Preface This volume is the first of two containing most of the papers presented at the Second International Conference ...
107 downloads
1311 Views
162MB 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
Preface This volume is the first of two containing most of the papers presented at the Second International Conference on Environment-Induced Cracking of Metals (EICM-2), which was held at The Banff Centre, Banff, Alberta, Canada in September 2004.' This conference was the fifth in a series of major international meetings sponsored or cosponsored by NACE in the area of environment-induced cracking. The previous four conferences, recognized as landmark events, were the: 9 International Conference on Fundamental Aspects of Stress Corrosion Cracking held at the Ohio State University, Columbus, OH, USA, 1967 9 International Corrosion Fatigue Conference held at the University of Connecticut, Storrs, CT, USA, 197 I 9 International Conference on Stress Corrosion Cracking and Hydrogen Embrittlement of iron Base Alloys held in Unieux-Firminy, France, 1973 9 First International Conference on Environment-Induced Cracking of Metals held in Kohler, Wi, USA, 1988 EICM-2 was jointly sponsored by the Committee TEG 186X on Environmentally Assisted Cracking of NACE International, the Working Party WP5 on Environment Sensitive Fracture of the European Federation of Corrosion and the Joint ASM/TMS Corrosion and Environmental Effects Committee in association with the Technical Committee of Hydrogen Embrittlement and Stress Corrosion of the Chinese Society of Corrosion and Protection and the Technical Committee TCI0 on Environmentally Assisted Cracking of the European Structural Integrity Society. This conference was international in scope and attracted a significant number of researchers and professionals. There were 115 participants from countries that included Australia, Japan, China, India, Saudi Arabia, Israel, Russia, Ukraine, Poland, Slovak Republic, Austria, Slovenia, Italy, Sweden, Germany, Belgium, France, the United Kingdom, Brazil, the United States and Canada. The conference provided a forum for dialogue, assessment and critique among researchers committed to different methodologies and actively promoted discussion and cross-fertilization of ideas among established and emerging researchers working in different areas related to the problem of environment-induced cracking in materials. The technical program consisted of three plenary sessions, 12 technical sessions with 72 oral presentations, a poster session and a luncheon talk. in total, 87 papers that described both fundamental research studies and more practical engineering applications were presented. Topics included but were not limited to: SCC, hydrogen embrittlement, corrosion fatigue, liquid metal embrittlement, localized environmentinduced attack, modeling environmental effects, crack growth mechanisms, hydrogen permeation and transport, hydrogen-plasticity interaction, test methods and interpretation of test data, materials degradation in service, failure analysis, life prediction of corrodible structures, and the history of EICM research. Together the papers provided a comprehensive account of advances in research on environmentinduced cracking and constituted the most recent fundamental and industrial survey of ' Volume 2 is published under the title "Environment-induced Cracking of Materials: Prediction, Industrial Developments and Evaluation," S.A. Shipilov, R.H. Jones, J.M. Olive, R.B. Rebak (Editors), Elsevier, Oxlbrd, 2007. 508 pages.
x
Preface
the subject. Plenary lectures were presented by Richard Gangloff, Digby Macdonald, Roger Newman, Roger Staehle, Alan Turnbull and Robert Wei. A luncheon lecture entitled "Growing Old Gracefully? A Perspective on Pipeline Safety" was given by Alan Murray of the National Energy Board Canada. A highlight of the conference was the General Discussion which featured a panel comprising the plenary speakers with John Scully as a moderator. During the discussion, recent developments and a number of key questions in environment-induced cracking research were considered, including how to move from theory to validation to practice, needs and opportunities for mechanistic advances, progress towards life prediction/prognosis/damage evolution, the challenge to mitigate and control EICM, and needs and opportunities for technological and practical progress. As the potential audience for the proceedings represents a wide spectrum of professionals, including researchers, engineers, practitioners and consultants with different background and specific research and/or industrial interests, it was found feasible to divide the papers into two volumes, each with a specific focus- as indicated in their titles - which can then be sold and/or used together or separately. This necessitated a different arrangement of the papers in the books than the order in which they were presented at the conference, in each volume, the papers are organized to provide a reasonable reflection of research development in the subject areas. Either volume may be used as an independent reference source that reviews the current state of the fundamental and/or industry-oriented research and provides a comprehensive introduction to the field. Applied researchers and specialists in industry may find volume two highly relevant, especially if they are engaged in the areas related to nuclear power generation, oil and gas transportation, the disposal of radioactive waste, aircraft maintenance, chemical and marine applications, failure analysis in industrial equipment, and assessing the SCC resistance; while volume one may find a ready audience among researchers who focus upon investigating the phenomenological aspects of SCC, hydrogen embrittlement, liquid metal embrittlement and corrosion fatigue, including environmental, microstructurai, electrochemical and mechanistic aspects, in high performance steels, nonferrous alloys, ceramics and glasses. Whatever the individual reader's preference, there are synergies and connections across the volumes which make their simultaneous publication an important occasion for researchers and practitioners alike. Of the 87 papers that were presented at the conference, 81 are offered to the reader across the two volumes. The papers were rigorously peer-reviewed, revised and edited extensively to meet the high standards for scientific publications intended for an international and discerning audience. Volume one contains 43 papers divided'into nine sections. These papers provide thorough coverage of the environmental, microstructural, electrochemical and mechanistic aspects of SCC, hydrogen embrittlement, liquid metal embrittlement and corrosion fatigue for a wide range of materials. Section I titled "Modeling environmental attack" contains three plenary papers: "Science based probability modeling and life cycle engineering and management" by R.P. Wei and D.G. Harlow, "A model to predict the evolution of pitting corrosion and the pit-to-crack transition incorporating statistically distributed input parameters" by A. Turnbull et al., and "Revisiting the film-induced cleavage model of SCC" by R.C. Newman et ai. which review the achievements of the science of EICM during the last decade or so and describe some of the models that simulate damage evaluation and prediction. The remaining seven papers of this section deal with new theoretical and experimental
Preface
xi
approaches to modeling localized environmental attack that allow researchers to investigate and predict environment-induced cracking in metals. Section 2 titled "Crack growth mechanisms" begins with a plenary paper titled "Critical issues in hydrogen assisted cracking of structural alloys" by R.P. Gangloff, followed by four papers addressing the mechanisms of environment-induced cracking of Ai, Fe, Mg, Ni, Ni-Cr and Ti alloys and amorphous Fes0B~Si9 exposed to aqueous solutions, hydrogen gas or liquid metals. Section 3 titled "Hydrogen permeation and transport" contains six papers which deal with the fundamental aspects of the strain-induced permeation and transport of hydrogen and hydrogen interactions with the microstructure of steels and alloys. The four papers in Section 4 titled "Hydrogen-assisted cracking and embrittlement" address hydrogen effects on the properties and fracture of ferrous alloys. New insights into the issues associated with SCC and corrosion fatigue of AI, Cu, Mg, Ni and Ti alloys, HSLA and stainless steels are provided in twelve papers, which are included in Sections 5 and 6 titled "Nonferrous alloys" and "Iron and nickel based alloys" respectively. Section 7, titled "Ceramics and glasses, '' considers some of the aspects that control corrosion and environment-induced cracking in high performance ceramics and glasses. Recent results on the influence of liquid metals on the mechanical behaviour and fracture of steels and superplastic alloys are presented in Section 8 titled "Liquid metal embrittlement." The final section, titled "History of SCC research," reviews experimental results and mechanistic models for SCC and corrosion fatigue in the century after the first investigation on record was carried out in 1873. As an indication of the scholarly and scientific rigour that characterise these papers, the total number of literature references to the papers in this volume is I, 107. Many individuals and organizations contributed to the success of EICM-2 and these volumes, and ! would like to acknowledge their assistance. The conference and its proceedings would not have been possible without the generous support of my coeditors, Russ Jones, Jean-Marc Olive and Rafil Rebak. These dedicated individuals contributed greatly of their time and skills to arrange various aspects of the meeting. They were instrumental in helping to develop the program, with organizing sessions and assisting with reviewing the manuscripts. ! also wish to thank Winston Revie for his assistance at the early stage of the planning of the conference. Special acknowledgments are extended to chairs of sponsoring committees for helping to organize international participation: Jorge Perdomo, Jean-Marc Olive, RaOl Rebak, Lijie Qiao and Wolfgang Dietzel. Pierre Crevolin, President of NACE in 2003/2004 and Interim Executive Director in 2004, his staff and especially Gretchen Jacobson, Publishing Director, provided invaluable assistance in publicizing EICM-2. Special thanks go to my students, Don Boll, Vicky Carathanassis, Christie Millington, Enoch Ng, Krishna Panchalingam, Feng Wang and Hong Wang, who helped with registration and assisted me through all stages of the conference, including the preparation of several graphs for this publication. Sincere thanks also to Kim Allan for providing technical support for the conference website, i would like to extend my special thanks to the session chairs, Jenny Been, Anne-Marie Brass, Steve Bruemmer, Jacques Chine, Wolfgang Dietzel, Noam Eliaz, Russ Jones, Rob Kelly, Fraser King, Amar Kumar, Graham Lobley, Stan Lynch, Jean-Marc Olive, Alan Plumtree, Rafil Rebak, John Scully, Raman Singh, Brian Somerday, Bob Sutherby, Mirna Urquidi-Macdonald and Marc Vankeerberghen, who worked very effectively to keep the program organized and the floor discussions effective. I greatly appreciate the work of the technical experts who joined in reviewing the manuscripts for these volumes, and who approached the
xii
Preface
manuscript reviews with the detail and exacting standards that mark the review process at technical journals. The names of these reviewers are listed on the next page. Without the generosity of our financial sponsors, and the individuals who secured sponsorship funds, the conference would not have happened and these papers would not have seen the light of day. Hence, my heartfelt thanks to NACE International (Pierre Crevolin), National Energy Board Canada (Alan Murray), Metallurgical Consulting Services Ltd. (lain Le May), the University of Calgary (Ron Bond), Canadian Energy Pipeline Association (Jake Abes), NOVA Chemicals Corporation (Fraser King), ASM Canada Council (Steve Yue), ASM Calgary Chapter (Sammy Tang), and Broadsword Corrosion Engineering Ltd. (Pat Teevens). During the preparation of the proceedings, ! was very ably assisted by Michael Aleksiuk who provided diligent and insightful proofreading of several of the manuscripts, and by Lucy Dickinson, Nicola Jones and Kristi Green of Elsevier Ltd. for their kind co-operation at all stages of the work. My sincere thanks go to the authors of the papers for making time in their busy lives to put their work in the public domain. To all those who have helped, ! express my sincere "Thanks!" Last, but certainly not least, 1 would like to acknowledge my thanks and appreciation to my wife, Hyacinth, whose support, encouragement and companionship sustained me through both the organizing of EICM-2 and the preparation of this publication.
Sergei Shipilov Editor and Conference Chair
xiii
List of Reviewers The quality of the papers that appear in this volume reflects not only the obvious efforts of the authors but also the unheralded, though essential, work of the reviewers who joined with the editors in reviewing the manuscripts for the proceedings. The editors sincerely acknowledge the following reviewers who provided comments and constructive suggestions for the revision of manuscripts. T.M. Ahmed, University of British Columbia, Vancouver, Canada P.L. Andresen, General Electric Company, Schenectady, USA K. Arioka, Institute of Nuclear Safety Systems, Fukui, Japan I. Aubert, Universit6 Bordeaux l, Talence Cedex, France J. Been, NOVA Chemicals Corporation, Calgary, Canada G.A. Cragnolino, Southwest Research Institute, San Antonio, USA J. Congleton, University of Newcastle upon Tyne, UK R. Eadie, University of Alberta, Edmonton, Canada F.P. Ford, General Electric Company, Schenectady, USA G.S. Frankel, Ohio State University, Columbus, USA H.E. H~tnninen, Helsinki University of Technology, Helsinki, Finland R.H. Jones, Exponent Failure Analysis Associates, Bellevue, USA J.-M. Olive, Universit6 Bordeaux 1, Talence Cedex, France J.J. Perdomo, Smurfit-Stone Container Corporation, Carol Stream, USA M. Puiggali, Universit6 Bordeaux I, Talence Cedex, France R.B. Rebak, Lawrence Livermore National Laboratory, Livermore, USA R.E. Ricker, National Institute of Standards and Technology, Gaithersburg, USA P.R. Roberge, Royal Military College of Canada, Kingston, Canada H.-P. Seifert, Paul Scherrer Institute, Villigen PSI, Switzerland S.A. Shipilov, Metallurgical Consulting Services, Calgary, Canada B.P. Somerday, Sandia National Laboratories, Livermore, USA M. Touzet, Universit6 Bordeaux I, Talence Cedex, France W.-T. Tsai, National Cheng Kung University, Tainan, Taiwan, China S. W~stberg, Det Norske Veritas (DNV), Havik, Norway W. Zheng, CANMET Materials Technology Laboratory, Ottawa, Canada
"Science based probability modeling and life cycle engineering and management R.P. Wei, D.G. Harlow Dept. of Mechanical Engineering and Mechanics, Lehigh University, Bethlehem, PA 18015, USA Abstract
Life cycle engineering and management (LCEM) of modem, high value-added engineered systems demands the formulation and use of science based probability models (SBPM) for predicting the evolution and distribution of damage, and their impact on structural reliability. In this paper, the need for SBPM is discussed within the context of LCEM. The approach and its efficacy are illustrated and demonstrated through selected examples on the formulation and use of these models. They include modeling of corrosion and corrosion fatigue of aluminum alloys, and its application to aging aircraft, and the impact of residual stress/environment on fatigue (S-N) response into the very high cycle (up to 10~~cycles) domain. I. Introduction
Material aging, through the evolution and distribution of damage (for example, by localized corrosion and corrosion fatigue), is one of the principal causes for the reduction in the reliability and margin of safety of engineered systems. It can contribute significantly to the cost of maintenance and operation and, thereby, their overall life cycle cost. To quantify materials aging and facilitate the overall optimization of the performance, reliability, and life cycle costs of these systems (i.e., for life cycle engineering and management, LCEM) new modeling approaches are "needed. Traditional (and current) approaches to engineering design are no longer adequate. The reason lies in the fact that these approaches are based largely on the use of experientially based statistical methodologies and accelerated testing over periods that are well short of those of the intend service. The models developed from them are essentially parametric representation of statistical fits to the experimental data, and are effective only over the range of the underlying data. They capture, at best, the influences of the limited number of controlled (external) variables used in testing. Furthermore, variability associated w i t h measurement errors (which cannot be separated from the experimental data) are incorporated into the statistical analyses, and can lead to overestimations of the uncertainty bounds. As such, simple application of known statistical techniques cannot provide the necessary tools for LCEM of
4
Modeling Environmental Attack
engineered systems, and a different approach needs to be adopted. In this paper, a science (mechanistically) based probability modeling approach that has been used successfully over the past decade [ 1-7] is presented to illustrate the modeling process and its efficacy. The overall framework and approach are described. Its use and efficacy are illustrated through two examples: the first, on modeling of pitting corrosion and fatigue crack growth in aluminum alloys and its application to aging aircraft, and the second, in considering the fatigue (S-N) response of a bearing steel into the very high cycle domain (i.e., up to 10 l~ cycles). 2. Framework
Materials aging is considered in the context of its influence on the assessments of reliability, safety, availability and maintenance of engineered systems. The framework for these assessments is depicted in Fig. 1. Within it, the materials aging process is reflected specifically in the evolution and distribution of damage that compromise functionality, reliability and safety. The key issues, therefore, pertain to the assessment of such a system under given sets of projected operating conditions (i.e., in terms of forcing functions and environmental conditions) in relation to its current state or its initial state (either new, or after major maintenance service) and its future state. Such assessments are typically made through the use of a set of analysis tools, in conjunction with a comprehensive suite of diagnostic or nondestructive evaluation (NDE) tools that provide information on the current state (sizes and distribution) of damage in the system. Assurance of reliability and continued safety, and availability, requires a quantitative assessment of the system in its 'projected future state '. For this assessment, appropriate quantitative models are needed for estimating the accumulation of damage (in size and distribution) over its projected period of operation. The outcome of this assessment then serves as the basis for decisions on its suitability for continued service as reflected in Fig. 1 by the labels Reliable, Conditioned Reliability, and Not Reliable. A system judged to be reliable would be accepted for unrestricted operation until the
Based on a damage function D(x~, yj, t), that is a function of the key intemal (x.~ and extemal (y~ variables Fig. 1. A simplified flow diagram for life prediction, reliability assessment and management of engineered systems [6].
Volume 1: Chemistry, Mechanics and Mechanisms
5
next scheduled maintenance, the one with conditioned reliability would be subjected to operational constraints, and the one deemed to be unreliable would be sent for overhaul or be retired. The process labeled as Probabilistic Estimation of Damage Accumulation in Fig. 1 is the key element of this process. The confidence that can be placed on this assessment depends importantly on the robustness of the underlying models for damage evolution and distribution within a component or system. It requires the development of methods that are predictive and that can provide accurate estimates of the evolution and probabilistic distribution in damage over time that can be used for reliability and safety assessments and service life prediction.
3. Science based probability approach
3.1. Methodology The requisite methodology must provide the following capabilities: (i) projection beyond typical underlying data, (ii) analyses for critical variable response, (iii) investigation into the reliability and availability of components and systems, and (iv) life cycle engineering and management of systems. Science (mechanistically) based probability modeling, vis-?t-vis experientially based statistical modeling, provides the structure to meet this need. A comparative assessment of these two approaches is given in [7]. The essence of science based probability modeling of damage evolution and distribution is the formulation of a time dependent damage function D(x,,yi,t) that captures its functional dependence on all of the key internal (x,, e.g., materials) and external (y,, e.g., loading) variables, and their variability. As such, this damage function D(x,,y~,t) accounts for its mechanistic and statistical dependence on the key random variables. It is, thereby, the foundation for time-dependent probability analyses for estimating the distribution of damage, or the distribution in service lives, that are essential for system design and management. The development of D(x~,y,,t) is based upon scientific understanding and modeling of the mechanisms of damage nucleation and growth. The essential process for model development is shown schematically in Fig. 2, and is iterative. It involves the identification and confirmation of a set of key external and internal variables, and the formulation of an appropriate mechanistic (deterministic) model for D(x~,y,,t) that express its functional dependence on these variables. The next step is to determine the probability distribution for each of the key variables in terms of either the probability density function (pdjO, or the cumulative distribution function (cdjO. From these functions, say the pdfs, a joint probability density function (j]odj9 is constructed. The jpdf is then integrated with the mechanistic model to yield a science based probability (stochastic) model. In practice, however, the stochastic results are to be derived through simulation; e.g., through the use of Monte Carlo methods. The experientially based statistical methods, on the other hand, bypass the identification and quantification of the role of internal variables, and model development is by and large limited to establishment of empirical fitting functions to the experimental data.
3.2. Comparison of approaches The philosophical and practical differences between the two approaches to modeling are given by Harlow and Wei [7], by using a tensile ligament instability model for creep
6
Modeling Environmental Attack
pay. External Variables
pay. Internal Variables
I
Design of Experiments (Statistical Modeling)
i
Physics Modeling
jpay. Extemal & Internal
Variables
(Deterministic)]
& Mechanistic I
I
Probability & Stochastic
Models
Fig. 2. Simplified flow diagram for the development of mechanistically based probability models.
controlled crack growth (see Ref. [7]) and by statistical least squares fit to the experimental data. The mechanistic model is based on the recognition that crack growth is governed by the "tensile instability" (or necking failure) of ligaments in the crack-tip process zone ahead of the crack tip. These ligaments are identified with the regions of material isolated by the growth of voids nucleated at nonmetallic inclusions in high strength steels (see Ref. [7]). In this model, the steady-state creep crack growth rate (da/dt)s= is related to the steady-state creep rate in the tensile ligament within the process zone and the crack growth life through the Hart-Li model for creep (see Ref. [7]). The statistical model, on the other hand, is a simple two-parameter exponential equation that fits the data in semi-logarithmic space. The comparison is shown in Fig. 3. Note that, because the data were obtained from a very small sample of material, most of the uncertainty in the statistical model reflected errors in crack length measurements, rather than "true" variability in material properties. Note also the significant difference between the two models in the low growth rate region, which is of paramount importance for life prediction. The difference in approach is self evident. In the mechanistically based model, key internal and external variables are identified. Their variabilities are readily incoi'porated into the model to assess the overall variability in response. The contribution of each of the random variables on the variability in response may be readily assessed. Given the explicit functional dependence, when duly validated, it can be used to predict response beyond the range of the experimental data. The experientially based statistical model, on the other hand, represents a statistical fit to the data in which the key internal variables could not be identified. As such, it is incapable of capturing the functional dependence on these variables, and its usefulness is limited to the range of the experimental data. As experimental (including measurement) errors are lumped into estimates of the fitting parameters and their variability, the quality the subsequent reliability analyses may be overly conservative, or uncertain. A more detailed discussion of these approaches may be found in Ref. [7].
Volume 1: Chemistry, Mechanics and Mechanisms 10-6
Predictions 95% confidence bounds
10-7 10-8 10-9
7
-
statistical model " (least squares) i
~ ~ ~
,,~// ~7~f- 1
J /
~ ~
1
j=ll==Jl~f
~
=n 10-10 E v 10_11 m 10-12 -o 10-13
J~--mechanistically based i~ w probability model
10-14
AISI 4340 Steel in dehumidified Argon at 297K (data from Landes and Wei)
10-15 10-16
=
20
=
=
I
40
i
=
=
I
=
=
=
60
i
80
=
=
=
I
100
,
,
,
120
K (MPa-mlt2)
Fig. 3. Comparison between mechanistically based probability and statistically based models for crack growth kinetics [7].
