvent density to the 1.13 power,['] has almost no effect on the cleaning efficiency. 1.13

6 2 =6,(;)

where

ai

=

pi

=

solubility of species i density of species i

This indicates that the scaleup process should focus more on flow factors (impeller design, internal reactor geometry, baffling, etc.) than on temperature and pressure in the system.

Impeller RPM vs % of Contaminant Removed

""i..'

I

70

8 I I

0

I

m

Silicone Oil

I

I

I

I

I

500

1000

1500

2000

2500

Impeller RPM Figure 3. Plot of test data: average contaminant removal for 5-minute runs vs. impeller speed.


237

Silicone Oil Removal 63 1200 psi, Varying Temperature

Cleaning Efficiency 1.0 =fill scale

Figure 4. Removal of silicone oil (polydimethylsiloxane)from a stainless steel coupon at 1200 psi.

5.6

Discussion of PNL Scaleup Method

The test data show that the conditions required to completely clean the coupon of silicone oil are a pressure of 105 bar at a temperature of 45OC with an impeller speed of 1000 rpm. To scale up this process for use with actual parts contaminated with a similar material, we must first decide which criteria to include for achieving similitude. Since temperature and pressure are intrinsic values that do not rely on any other aspect of the prototype design, except that it must be able to hold the desired pressure at the desired temperature, the mechanisms governing mass transfer and agitation become paramount. The movement of a contaminant near the surface of a part to the bulk fluid is governed by mass transport mechanisms. The molar flux of a species A from a surface may be expressed in terms of a composition driving force and a mass transfer coefficient:


where

NA = molar flux of species A from the surface (moVs/cm2)

kx = local mass transfer coefficient (moVs/cm2) xAo = mole fraction ofA near the surface, usually taken to be the equilibrium concentration xAoo= mole fraction ofA in the approaching bulk fluid. Mass transfer coefficients are typically correlated using a dimensionless quantity known as the Nusselt number for mass transfer shown below:

where

Nu,

Nusselt number for mass transfer, dimensionless k, = local mass transfer coefficient (moVs/cm2) x = characteristic dimension of the system (cm) c = total concentration of chemical species (mol/ cm3) DAB= binary diffisivity of species A and B (cm2/s). =

For flow in a pipe, the pipe diameter is conventionally used as the characteristic length. For flow around a submerged sphere, the sphere diameter is used as the characteristic length. For flow past a


239

flat plate, the length along the plate is used, and the Nusselt number changes along the length of the plate. In general, the Nusselt number is a fbnction of the Reynolds system number (Re = vxp/p), the Schmidt number (Sc = p/pDAB), geometry, and the magnitude of the mass flux. For the case of smaller mass-transfer rates, the Nusselt number depends less on the magnitude of mass flux. For small mass-transfer rates, heat and mass transfer become analogous;in this case similar forms of heat and mass transfer correlations can be used to describe the same geometry:

j,,,, z j,,,,

z

1

~ h= ,a function of Re

where JD = Chilton-Colburn j-factor for mass transfer, JH =

f

=

dimensionless Chilton-Colburn j-factor for heat transfer, dimensionless friction factor, dimensionless

The Chilton-Colburn factors can be expressed as follows: L

2

where


v

= fluid bulk velocity, c d s

p

=

fluid bulk viscosity, g / c d s

p

=

fluid bulk density, g/cm3

bloc = Cp

=

k

=

local heat transfer coefficient, caV(s cm2 K) fluid bulk heat capacity, caVg/K fluid bulk thermal conductivity, caV(s cm K).

For turbulent flow (2100

supercritical fluid cleaning

Supercritical

Supercritical Fluid Technology for Drug Product Development

Supercritical

Supercritical Fluid Methods and Protocols (Methods in Biotechnology)

Supercritical Fluid Extraction of Nutraceuticals and Bioactive Compounds

Metal Complex Catalysts Supercritical Fluid Polymerization Supramolecular Architecture

Hyphenated Techniques in Supercritical Fluid Chromatography and Extraction

Supercritical Fluid Technology in Materials Science and Engineering

Spring Cleaning

Cleaning Up

Spring Cleaning

Chemical Synthesis Using Supercritical Fluids

Advanced cleaning product formulations

Laser Cleaning II

Advanced Cleaning Product v5

Green Cleaning For Dummies

Fluid-fluid interactions

Fluid

Solubility in Supercritical Carbon Dioxide

Advanced cleaning product formulations

Blast Cleaning Technology

Advanced cleaning product formulations

Blast Cleaning Technology

Advanced cleaning product formulations

Chemical Synthesis Using Supercritical Fluids

The Cleaning Lady

Surface Contamination and Cleaning

Supercritical Fluid Methods and Protocols (Methods in Biotechnology) (Methods in Biotechnology)

Cleaning Windows XP For Dummies

ICS Cleaning Specialist June 2011

supercritical fluid cleaning

Supercritical

Supercritical Fluid Technology for Drug Product Development

Supercritical

Supercritical Fluid Methods and Protocols (Methods in Biotechnology)

Supercritical Fluid Extraction of Nutraceuticals and Bioactive Compounds

Metal Complex Catalysts Supercritical Fluid Polymerization Supramolecular Architecture

Hyphenated Techniques in Supercritical Fluid Chromatography and Extraction

Supercritical Fluid Technology in Materials Science and Engineering

Spring Cleaning

Cleaning Up

Spring Cleaning

Chemical Synthesis Using Supercritical Fluids

Advanced cleaning product formulations

Laser Cleaning II

Advanced Cleaning Product v5

Green Cleaning For Dummies

Fluid-fluid interactions

Fluid

Solubility in Supercritical Carbon Dioxide

Advanced cleaning product formulations

Blast Cleaning Technology

Advanced cleaning product formulations

Blast Cleaning Technology

Advanced cleaning product formulations

Chemical Synthesis Using Supercritical Fluids

The Cleaning Lady

Surface Contamination and Cleaning

Supercritical Fluid Methods and Protocols (Methods in Biotechnology) (Methods in Biotechnology)

Cleaning Windows XP For Dummies

ICS Cleaning Specialist June 2011

Recommend Documents