It is worth noting that there is a crucial difference in the role of experimentation between the science based probability and the experientially based statistical approaches. For the science based probability approach, experimentation is one of discovery and hypothesis testing to guide model formulation. For the experientially based statistical approach, on the other hand, the goal is to establish the best parametric fit to the experimental data in terms of a limited set of identifiable external variables. In the first case, variability arises naturally out of the randomness in the key internal and external variables, whereas the other simply captures the scatter in experimental data. In the following sections, modeling of pitting corrosion and corrosion fatigue of aluminum alloys is used to illustrate the process, and to demonstrate the efficacy and utility of the approach for estimating the evolution and distribution of damage for LCEM of engineered systems. The applicability of this approach in understanding the dichotomy between S-N and fracture mechanics approaches to corrosion fatigue is discussed. The use of this approach to understand S-N response of a high strength bearing steel in the very high cycle regime (up to 10 l~ cycles) is discussed. 4. Corrosion and corrosion fatigue in aluminum alloys, and applications 4.1. Particle induced pitting in an aluminum ahoy
A simplified model for pit growth was first proposed by Harlow and Wei [2] and was used successfully to account for damage evolution in airframe aluminum alloys. For simplicity, the model assumed the pit to be hemispherical in shape, with radius a, and its growth (driven by an external constant-current source) would be at a constant volumetric rate, obeying Faraday's law. Specifically, the pit volume (V) is represented by one-half of a sphere, with V = (2/3)~ra3. The rate of pit growth, the time evolution of pit size and the time required to reach a given pit size are as follows"
8
Modeling Environmental Attack
da dt a =
da dV dV all
1 2~a 2
dV dt
Mlp 1 2~rnpF a 2
3MI, 3 2rcnpF t + ao
(1)
2~rnpF 3 3Mlp ( a 3 - a ~
t ~ _ _ _ ~
In Eq. (1), M is the molecular weight; Ip is the pitting current; n is the valency; p is the density; F is Faraday's constant (9.65 x 107 C/kg-mol), and ao is the initial pit size, or the size of the initiating particle or particle cluster. For aluminum, M = 27 kg/kg-mol; n = 3; and p = 2.7 x 103 kg/m 3. For particle induced pitting, the pitting current is determined by the cathodic current density that can be supported by the particle (or cluster of particles) and its surface area. Based on studies of pitting corrosion in 2024-T3 aluminum alloy (see Fig. 4, for example), it is recognized that pitting resulted naturally from dissolution of the aluminum matrix through its galvanic coupling to the constituent particles [8]. Based on this recognition, a simple, science based model was proposed [9]. The model envisioned that a pit would be nucleated at a surface particle, in a 'contiguous cluster' of constituent particles, by galvanic corrosion of the matrix. Its continued growth would be sustained by galvanic current from other particles in the cluster that are progressively exposed at the surface of the growing pit [ 10-12].
Side View
Fig. 4. SEM micrographs of (a) a particle induced corrosion pit, and (b) the epoxy replica of a severe corrosion pit in plan (bottom) and side (elevation) view relative to the original pit in a 2024-T3 aluminum alloy sheet [9]. For modeling, the particles are again approximated by spheres of different radii. The rate of pit growth around the surface particle of radius ao (regime 1), and the time evolution of pit size are identical to those given in Eq. (1). The pitting current Ip, however, is explicitly taken to be the product of the limiting cathodic current density i~o that can be supported by the particle and the surface area of the particle Apo, i.e.,
Volume 1: Chemistry, Mechanics and Mechanisms
9
Ip =icod ~ = t~o " (2x G2). The particle area is taken to be one-half of the surface area of a
sphere to approximately account for the increase in exposed surface as the pit grows. The extent of this initial stage of growth depends on the point at which pitting separates the particle from the alloy matrix and the time when sufficient subsurface particles are exposed to sustain continued pit growth. This "transition size" is taken as ate, and was estimated to be about 3 times ao. The initial stage of growth, therefore, is explicitly given (for ao < a < a,,) in terms of Eq. (1) as follows:
da dt a-
MIp 1 2xnpF a 2
Mico (2~a 2) 1 2zrnpF a 2
I3Mtcoao 9 2 31113 npF
t + a~
Mi~oa2o 1 npF a 2 ,; a o _
and
(do/dtLock >_(da/dr),,,,
(7)
The time-based fatigue crack growth rate is simplyflda/dN), where f i s the frequency of cyclic loading.
12
Modeling Environmental Attack
Experimental data suggest that, in practice, transition from pitting to fatigue crack growth is determined by the second criterion in Eq. (7). From Eqs. (4) and (7), the transition crack size atr may be determined by solving the following equality; viz." m
(8)
npF
The number of fatigue cycle associated with pit growth and for fatigue crack growth, and the overall fatigue life in a smooth specimen are as follows: npF 1 (atr N p,, = tip, = f Mi,.o~ p ~ (2xa_2)
-ao)
]"
2 (nc - 2)AK,h Njr ~ (no _ 2)CF,B2Ao.2 (flAera]i z _AK,h)(,, _2) 1+ (n~ -l)(flAcra,'/. z - AK,h ) '
nc > 2
N v ~ Np~, + N/~g
(9) Representing each of the internal random variables by a Weibull distribution, and using reasonable estimates for these values (see Table 1), the fatigue life sensitivity to each of the variables was determined through Monte Carlo simulation, and is shown in Fig. 6. Their collective impact on the distributions in fatigue lives at various stress levels is shown in Fig. 7. Without corrosion, the corresponding fatigue lives would have been up to three orders of magnitude longer, depending on the applied stress Acr, see Eq. (8). This example illustrates the importance of a mechanistically based probability approach in identifying the key random variables, and in assessing their influences on service life, and structural integrity and reliability. Table 1 Key random parameters and the associated Weibull cdf parameters Random variable a fl y Initial pit radius, ao (lam) 1.29 11.78 5.7 Pitting current density, I (A/m2) 2.6 0.56 0.5 Coefficient, CF(m/cyc)/(MPa~lm) T M 15.0 9.9E-12 3.0E-I1 Threshold driving force, AKth (MPa,/m) 2.1 0.34 0.2
/t 16.6 1.0 3.95E-11 0.5
cv (%) 78 41 8 50
4.3. S-N versus F M approaches to corrosion fatigue, and resolution o f a dichotomy
From the foregoing analyses, it is clear that S-N response is significantly affected by pitting, which serves, principally, to truncate the early stage of fatigue crack growth and shorten fatigue life. In other words, conventional corrosion fatigue response reflects the foreshortening of corrosion-fatigue crack growth life by pitting corrosion. Because electrochemical variables strongly influence pit growth, these variables would also affect the conventional S-N data. Crack growth, on the other hand, occurs by hydrogen
Volume 1" Chemistry, Mechanics and Mechanisms
0.990 0.900 . - 0.500 -o o 0.250
Ac=200MPa all rvs I
-
aco
-~- 0.100 . . . . . . . . . ..Q r 0.050 - -
13
~
....
j/~
rE -.-~ ....
o t,_
n
0.010 0.005 0.001 2e+5
I/
/I
3e+5
I
I
I
|
I F
5e+5 7e+5 le+6
I
I
2e+6
3e+6
In(N F)
Fig. 6. Single simulation showing the sensitivity of fatigue lives to variability in each of the internal random variables (see Table 1 at Acr = 200 MPa).
0.990 Ar (MPa) 400 0.900 0.750 300 .~ 0.500 - ~ 200 o 0.250-
.......
lOO
S/
:--_- o.loo
/
/
//
r 0.050 .Q
/
,f
,
.o
n
! f
0.010 0.005
d 0.001 104
105
106
107
cycles; N F
Fig. 7. Variability in fatigue lives attributed to the internal random variables (see Table 1) at different stress levels.
embrittlement and would depend on the crack-tip environment, which is, by and large, shielded from changes in external electrochemical variables. As such, it would be essentially independent of these variables. From this perspective, therefore, the perceived dichotomy (i.e., the inconsistency in electrochemical response) between the conventional and fracture mechanics approaches to corrosion fatigue (and stress corrosion cracking) is resolved. Although the discussion here is focused on the influence of pitting corrosion on corrosion fatigue, it may be generalized to include other forms of localized corrosion, as well as stress corrosion cracking. In light of this resolution, it would be reasonable and worthwhile to reexamine the wealth of research data on corrosion fatigue over the past three decades to broaden the understanding of corrosion fatigue.
14
Modeling Environmental Attack
4. 4. Evolution and distribution of damage in aging aircraft From an engineering perspective, to demonstrate the efficacy and utility of this modeling approach, a comparison was made between the model predictions and damage measured on a transport aircrat~ that had been in commercial service for about 24 years. Instead of predictions of corrosion and corrosion fatigue lives, the models were exercised through 'Monte Carlo' simulation to determine the evolution and distribution in damage size as a function of time. The results are shown in Fig. 8. The specifics of the analyses and comparisons are detailed elsewhere [14]. The essence of the finding is that, by using short-term laboratory data, the model was able to capture the essence of the size and distribution in damage for an aircraft that had been in service for about 24 years (i.e., for well over two orders of magnitude extrapolation in time). Through this process, the model may be used also for estimating the evolution and spatial distribution in damage over time, Fig. 9, either over different locations in a given structure or component, or for a single location in a group of structures or components. 0.9990 0.9000
0.7500
N%.
- ~ 0.5000 A
~.
0.2500 -
~,~ "~
N 0.1000 tO 0.0500 I1)
~ ~ ~
k
, ~I
~
9
E 0.0100 0.0050
t~ "10 i._
"-,
~~.~k".
", ,
~5,ooo---'\'%_'. 20.00o----'t--'~
",t..
3o,ooo....
-',- ....
........
k
X " ,.
~-'~---~r,-'r. -. - - - ~ - . . . . ~ \
, i ...... I
0.10
,,
-~--~------~
Jri . . . . .
I
\.
x\
..
25,000 . . . . ~ - - - - ~ - - ~ N - 35 ' 000 . . . .
0.0001 0.01
X
',. -.
9
22,533___ ~.__..~_~..___~_ 0.0010
586 measured: s e c 2; stiff 4
~ "~
c,es
t~
ft.
CZ-184 (B707-321B) 57,382 flight hours 22,533 flight c y c l e s ~ 24 years in service
, i%,,,~,1
1.00
10.00
....
--'~
I,=,1 100.00
d a m a g e size, a ( m m )
Fig. 8. Estimated evolution and distribution of damage versus observations in the CZ- 184 aircrat~ [ 14].
5. S-N response for very high cycle fatigue (VHCF). Considerable interest has been developed over the past five years by the observation of unexpected S-N fatigue response at lives in the 108 to l 0 ~~cycle range; see the papers in the special session on giga-cycle fatigue in [15]. The response is reflected by a lower "endurance limit" relative to that observed by the conventional procedure of testing to only 107 to 108 cycles, and by the prominence of subsurface crack nucleation at internal inclusion particles in the high cycle domain. These internal nucleation sites have been dubbed "fish eyes" because of their darkened appearance in optical microscopy, see Fig. 10. The precise mechanisms for this high-cycle response are not fully understood. In Ref. [16], Murakami summarizes the view that attributes the behavior to the local
Volume 1: Chemistry, Mechanics and Mechanisms
15
(One Realization) 1.2 0.8 0.4 0.0
MSD - PoO (20,000 cycles)
~HYsD-POe~22~c~c'e~'
~
2
;
80
MSD-Po'0"i2'5,000 cycles)
-""1 t~E 20
MSD-PoO(30,000oyoIes)
~,o H 'MSD Poe"' ~176176 c'c =0.
j._,,.,l ...... L.,,,t,, ....
0
[ Time
I I11
250
500
J .....
IL I1~ IlltL
~1
IL
I
,,j , . . ,. ] , J ,_ , ]1,, d I. ] . . , ..... ~ ~ ~1, .... L[ ~j"b .... 750 1000 1250 1500 consecutive numbering of hole sides
I L ' " ' L .......
J ..... ~ l ~[t~ l"
1750
2000
Fig. 9. Successive simulation showing the evolution and distribution of corrosion and fatigue damage, and the formation of significant areas of MSD over 1000 fastener holes for the CZ- 184 aircraft [ 14].
Fig. 10. Scanning electron micrographs of crack nucleation at a typical inclusion, high and low magnifications, respectively [ 15].
concentration of dissolved hydrogen at the crack-nucleating inclusions. Other possible contributors include the influences of residual stresses (surface versus interior), environment (external versus internal), or both. Based on the studies summarized above (see Section 4), it is reasonable to assume that the conventional S-N response for steels can be related also directly to the crack growth life (i.e., the number of fatigue cycles required to grow a crack from its nucleus to failure). This approach was applied to assess, computationally, the influences on S-N response by "surface" residual stress and its surface-to-interior distribution or the effects of environment, as well as, the probabilistic influences of the variability in the size of crack nuclei and in other material properties. In Ref. [ 17], a crack growth based probability description for fatigue life prediction into the giga cycle range that explicitly
16
Modeling Environmental Attack
incorporates the effects from internal damage as well as external damage is proposed. A connection between the S-N and crack growth behaviors was established and demonstrated. Through this description, the S-N response and the associated variability in fatigue lives are linked to key random variables that are explicitly identified in the crack growth model; namely, the initial surface damage sizes, the initial internal damage (inclusion) sizes, the fatigue crack growth rate (or power-law) coefficient, and the fatigue crack threshold AK (AKth). The identification and quantification of these rvs are vital for probabilistic estimation and prediction of fatigue life. The model is assessed through comparisons with an extensive set of fatigue life data for SUJ2 steel, Figs. 11 and 12. 0.999 0.990 0.900 0.750 0.500 v
~ 0.250
(
"5 0.100 >, 0.050 0.010 ~ 0.005 0.001 i-i 1e+0
le+1
le+2
le+3
le+4
le+5
le+6
le+7
le+8
le+9
Number of Cycles to Failure, Nf Fig. 11. S-N data for SUJ2 steel along with cdfs computed from a fatigue crack
growth model.
4000 Environment (solid) 3500
~~n~rt
0
(dashed)
3000
v D ....
surface (261) fisheye (63) run out (8) median
2500 "o 2000
1800~
.m
9
ca. c~ 1500 1000
1600 ~
v ""~~'v~
SUJ2 Steel
1400 m ~ s 1200 "1o ooo t~
, , , ,~.11~ l o o.8 o.6 o4 o2 o o Probability, PSD
500
100 101
102
103
10 4 10 5 10 6 10 7 108
109 1010
cycles
Fig. 12. Schematic of the median characteristics for an alternative interpretation of the very high cycle fatigue S-N behavior.
Xm~xthen the value is rejected and another sample is taken. It should be noted that the values generated by Eq. (2) are such that x > Xo. The adoption of a random number generator in calculating the initial set of corrosion pit depths is necessary to reflect the statistical variability of the data, recognising that every exposure test would generate a slightly varying distribution of pit depths. In applying this method, we adopted two approaches: fitting the Weibull distribution to the initial set of measurements using the random number generator, with a~ and a2 as fitting parameters; alternatively, adopting a distribution using arbitrary but constrained values of al and a2 assumed nominally to relate to short exposure time measurements. In the latter case, the range of possible pit depths would be limited so that the distribution is relatively tight. In both approaches, the total number of pits is fixed in relation to the experimental measurements. The value of Xm~xdoes not tend to be critical so long as it is set sufficiently large that it does not sharply trtmcate the distribution. The adoption of an arbitrary initial distribution is justified because the distribution of pit depths at longer times is determined predominantly by the growth rate parameters; the "memory" of the initial distribution fades with time.
2.2.2. Evolution of pit depth distribution It can be shown that the evolution of the pit size distribution function P(x,t) from an initial stable population is governed by the partial differential equation (see Section 7 of
Volume 1: Chemistry, Mechanics and Mechanisms
23
Ref. [9] for an analogous approach considering the evolution of defects that are distributed in a volume rather than on a surface) OP(x, t) = _ O[g(x)P(x ' t)] + S(P(x, t),x, t), Ot Ox
(3)
where g(x) is the growth rate of the pit and S(P(x,t), x, t)Sx is the number of pits of size in the range x ~ x + 5x which are nucleated or deactivated in the given area of the exposed surface during the time interval &. The latter term is included for generality but since we are considering a stable pit size distribution, for which pits do not deactivate and no new pits grow into this size range, it would be neglected for this application. (Note: Although the S term in Eq. (3) was not used in this analysis, we are grateful to the reviewer for pointing out the need to include the dependence on P(x,t) so that it is formally correct; it was excluded in the original version.) This 'continuity' equation is similar in form to that for the rate of change of density in a fluid where g(x) would now be the fluid velocity, as implied by Section 7 of Ref. [9]. The way in which the distribution of pit velocities are treated is to regard the pit distribution being split into various sub-distributions where, for the Ith distribution, the parameter ot of the pit growth law has values in the range r ~ tXI 4- AIX. Provided that Atx is small enough, there is minimal error in assuming that all pits in the given subdistribution have the same value of ix. Equation (3) can then be applied to each pit subdistribution without difficulty. This is the approach that has been taken when developing the numerical simulation methods for pit growth. 2.2.3. Pit growth rate
It is assumed that the depth can be described by x = oft p
(4)
In contrast to the approach of Wei et al. [6], who assumed that the pit maintained hemispherical geometry and grew at a volumetric rate determined by Faraday's law, we do not make any a priori assumption of the value of the exponent, 13. For steam turbine disc steels in relevant environments, the aspect ratio of the pits varied with the environment and with pit depth, and at any pit depth showed a wide spread in values [10]. Correspondingly, it was not pertinent to assume a specific model for the pit growth process, the parameters of relevance being those that best represented the experimental data. From Eq. (4), the growth rate has the form: dx = g(x) = 13a'~x ~ ' - ~
(5)
dt It is readily apparent that the pit growth rate must be statistically distributed. Consider an initial pit size distribution, perhaps generated from an initial distribution of MnS particle size. From Eq. (5), for constant values of tx and 13 (> 1 samples of the surface having the same area. For large enough areas, each sample will have a different number of pit depths whose sizes lie in the range x ---}x + ~Sx. It is vital to understand how the pit distribution evolves with time. Because of aggressive environmental conditions, the corrosion pits will grow in size during service conditions. It is assumed that the time dependent growth of a pit of size x(t) is governed by the relation
dx d--t-
g(x),
(Al)
where g is some function of the pit size that may also be dependent on other quantities which are fixed in time such as the applied stress and temperature. The objective now is to develop the relationships that define the distribution of pits as a function of exposure time. Atter time t has elapsed, a pit that has size x will have grown from an initial size X where ) d~ x gi~) = t ,
so that
T(x)-T(X)=t
~ d~ where T(Y)=0 g(~)"
(A2)
The relation (A2) is valid only if the behaviour of the function g(x) as ~ ~ 0 is such that the integral defining the function T(y) is well defined. The approach taken is suitable for the function to be used in the subsequent analysis. It can be shown that the pit size distribution function at time t > 0, denoted by P(x,t), is governed by the partial differential equation, having the form of a continuity equation (see Ref. [9] for an analogous formulation involving a volumetric distribution of defects) OP(x,t) = - - - , u ( x ) P(x,t),' at 0x LC, X
J--x
(A3)
Ij
where P(x, t) = g(Xt(x,t)) p(Xt(x,t)) g(x)
Xt(x,t) = T - t ( T ( x ) - t) '
(A4)
where Xx(x,t) is the initial pit size that requires time t to grow to size x, and where p(x) = P(x,0) is the initial pit size distribution function. The function T-~ is the inverse of the function defined in Eq. (A2).
A.2. Corrosion pit growth law It is observed that pit growth laws are of the form x = ctta. The function g(x) appearing in the growth law (Eq. (AI)) is then of the form dx = g(x) = 13ctt/~ x H/~, dt
(A5)
so that g(x) and the pit growth rate decrease as the pit depth increases. It should be noted that, for the selected values of 13the behaviour of the function g(x) as x ~ 0 is such that the integral in Eq. (A2) defining the function T(x) is well defined. The function T(x) appearing in Eq. (A2), its inverse T-~(x) and the function X~(x,t), defined by Eq. (A4), are then given by
Volume 1: Chemistry, Mechanics and Mechanisms
T(x) =
(x)
,
43
T-l(x) = a x 13, (A6)
X,(x, t ) - [x '/13 - ot'/13t] ~ .
A.3. Inclusion of corrosion pit nucleation effects When pits are nucleating during service it can be shown that the evolution of the pit size distribution function P(x,t) is governed by the partial differential equation 3P(x, t_._____~) 3 - - ~---[g(x)P(x, t)] + S(P(x, t), x, t),
(A7)
where S(P(x,t),x,t) is the number of pits of size in the range x ---> x + 15x which are nucleated in the given area of the exposed surface during the time interval 1St. The same approach can be used for pits that 'die'. The solution of Eq. (A7) will require numerical methods. A. 4. Pit to crack transition and crack growth law Corrosion pits, when they have grown sufficiently, often initiate cracks which can grow in time due to stress-corrosion mechanisms. It is assumed that a corrosion pit can transform to a crack of the same size when the following conditions are both satisfied
x>xt
! > crack
pit
The parameter Xth is a threshold crack size that must be exceeded before a crack can exist. The pit growth rate is defined by Eq. (AI) in general, or by Eq. (A5). The crack growth rate is assumed to be governed by the specific relation dx
= C 6 pxq,
(A9)
dt where a stress dependence has been introduced to represent its effect on stress corrosion cracking. It can be shown from Eqs. (A5) and (AI0) that the second of the crack transition conditions (Eq. (A9)) is satisfied if
p X > Xcrit =/Pctl/13/l+(q-l)13Cl~p
(alo)
Thus from Eq. (A9) a growing pit will convert to a crack if both x > Xth and x > Xcrit. It is useful to consider at any time t the percentage of pits that have converted to cracks. If a time incremented analysis was performed, then this percentage could be tracked as time progressed. The approach made here is to consider pit and crack distributions only at time t without having determined the state of the system at intermediate times. It is useful to consider the history of one particular corrosion pit having randomly selected values of a and of C. A unique curve can be plotted for the pit growth rate as a function of size that will depend upon the
Modeling Environmental Attack
44
value of a, as shown schematically in Fig. AI. Similarly a unique curve can be plotted for the corresponding crack growth rate that will depend upon the value of C. At time t = 0, the pit will have a size that is greater than the value x0. As time progresses, its size will increase, but at a decreasing rate. At time t = ttra,s the critical size Xcrit is reached at which the pit can convert to a crack (assuming that the threshold size Xth has already been exceeded). It is emphasised that the features of Fig. A I are known as soon as the pit size and the parameters a and C have been selected at random from statistical distributions, and are independent of the exposure time t, which controls only the size of a particular pit/crack. At time t > tt~a~sthe crack size will have increased to the point A shown on Fig. Al. If the pit did not convert to a crack, then it would have reached the point B. As t --> ttransthe points A and B would converge to a single point that is at the intersection of the pit and crack growth curves.
,
Crack
dx dt
! I
! i
X0
Xth
Xcrit
Size (x)
Fig. A1. Illustration of pit to crack transition for a single pit having a specific value of the growth rate parameters tx and C (diagram assumes Xcrit > Xth ).
When simulating a set of corrosion pits, and examining their sizes only after a time t has elapsed, it is possible to determine the number of pits (if any) that have already converted to cracks. For specific values of ot and C attributed to a particular pit and defining the value Xent, if the current pit size exceeds the critical value X~it for a transition, and if it also exceeds the threshold size Xth , then the pit will have converted to a crack at an earlier time ttr~s when its size had the value Xcnt. It is thus possible to determine the number of cracks at time t, but only if, before transition the crack growth rate is less than the pit growth rate for all sizes x < Xcnt, and following transition, the crack growth rate exceeds the pit growth rate for all sizes x ___X~nt, ensuring that cracks are unable to convert back to pits.
A.5. Implementing a numerical simulation of corrosion pitting and cracking In order to test the usefulness and validity of the statistical analysis that has been described above, it necessary to develop a numerical method of simulating the occurrence and growth of corrosion pits. First of all it is necessary to have access to a sub-routine that generates a random number x in the range 0 < x _< 1. Such a routine is readily available when using FORTRAN compilers. The random number generator is initialised using the current date and time so that the first number generated is random. If such 'seeding' is not undertaken then the simulation will
45
Volume 1: Chemistry, Mechanics and Mechanisms
generate the same results whenever the simulation sottware is first used. By repeatedly selecting random numbers and taking the average of values selected, the random number generator can be tested. As the number of samples increases the average should tend to the limiting value 0.5. For such a test using 100000 samples, the average value found was 0.50045, which is very close to the limiting value. The next step is to use the random number generator to calculate a sample set of corrosion pit depths. It is useful to restrict these values so that they lie in the range Xo < x < Xm~x,where x0 and Xmx are prescribed values known at the outset. The size distribution for the pit sizes is assumed to be defined by the three-parameter Weibull distribution as this is a reasonable choice that leads to relatively simple relationships. Other distributions could be substituted if required. The probability that the size of a corrosion pit lies in the range 0 ~ x (i.e. the cumulative probability) is then given by the relation
F(x)- ip(~)d~ = 1 - exp[-at(X-Xo)'~]
'
(All)
0
where a~ and a2 are the Weibull parameters that are assumed to be given. The relation (AI2) may be inverted so that
x=x 0 +
~ln a~
(Al2) I - F(x)
The random samples of the pit size x are found using Eq. (A12) where the value of F(x) is selected using the random number generator. If the value of x > Xm~xthen the value is rejected and another sample is taken. It should be noted that the procedure always generates values of x such that x > x0. It follows from Eq. (AI 1) that the corresponding probability density function p(x) is given by p(x) = ata2 (x - x o
a2-1
exp [ - a , ( X - X o f ~ ] 9
(AI3)
The next step is to select random values for the parameters ot and C appearing respectively in the growth l a w s - Eqs. (A5) and (AI0). These parameters are assumed to belong to normal distributions, and the Box-Muller method for selecting random values is used. Two random numbers x~ and x2 are first generated lying in the in the range 0 _<x < 1. The parameter y is then calculated using
I 2 cos(2nx~)'
(AI4)
y = l.t + o" ln(1/xl)
leading to a random value y that has been selected from normal distribution having mean value la and standard deviation or. This method is used to generate random values of ot and C, each distribution having different mean values and standard deviations. Thus, each corrosion pit that is included in the simulation has randomly selected values for its size x, for its pit growth parameter a, and for its crack growth parameter C which takes effect only aider the pit to crack transition has occurred. Having selected these values, the time dependent behaviour of individual pits is deterministic being governed by the pit and crack analysis presented above.
47
Revisiting the film-induced cleavage model of SCC Andrew Barnes a,b, Nicholas Senior a, Roger C. Newman a Corrosion and Protection Centre, UMIST, P.O. Box 88, Manchester, M60 1QD, UK b Present address: Invista Performance Technologies, Wilton Centre, Cleveland, TSIO 4X~, UK r Department of Chemical Engineering and Applied Chemistry, University of Toronto, 200 College Street, Toronto, Ontario M5S 3E5, Canada
Abstract Following a brief historical review, the present state of the film-induced cleavage (or filminduced fracture) model is presented using data on thin wires and foils of noble-metal alloys subject to surface de-alloying. A definitive demonstration of totally dry, single-shot transgranular film-induced cleavage is still lacking, but recent progress has been promising, lntergranular substrate fracture from a surface de-alloyed layer was obtained at liquid nitrogen temperatures as early as 1989, although recent experiments have shown why this may not be a readily reproducible result (the formation of ice within the de-alloyed layer can inhibit crack propagation). Convincing results were obtained on thin de-alloyed wires of Ag-20a/oAu, left in the aqueous solution where they were de-alloyed, but fractured at controlled electrode potentials at which no Faradaic reactions were possible- but any wet environment carries the theoretical possibility of some kind of unknown environmental effect. Nevertheless such data do demonstrate that there is a real low-energy substrate fracture, initiated from a de-alloyed layer and resembling cleavage, that under normal loading conditions occurs repetitively to grow a macroscopic crack.
I. Introduction The term 'film-induced cleavage' (FIC) was introduced by Sieradzki and Newman to describe a mechanism of stress corrosion crack growth in which successive small substrate fractures are initiated by a thin surface film, usually a nanoporous de-alloyed layer. The term tends to be used for both transgranular and intergranular stress corrosion cracking (SCC), but a more accurate term for the latter is film-induced intergranular cracking [1]. Experimentally, this phenomenon can be traced back to work of Edeleanu and Forty [2] and Pugh and others [3-12] on o~-brass. Edeleanu and Forty thought the role of dezincification, or corrosion in general, was to attack slip bands, somehow exposing intervening material that was embrittled by short-range order. Pugh thoroughly explored the possibility of substrate embrittlement by hydrogen, and eventually rejected this on electrochemical grounds, for brass but especially for
Modeling Environmental Attack
48
more noble alloy systems such as Au-Cu, which show a similar fracture morphology and unexpectedly rapid crack propagation [ 13-15]. The experimental evidence for FIC (or intergranular fracture) rests chiefly on a type of single-shot fracture experiment [16-22] pioneered by Shahrabi [16]. A thin foil or wire of alloy is de-alloyed, then fractured either: 9 Very rapidly in the corrosive solution (making the assumption that such a high strain rate would eliminate conventional SCC or at least require it to propagate at extraordinary velocity that would itself require a new explanation). 9 Atter rapid freezing in liquid nitrogen (exploiting the fact that fcc metals are normally ductile at such temperatures) [16,17]. 9 After applying a potential at which no further Faradaic reactions are possible, and conventional SCC would not occur (this was not possible for brass but works well with gold alloys) [19-22]. In every case it is necessary to confirm that the substrate embrittlement is reversible - in the case of brass, where the de-alloyed layer is very thin, this required only a few seconds exposure to ambient air. For the gold alloys held at constant potential, substrate embrittlement may disappear over ageing periods of minutes to hours. The main phenomenon implicated in such ageing is the coarsening of the nanoporosity within the de-alloyed layer by surface self-diffusion of the noble metal [18]. In this paper we report recent progress in this area.
2. Experimental 2.1. Recent experiments on thin wires of Ag-Au alloys (A. Barnes, UMIST) An Ag-20a/o Au alloy was cast using 99.95% Au and 99.999% Ag, then drawn into wire of 0.5 mm diameter. 'Intergranular' cracking samples were annealed at 700~ for 1 h, giving fully recrystallized grains of 50-150 ~tm diameter. 'Transgranular' cracking samples were annealed under load at 900~ for 72 h. This resulted in a 'bamboo'-type structure, where the grains covered fully the cross-section of the wire, and were distinguishable by thermal etching of the grain boundaries. All samples were annealed under a hydrogen atmosphere. The wires were placed within a home-built tensile machine with a 0.225 kg load cell. Short lengths were electroetched down to 20-100 ~tm diameter using a single drop of cyanide-containing solution [23] suspended from a looped platinum wire cathode that encircled the sample (Fig. 1). A voltage of 1800 mV was applied until,the desired diameter was reached. Control tensile tests with no environment present showed >85% reduction in area (RA) for all sample treatments. The apparent ultimate tensile strength (UTS) values for the conventional and bamboo structures are displayed in Table 1.
Table 1 Comparison of mechanical properties for Ag-Au material Condition As-received 'lntergranular' samples 'Trans~anular' samples
Max UTS (MPa) 314 180 130
Min UTS (MPa) 270 124 85
%RA 60-75 80-95 90-95
49
Volume 1: Chemistry, Mechanics and Mechanisms
500 Bm .~
~ L /\ ..........
Grain J
'Bamb '
/,,
' " ~~>20Bm '"l I Electrolyte droplet Appliedstress Fig. 1. The electroetching technique.
The prepared wire was washed with deionized water, and the cell filled with 1 M HCIO4, prepared from 15 Mf~ cm deionized water and Analar grade acid. Immersion of counter and reference electrodes (SCE, via a salt bridge) allowed the potential to be controlled by a potentiostat; generally de-alloying was carried out at one potential and fracture at another, with a number of variants that will be described later on. Typically de-alloying was performed at 1050 mV vs. SCE, where the de-alloying rate was very high (initially more than 100 mA/cm2). All experiments were conducted at room temperature. For experiments where the wires were stressed during de-alloying, the conditions were close to constant displacement, though with some elastic displacement stored in the unreduced length of the wire. Various types of video camera were used to record the fracture of the samples; these results will be presented elsewhere. Briefly, conventional video showed that fracture was certainly fast, taking less than 20 ms, but high-speed video did not add anything to this finding because a certain crack opening was required to be sure of detecting a crack. Micro-indentation studies are also planned. 3. Results and discussion
3.1. Intergranular fracture under low loads
The "classical" film-induced fracture experiment uses normally processed material and produces an intergranular fracture [17]. A surprising result was obtained in experiments where the samples were stressed lightly during de-alloying, initially just to keep them straight. In 17 experiments, 14 spontaneous intergranular fractures were obtained at stresses (referred to the reduced cross-section)of 5 to 15 MPa. Sometimes
Modeling Environmental Attack
50
(a)
(b) Closeup
Fig. 2. SEM micrographs of an intergranular fracture, de-alloyed at 1050 mV vs. SCE whilst at 5% UTS.
the fracture surface was entirely brittle (Fig. 2), while some samples showed a combination of intergranular fracture and ductile tearing (Fig. 3(a)). In no case was the de-alloyed layer thickness more than 10% of the reduced section thickness. The depth of de-alloying is clearly visible on all fracture surfaces. There was no evidence of gradual crack propagation- video evidence and current measurement showed a sudden fracture taking place in less than 0.1 s. A further 19 experiments were carded out on similar samples with a slower rate of de-alloying (potential of 975 mV vs. SCE). 14 samples failed by sudden intergranular fracture at reduced-section stresses of 2 to 21 MPa and de-alloyed layer thicknesses of 0.4 to 2.7 lam. The amount of load seemed to have little effect on the time to fracture- 11 experiments were carried out at 1050 mV with higher loads of up to 39 MPa and failure times remained in the range of 10 to 20 s. If anything, higher loads produced longer times to fracture and may have required greater de-alloyed layer thickness (up to 4 ~tm). Such experiments can easily be dismissed as SCC experiments and not FIC experiments- the stress was applied during corrosion, and the samples failed by spontaneous cracking. But the extremely low value of the stress, the small depth of dealloying, and the sudden nature of the fracture (in a material that does not fracture suddenly) strongly suggest that we are observing a film-induced fracture phenomenon. This will be explored further below.
3.2. Transgranular fracture under low loads A further 33 experimems were carried out on the samples with a bamboo grain structure. These were not as reproducible as the intergranular cracking experiments, and sometimes the fracture found a grain boundary outside or at the edge of the reduced
Volume 1: Chemistry, Mechanics and Mechanisms
(a) SEM micrograph of intergranular material de-alloyed at 975 mV vs. SCE under high tensile stress-20% UTS. Predominantly ductile fracture, with some intergranular.
51
(b) SEM micrograph of transgranular material. De-alloyed at 975 mV vs. SCE and fractured at 0 V vs. SCE. 2 I,tm layer.
Fig. 3. Intergranular/ductile and transgranular fracture morphologies.
section. Nevertheless, out of the 33 experiments, 12 produced cleavage-like transgranular failures at reduced-section stresses of less than 23 MPa, and de-alloyed layer thicknesses of up to 4.3 pm. A typical example of transgranular cracking is given in Fig. 3(b). It should be mentioned that these samples were otten thinner than those used for intergranular cracking experiments, but the depth of de-alloying was more uniform and there is no possible issue of local penetration as might be surmised for grain boundaries. Furthermore, the cracking of the de-alloyed layer and the cracking of the substrate have quite different morphologies. All the transgranular cracks propagated suddenly, with no evidence from the load or the current that there was any gradual crack propagation. In future work EBSD (electron backscatter diffraction) analysis should be used to determine the orientations of the grains in the bamboo structure. 3.3. Fracture by overload at the de-alloying potential These experiments produced the most consistent brittle fractography. 16 of 20 samples showed intergranular fracture atter de-alloying to a depth of 3 lam at 975 mV vs. SCE, and 9 of 17 bamboo samples showed transgranular fracture under the same conditions. The overloading time in these tests was less than 0.1 s, so if one were to interpret the behaviour in terms of rapid SCC propagation (rather than a nonenvironmentally dependent brittle fracture), the velocity would have to be on the order of millimetres per second - not impossible, but a more consistent explanation is that the crack just jumps through the sample without further influence of the environment.
52
Modeling Environmental Attack
3.4. Crack propagation at reduced potentials after de-alloying at high potential 3.4.1. lntergranular fracture Twelve experiments were conducted in which de-alloying was carried out at very low load, producing no cracking; then the potential was dropped to 600 mV before fracturing the sample after a short delay of around 5 s. Current measurement showed that this delay was sufficient in every case to allow capacitive discharge of the double layer within the de-alloyed material, so there is no IR drop issue to complicate the behaviour. Eleven of these samples showed complete or nearly complete intergranular fracture from de-alloyed layers of 2 to 4.1 pm thickness and at reduced-section stresses of less than 8 MPa. Thus the de-alloying reaction does not have to occur simultaneously with tensile loading to cause substrate fracture. In fact these samples fractured at lower loads than those that were loaded during de-alloying (possibly premature fractures of the de-alloyed layer sap its ability to inject a crack). Longer ageing times at 600 mV produced progressive healing of the substrate embrittlement caused by the de-alloyed layer. After 5 min ageing at 600 mV, 3 out of 3 samples fractured in a ductile manner and sustained 18 to 46 MPa of reduced-section stress. Apparently the notching effect of the de-alloyed layer accounts for the reduction in apparent UTS value - this aspect was not investigated further. Since one could postulate some continued reactivity of silver at 600 mV (though this was not indicated by any sustained anodic current), a second series of experiments was conducted with a more radical reduction in potential, to 0 V vs. SCE. For a short delay time before fracture (5 s), 5 out of 21 samples showed complete intergranular fracture from de-alloyed layers 0.4 to 4.1 pm thick and at reduced-section stresses less than 21 MPa. 3 of 11 samples failed intergranularly for a longer delay time of 1 min. When the de-alloying time was reduced (to give thicknesses less than about 1 lam), there was (surprisingly) a more consistent intergranular cracking with the 5 s delay time; this is not understood at this point, but such results are useful as they dispose of any suspicion that the whole sample might be penetrated by fine intergranular de-alloying (which anyway would not explain the ageing effects).
3.4.2. Transgranular fracture The bamboo structured samples showed plenty of cleavage like fracture with the 600 mV hold potential and the short (5 s) delay time - 4 out of 12 samples showed cleavage failure from de-alloyed layers of 0.9 to 2.7 pm thickness and reduced-section fracture stresses of less than 29 MPa (it was noticeable that more stress was required to cause transgranular than intergranular fracture). Ageing for 1 min at 600 mV did not alter these statistics - 2 out of 4 samples showed transgranular fracture. However for ageing at 0 mV vs. SCE, transgranular fracture was only obtained if the sample was fractured instantly, otherwise ductile behaviour ensued. The ageing process appears to be rapid at 0 mV (or the layer is disrupted in a time-dependent manner), and the transgranular film-induced fracture is more demanding than the intergranular version in terms of the quality of the de-alloyed layer.
Volume 1: Chemistry, Mechanics and Mechanisms
(a) Surface of a sample de-alloyed at 1000 mV vs. SCE.
53
(b) Surface of a sample de-alloyed at 1100 mV vs. SCE.
Fig. 4. Secondary electron SEM micrographs of the effect of de-alloying potential upon surface fracture of unstressed Ag-Au foils.
3.5. New approaches for transgranular fracture 3.5.1. Work carried out on Ag-Au foils (N. Senior, UMIST) Foil samples of Ag-22.7a/o Au were prepared by cold rolling an ingot to a 100 ~tm thickness and annealing at 975~ for 1 h under a hydrogen atmosphere. The resultant grain sizes were 50-200 ~tm. The main focus was to enhance the likelihood of transgranular FIC whilst avoiding the use of thin wires or complex grain growth techniques. It was found that the de-alloying and subsequent tensile fracture of unmasked foil samples would routinely result in a combination of intergranular and ductile fracture. As with Barnes' work, it was noted that a lower de-alloying potential resulted in greater propensity for brittle substrate fracture. Analysis of the surface morphology for samples de-alloyed at 1000 mV vs. SCE, but not subjected to any tensile stress, showed shallow cracking, conf'med to the de-alloyed layer, along all grain boundaries, and occasionally along twins, as shown in Fig. 4(a). This is attributable to the induced tension from dealloying, which is exacerbated by prior thermal etching of the boundaries. Consequently, it is difficult to obtain any transgranular fracture with such samples. Higher de-alloying potentials, (1100 mV and above), resulted in not only intergranular fracture across the de-alloyed layer, but also a network of transgranular cracking of the layer (Fig. 4(b)). This removed completely the ability of the layer to inject any kind of brittle fracture into the bulk material underneath. Possibly this is associated with some kind of superficial gold oxidation, which destroys the integrity of the layer, or perhaps the layer needs some residual silver to keep its morphology intact, and a too rapid removal of all the silver creates too much internal stress.
54
Modeling Environmental Attack
To bypass this problem, the foils were masked by coating with lacquer and then damaged in a controlled manner. An array of bare spots, of 40 lam diameter, could be produced, and could be de-alloyed such that many de-alloyed spots lay entirely within a single grain. This technique removes the need to produce an extremely large grain size. Preliminary work has shown that the de-alloying of these spots is prone to the same effect of internal stress witnessed at high de-alloying potentials; a fine network of cracks is immediately injected across the de-alloyed layer from the interface with the masked bulk alloy. These cracks de-alloy at an enhanced rate, and the resultant fractured structure simply deforms under applied stress, rather than generating filminduced cleavage. We feel that this problem might simply be overcome by fine-tuning the variables of spot size, de-alloying potential, solution concentration and temperature. As such, it still holds promise for the future. As Friedersdorf [1 ] found, it is not necessarily easy to reproduce the behaviour of the 20% Au alloy with higher gold-content alloys. Even at 23% Au, the de-alloying has a different character which probably reflects the higher potential required to give the same reaction rate; this higher potential affects both the silver dissolution and eventually the gold oxidation reactions. The 20% Au alloy more nearly conforms to the 'ideal' situation of a nearly-pure, non-oxidized gold layer and a very rapid dissolution rate. 3.6. 'Dry'fracture
It has already been demonstrated that FIC occurs in Ag-Au alloys under conditions in which there is no sustained silver dissolution, and hydrogen embrittlement can be ruled out on a thermodynamic basis. However rinsing and drying of the de-alloyed layer usually destroys its ability to inject cracks into the substrate. One way to approach the problem of 'dry' FIC is to freeze the de-alloyed sample in liquid nitrogen and strain it at low temperature. Whilst this technique has been shown to produce both intergranular and transgranular fracture in thin foils [16,17], the statistics are not as convincing as for the wet fractures. It might be that the thermal shock of immersion into liquid nitrogen damages the layer, although microscopy and impedance measurements both failed to detect a difference prior- and post-immersion. Immersion of the dealloyed specimen into various water/alcohol mixtures at-20~ resulted in identical fracture behaviour. It is therefore likely that it is the formation of ice within the pores that can damage the ligament structure and reduce the likelihood of injecting a crack into the substrate. Work is currently underway to de-alloy using an ethanol-based perchloric acid solution, which should allow the cooling of the sample to below-100~ without ice formation. Very recently we have found that the combined use of an elevated temperature (60~ and a relatively low de-alloying potential gives intact, uncracked layers that age much more slowly than the usual layers formed at room temperature; these new layers easily inject cracks into the substrate after rinsing with deionized water, and to some extent retain this ability after drying. A publication on this work is in preparation. 3. 7. Platinum alloys
Cu-Pt alloys bear advantages over Ag-Au alloys with regards to F|C studies; the higher melting point of Pt results in a greatly reduced surface diffusion coefficient
Volume 1: Chemistry, Mechanics and Mechanisms
55
within the de-alloyed residue compared with that of Au, and consequently, the dealloyed structure has both a much reduced pore/ligament size (as little as 3 nm [24]), and is less prone to the effects of ageing. These advantages are unfortunately outweighed by the smaller grain size, (20-50 ~tm obtained at 1000~ for 12 h under hydrogen), and a very strong tendency for intergranular de-alloying- the material disintegrates along grain boundaries without stress, allowing the formation of internal de-alloyed layers (Fig. 5). There is thus little chance of studying the injection of cracks without taking steps to prepare large-grained material that can be masked. The intergranular nature of the de-alloying reaction in this material may reflect the difficulty of ionic motion within the exceptionally fine pore structure; the greater mobility of the lattice at grain boundaries allows easier transport.
Fig. 5. Formation of internal de-alloyed layers within a Cu-25a/o Pt alloy, de-alloyed in 1 M H2SO 4 at 1200 mV vs. SCE for 24 h.
4. Further comments and conclusions
Like many microscopic issues in fracture, FIC is best studied using specially prepared materials such as monocrystals and large-grained wires. Many of the characteristic phenomena disappear or become elusive when investigators use routine sheet material. Spontaneous fracture of the de-alloyed layer may occur, blunting (literally) its ability to inject cracks into the unattacked substrate. Such fracture appears to be potential-dependent, perhaps reflecting the enhanced removal of Ag or the onset of Au oxidation. Dry, transgranular FIC remains elusive in the Ag-Au system and is the focus of successful ongoing work. Polycrystalline Pt alloys suffer from extreme intergranular de-alloying, which though interesting in its own right necessitates the use of special materials in order to explore the mechanics of transgranular crack injection. Possibly these materials will respond better at higher temperatures.
56
Modeling Environmental Attack
A 'piezoelectric' effect has been reported in nanoporous metals and attributed to surface charge-induced stress [25]. Whilst not directly related to the stress corrosion mechanism, this does indicate the richness of phenomena that may influence the mechanics of a nanoporous layer during the injection of a fast crack into the substrate. We have known for many years that foils de-alloyed on one side show peculiar reversible bending and straightening when dried and rewetted, and this property is also being exploited by others in ongoing work related to sensors.
References [ 1] F.J. Friedersdorf, K. Sieradzki, Film-induced intergranular cracking of binary noble alloys, CORROSION/95, NACE International, Houston, 1995, paper no. 166. [2] C. Edeleanu, A.J. Forty, Phil. Mag. Series 8 (1960) 1029-1040. [3]. A.J. Bursle, E.N. Pugh, On the mechanism of transgranular stress-corrosion cracking, in: P.R. Swann, F.P. Ford, A.R.C. Westwood (Eds.), Mechanisms of Environment Sensitive Cracking of Materials, The Metals Society, London, 1977, pp. 471-481. [4] A.J. Bursle, E.N. Pugh, An evaluation of current models for the propagation of stresscorrosion cracks, in: Z.A. Foroulis (Ed.), Environment-Sensitive Fracture of Engineering Materials, TMS-AIME, Warrendale, 1979, pp. 18-42. [5] J.A. Beavers, E.N. Pugh, Metall. Trans. A, 11A (1980) 809-820. [6] E.N. Pugh, On the propagation of transgranular stress-corrosion cracks, in: R.M. Latanision, J.R. Pickens (Eds.), Atomistics of Fracture, Plenum Press, New York, 1983, pp. 997-1010. [7] P.W. Slattery, J. Smit, E.N. Pugh, Use of a load-pulsing technique to determine stress corrosion crack velocity, in: S.W. Dean, E.N. Pugh, G.M. Ugiansky (Eds.), EnvironmentSensitive Fracture: Evaluation and Comparison of Test Methods, ASTM STP 821, ASTM, Philadelphia, 1984, pp. 399-411. [8] D.V. Beggs, M.T. Hahn, E.N. Pugh, Recent observations on the propagation of stress corrosion cracks and their relevance to proposed mechanisms of stress corrosion cracking, in: R. Gibala, R.F. Hehemann (Eds.), Hydrogen Embrittlement and Stress Corrosion Cracking: A Troiano Festschrifl, ASM, Metals Park, 1984, pp. 181-205. [9] E.N. Pugh, Corrosion 41 (1985) 517-526. [10] U. Bertocci, E.N. Pugh, Modelling of the potential at the tip of a transgranular stress corrosion crack in the alpha brass-ammonia system, in: A. Turnbull (Ed.), Corrosion Chemistry Within Pits, Crevices and Cracks, HMSO, London, 1987, pp. 187-198. [ 11] U. Bertocci, E.N. Pugh, Modeling of electrochemical processes during transgranular stress corrosion cracking of copper-base alloys, in: Proc. 10th International Congress on Metallic Corrosion, vol. 5, Trans Tech Publications, Zurich, 1989, pp. 219-229. [12] U. Bertocci, E.N. Pugh, R.E. Ricker, Environment-induced cracking of copper alloys, in: R.P. Gangloff, M.B. lves (Eds.), Environment-Induced Cracking of Metals, NACE, Houston, 1990, pp. 273-286. [13] B.D. Lichter, Microstruct. Sci. 13 (1986) 361-378. [ 14] T.B. Cassagne, W.F. Flanagan, B.D. Lichter, Metall. Trans. A, 17A (1986) 703-710. [ 15] T.B. Cassagne, W.F. Flanagan, B.D. Lichter, Metall. Trans. A, 19A (1988) 281-292. [ 16] R.C. Newman, T. Shahrabi, K. Sieradzki, Scripta Metall. 23 (1989) 71-74. [17] R.G. Kelly, T. Shahrabi, A.J. Frost, R.C. Newman, Metall. Trans. A, 22A (1991) 191-197. [ 18] R.G. Kelly, A.J. Young, R.C. Newman, The characterization of the coarsening of dealloyed layers by EIS and its correlation with stress-corrosion cracking, in: D.C. Silverman, M.W. Kendig (Eds.), Electrochemical Impedance: Analysis and Interpretation, ASTM STP 1188, ASTM, Philadelphia, 1992, pp. 94-112. [ 19] M. Saito, G.S. Smith, R.C. Newman, Corros. Sci. 35 (1993) 411-417.
Volume 1: Chemistry, Mechanics and Mechanisms
57
[20] R.C. Newman, M. Saito, Anodic stress-corrosion cracking: slip-dissolution and filminduced cleavage, in: T. Magnin, J.M. Gras (Eds.), Corrosion-Deformation Interactions, Les Editions de Physique, Les Ulis, 1993, pp. 3-26. [21 ] A. Barnes, R.C. Newman, Met. Mater. Process. 8 (1996) 211-217. [22] R.C. Newman, Stress corrosion cracking mechanisms, in: P. Marcus, J. Oudar (Eds.), Corrosion Mechanisms in Theory and Practice, 2nd ed., Marcel Dekker, New York, 2002, pp. 399-450. [23] L. GaI-Or, Electropolishing of gold alloys, in: D. Schneller (Ed.), Proc. 2nd Santa Fe Symposium on Jewellery Manufacturing Technology, Met-Chem Research Inc., Boulder, 1988, pp. 173-188. [24] D.V. Pugh, A. Dursun, S.G. Corcoran, J. Mater. Res. 18 (2003) 216-221. [25] J. Weissmtiller, R.N. Viswanath, D. Kramer, P. Zimmer, R. Wtirschum, H. Gleiter, Science 300 (2003) 312-315.
59
Crack tip strain rate equation with applications to crack tip embrittlement and active path dissolution models of stress corrosion cracking M.M. Hall, Jr. Bechtel Bettis, Inc., West Mifflin, PA 15122, USA
Abstract
Analytical equations for crack tip strain rate are needed for development of stress corrosion crack growth rate models. Considered here is the quasi-static case of a crack propagating at constant velocity under constant applied stress intensity factor. Crack tip solutions obtained by Hui and Riedel for high temperature creep crack growth are adapted for use at the low deformation temperatures associated with aqueous stress corrosion cracking. The low temperature dislocation creep model developed by Kocks and Mecking is used to establish the temperature dependence of the crack tip strain rate equation parameters. A creep damage rate criterion is introduced to obtain crack growth rate expressions for cracks propagating by crack tip embrittlement and active path dissolution mechanisms of stress corrosion crack advance. I. Introduction
There are models in the literature on aqueous stress corrosion cracking (SCC) that assume crack growth rate is dependent on crack-tip strain rate [ 1,2]. Although aqueous SCC occurs below temperatures ordinarily associated with creep, low-temperature dislocation creep [3] arguably makes a significant contribution to crack-tip strain rate as crack tip stresses are well above the yield stress. Therefore, low temperature creep must be considered in the development of SCC models. Perhaps due to a common perception that creep is a high temperature phenomenon, only, analytical expressions from the high temperature creep crack growth rate literature have not been considered for use in development of SCC models in the past. The crack tip strain rate solution derived by Hui and Riedel (HR) [4] for high temperature creep crack growth is adapted here for use in SCC modeling. The HR solution was obtained for creep cracks propagating at constant velocity in a spatially uniform and time-independent far field stress. This condition is attained approximately for small stress corrosion crack extensions in constant-load test specimens having a uniform or slowly varying far field stress. Of more practical interest are large crack
Modeling EnvironmentalAttack
60
extensions in situations where the applied stress intensity factor (K) is not constant, due to gradients in the far field stress and stress reductions that can occur with increasing crack length under fixed displacement conditions. As the HR solutions are valid for both transient and quasi-static crack growth, the HR solutions adapted here can be used to analyze situations where K is not constant if the crack growth analysis is approximated as a sum of a number of small crack extension steps at constant K. In order to apply the HR crack tip solution, we adopt the low temperature dislocation creep constitutive model developed by Kocks [5], Kocks and Mecking (KM) [6] and Estrin and Mecking [7]. We argue that the KM constitutive equation for high stress, low temperature dislocation creep is mathematically similar to the high temperature constitutive equation used by HR. As a consequence of this similarity, the HR solutions can be applied to development of SCC models that relate SCC crack growth rate to crack-tip strain rate. We argue further that in addition to a crack-tip strain rate expression, a fracture or damage rate criterion for crack advance is required to complete development of an SCC crack growth model. Crack growth rate equations are developed and applied to crack tip embrittlement mechanisms of crack advance, such as hydrogen embrittlement, and to active path dissolution mechanisms, such as the film rupture oxidation mechanism.
2. Equation development 2.1. The Kocks-Mecking (KM) const#utive model for low temperature creep Steady state creep occurs when strain hardening and recovery mechanisms occur at equal rates. At the low temperatures associated with aqueous SCC of nuclear reactor structural materials (T < 0.35 Tin) strain recovery and secondary creep occurs predominately at crack tips where stresses are well above yield. Although due to different mechanisms, both high and low temperature steady state creep rate can be described by mathematically similar power-law creep constitutive equations given by
~ = ~e + ~crp =---E.q-~o
(1)
In this equation, o~e and ~c,.p are the elastic and creep plastic components of strain rate, respectively, cr is the applied stress, Cro is a reference stress that for high temperature creep models is taken as the yield stress and in the KM low temperature creep theory is the fully strain hardened value of a structure parameter that is proportional to the dislocation density, ~o is the creep rate for an applied stress equal to Cro and m is the stress exponent. In the high temperature range, diffusion driven mechanisms of creep dominate [8] so that m is constant (-4) while o~o has an Arrhenius dependence on temperature with an activation energy equal to the activation energy for self-diffusion. In the low temperature range, dislocation mechanisms of creep dominate [8] so that o~o and or0 are independent of temperature and m = AH/RT where AH is an activation enthalpy that is characteristic of dislocation glide and recovery mechanisms [5].
Volume 1: Chemistry, Mechanics and Mechanisms
61
2.2. Hui and Riedel solutions Hui and Riedel derived crack-tip stress, strain and strain rate expressions for application to high-temperature, steady state creep crack growth. They assumed the non-hardening creep constitutive model given by Eq. (1) and formulated the crack tip problem in a coordinate system having an origin located at the tip of a crack moving with velocity ~i measured relative to a stationary coordinate system. Of interest is the total crack tip strain rate at a material point located at a distance r ahead of the approaching crack and having a coordinate x in the stationary coordinate system. The total strain rate referred to the stationary and moving coordinate systems, respectively, is given by
~(x,t)_(Og(x,t)] at
x'
tgt
/
r
tgt
/
,"
HR assumed quasi-static crack growth, that is, constant crack growth rate under constant applied K, and derived multi-axial expressions for the crack tip stress and strain distributions. For convenience of expression, their multiaxial equation notation is simplified here and the angular functions are absorbed into "equivalent" stress and strain variables, which we take here to represent the tensile components of the HR stress and strain directly ahead of the propagating crack. The HR stress and strain solutions can then be written as
/,,
,~l/(m-1)
or(r) - a O - o ~ , ~ )
;
/
s(r) - ( m
-l)~m
Oo T
/ Ekor )
l/(m-l) (3)
The crack tip strain rate can be obtained either by substituting the expression for crack tip stress into Eq. (1) or substituting the strain expression into Eq. (2):
~(l') - r
~ E~or 0"~ I m/(m-l),
(4)
where a = [0.29(m - 3)] l/(m-0. Notable features of the HR solutions are thot they are independent of time and that the stress, strain and strain rate in the near vicinity of the crack tip are not explicit functions of the far field stress (K). These results are a consequence of the quasi-static, constant velocity assumption and the fact that a is prescribed in the HR solution methodology. As discussed below, additional information specific to the crack advance mechanism is required to relate the HR solutions to the applied, K-dependent, stress field. The time-independence of the HR crack tip solutions means that there is no elastic component of the strain rate referred to the moving frame of reference, that is, s (r(r)/E = 0. Then the quasi-static strain rate is due entirely to creep, that is, ~i = 0 if O~rp = 0. Using Eq. (2) and given c3c(r)/c3t = 0 we must have
62
Modeling Environmental Attack
06(x,t)) _[06(r)I
(s)
The meaning of the first equality in this equation is that the time-dependent strain rate at a material point x in the stationary frame of reference is equal in magnitude to the quasi-static strain rate due to crack advance, the latter determined at a distance r = x - ~it from the moving crack tip. Moreover, considering the second equality, as a material point approaches the crack tip the strain rate due to crack advance is equal to the crack tip creep rate. This means that the rate of stress elevation due to crack advance is just balanced by the rate of stress relaxation due to creep. A further meaning of the latter equality is that sustained subcritical SCC crack growth under constant applied K is not possible in the absence of creep. 2. 3. Application of HR solutions to crack growth
To apply the HR equations to development of SCC crack advance models, the HR crack tip stress and strain must be related to the remotely applied loads. Through analytical and numerical analyses, Hui [9] and others [10,11] have considered this relationship for "creep brittle" crack growth, that is, for cracks growing at a rate comparable to the rate of growth of the creep zone. They showed that the distributions of crack tip stress and strain can be approximated by a "nesting" of crack tip solutions with the K-independent HR creep zone nearest the crack tip. Beyond the HR creep zone, the stress and strain fields approach solutions of the type derived by Hutchinson [12] and Rice and Rosengren [13] (HRR) for an elastic, strain-hardening plastic material, and beyond that is the linear elastic K-field. Figure 1 shows schematically this distribution of crack tip stress. 2.3.1. Criteria for sustained crack growth due to crack tip embrittlement Two criteria for sustained crack growth are introduced here to relate the HR solutions to the far field stress. First, as creep is necessary to sustain crack growth, the 1
L n cr
/ ,l ol K
HRR Ln r
Figure 1. Showing the distribution of crack tip stress within the creep, elastic-plastic and linear elastic zones.
Volume1:Chemistry,MechanicsandMechanisms
63
boundary between the HR and HRR zones is constrained to be near the outer boundary of the crack tip process zone, having radius rc, where the embrittlement and creep fracture processes operate. Creep in the zone beyond rc will be limited by crack advance, which in our assumption maintains pace with spread of the creep zone. We then choose to equate the HR and HRR stresses at r~. The stress and strain distributions within the HRR [ 13] elastic-plastic zone are given by
2N ( KJ IN+I cr( r ) = cr~ x ~
cr~
2 ( KJ IN+I
" c ( r ) = 6o x ~
cr~
(6)
where K j = ~/EJ and d is Rice's J-integral, N is the strain hardening exponent and fl isdependent on N but has a typical value in the range 1.4~rto 1.8st. The second criterion is that the creep rate within the creep fracture zone must exceed a sustained minimum or "critical" value in order to sustain crack growth. Applying the second equality of Eq. (5) and using the second of Eq. (6), this criterion can be written as
a ae(r)] >
a
e~
N+l
r~
(7)
where 6c is minimum strain required to fracture the crack tip volume of radius ro A further meaning of Eq. (7) is that, to sustain crack growth, the crack tip strain gradient must exceed a critical value. 2.3.2. Crack tip strain rate and crack growth rate equations
With these criteria and the HR crack tip equations we derive an expression for the crack growth rate. Equating the HR stress, Eq. (3), and the HRR stress, Eq. (6), at rc, using Eq. (4) and m = AH/RT, we obtain an expression for the crack tip strain rate at rc: 2N AH
where K o -~[flrr
o is a reference stress intensity factor. Using Eqs. (7) and (8) we
obtain an expression for the crack growth rate: 2N AH
N+l, or
,9,
This is a relatively simple, power-law crack growth rate expression in which the stress intensity factor exponent is a function of both temperature and strain hardening exponent, N, the latter of which is a function of the level of pre-strain for cold worked materials. A further feature of Eq. (9) is that the apparent activation energy Q is a function of the strain hardening exponent and applied stress intensity factor:
64
Modeling Environmental Attack
Olnh ] Q - -R
0lIT
K =
2N N +l
(10)
2. 4. SCC crack growth rate models for CTE and APD mechanisms o f crack advance
Turnbull [14], Ford [15] and Parkins [16] have reviewed the literature on stress corrosion crack advance mechanisms. The high temperature aqueous SCC mechanisms that have received greatest modeling attention can be classified broadly as crack tip embrittlement (CTE) and active path dissolution (APD) mechanisms [15]. Crack advance by hydrogen embrittlement (HE), which is the most extensively studied example of a CTE mechanism, is due to the fracture of a hydrogen-embrittled crack tip volume. Crack advance by the APD mechanism is due to metal dissolution, which may require fracture of crack tip oxides. 2.4.1. HE-CTE models In the HE SCC example considered here, thermally activated creep strain is required to initiate and sustain crack advance. This is because the hydrogen concentration within the fracture process zone is below the critical value necessary for the spontaneous brittle fracture commonly considered in stress-controlled HE fracture models of SCC [17]. Then application of the strain-controlled power-law crack growth rate equation, Eq. (9), to a strain-controlled HE fracture mechanism of SCC crack advance is straight forward with the addition of an expression for the effect of corrosion generated hydrogen on the crack tip fracture strain, c~. An example of a HE fracture strain model relevant to aqueous SCC is the one developed by Hall and Symons [2] as part of the hydrogen assisted creep fracture (HACF) model development. 2.4.2. APD models Crack advance by active path dissolution occurs due to electrochemical dissolution of metal along a preferred corrosion path. The usual assumption is that the rate of corrosion penetration (crack growth rate) is obtained by application of Faraday's law: ~t = ~ zpF
(1 l)
where M is molecular weight, ia is the anodic current density, z is charge of the metal cation in electron equivalents, p is density and F is Faraday's constant. The current density is ia - iaA* where A* is the effective area fraction of the crack tip surface that is fi'ee of oxide and ta is the anodic current density for the active, bare metal surface. As the crack propagates, the instantaneous current density will at once decrease with time due to repassivation of the bare metal surface and increase with strain due to strain induced fracture of the corrosion film. During an increment of time dt, the change in the crack growth rate, which is related to the net change in current density, is given by
Volume 1: Chemistry, Mechanics and Mechanisms
65
.,,,, ,,o= ,,,, zpF
+
lk
(oio/ k
a,,-]
j, u
dr.
(12)
Assuming that the instantaneous repassivation rate (first term, Eq. (12)) is proportional to the instantaneous current density and using Eq. (11), we have
(8i=] _ -k* zpF ?~. Ot )~ -k*i~
(13)
M
The oxide disruption rate (second term, Eq. (12)) is given by
] *':") :'" k.
(14)
j,
The repassivation time constant, which is the inverse of k* defined by Eq. (13), is a measure of the film repair rate in the absence of straining of the crack tip and y = c~A*/ c3c _... ~.~
5 M HzSO 4
..-
--..
0.01 M H2SO 4
EH~H*.5 M rt
-0.50
.:.. ~'....
EH/H,, 1 M
~.
'~"
...
a----- E ~ . . o.1.
9 .=.~...~. ~
H/H,, 0.01 M
-1.00
.
le-7
.
.
.
.
.
.
~ .
.
le-6
.
.
.
.
.
.
.
.
.
.
le-5 Current
.
.
.
.
.
".. ' ~ .
.
.
.
.
.
.
le-4
.
.
.
.
.
le-3
.
.
.
.
.
.
le--2
Density (Mcm2)
Fig. 1. Anodir E-log(i) data for PH 13-8 Mo exposed to various deaerated H2SO4 plus 0.1 M Fe§ 0.018 M Cr.3 and 0.01 M Ni§ solutions at 25~
86
Modeling Environmental Attack -1
icrit
~
-2
__1
-3 i~ass
~
-4
. . . .
-0.9
,
-0.7
. . . .
,
-0.5
. . . .
,
. . . .
-0.3 Potential
,
. . . .
-0.1
,
. . . .
0.1
,
0.3
. . . .
0.5
(VH~Hg2SO4))
Fig. 2. Polarization data entered in CREVICERv2 program. The curved line represents experimental data from PH 13-8 Mo adjusted so that ic~it,Epp, and ip~s are representative of the 5 M H2SO4plus metal salts E-log(i) data.
where ill2 is the cathodic HER current density (A/cm2), [H +] is the hydronium ion concentration (mole/L) and r/r is the H overpotential (V). The Tafel slope was ~160 mV/decade. 3.2. Hydrogen uptake on planar electrodes as a function of IR-corrected potential
Planar samples of PH 13-8 Mo charged in 5 M H2SO4 plus dissolved metal ions were analyzed to determine CH, adsorbed atter charging to achieve a homogeneous global H distribution (Fig. 3). The scatter bars in TDS results shown were determined by analyzing multiple samples charged in the same bulk solution under the same electrochemical conditions. CH is greatest at potentials cathodic to the reversible H potential as expected. CH increases with H overpotential; uptake at a charging potential of-1.10 Vngmg2so4 is 40 times greater than uncharged levels (uncharged levels were a fraction of a wppm). Generally, CH decreased as the applied potential became more noble. It is interesting to note that H absorption is observed above the reversible H potential. Figs. 1-3 also illustrates that considerable H uptake could occur in pits or crevices when ohmic voltage drop places the pit surface at a potential of-0.75 V even if the pit mouth were polarized to a more positive potential. 3.3. Pit scaling relations and H uptake in rescaled model pits
An example of local potential profiles calculated using the crevice modeling program (CREVICERv2) is shown in Fig. 4 at an applied potential o f - 0 . 4 7 5 Vag/ng2SO4. The electrochemical parameters used in the model are shown in Fig. 2. Fig. 4 shows that the local potential, E, is equivalent to the applied potential, Eapp, at the crevice mouth. However, ohmic potential drop decreases the local potential towards the
87
Volume 1: Chemistry, Mechanics and Mechanisms
deaerated OCP. This occurs over shallower depths as the crevice mouth dimension (gap) decreases. The local potential drops below EH/H+ at Xcntn = XHER and reaches the primary passivation potential, Epass, at a further depth xcnt2 = Xpass.
le+O
~E o n
E
Xcr~t,, a passive stainless steel is subjected to H production and absorption, but remains in the passive state. When the local potential drops far enough, the material undergoes a passive-active transition and H production is thermodynamically favorable on the actively dissolving surface (e.g,, x > xcnt2). From these results, relationships for two critical distances from the mouth of the crevice were developed. The first critical distance is the distance from the crevice mouth at which the pit potential equals EH/H+, Xeritl(EH/H+) = XHER-The second critical distance, xr was designated as the distance from the mouth of the crevice where the local crevice potential drops to Epass. Fig. 5 shows the anodic and cathodic half-cell reaction rates vs. depth x. The anodic reaction rate peaks due to the active polarization nose while the HER increases with potential decrease, and levels at the OCP of the pit bottom. However,/an can decrease with pit depth, as the active polarization curve indicates, when potentials reach the OCP
le-2 voO'ee "~ ~OoOO o
_ ~ ~ ~
~
o
i
0
v
9 o
v
,=
8
o
vv m Vv 9
o 1500pmGap 1000I~m Gap 500 I~m Gap 400 p.m Gap 300 p,m Gap 200 pm Gap
9
[]
~ O
v_Vv eem"vvVv
000
~
Oo ~
o go
o o
vvVv
mm
no
O
0 0
9 v o
le-4
o
m v v o
9 O 9 v 9 o
O
o
9 v v o
le-3
9
v 9
9
A E
9
9 r f ~ , " 6"wv
09o;o ~n~
(a)
II
O
~ v
VVVV 9 v V 9 VVVVVVvv VV 000000000001 9 9 0 0 IlullNli
u
,
,
5
10
15
Depth (ram) le-2
(b)
le-3 F
,.,oOQe e
"
E.
9 O 9 V 9 G
le4
1500pm Gap 1000prn Gap 500 pm Gap 400 pm Gap 300 pm Gap 200 pm Gap
le-5 0
5
10
15
20
Depth (mm)
Fig. 5. Current density vs. depth data showing (a) ia~o,iir and (b) the absolute value of/cathodic-The externally applied potential (Eapp)was-0.475 Vngmg2so4.
89
Volume 1: Chemistry, Mechanics and Mechanisms
of the pit. The relationships between Xcritl, Xerit2 and G for PH 13-8 Mo in 5 M H2SO4 plus metal salts are shown in Fig. 6. The square of each critical distance is, approximately a linear function of G. A plot of x vs. G (not shown) is not linear, confirming that x2/G is the correct geometrical scaling relationship for both the case of H evolution and passive-to-active transitions in this system. The x-G relation is critical to translating actual pits of micron scale to manufactured pits of millimeter scale [ 10].
120
I O0
9
x=e
0
Xeatl ' .......
Xca
t~
E r
EH/H.9 Ep,,ss > E
,,,,
X~
~
EH,IH+> E > Ep==
E > EH~H.> ,,,,
0
.
.
.
.
,
300
.
.
.
.
l
.
.
.
.
600
l
900
.
.
.
.
E~= l
1200
.
.
.
.
1500
Gap (~m)
Fig. 6. Xerit2 VS. gap data. Two x~it parameters are plotted for conditions: icrit= 10-2 A/cm2, EHm+= -0.59 Vng/Hg2S04,Epass =-0.70 VHg/Hg2SO4, Eapp --0.475 VHg/Hg2SO4.
Guided by Fig. 6, model pits/crevices of millimeter scale (e.g,, 1500 Bm inside cylindrical diameter, G, and 40 mm length, x) were produced using PH 13-8 Mo. Figure 6 shows that this is equivalent to a 60 ktm diameter crevice with x > 3 mm, etc. It can be seen that rescaled model pits possess xc~itland xc~it2over macroscopic lengths from 5-10 mm. Hence, different corrosion and H uptake behavior would occur over lengths of millimeters instead of micrometers typical of actual pits. These samples were held in 5 M H2SO4 plus dissolved metal ion under potentiostatic control. Local potential was measured vs. depth x. Potentials were held at potentials ranging from -0.4 to --0.60 VH~g2so4. IR drop was insufficient at high passive potentials such as -0.40 Vugmg2so4 to produce significant potential drop inside one-dimensional pits because pits walls exhibited low anodic current densities that did not produce enough ohmic voltage drop to activate stainless steel surfaces. However, when Eapp was decreased to -0.60 VHg/Hg2SO4, significant potential drop was seen vs. x such that the bottom of the pit approached the limiting or OCP near -0.76 V (Figs. 1 and 2). Rescaled pit specimens were exposed, removed, rinsed, stored in LN2 and sectioned radially for H analysis versus depth x. Fig. 7 shows the potential profile experimentally measured using the micro-reference electrode for a sample held in 5 M H2SO4 plus dissolved metal ions at -0.60 Vngmg2so4 (applied potential) with dimensions of 1500 ~tm inside diameter and 40 mm length. The pit exhibits a 150 mV potential drop from its mouth to its bottom. Most
90
Modeling Environmental Attack
of the potential drop occurs in the first 10 mm of depth, consistent with the modeling results (Fig. 4). An indication of anodic dissolution as a function of x, is recorded by examining the increase in inside diameter of the pit. This increase in pit diameter indicates increased anodic dissolution as a function of x due to activation, The decrease in diameter back towards the original value at the pit bottom is consistent with a decrease in ian at the OCP near the bottom of the crevice (Fig. 5).
-0.58
"t
: ]~~
EH,H*
-0.60
9
Potential (VHc~SO4)
9
TDS Measured CH
9 10 9
o Predicted C H --D-- Pit Inner Diameter
~&
2400
2200
-0.64
9
o00OO
OQ
0000
O0
0000
~
2000
~=
E a 1800 ~
|
-0.68 "~
v
6
>i -0.66 ID
0
9
O
-0.70
5
._
-4
O "l-
Ep==
-0.72
z
43.74 0 0
10
20
1600
1400
30
Depth (mm)
Fig. 7. Potential, Ca and pit diameter from a rescaled experimemalpit for a 1500 Hm gap by 40 mm depth cylinder exposed to 5 M H2SO4plus dissolved metal ions. Eapp=-0.60 Vagmg2so4.
The decrease in E and increase in ica produce an increase in CH with depth in the crevice (Fig. 7). The local level of CH within pits increases even when uptake is minimal on external passive surfaces. Independent corroboration is obtained when CH predicted from the planar electrode studies at each interfacial potential is plotted (Fig. 7). The predicted values from the bulk electrode studies are consistent with TDS measured values. The local CH within the model pit increases nearly 100% from approximately 4 times an uncharged specimen near the pit mouth to almost 7 times that of an uncharged specimen at x greater than 10 mm within the pit where potential is lowest and cathodic current density isat its highest value. Therefore, a planar electrode held just in the passive region at -0.6 V would be expected to produce and concentrate H in pits or crevices at depths greater than x > XHER- The analysis described can be repeated at various Eapp to provide indication of the potential dependency of CH as a function of x for occluded sites of different geometry. 4. Discussion
PH 13-8 Mo exhibits an active-passive transition in a simulated pit/crevice solution. Moreover, this alloy exhibits a deaerated OCP below EH/n+ (Figs. 3 and 5). Given this behavior, there are a variety of ways that a pit, crevice or crack that experiences IR
Volume 1: Chemistry, Mechanics and Mechanisms
91
drop, acidification and 02 depletion could lead to H uptake under conditions where boldly exposed planar PH 13-8Mo surfaces would be expected to remain passive and above EWH+. H uptake was observed in 5 M H2SO4 plus dissolved metal salts at the deaerated OCP (Fig. 7) as well as when a rescaled pit was held in at -0.6 V in the passive region. Here, ohmic voltage drop decreased the local potential below EWH+at x > XHEReven though the applied potential was above the passive-active transition. These data help to explain the finding that PH stainless steels are susceptible to H embrittlement at OCP or under anodic polarization in chloride solutions, especially when pitting or crevice corrosion occurs [17,18]. H production has also been observed in model pits in iron [19], carbon steel [19] and a duplex stainless steel [20] when x > XHERdespite external polarization to noble potentials in the passive region. H uptake has also been reported in the literature during permeation experiments on anodically polarized planar electrodes [21] under pitting conditions [22], as well as during intergranular corrosion [23]. At issue is accurate quantification of local H concentrations in such corrosion sites. The use of rescaled occluded sites in conjunction with TDS provides the opportunity for quantification of H levels over a range of applied potentials. Ultimately, C. values determined in this manner could be compared to CHcrit values that produce a large drop in KIr Moreover, the potential and x2/G combinations that produce HEAC could be predicted from identification of the conditions where C. > CHcrit. Modeling shows that IR drop in a crevice can produce local crevice potentials that enable activation of the metal surface within the crevice concurrent with H production and uptake. The relationship x2/G = constant applies for Ep~ss and EWH+ albeit with different constants. Such scaling relationships have been seen before [10,14,15]; the exact values of the constant in this paper are specific to PH 13-8 Mo. The x2/G = constant relationship has been previously used to describe activation phenomena (similar to the IR* model of Pickering) [10,16,19,20]. In this study both XcritI and Xcrit2 have been specified, although such relations should hold at any potential. Different regimes have been identified for crevices or pits in PH 13-8 Mo. There is a combination of x and G in a crevice or crack where the stainless steel would be passive and remain above EH/H+, a region where H production could occur but not activation, and a deep crevice depth (xc~it2)beyond which both activation and net H production occur. Here, H absorption was greatest (Fig. 7), charging efficiency was greatest (not shown) and nuclear reaction analysis (not shown) indicated very high local hydrogen concentrations within fractions of micrometer depths from a corroding surface undergoing concurrent hydrogen production. A variety of alloys could exhibit such generic behavior albeit with different details. The requirements include pit susceptibility, a passive-to-active transition in a solution analogous to the pit chemistry, an OCP in the active deaerated state that is below the EHm+ and x > XHERat appropriate G. The relationship described can provide guidance for construction of model pits of a large enough scale to measure local potentials and CH electrochemically identical to an actual crevice of micrometer scale too small for interrogation for CH. The model pit results in Fig. 7 show that H uptake can occur under conditions where the rescaled pit mouth is held in the passive range when x > xmt. The local CH values (with mm scale spatial resolution) increase to 7 times the value in an uncharged specimen exposed to humid air and could not be predicted from global H measurements on planar specimens containing pits or crevices.
92
Modeling Environmental Attack
5. Conclusions Exposure of a PH stainless steel to a simulated pit environment confirms that H uptake occurs locally on anodically polarized planar electrodes containing an occluded site when x > XHER. Such occluded sites experience sufficient IR drop to shift local potentials below EWH+ even when external surfaces are polarized to passive potentials. Scaling laws can be used to rescale pits to sizes that enable spatially resolved quantification of local H concentrations as a function of both Eapp and pit depth. The x2/G scaling relationship described both passive-to-active potentials and EWH+. The rescaling technique can be applied to study HEAC susceptibility vs. applied potential for various occluded site geometries with and without candidate crack tip inhibitors.
Acknowledgments This work was funded by the Office of Naval Research under Grant No. N0001403-1-0029. Jason Lee was supported by NRL J.O. No. 73-5052-13, NRL publication NRL/PP/7303/04/0003. Helpful discussions with R.G. Kelly are gratefully acknowledged.
References [1] C.F. Baes, Jr., R.E. Mesmer, The Hydrolysis of Metal Cations, R. Krieger Publishing Co., Malabar, FL, 1986. [2] B.F. Brown, in: R.W. Staehle J. Hockmann, R.D. McCright, J.E. Slater, (Eds.), Stress Corrosion Cracking and Hydrogen Embrittlement of Iron Base Alloys, NACE, Houston, TX, 1977, pp. 747-750. [3] K.R. Cooper, R.G. Kelly, J. Chromatogr. A850 (1999) 381-389. [4] H. Okada, in: J. Hockmann, J.E. Slater, R. W. Staehle (Eds.), Stress Corrosion Cracking and Hydrogen Embrittlement of Iron Base Alloys, NACE, Houston, TX, 1977, pp. 124-134. [5] E.A. Taqi, R.A. Cottis, in: A. Tumbull (Ed.), Corrosion Chemistry Within Pits, Crevices and Cracks, HMSO, London, 1987, pp. 483-494. [6] C.T. Fujii, in: H.L. Craig, Jr. (Ed.), Stress Corrosion Cracking-New Approaches, ASTM STP 610, ASTM, Philadelphia, PA, 1976, pp. 213-225. [7] W.R. Cieslak, R.E. Semarge, F.S. Bovard, in: A.D. Romig, W.F. Chambers (Eds.), Microbeam Analysis 1986: Proc. 21st Annual Conference of the Microbeam Analysis Society, San Francisco Press, San Francisco, CA, 1986, pp. 303-306. [8] L.M. Young, M.R. Eggleston, H.D. Solomon, L.R. Kaisand, Mater. Sci. Eng. A203 (1995) 377-387. [9] P.S. Tyler, M. Levy, L. Raymond, Corrosion 47 (1991) 82-87. [10] H.W. Pickering, in: G.S. Frankel, J.R. Scully (Eds.), Research Topical Symposium on Localized Corrosion, NACE International, Houston, TX, 2001. [11] S.W. Smith, Ph.D. Dissertation, University of Virginia, Charlottesville, VA, 1995. [12] S.W. Smith, J.R. Scully, Metall. Mater. Trans. A, 31A (2000) 179-193. [13] J.R. Scully, J.A. Van Den Avyle, M.J. Cieslak, A.D. Romig, Jr., C.R. Hills, Metall. Mater. Trans. A, 22A (1991) 2429-2444. [ 14] R.G. Kelly, in: G.S. Frankel, J.R. Scully (Eds.), Research Topical Symposium on Localized Corrosion, NACE International, Houston, TX, 2001. [ 15] Y. Xu, H.W. Pickering, J. Electrochem. Soc. 140 (1993) 658-668. [16] B.G. Ateya, H.W. Pickering, J. Electrochem. Soc. 122 (1975) 1018-1026.
Volume 1: Chemistry, Mechanics and Mechanisms
[17] [18] [19] [20] [21] [22] [23]
93
C.S. Carter, D.G. Farwick, A.M. Ross, J.M. Uchida, Corrosion 27 (1971) 190-197. G. Bumell, D. Hardie, R.N. Parkins, Brit. Corros. J. 22 (1987) 229-237. K. Cho, M.I. Abdulsalam, H.W. Pickering, J. Electrochem. Soc. 145 (1998) 1862-1869. J.N. AI-Khamis, H.W. Pickering, J. Electrochem. Soc. 148 (2001) B314-B321. C.F. Barth, E.A. Steigerwald, A.R. Troiano, Corrosion 25 (1969) 353-358. R.J. Gest, A.R. Troiano, Corrosion 30 (1974) 274-279. A. Sehgal, B.G. Ateya, H.W. Pickering, Acta Mater. 45 (1997) 3389-3399.
95
Modelling of the effect of hydrogen ion reduction on the crevice corrosion of titanium Kevin L. Heppner, Richard W. Evitts Department of Chemical Engineering, University of Saskatchewan, 57 Campus Drive, Saskatoon, SK S7N 5C5, Canada Abstract
In this research, the effect of the level of hydrogen ion reduction on a titanium crevice immersed in 0.5 M NaC1 solution at 70~ was investigated. The fraction of the total dissolution current supplied by hydrogen ion reduction, qJ, was varied from 0 to 0.8 in increments of 0.2 and the steady state pH, conductivity (to), and iR drop (O) profiles in the crevice were calculated. The mass transport model of Watson and Postlethwaite was employed and the crevice was assumed to be passivated. As q~ increased, the solution conductivity and the iR drop along the crevice were significantly reduced while the pH showed only a slight increase. A plot of total iR drop along the crevice length, pH at the crevice tip, and solution conductivity at the crevice against qJ was constructed and the following relationships were obtained (R 2 > 0.995):
pH = 0.0091W + 1.27 = -0.919q j + 97.0 ~c= -0.0044q j2 - 0.399qj + 157 Of the three solution properties studied, the pH showed the weakest dependence upon the hydrogen ion reduction fraction, varying linearly from 1.27 to 2.01 as q~ varied from 0% to 80%. The iR drop along the crevice length showed the greatest sensitivity to the hydrogen ion reduction level. I. Introduction
Crevice corrosion is a dangerous form of localized corrosion often initiated by environment induced cracking. Once a crack is initiated, localized attack is at least partially sustained by crevice corrosion or similar phenomena. Titanium is a metal which, under normal conditions, will resist corrosive attack quite well. However, it is susceptible to crevice corrosion. It is known that hydrogen evolution is a significant cathodic charge contributor in a titanium crevice system [1,2]. The extent to which
Modeling Environmental Attack
96
hydrogen evolution occurs in a titanium crevice is unique to this metal and can account for up to 80% of the total cathodic charge [1]. Heppner et al. [3] recently published a crevice corrosion model which accounted for the effect of internal hydrogen evolution on the crevice corrosion incubation period but did not investigate the effect of H+ reduction in detail. The focus of the current study is to predict the overall effect of hydrogen ion reduction upon the crevice corrosion process. By varying the percentage of cathodic charge contributed by hydrogen evolution, the magnitude of the effect of hydrogen evolution upon mass transport through the crevice solution can be determined. The rate of anodic metal dissolution is equal to the combined rate of external (to the crevice) oxygen reduction and internal hydrogen ion reduction in a differential aeration cell. In the absence of internal crevice cathodic reactions, the entire current leaving the anodic crevice walls and entering solution as an ionic mass flux would induce electromigration transport through the crevice solution. However, hydrogen evolution near the anode consumes a portion of the current, thereby reducing electromigration transport rates in the solution. Reduced mass transport decreases the rate of chloride ion influx. However, reduced mass transport rates also hinder the transport of hydrogen ions out of the crevice. Yet, hydrogen evolution consumes hydrogen ions in solution and counteracts the effects of chemical hydrolysis of Ti 4+ ions which produces H+. Overall, hydrogen ion reduction increases the crevice solution pH and therefore reduces the magnitude of the passive current in the crevice. The coupling between passive current and pH is autocatalytic and is a primary cause of the development of aggressive solutions in a crevice. Because hydrogen ion reduction directly affects the system pH, it has a large impact upon the entire crevice corrosion process. In this work, the effect of hydrogen evolution on mass transport in the crevice solution will be examined in detail. 2. The mathematical model
Watson and Postlethwaite [4] developed a model of mass transport applicable for dilute solutions. This model has been shown previously to adequately represent transport in a crevice solution [5,6] and is therefore used in this study: c3G DiV =2Gat
ziDiF ( vG " i +tr + 6--Ci) +
(1)
where the charge density (6) is defined as: 6 = F~,zjCj
(2)
J and the diffusion potential current density (idp) is:
idp = F~_,zjDjVCj J
(3)
The current density (i) is determined from the passive current integrated along the crevice length:
97
Volume 1" Chemistry, Mechanics and Mechanisms
,~pip (x)dA m(x) i(x)
(4)
Acs(X)
The source term G, in Eq. (1) represents the influence of chemical reaction kinetics upon the concentration profile in space and time. In this study, reaction kinetics is assumed to occur much faster than mass transport processes. Therefore, chemical equilibrium is assumed at all times, an assumption which mathematically decouples mass transport and chemical reaction processes. With this assumption in place, the effect of chemical reaction is represented as a set of chemical equilibrium expressions for the hydrolysis of Ti (IV) [7]:
Ti 4+ + H 2 0 --->TiOH 3+ + H +
(5)
TiOH 3+ + H 2 0 --->Ti(OH)~ + + H +
(6)
vir
(7)
+ + .2o
vir
+ 14 +
Ti(OH)~ + H 2 0 --->Ti(OH)4 + H +
(8)
Chemical equilibrium expressions, Eqs. (5)-(8), are augrnented with a set of species mass balances. The resultant set of algebraic equations is solved using the NewtonRaphson method. Electrode reactions which are assumed to occur inside the crevice are the dissolution of Ti (IV) (anodic) and the reduction of hydrogen ions (cathodic):
Ti ~ Ti 4+ + 4e-
(9)
2H § +2e- ~ H 2
(10)
At the bold surface surrounding the crevice mouth, oxygen reduction is assumed to OCCur."
02 + 2 H 2 0 + 4e- ~ 4011-
(ll)
In this model, only the crevice interior is modelled; the rate of mass transfer of oxygen to the bold surface is considered to be non-limiting. The time required to deoxygenate the crevice is calculated via the following equation:
ldeoxy =
4 FAcs LCo2
ipAm
(12)
98
Modeling Environmental Attack
The rate of anodic and cathodic processes must be equal by mixed potential theory [8] and are proportional to the passive current. The passive current is adjusted for temperature using an Arrhenius type expression and pH using a Freundlich isotherm [9]: i p = i p,ro exp --~-~-
I E~]
(13)
log(i p ) = log(k ) - n. p H
(14)
The activation energy for Ti (IV) is 12700 J/mol [2]. Values of n are based upon the electrical charge of the dissolving metal ion and can be interpolated from values given in Ref. [9]. Use of both equations requires the specification of a reference passive current density, temperature, and pH. For Ti, a reference passive current of 0.1 ~tA/cm2 is used at 25~ and a pH of 7. This value is interpolated from experimental data obtained by Griess [2]. 2.1. N u m e r i c a l solution o f the m o d e l
The mass transport model, Eq. (1), was solved using a f'mite volume method. The electromigration term was discretized using an upwind formulation, i.e.:
dr
2
xi+1 - x ,
2
x i -x,_ l
where:
zli+i .)
-Iz(i+i ,l
(16)
Thus, the direction of the one-sided finite difference expression depends upon the direction of ionic electromigration. Use of an upwind formulation avoids the possibility of negative transport coefficients and therefore ensures that a more realistic solution of the transport equation will be obtained [ 10]. One boundary condition for the mass transport model was provided by the magnitude of the passive current at the crevice tip: i + idp ziuiFC i ~ x
dC i
Di ~
d.x
ip
=~
ziF
(17)
Eq. (17) is a pseudo steady state mass balance at the metal-solution interface. The right hand side of Eq. ,(17) is the corrosion-induced ionic flux. At the crevice mouth, the bulk concentration was assumed: C i =C~ utk
(18)
99
Volume 1" Chemistry, Mechanics and Mechanisms
Fig. 1 shows the finite volume grid over which the transport equation is solved. Fig. 2 displays the solution algorithm via a flow sheet. i+id~
z~u~FC~
K
c,-c~ ~
ac,_ i~ - ~
......
d~
%F @aaemo6
.
.
:
:
:
:" 0
.: 0
.: 0
~9 0
i9 0
9:
9:T
"l: T . , i.:, , :~ li
aaeooo
t
:
:
-
~
.:
:.9 0
.:9 0
a
" 9
: 0
" 0
':
T"I9"
I " ."9
: 0."
:
Moluth
9
..... :: Tip
Each Small Arrc~ R e p r e s e n t s a C~rosion-lnduec~t Mass Influx of Metal Ions
Fig. 1. Finite volume grid over which transport equation is solved.
G e t I n p u t from U s e r
I
~ -'~'~o~o-y
-~ ~no
1
~
,]
~.~,~o~,~o~
,"'!
'"
'-!
]
,iLlr
~~
~=~ § ~,
I..........SoI~oM--
'
I
~
ou~ ~ = , ~
"'
,,
D~a
"t
No
Fig. 2. Flow sheet describing the computational algorithm.
3. Results and discussion He et al. [ 1] recently performed an experimental study upon a high purity titanium crevice immersed in an acid chloride solution at 70~ By comparing the weight loss
1O0
Modeling Environmental Attack
measurement with the integrated current-time graph obtained during the course of the experiment, the relative influence of hydrogen evolution upon the anodic current was determined. He et al. [ 1] determined that approximately 70% of the cathodic charge is carried by hydrogen ion reduction. This measurement was the basis of selection of the value of q~, the fraction of cathodic charge carried by hydrogen evolution, for the present model. While the effect of hydrogen ion reduction upon metal dissolution is experimentally measurable, its effects upon many properties within the crevice are not readily obtained. For this reason, the current study was undertaken. At the start of the crevice corrosion process, the deoxygenation of the crevice solution causes the formation of a differential aeration cell. Once a differential aeration cell exists, a strong acid-chloride solution may develop in the crevice. The measurement of He et al. [1] gave an indication of the time-averaged influence of hydrogen ion reduction. However, at the beginning of the experiment when the solution pH was relatively high and oxygen was available, the portion of the cathodic charge carried by hydrogen ion reduction was likely quite low. As the crevice solution became more acidic and oxygen was depleted, the dependency of the anodic current upon hydrogen evolution, due to the mixed potential, would have increased substantially. Therefore, the portion of the anodic current supplied by hydrogen ion reduction likely varied between approximately 0 and 80%. This is the range of values of W used in this study.
For each simulation in this work, a titanium crevice measuring 1 ~tm across the gap, 1 cm wide, and 1 cm deep was used. It was immersed in 0.5 M NaCI solution at 70~ Fig. 3 illustrates the effect of qJ on the pH, iR drop, and conductivity of the crevice solution. From Fig. 3, as the portion of the anodic current supported by hydrogen ion reduction was increased from 0 to 0.8, the iR drop in solution along the crevice length decreased. This is because the net current density or net charge flux flowing from the anodic crevice wall decreases as qJ increases. The net current density at any point in the crevice is the difference between current supplied by metal dissolution and current consumed by hydrogen ion reduction, i.e.:
lp = ip(1- V) 9n e t
(19)
The current density leaving the crevice wall adds to the current density flowing through solution, and thus, the magnitude of the current density in solution depends on the net anodic current. The addition of current released from the crevice wall to the current flowing through solution is analogous to tributaries adding to a river. Further examination of Fig. 3 reveals that as the iR drop decreases, the solution conductivity also decreases. This indicates that the magnitude of the current density flowing through solution is more sensitive to W than conductivity is to W. This can be clarified using Ohm's law for a system of uniform composition: w,
= --
i K"
(20)
Because conductivity is in the denominator of Eq. (20), as it decreases it will tend to increase the potential gradient. Therefore, for the potential gradient to decrease, the current density must decrease to a greater extent than the conductivity.
Volume 1" Chemistry, Mechanics and Mechanisms
101
Fig. 3 also shows that as the rate of hydrogen ion reduction increases inside the crevice, the hydrogen ion concentration decreases and, thus, the pH increases. While hydrogen evolution consumes H § thereby increasing the pH, it also reduces the rate of electromigration which tends to increase the H § concentration in the crevice solution. Due to this competing effect, the pH is the property (of those studied in this work) which shows the least dependency upon the rate of hydrogen ion reduction. Furthermore, the iR drop is the solution property that is most sensitive to changes in q~.
200
8 7 6 5
(a) ~ ...............
160
8 7-
150 E to
'7, 9"
6-
(b)
120 E to ~ . "7, >
m
'
100 ~v vie Ica.4
~ 4 ~--" 3 2
>'~-
3-
50
~=>c~ ~rr
21-
0
c 0 0
0
1 !
i
5
i
i
10 15 20 Time (h)
25 160
(c)
_
.
65-
120 to IE -r, ~-
# i
~4-
~-*
.
.
.
.
.
.
.
.
.
.
~,Z" ~ ~= >c~ ~ t. m
40 i
0
5
i
i
0
i
10 15 20 Time (h)
O
O
25
E to
~4-
10 15 20 Time (h)
8 765 3 2 1 0
> 121 "~=
0
c 0 0
25
Yo-
140
(d)
,q-,
E
-105
to
-70
~ E
-35
> 121 ~ := ,,~
i
5
i
i
i
10 15 20 Time (h)
0
25
o~
c 0
0
pH ---Conductivity
- 30 i
0
i
40
E
"7
90
----7
3210
5
i
~
120
(e)
_
65-
0
i
80
i
3210
i
80
5
i
i
0
i
10 15 20 Time (h)
25
:= > a""
Ya:
- -iR Drop
c 0 0
Fig. 3. Predicted transient iR drop along the crevice length, pH at the tip, and conductivity at the tip in a titanium crevice measuring 1 lam across the gap, 1 cm deep, and 1 cm wide and immersed in a 0.5 M NaCI solution at 70~ (a) 0% hydrogen evolution; (b) 20% hydrogen evolution; (c) 40% hydrogen evolution; (d) 60% hydrogen evolution; (e) 80% hydrogen evolution.
Modeling Environmental Attack
102
In addition to its direct affect upon solution properties, the value of ~F also affects the kinetic behaviour of the system. The conductivity is calculated from the concentration of each ionic species in solution and is therefore an indicator of the rate of change of the crevice solution composition with time. Conductivity is therefore used as an indicator of the time to steady state onset. Examination of the conductivity profiles displayed in Fig. 3 reveals that, as ~F increases, so does the time required for the system to reach steady state. As ~F increases from 0 to 0.8, the time required for the system to reach a steady state condition increases from approximately 5 to approximately 18 h. Fig. 4 shows the variation of pH and conductivity at the crevice tip, as well as iR drop along the crevice length as the value of qJ is increased. Each solution property showed a distinct linear or parabolic variation with q'. The data points were fit to a best fit line and each correlation fit the data with excellent accuracy (R 2 > 0.995):
pH = 0.0091q~ + 1.27
(21)
r =-0.919q ~ + 97.0
(22)
(23)
x = -0.0044W 2 - 0.399qj + 157
a~
2.5
(a)
2
(b)
6
~- 1.5
a0 60
~e
cl:i
-r"
120
o
1
.,,...
40
-....
0.5 1
i
i
a m
i
0
. g
20 40 60 80 % Hydrogen Evolution, ~t'
0
100
200
i
!
i
!
20
40
60
80
% Hydrogen Evolution, ~
100
(c)
.-.150 9"='- 100 >'7 "~ o "o rO
o
v ~u 0 0
i
I
i
i
20
40
60
80
1 O0
% Hydrogen Evolution, 9 Fig. 4. Variation of steady state solution properties with qJ: (a) pH at the crevice tip; (b) iR drop along the crevice length; (c) conductivity at the crevice tip.
Volume 1: Chemistry, Mechanics and Mechanisms
103
While the pH and the iR drop varied linearly with ~F, conductivity showed a parabolic dependence. For pH to have linear dependence upon ~F, the ratio of the rate of production of H+ via chemical hydrolysis to the rate of transport of H+ out of the crevice must remain constant for varying values of W. However, for the conductivity to show concave-down parabolic dependence upon ~F, the overall rate of transport of ions out of the crevice must outpace the rate of internal cationic production as ~F increases. Thus, the dependency of these parameters upon ~F gives insight into how the overall system adjusts to internal hydrogen ion reduction.
4. Conclusions Based upon the results of the preceding work, the following conclusions can be made: 1. Internal hydrogen ion reduction greatly impacts the crevice corrosion process in a titanium crevice. It has a marked effect upon the pH, conductivity, and iR drop in the crevice solution. 2. A linear dependence of pH upon q~ is observed. This implies that, as q~ is varied, the ratio of the rate of H + production due to chemical hydrolysis of Ti 4+ to the rate of H+ transport out of the crevice remains approximately constant. 3. A concave-down parabolic dependence of conductivity upon q~ is also observed. This result suggests that the net rate of ionic transport out of the crevice is more sensitive to changes in q~ than the rate of anodic metal dissolution is to changes in q~. Because mass transport out of the crevice decreases the conductivity while metal dissolution increases the conductivity, the conductivity decreases significantly with increasing q~. 4. Changing the value of q~ causes the transient behaviour of the system to change. As the value of q~ increases from 0 to 0.8, the time to steady state onset of the system changes from approximately 5 to approximately 18 h. 5. As q~ increases, both the iR drop and the conductivity in the crevice decrease. Therefore, the dependence of current density flowing through solution upon W must be greater dependence than the dependence of the conductivity upon q~. 6. Contrary to expectations, of the solution properties studied in this work, pH shows the least dependency upon the rate of hydrogen ion reduction. This is due to a competing effect. As q~ increases, the rate of consumption of H+ at crevice-internal cathodic sites increases while the rate of electromigration mass transport decreases. While H+ consumption obviously increases the pH, decreased mass transport rates reduce the rate at which H+, which is produced via Ti4§ hydrolysis, electromigrates out of the crevice solution, thereby decreasing the pH. Overall, as ~F increases, the crevice solution pH increases.
Acknowledgements The authors are grateful for financial support from the Natural Sciences and Engineering Research Council (NSERC) and the University of Saskatchewan and for the use of computer facilities at the University of Saskatchewan.
104
Modeling Environmental Attack
References [ 1] [2] [3] [4] [5] [6] [7] [8] [9]
X. He, J.J. NoEl, D.W. Shoesmith, Corrosion 60 (2004) 378. J.C. Griess, Jr., Corrosion 24 (1968) 96. K.L. Heppner, R.W. Evitts, J. Postlethwaite, Corrosion 60 (2004) 718. M.K. Watson, J. Postlethwaite, Corrosion 46 (1990) 522. K.L. Heppner, R.W. Evitts, J. Postlethwaite, Can. J. Chem. Eng. 80 (2002) 849. K.L. Heppner, R.W. Evitts, J. Postlethwaite, Can. J. Chem. Eng. 80 (2002) 857. A. Liberti, V. Chiantella, F. Corigliano, J. Inorg. Nucl. Chem. 25 (1963) 415. M. Stem, L. Geary, J. Electrochem. Soc. 104 (1957) 56. L.L. Shrier, R.A. Jarman, G.T. Burstein (Eds.), Corrosion, vol. 1, 3rd ed., ButterworthHeinemann, Oxford, 1994, p. 2:20. [ 10] S.V. Patankar, Numerical Heat Transfer and Fluid Flow, Hemisphere, Washington, 1980.
Nomenclature C D F G
Molar concentration, (mol/m 3) Diffusion coefficient, (m2/s) Faraday's Constant, (96487 C/mol) Rate of production or consumption via reactions, (mol/s) Current density, (A/m2) Parameter in Eq. (14) Parameter in Eq. (14)
R T t u
x z
Universal Gas Constant, 8.3145 (J/mol. K) Temperature, (K) Time, (s) Mobility, (m2mol/J-s) Spatial coordinate, (m) Charge number
Greek letters a 6 9
Upwind parameter Charge density, (C/m3) Permittivity, (Farad/m) Electrical potential, (V)
K"
q~
Conductivity, (~-lm-i) Fraction of cathodic charge carried by hydrogen evolution
Superscripts and subscripts bulk dp
i
Bulk solution Diffusion potential Index
J net P
Index Net Passive
105
Transport effects in environment-induced cracking A.I. Malkin Institute of Physical Chemistry, Russian Academy of Sciences, 31 Leninsky Prospect, Moscow 119991, Russia Abstract
Hydrodynamic effects in environment-induced cracking were subjected to theoretical analysis. In order to explain for kinetic regularities of crack evolution, semi-empirical models were suggested. In addition, two possible rate-determining mechanisms of crack growth during liquidmetal embrittlement and stress-corrosion cracking were considered: (i) the cavitation mechanism, and (ii) the mechanism of reduction of local stress at the crack tip. The crack growth rate at the steady-rate stage was calculated for both mechanisms. Mathematical representations of the delay time and tensile stress of crack initiation were proposed. I. Introduction
In the past few decades, considerable advances were made in identifying kinetic regularities of environment-induced crack growth. The problem has been approached both theoretically and experimentally [ 1,2]. Nevertheless, the present understanding of rate-determining processes is far from complete. In theoretical studies of environmentinduced cracking (EIC), two aspects must be considered. The first concerns the micromechanism(s) of crack growth, i.e., description of the elementary fracture event and the preceding processes localized at the crack tip. The second aspect entails a description of the transport of liquid environment to the crack tip. Although the first aspect of the problem is least understood and most important (from a chemical perspective), certain experimentally-observed regularities are apparently functions of the transport of environment. Crack growth rates depend on the local stress and fluid conditions in the vicinity of the crack tip. The local stress depends on stress distribution at crack faces, whereas fluid parameters (including the distribution of fluid pressure inside the growing crack) are controlled by the rates of transfer processes and crack growth. Therefore, the problem of crack growth due to the effect of the liquid environment must be considered self-consistent. Surprisingly, the self-consistent approach is usually ignored. Transport processes, as applied to EIC, have been discussed for liquid-metal embrittlement (LME), stress-corrosion cracking (SCC) and surface crack growth under
Modeling Environmental Attack
106
the influence of cutting fluids [1-4]. A tentative analysis within a self-consistent context has already been made [5]. Here the objective is to evaluate the effect of non-local hydrodynamic interactions between liquid environment and solid on crack kinetics (Note: "Non-local interaction" means a contact interaction outside the tip of crack.) Using the regularities that were observed experimentally, semi-empirical models for liquid environment-induced crack propagation are developed in this study. Within the framework of these models, a phenomenological description of fractures was combined with an accurate description of transport processes. Two possible rate-determining mechanisms stemming from nonlocal interaction are considered: the cavitation mechanism (i.e., loss of continuity of liquid), and the mechanism of reduction of local stress at the crack tip.
2. Experimentally-observed regularities In the majority of kinetic experiments, the parameter being measured is timedependent crack length. Experimental results are usually presented as a "kinetic diagram" (Fig. 1), which represents the relation between the crack growth rate and stress intensity factor (SIF) at given temperature, pressure and chemical composition of the environment [ 1]. The SIF-dependence of crack growth rate usually incorporates four clearly-defined regions. Region I is the zone of subliminal values of SIF corresponding to the lack of crack growth. Region II corresponds to the environment-induced crack growth on the right of the threshold SIF for SCC (K1scc). Region III is characterized by the crack growth that depends on SIF only very slightly (if at all). Finally, region IV represents crack growth without an impact of environment. It is commonly supposed that EIC and the transport of environment to the crack tip are the processes that determine crack growth rate at regions II and III respectively. The dependence of the crack growth rate on temperature is usually expressed by the Arrhenius law for regions II through IV. This means that the activation energy at region III is much less than at region II and region IV. This distinction indicates a change in the rate-determining process with increasing SIF.
lgLc I I
II
1 I I I I
III
iV/ I
I
,, ,, ! !
glSCc
! I I gl
~//0)
Fig. 1. The typical shape of a kinetic diagram.
Volume 1: Chemistry, Mechanics and Mechanisms
107
Surprisingly, kinetic diagrams like that shown in Fig. 1 were obtained for a broad spectrum of solid-environment combinations, including "metal-liquid metal" (LME of metals [1,2]) and "silicate-aqueous solution" (the embrittlement of silicate solids in aqueous solutions [6]). Because fracture mechanisms for such dissimilar solidenvironment combinations are chemically different, the similarity between kinetic diagrams supports the idea that the general shape of the diagrams is determined by transport effects. This idea raises two questions: what are the specific rate-determining mechanisms at the steady-rate region III, and what are the reasons for the change of crack growth regions in Fig. 1? It is necessary to keep in mind that representing experimental data in terms of a kinetic diagram, while convenient, does not reflect the actual situation at the crack tip. The construction of a kinetic diagram is performed at specified external conditions, whereas the processes responsible for crack growth depend on local (internal) conditions in the vicinity of a crack tip. In most cases, the extemal conditions affect crack growth rate indirectly. The thermodynamic parameters and even the phase condition of fluid at the crack tip can differ significantly from external conditions. Furthermore, when constructing and inspecting the diagrams, the difference between the actual and "geometrical" values of SIF (the latter calculated with unloaded crack faces) is ignored, whereas some evidence suggests the difference is important. As an example, one can refer to the experimental results of LME crack propagation [ 1]. It was found that the crack growth rate at steady-rate region III depended strongly on an external load. That is, crack growth rates did not depend on crack length and, therefore, on local stress calculated for load-flee crack faces. At the same time, the crack growth rate depended strongly on stress far from the crack tip. These facts were apparently recorded for the first time in early studies of LME. Experiments with amalgamated aluminium and brass revealed the independence of crack-initiating stress on the length of the initial crack [4]. It may appear that crack initiation does not correlate with the local stress at the crack tip, yet it is controlled by the external stress. These observations testify to the importance of solid-environment interactions that take place far from the crack tip. As shown below, these special features of crack evolution are likely caused .by decreasing SIF and/or losing the continuity of liquid owing to a drop of fluid pressure within an opening and/or growing crack. 3. Semi-empirical models Theoretical studies of hydrodynamic effects on the kinetics of crack propagation are based on over-simplified semi-empirical models [5]. We consider a surface crack with the length Lc(t) in a half-space (x > 0) bordered by incompressible Newtonian fluid. Let the half-space at y --->+ oo be loaded by a tensile stress Croo.The number of not-tooexacting restrictions allows use an approximation of inertialess viscous flow with a local velocity profile approaching Poiseulle's one. As a result, the cross-average velocity of fluid flow, u(x,t), and the pressure gradient along the crack growth direction are governed by the relation
u(x, t)=
h2 c)P 3/1 c)x
(1)
Modeling EnvironmentalAttack
108
where ~t is the viscosity of the fluid. The displacement of crack faces h(x,t) satisfies the continuity relation
dh
dhu + ~
dt
= 0
(2)
dx
resulting from averaging the continuity equation for incompressible fluid over the cross-section of a crack. Eqs. (1) and (2) are supplemented by the specification of external pressure P0 and the adhesion (impermeability) condition at the crack tip. To relate the crack-opening displacement (COD) to the liquid pressure, a crack model must be specified. For this purpose, one can use the integro-differential relations of the Barenblatt-Dugdale model [2]. However, the quasi-brittle character of EIC enables us to consider the simplified approach, which is not associated with a specific model of the plastic zone. We replace the COD in Eqs. (1) and (2) by its asymptotic relation
h=
2(]-vp)KI ffL c +Rp -X 2~G
(3)
Rp =s K--Li2 8(a=) where KI is the SIF, ~ the effective yield stress limit (almost constant), G the shear modulus, VpPoisson's ratio, and Rp the plastic zone size. This simplification is justified by the fact that the absolute value of fluid pressure peaks in a crack tip. The fluid pressure variation is relatively small outside the vicinity of the crack tip. It is then expected that the x-dependence of fluid pressure is affected by an error in the COD measurement. It is important to note that the plastic zone size (Rp) obtained using Eq. (3) does not need to be determined by the Barenblatt-Dugdale model. The intermediate asymptotes of Eq. (3) are valid for a broad spectrum of elasto-plastic models [2]. The difference from the Barenblatt-Dugdale model involves only the factor of the order of unity in the relation to crack tip opening displacement. Because here fluid pressure at the crack tip is def'med with logarithmic accuracy, this difference is not essential. Using Eq. (3) instead of integro-differential relations facilitates solution of the problem. Given the COD as defined by Eq. (3) with the beforehand-unknown timedependent SIF, the fluid velocity is determined by a straightforward integration of Eq. (2). The fluid pressure is then calculated by integration of Eq. (1). To relate the distribution of fluid pressure with the level of SIF, we used the approximate relation [7]
K I - K ~~) = - S ~f-~ r I ~IG 2 ~ c
gc
2 xt~ c 2 /'l G 2 L//21( I
(4)
Volume 1: Chemistry, Mechanics and Mechanisms
109
where time derivatives are labelled by points at the top, K~~ _=1.12~/x Lc o'oo,
Ci =
1.51, and c2 = 1.07. The fluid pressure at the crack tip can be calculated with the same accuracy from the relation
Pc - PO = 3x~t G2 Lc In Rp 2 (1-Vp) 2 K / Lc
zrltt G 2Lc [r (1-Vp) 2 KI3
(5)
Eqs. (1)-(5) with the adopted semi-empirical approach must be supplemented with a relation for crack growth rate as a function of SIF, pressure (PC) and temperature (Tr of fluid in the vicinity of the crack tip. The consideration below is restricted by isothermal problems, so that Tr is taken to be equal to external temperature. The role of temperature increases appear to be of little importance in most cases. Therefore, the system of ordinary differential equations descriptive of the fluid-filled crack growth can be rewritten in the ultimate form
[(i +
( I - v p ) 2 K / ( K I -K~ 0)) 3 c I vo K I =0 T(KI,Pc, T)+ 2 ~ c2 It a 2_cL2/3 2c 2 L~
(6)
where Pc is determined from Eq. (5). The distinction between f ( KI, Pc, T) and an experimentally-obtained kinetic diagram is that the former incorporates the actual value of SIF depending on the fluid pressure distribution within the crack and fluid conditions at the crack tip, whereas the latter is a function of the "geometrical" value of SIF and fluid conditions outside the crack. To complete the above models, we must specify the function f ( KI, Pc, T) that relates crack growth rate to SIF and fluid pressure at the crack tip. Let us assume that the impact of non-local "liquid-solid" interaction on the crack growth rate is small. Then, first-order corrections to the SIF value and fluid pressure from Eqs. (4) and (5) are given by the relations AK, ~-12.5x/-~- / ~ G 2 ~ j - ' ~ (1- Vp)2K~~ (7)
APc =
-
2"68zrpG2L~ In o', (1- vp )2 o" L~ o',~
As evident from Eq. (7), in the event of strong SIF-dependence of function f ( KI, Pc, T), non-local hydrodynamic effects must be more pronounced in the case of longer cracks. At the initial crack growth stage - when the crack is short- the effects are of little importance. Hence, for the SIF-dependence of function f ( K~, Pc, T) in Eq. (6), conceptually one can use the empirical interrelations approximating the crack growth rate at region II of a kinetic diagram.
110
Modeling Environmental Attack
The question of the pressure-dependence of function f ( KI, Pc, T) is more complex. The problem is that in order to complete the above models, we must to have these relations for the negative pressure range corresponding to metastable liquid conditions. Experiments under such conditions appear to be too difficult to be practical. As for "microscopic" models, the question of the pressure-dependence of crack growth rate has apparently not yet been examined. Nevertheless, situations exist where the role of the dependences is moderately important. This is true when the cavitation microcavities originate before significant pressure-related variation in the crack growth rate. Below we assume that the function f ( KI, Pc, T) is independent of the fluid pressure at the crack tip. In the subsequent analysis, we use exponential or threshold power-type functions [1,2]
Lc=voexpi
Lc = v 0
KI K lscc
1
rl
E - kb rK I , )
KI K lscc
1 exp ----~ E
(81
(9)
A remark is necessary at this point. Since we assume that the initial crack is completely filled with liquid, the initial COD must not equal zero. Clearly, this can be _(0) 9 This the case only with nonzero SIF and, therefore, with nonzero initial stress croo corresponds to the situation of provisional wetting of solid under small subliminal loading. An alternative treatment of the quantity 0-~0) relates it to the initial value of crack tip opening displacement. The nonzero initial stress in this instance is introduced to accommodate the residual stress at the crack tip remaining after the creation of artificial crack (notch). 4. Mechanism of the reduction of local stress
According to the deterministic approach adopted herein, an environment-induced crack begins to grow as soon as SIF exceeds its threshold (Ki~cc). To consider the problem of initiating crack growth, one must set LC = 0 in Eq. (6). The condition for the initiation of crack growth is expressed, in terms of Eq. (8), as KI = K~scc. We defined the delay time (q) in the start of crack growth as the time needed for SIF to attain its threshold value Ki~cr during instantaneous loading from a small subthreshold stress 0-~0) to a super-threshold stress 0-00that exceeds the threshold external stress. The delay can be expressed as an expansion of small parameters 0-~0) / 0-00 and
zcL c 0 -(0)2 / K~scc. A reasonably-accurate relationship is given by the relation
Volume 1: Chemistry, Mechanics and Mechanisms
ts:
0.88
111 ,u G 2
(10)
;'r ( l _ v p ) 2 0"oo0 .(0)2
As evident from Eq. (10), the delay time depends, as a first approximation, neither on threshold value of SIF nor on the initial crack length. This feature is clearly dictated by the strong cubic SIF-dependence of the rate of change of SIF (see Eq. (6)). For this reason, a large part of the delay time occurs when SIF and COD are small. Spasmodic increases in SIF take place at the end of the delay period. In the case of loading with constant strain rate, the qualitative pattern is the same as when the strain rate is moderately low. As was the case above, the prolonged stage of insignificant variation of SIF alternates with its spasmodic increasing to the actual "geometrical" value. When the strain rate meets the condition lu G2 ~
>> 1,
(11)
( l - V p ) 2 0 .(0)3
it is not difficult to obtain an asymptotic interrelation for the external stress (%) corresponding to initiation of crack growth. Retaining only the dominant part of the asymptotic expansion, we have 0.89 K lscc
i: < ?.,
~/zc Lc (O)
(12)
t~g = 1.32G IPcr(O ~ > o~* ~f"~ ( l - v p ) ) '
where the critical value of strain rate ( ~, ) is given by the relation 2 cr(~176 ~. = 0.46(1 - Vp)2 glscc
(13)
l.t G 2 L c (0)
The results suggest that, with subcritical strain rates, the non-local effect of a liquid environment on the starting stress is negligible. In contrast, with supercritical strain rates, such an effect is found to be crucial. In this instance, the starting stress does not depend on the threshold value of SIF and the initial crack length. The level of the crack initiation stress is governed by the viscosity of the liquid, the strain rate and the initial COD specified by the quantity cr~~ It should be noted that Eq. (11) represents apparently a not-too-limiting condition, such that the validity domain of Eqs. (12) and (13) seems to be large. The non-local hydrodynamic interaction of a liquid environment and a stressed solid appears to be essential to crack growth kinetics. The peculiarities of crack growth
112
Modeling Environmental Attack
kinetics for constant loading condition are considered below. Qualitative examination of Eqs. (6) and (8) shows that, with the constraint fl = (1 - Vp) 2 e E/kT k2T 2 croo/gtGvo b2 > 9.02
(14)
qualitatively-distinct stages of crack growth exist, including the crack growth corresponding to the steady-rate region III. Depending on the initial condition, SIF approaches some constant value with increasing crack length or, alternatively, zero. In such a situation, the initial value of SIF is dictated by the previous stage of unsteady loading. This result implies that, with a sufficiently large value of fl, the kinetic behaviour of the crack can be qualitatively distinct, depending upon the pre-history of loading to a constant stress croo.In particular, with "instantaneous" provisional loading, the growth of an initially-long crack would follow the steady-rate asymptote with nonzero and constant SIF, whereas a short crack would retard its growth. It must be noted that values of fl typically range from 105 to 10 ~~ Because fl >> 1, the limiting value of SIF can be evaluated by using asymptotic expansion. Consequently, the crack growth rate at the steady-rate stage takes the form
Zc -- 0.10
(1-Vp)2Croo ( E ) 2 ~f-~u -~
(15)
which is applicable in the case of E/kT>> 1. The qualitative behaviour of a crack within the model that includes Eqs. (6) and (9) is similar to the behaviour described above. When n > 2, the length-dependence of crack growth rate stabilizes or crack growth terminates. In this case, we obtain the following approximation for the crack growth rate at the steady-rate stage: n
Lc z K2scc 0.10 (1-Vp) 2 croo n - 2 pG G
2
2 n-2 ,uGv 0
(n-2)k T
This relation differs from Eq. (15). The temperature dependence of crack growth rate given by Eq. (15) is dictated by the activation energy of fluid viscosity (E0, which is generally small in comparison to the "chemical" value of the activation energy of disruption of interatomic bonds. The temperature dependence of crack growth rate given by Eq. (16) corresponds to the activation energy given by
Ec=nEv - 2 E n-2 '
(17)
of which the value and the sign depend on exponent n in Eq. (9). The conclusion is certainly valid in the case of weak temperature-dependence of exponent n and threshold SIF value.
Volume 1: Chemistry, Mechanics and Mechanisms
113
By this means, the mechanism of the reduction of local stress can serve conceptually as the reason for experimentally-observed steady-rate crack growth. However, the steady-rate regime can be realized in the event that fluid continuity is not violated at the preceding period of crack growth. On the other hand, the origination of microcavities in the vicinity of a crack tip may first occur at the steady-rate period. In any case, the general shape of empirical kinetic diagrams points to the need for considering cavitation effects. 5. The cavitation mechanism
Because the drop of the maximum liquid pressure is localized in the immediate vicinity of the crack tip, one should expect a loss of liquid continuity. To account for the loss of immediate liquid-solid contact, a boundary condition at the crack tip should be renewed. The adhesion condition for the velocity of liquid should be replaced by the evident condition for the liquid pressure (we use the conventional notations and restrict the evaluation to the case of complete wetting) Pc = Pv
(18)
YLv hc
The velocity of the liquid-vapour interface (LVI) can be calculated from Eqs. (1) and (2). During the initial period of crack growth, the LVI velocity is higher than the crack growth rate. Therefore, an impermeability condition must be used. However, in the subsequent evolution, the crack tip opening displacement increases and the LVI velocity decreases. The condition of critical microcavity origination should thus be formulated as equality between the velocity of LVI and the rate of crack growth. We assume that without the immediate liquid-solid contact, environmental influence on crack growth rate is neglected. If this is the case, concurrent with the critical microcavity origination, the crack growth rate drops suddenly and the liquid-solid contact is re-established. Therefore, the crack growth rate must be equal to a critical value. This must be the case whenever the critical crack growth rate is higher than the crack growth rate without an environmental effect. As soon as this condition is violated, the critical microcavity evolution no longer results in a decreasing crack growth rate. The cavitation mechanism of crack growth stabilization therefore becomes invalid. For the case of small decreasing local stress, we obtain the following relation for the crack growth rate at the steady-rate stage:
Lc
0.59 yLv 1+ - ~2 /1G lfi(Crs/Croo)
.
(19)
2 yLvo'sG
The quantitative estimation of Eq. (19) shows that the factor in square brackets is close to unity for most cases. Thus, the crack growth rate does not depend substantially on the local stress. We can conclude that both the cavitation mechanism and the mechanism of decreasing local stress create the steady-rate stage of crack growth. The steady-rate crack growth can be sustained sequentially by both mechanisms: first by the
114
ModelingEnvironmentalAttack
mechanism of the reduction of local stress and then by the mechanism of alternating origination and restoration of microcavities. It was useful to evaluate the critical crack size (L,), which results in the replacement of the stabilization mechanism. We obtained the following relationship
zc e K2~,. exp
L, --- 8 o.2
I2(1-Vp) yLvtYSll + (1-vp) K2(p~ 3 rc ,u G L c 2 ZL,o. G s
(20)
where Lc and KI, correspond to the steady-rate stages described either by the model represented by Eqs. (8) and (15) or by the model represented by Eqs. (9) and (16). For some cases of LME [1 ], the critical crack size (Lo) obtained using Eq. (20) appears to be exceptionally large (e.g., >1 m). However, this does not contradict the experimental data. 6. Conclusion
The analysis of transport effects on EIC was restricted to a non-local hydrodynamic interaction between solid and environment. That analysis offers a qualitative explanation of observed regularities of LME crack growth such as (i) the independence of delay time and crack starting stress of crack size, (ii) the existence of a steady-rate region of crack growth, and (iii) the relatively small activation energy of crack growth rate at a steady-rate region. Two non-local rate-determining mechanisms- namely the reduction of local stress in the vicinity of the crack tip and the alternating origination and restoration of microcavities in the liquid- were considered. Both mechanisms were found to be associated with the negative fluid pressure within the growing and opening surface cracks. References
[1] C.L. Briant, S.K. Banerji (Eds.), Embrittlement of Engineering Alloys, Academic Press, New York, 1983. [2] G.P. Cherepanov, Mechanics of Brittle Fracture, Nauka, Moscow, 1974 (in Russian). [3] A.R.C.Westwood, N.S. Stoloff(Eds.), Environment-SensitiveMechanical Behavior, Gordon and Breach, New York, 1966. [4] W. Rostoker, J.M. McCaughey, H. Markus, Embrittlement by Liquid Metals, Reinhold, New York, 1960. [5] A.I. Malkin, E.M. Podgaetsky, Physics-Doklady 43(1) (1998) 1-5. [6] S. Wiederhom, J. Amer. Cer. Soc. 55 (1972) 81-85. [7] D.P. Rooke, D.A. Jones, J. Struc. Anal. Eng. Des. 14 (1979) 1-6.
115
Will finite-element analysis find its way to the design against stress corrosion cracking? M. Vankeerberghen SCK.CEN, Boeretang 200, B-2400, Belgium Abstract
Stress corrosion cracking (SCC) is a potentially significant material degradation mechanism for structures in a variety of material-environment combinations. Therefore, an adequate design methodology is pertinent to many industries. Major developments in corrosion science and multiphysics computing occurred in the 20th century. Those advances should make it possible to substantially increase our ability to predict SCC. Finite-element stress analysis and computational fluid dynamics have long been applied to industry. Their advantage lies in a clear separation of the geometry, which can be arbitrary, and the mathematical description of the underlying science. Will these numerical simulation techniques also be applied to the design against SCC? Recent developments in finite-element predictions of electrochemical conditions within cracks and crevices, together with finite-element predictions of the mechanical crack-tip loading and a constitutive law describing the material-environment interface at the crack tip, should point to a design methodology against SCC. Such a design methodology consists of first calculating the electrochemical conditions and the mechanical loading at the crack tip, and then determining the crack propagation rate via a mechanico-electrochemical diagram. The approach is demonstrated for SCC of Type 304 stainless steel in high-temperature water. It makes use of (i) finite-element pre-calculated electrochemical crack-tip conditions, (ii) an analytically calculated mechanical crack-tip loading, and (iii) a mechanico-electrochemical diagram describing the constitutive behaviour of the interface. The constitutive behaviour can be obtained experimentally and, in the future, it might be calculated by multi-scale modelling. 1. Introduction: finite-element calculations in other industries
Finite-element computational techniques were developed in the early 1970s [1,2]. Interdisciplinary finite-element analysis [3] has been considered since the early 1980s. Since their introduction, they have been applied to many industries (e.g., automotive and aerospace). In some fields, e.g., stress analysis, the calculations have become commonplace and have indeed reached maturity. The advantage of f'mite-element calculations lies in a clear separation of the geometry, which can be arbitrary, and the mathematical description of the underlying science. Table 1 illustrates the equations used and the variables one solves for in some fields where finite-element analysis has
116
Modeling Environmental Attack
Table 1 Application of the finite-element method in various engineering disciplines Discipline Finite-element stress analysis Computational fluid dynamics Finite-element heat transfer Computational electrochemistry
Unknowns 3 unknowns displacement(vector) 4 unknowns velocity (vector) pressure (scalar) 1 unknown temperature (scalar) n + 1 unknowns concentration1
Equation
Conservation Conservation of energy/momentum Conservation of mass and momentum
Navier-Stokes equations
Conservation of energy Conservation of mass and charge
Heat conduction equation Nernst-Planck and Poisson equations
Conservation of chargeand magnetic flux
Maxwell equations
o . .
Computational electromagnetics
concentration n potential (scalar) 6 unknowns magneticflux (vector) electric field (vector)
been employed. The disciplines relevant to stress corrosion cracking (SCC) are electrochemistry, fluid flow and stress analysis. 2. Finite-element calculations related to localised corrosion
The emergence of personal computers in the early 1980s, which provided a means of solving industrial problems numerically, created enormous growth in this field of research. One of the earliest attempts to apply finite-element modelling to localised corrosion was made by Fu and Chan [4]. The geometry of their localised corrosion cell was modelled using a finite-element mesh. The conductivity of the environment was assigned to each element according to its chemistry. The polarisation curves of the materials in the cell were designated as boundary conditions. The cell current distribution was calculated on the basis of charge conservation. In the field of cathodic protection, boundary element modelling was introduced at about the same time by Stremmen [5]. The construction of large and complex structures in deep waters increased the demand for computer-assisted cathodic protection design and evaluation. A boundary element code was adapted (i) to include modules for the definition of the geometry, (ii) for the handling of electrochemical boundary conditions, (iii) to assist in the numerical solution of the nonlinear discretised equations, and (iv) to present the results visually. In the field of crevice corrosion, the finite-element approach was later applied by Xu et al. [6] to model the IR drop believed to be associated with many forms of crevice and stress corrosion cracking. The conductivity was assumed to be constant, so that the problem was reduced to solving the Laplace equation. Non-monotonous polarisation curves, corresponding to active-passive transitions, were used as boundary conditions. In the field of flow-assisted (erosion) corrosion, Nesic and Coles [7] demonstrated how computational fluid dynamics (CFD) could be used in the analysis of a corrosion problem. The correspondence between the CFD calculated turbulent kinetic energy and
Volume 1: Chemistry, Mechanics and Mechanisms
117
the measured corrosion damage in a heat exchanger head pointed to erosion corrosion or flow-assisted corrosion. Some design changes suggested by their CFD calculations are currently being evaluated in the field. Today, the boundary and finite-element approaches are being developed further in the fields of galvanic corrosion and cathodic protection [8] and crevice corrosion [9,10] respectively. In the multidisciplinary field of SCC, finite-element calculations were introduced by Shoji et al. [11-15], but only the mechanical aspects of crack propagation were included in detail. The calculation of the electrochemical aspects by finite elements was investigated by Gavrilov [16]. The framework for simultaneous mechanical and electrochemical calculations has been developed by Vankeerberghen [17] and is further treated here. 3.
The
mechanico-electrochemical
( M E C )
diagram
Environmentally assisted cracking (EAC) is the result of a synergistic interaction between a stressed material and its environment. An example of such an interaction is SCC of Type 304 stainless steel in high-temperature (288~ pure water. The crack growth rate depends on various parameters. Amongst them are: - the oxygen bulk concentration, the conductivity of the environment, the applied stress intensity factor, - the crack length, the yield strength, the corrosion potential in the given environment, the degree of sensitization of the material, the efficiency of the coating, the fluid velocity, and temperature. Some of those parameters do not actually drive the SCC process; rather, they create a resistance to it (or facilitate it). Furthermore, some of the parameters are bulk or system variables, and do not adequately relate to the process at the crack tip. In this respect, it is useful to identify the principal driving forces for SCC. It is postulated here that the necessary (but not sufficient) requirements for crack advance are the local conditions at the crack tip. These local conditions will load the crack in two ways: mechanically and electrochemically [17]. The dual l o a d i n g - mechanical and electrochemical - can in its simplest form be portrayed as a two-dimensional diagram where the abscissa depicts the mechanical loading and the ordinate the electrochemical loading, both at the crack tip (Fig. 1). In such a mechanico-electrochemical (MEC) diagram, contours of equal crack propagation rate (CPR) can be drawn. The diagram is fixed for a given material-environment combination, and can be looked upon as a constitutive behaviour for the interface between the local environment and the material at the crack tip. Although the driving forces for crack propagation are the crack-tip electrochemical and mechanical loading, the extension of the actual crack propagation in a real system is influenced by the resistance to crack propagation created by the internal and external environments and the material. These resistances will also result in the experimentally observed dependencies with respect to SCC, but must be clearly distinguished from the driving forces.
118
Modeling Environmental Attack "all 5.5 ~m (Fig. 2) there is no mutual interaction between the stress fields around the bubbles. Hence, the values of Lt in this investigation are chosen to be within the range !.5 to 25.5 jam, which reflects either interaction or no mutual interaction between adjacent stress fields. The mechanical properties of the amorphous alloy are taken as Young's modulus E = l l0 GPa [22], Poisson's ratio v = 0.36 [23], and fracture toughness Kk ~ l0 MPax/m [24]. The assumptions of linear elasticity and isotropy both apply to a metallic glass. The geometry is simplified to a two-dimensional form so that the bubble is represented as a cylinder. Finite element analysis was employed to obtain stress intensity factors as well as critical pressure, volume and number of moles based on a fracture mechanics approach. These analyses were performed under plane strain conditions, using the commercial finite element code ADINA version 7.5.0 [25]. In addition, the finite element method was used to calculate the hydrogen concentration profile, hydrogen flux and number of gas moles in a bubble as a function of crack length and diffusion time, based on a diffusion approach. In this case, the ADINA-T version 7.5.0 [25] and a post-processor were used. All models were subjected to mode ! deformation and symmetry conditions allow for modeling one-quarter of the body. The meshes consist of eight-node isoparametric elements. Quarter-point quadrilateral elements were employed at the crack tip to model stress singularity. To further improve the results, a convergence test was employed to select sufficiently fine meshes. To obtain stress intensity factors, the conservative J-integral and the virtual crack extension method were employed, as described in more detail elsewhere [21]. The stress intensity factors were presented in normalized form as K/ =
(I)
po
+ to)
Volume 1: Chemistry, Mechanics and Mechanisms
205
where K; is the mode ! stress intensity factor and a is crack length, in order to demonstrate accuracy of the finite element model, several comparisons were made to solutions in the literature of either stress intensity factors or diffusion profiles, as previously discussed [2 I].
3.2. Determination of the critical pressure, volume and number of moles based on a fracture mechanics approach The values of the normalized stress intensity factor, K / , were calculated as a function of a/ro and L~/ro for the three geometries shown in Fig. 2. Values are shown in Fig. 3(a) as a function of distance between bubbles for the case of a single bubble with cracks, and in Fig. 3(b) for the three cases and similar bubble spacing. The value of K! increases with either decreasing L~/ro for a constant crack length, or with increasing crack length for a constant L~/ro ratio. This is expected because, as crack length increases or body dimensions decrease, deviation from the infinite body solution occurs, in the case of periodic arrays of bubbles, mutual interaction between bubbles becomes more significant. Moreover, for a/ro values lower than 5, the value of K! is slightly larger in the case of a doubly periodic array of bubbles (Fig. 3(b)). If the calculation for a doubly periodic array of bubbles would also be performed for L~/ro -I I (which is the upper limit for mutual interaction conditions) instead of L~/ro = 12, one may expect K / to be even higher (Fig. 3(a)). For a/ro values larger than 7, on the other hand, K / is smallest in the case of a doubly periodic array of bubbles. This implies that the two lines of cracks shield one another at these crack lengths. By equating the calculated values of K~ with the fracture toughness value K~c - l0 MPaqm, the critical pressure, Pc, for crack propagation was calculated for each crack length. The dependence of pJpo on a/ro and on the distance between bubble centers is shown for a collinear periodic array of bubbles in Fig. 4. It is evident that p~. decreases as crack length increases, with the decrease being steeper for lower values of L~/ro. The critical volume, Vc, of a bubble with edge cracks was calculated from the deformed body for each crack length, at the critical pressure. For a constant value ofp~, higher values of Vc indicate that the ribbon is less susceptible to crack propagation. It is obvious that as pc increases (i.e., the critical crack length decreases), the critical crack volume decreases. However, the critical total volume depends largely on the critical volume of the bubble itself, as reported previously [21]. At short critical crack lengths, the critical total volume is essentially equal to the critical bubble volume, while at long crack lengths the relative contribution of crack volume becomes more significant. The critical number of moles for crack propagation, no, may be determined by substituting the values of pc and Vc into the equation-of-state (EOS) for real gases
pV = nRT
I+
B2(T)p
(2)
where R is the ideal gas constant (8.3145 J/(mol.K)), T the absolute temperature, and B2(T) the second virial coefficient. At 27~ Krom et al. [8] calculated the value by means of linear regression B2 - 5.02 x l0 -9 Pa -~. The results are plotted in Fig. 5 as a
206
Crack Growth Mechanisms
(a)
L1/ro
i #
' sI
3 "
"~''~
I
6
It i
-~
...............l l
," t
51
-
K, 2
:
," s
s S
! / r/i T/ /
.-
2L
I--
.-
! n. !2ro !
"" 2L
~iai~I
I
I
I
10
20
30
40
50
air0 (a) 3.0
2L,
2.5
-
2.0
-
tl~~~ /
rl II I,i
K~ 15
/
~
/
///
~ z~_." ....
/"
/
.," / .." /
1.0
II) 0.5
-
line o f bubbles (L{ro = 1 I)
.....
two 0.0
0
lines of bubbles
(L{ro = 12)
I
I
1
I
I
2
4
6
8
10
a/to (b) Fig. 3. Dependence of the normalized stress intensity thctor on alro and Ltlro (a) as a function of bubble spacing in the case of one bubble with two edge cracks in a rectangular body, and (b) for the three arrangements of bubbles with edge cracks and similar values of L~/ro.
Volume 1" Chemistry, Mechanics and Mechanisms
207
xlO 3 Po.,L2rqi 12
!-- 2Ll-~l :: L1/r0 3 6
o
-----
............... 11
51
....
0
1'0
I
I
I
20
30
40
50
a/ro Fig. 4. Dependence of the normalized critical pressure on the crack length to bubble radius ratio and on the distance between bubble centers.
xl0-11 xlO-,i
300
./~'---
~,12 200
~
rj~
,/
g
s
/
I I
9
8
m
O
E
9
II
Ll/ro
3
I
I
I
I
I
1
2
3
4
a/ro
sS
5." ..
~ "
=====
...............
m 100
....
6
II 51
.... 3;.T.,* .~'''%'-
0
lO
20
30
40
50
a/ro Fig. 5. The critical number of moles for crack propagation, obtained on the basis of a fracture mechanics approach, as a function of a/ro and L~/ro for a collinear periodic array of bubbles.
208
Crack Growth Mechanisms
function of a/ro and L~lro for the case of a single collinear array of bubbles. A minimum value followed by an increase in nc as a/ro increases was observed. In addition, for each value of L ~/ro and the highest value of a/ro considered, nc decreases. In order to predict failure, one must know under what conditions a crack will first propagate and then arrest, and under what conditions a crack will propagate to failure. To this end, a failure criterion is proposed in this study, based on the intersection point between the "deformation relation" curve and the EOS curve for a constant number of hydrogen gas moles [21]. The "deformation relation" is the relation between the pressure within the bubble and cracks and their volume, as obtained by finite element analysis. If the pressure value at the intersection point p, is lower than p,, cracks will not propagate, in addition, as a crack grows, the pressure decreases and may fall below p,., in which case the crack arrests. Hydrogen continues to diffuse into the bubble with cracks until a critical value of moles is reached. This process repeats itself until the glassy ribbon fails.
3.3. Determination of the hydrogen concentration profile, hydrogen flux and number of gas moles in a bubble based on a diffusion approach Following hydrogen absorption in the material during electrochemical charging, monoatomic hydrogen diffuses toward microvoids, where it precipitates as molecular hydrogen. This precipitation causes both pressure buildup within the bubble (so that it expands) and crack growth. The pressure buildup inside a bubble depends on the flux of hydrogen into it, which may be described under steady-state conditions by Fick's first law. The temperature dependence of the diffusion coefficient may be expressed through a simple Arrhenius relation. A diffusion coefficient D = i.22 x 10-14 m2/s was used in the analysis, as typical for amorphous Fe-Si-B alloys at 27~ [4]. Under transient conditions, the time-dependent diffusion of hydrogen can be described by Fick's second law. For the finite element analysis performed here, the boundary condition on the outer 3 The assumption of an surface of the ribbon was estimated as Cs = 1.2 x 105 mOlH/malloy .
irreversible hydrogen recombination reaction was made. The sample geometries shown in Fig. 2 were considered, and a time increment of 2 s was applied. The hydrogen concentration, C, at the edge of the bubble with cracks was determined in the finite element analysis and employed in Fick's first law to obtain the flux of hydrogen, J, across the bubble and crack surfaces. The number of hydrogen gas moles, n, inside the bubble with cracks was determined as a product of diffusion flux, surface area of bubble with cracks, and time increment. This procedure was repeated for increasing time periods to determine the accumulation of hydrogen gas in microvoids as a function of time. it has been shown [2 I] that for a given bubble spacing L~/ro, the number of gas moles increases monotonically with crack length. This is expected for the irreversible reaction of gas precipitation in a growing crack (i.e., growing volume). In addition, as the distance between bubbles decreases, the number of moles decreases as well. This can be explained in terms of the need to fill simultaneously several traps with mutual interactions.
209
Volume 1" Chemistry, Mechanics and Mechanisms
Pc). :2ro:
2500 !
2000 0
2LI/
s S~
1500
_
-"" "1"600/
~..,...'.'"
Sw s
/
~,/
15001"-
G"
,9,.y'""
Zd'"
|
Y"
1400~
-
1000
..... " ~ "
"
/ / /
13001-
-
500
0
Ln/ro 3 -.-
t"
6
...............11 . . . . 51
/ / /
I
I
I
i
0
2
4
6
8
a/ro
I
!
I
10
20
30
10
I 40
50
a/ro Fig. 6. The dependence of diffusion time on crack length for the case of a collinear periodic array of bubbles with cracks and various values of L~/ro.
3.4. Coupled fracture mechanics~diffusion analysis By coupling fracture mechanics and diffusion analyses, one can calculate the time necessary to attain the critical number of hydrogen gas moles inside a bubble with cracks, which in turn will lead to crack propagation. Thus, the time to failure can be predicted. The dependence of crack length on diffusion time for a collinear periodic array of bubbles with cracks for various values of Ln/ro is illustrated in Fig. 6. It can be observed that an incubation period exists before the crack is filled with sufficient hydrogen gas to initiate propagation. Subsequently, except for L~/ro = 3 (in which a constant slope may be observed), the crack propagation rate cannot be said to be constant, as claimed by Krom et al. [8]. The procedure employed for determining time to failure, accompanied by tabulation, has been described elsewhere [21]. This procedure yielded, in the case of a collinear periodic array of bubbles in an infinitely-long strip, the time to failure tr = 1345 sec when Li/ro - 3, tj = 1402 sec when Ll/ro-- 6, tl .= 1569 sec when Li/ro = II, and t l 2507 sec when L~/ro - 51. For comparison, Eliaz and Eliezer [2,4] charged FesoB~Si9 ribbons electrochemically. Interconnecting cracks on the surface of the ribbons were detected after 1800 s of charging at ic = 20 A/m 2. Thus, the new coupled diffusion/stress finite element model that is described here provides a reasonable fit with Eliaz and Eliezer's experimental data [2,4].
210
Crack Growth Mechanisms
4. Discussion and conclusions
In this study, a coupled fracture mechanics/diffusion approach was developed to model the failure of metallic glasses due to HE in the absence of external loads. The model can be employed to predict time to failure of amorphous FesoBt~Si9 ribbons in which high-pressure bubble formation and interconnecting crack propagation occur during electrochemical hydrogen charging. The crack propagation process clearly depends on a variety of parameters, including mutual interaction between bubbles, the geometry and dimensions of the specimen, and the length of the crack relative to bubble size. In the absence of mutual interaction between stress fields around adjacent bubbles, gas accumulation in bubbles is more rapid. The intersection point between the p-V curves for the fracture mechanics-based elastic solution and for the EOS of real hydrogen gas can be used, in conjunction with the value of the critical pressure, as a failure criterion. Crack propagation is characterized by the existence of an incubation period in which the hydrogen fills the bubbles and cracks until a critical pressure is reached. Subsequently, except when L~/ro -- 3, the rate of crack propagation to failure is not constant. The time to failure, as predicted by this model, is similar to that observed experimentally. Our model can be extended to other failure mechanisms, and may be useful for materials selection. For example, a similar approach can be utilized to predict the synergistic effects of helium and hydrogen in isotropic and anisotropic crystalline metals. In order to model these effects accurately, it is not sufficient to take into account only diffusion and trapping phenomena. The mutual effects of helium bubbles, or the effect of the stress fields around helium bubbles on the diffusion and degradation processes, must also be taken into consideration. Such a model may be useful in the selection process of materials for the first wall of thermonuclear fusion reactors. Various modifications can further improve the model. Firstly, additional arrangements of microvoids (e.g., zig-zag periodic arrays of voids with cracks) should be considered. Secondly, additional microvoid sizes (mainly a normal distribution of void diameters) should be investigated to better simulate the experimental microscopic observations. Thirdly, one may take into account equilibria between (i) diffusion and trapping of hydrogen, (ii) hydrogen diffusivity (which is concentration and pressure dependent), and (iii) hydrogen effects on the local values of the elastic constants as well as on stress relaxation. In addition, a three-dimensional model can be developed, although this would require an in-depth microscopic characterization of the orientation of cracks relative to the bubbles. Finally, electrochemical hydrogen p.ermeation experiments must be conducted to define the boundary condition of Cs more accurately. References
[I] S. Tetelman, W.D. Robertson, Trans. TMS-AIME 224 (1962) 775-783. [2] N. Eiiaz, D. Eliezer, Metall. Mater. Trans. A, 31A (2000) 2517-2526. [3] J.O'M. Bockris, A.K.N. Reddy, Modem Electrochemistry, vol. 2, Plenum Press, New York, 1970, pp. 1328-1344. [4] N. Eliaz, Hydrogen interaction with amorphous and quasicrystailine alloys, Ph.D. Thesis, Ben-Gurion University, Beer-Sheva, 1999. [51 N. Eiiaz, E. Moshe, S. Eliezer, D. Eliezer, Metall. Mater. Trans. A, 3 IA (2000) 1085-1093. [6] S.M. Schitigl, E. Van der Giessen, Scripta Mater. 46 (2002) 431-436. [7] S.M. SchlSgl, J. Svoboda, E. Van der Giessen, Acta Mater. 49 (2001) 2227-2238.
Volume 1: Chemistry, Mechanics and Mechanisms
[8] [9] [10] [! I] [12] [13] [14] [15] [16] Ii7] [18] [191 [20] [21] [22] [23] [24] [25]
211
A.H.M. Krom, A. Bakker, R.W.J. Koers, Int. J. Pres. Ves. Piping 72 (1997) 139-147. A. Taha, P. Sofronis, Eng. Fract. Mech. 68 (2001) 803-837. J. Lufrano, P. Sofronis, Acta Mater. 46 (1998) 1519-1526. P. Sofronis, J. Lufrano, Mater. Sci. Eng. A260 (1999)41-47. J. Lufrano, P. Sofronis, D. Symons, Eng. Fract. Mech. 59 (1998) 827-845. J.M. Lufrano, P. Sofronis, TAM Report no. 771, University of Illinois at Urbana-Champaign, Urbana, IL, 1994. P. Sofronis, R.M. McMeeking, J. Mech. Phys. Solids 37 (1989) 317-350. A.T. Yokobori, T. Nemoto, K. Satoh, T. Yamada, Eng. Fract. Mech. 55 (1996)47-60. P. Sofi'onis, I.M. Robertson, TAM Report no. 972, University of illinois at Urbana-Champaign, Urbana, IL, 2001. J. Lufrano, P. Sofronis, H.K. Birnbaum, J. Mech. Phys. Solids 46 (1998) 1497-1520. J. Lufrano, P. Sofronis, H.K. Birnbaum, J. Mech. Phys. Solids 44 (1996) 179-205. A.G. Varias, A.R. Massih, J. Nucl. Mater. 279 (2000) 273-285. R.V. Goldstein, A.V. Balueva, Fatigue Fract. Eng. Mater. Struct. 20 (1997) 1269-1277. N. Eliaz, L. Banks-Sills, D. Ashkenazi, R. Eliasi, Acta Mater. 52 (2004) 93-105. AlliedSignal Inc., Metglas 2605SA!, Technical Bulletin, Parsippany, N J, 1995. H.S. Chen, J. Appl. Phys. 49 (1978) 462-463. J.-J. Lin, T.-P. Perng, J. Mater. Sci. Lett. 10(1991) 1443-1445. K.J. Bathe, ADINA (version 7.5.0)- Automatic Dynamic Incremental Nonlinear Analysis: Theory and Modeling Guide, ADINA Engineering, Watertown, MA, 2001.
215
Quantification of hydrogen transport and trapping in ferritic steels with the electrochemical permeation technique A.-M. Brass Laboratoire de Physico-Chimie de l 'Etat Solide, CNRS UMR 8648, Bdt. 41 O, Universitd Paris-Sud, 91405 Orsay, France
Abstract
The quantification of hydrogen concentration in materials is important. This is especially true for ferritic steels with complex microstructures, where hydrogen can be trapped with various binding energies to microstructural defects. The objective of this paper is to evaluate the pertinence of computing the diffusible hydrogen concentration from steady-state flux values derived from electrochemical permeation or decay transients. Cr-Mo steels with different microstructural features were chosen for this purpose. The numerical data calculated from permeation and degassing experiments were compared with the hydrogen concentration measured with the fusion thermal conductivity method. According to these comparisons, the quantification of the permeation experiments in terms of hydrogen concentration must be validated by independent measurements of hydrogen concentration. I. Introduction
The modelling of time- and temperature-dependent hydrogen distributions in materials requires a thorough knowledge of hydrogen diffusion and trapping phenomena. In particular, the internal hydrogen concentration must be quantified and the respective contributions of hydrogen strongly trapped at microstructural defects and of diffusible (lattice + weakly trapped) hydrogen must be evaluated. This is a difficult task in the case of ferritic steels with complex microstructures, where numerous microstructural defects (e.g., dislocations, carbide or non-metallic phase interfaces, grain boundaries, microvoids) can trap hydrogen with different binding energies [1]. Moreover, the dependence of trapped hydrogen concentration on the external or internal hydrogen activity cannot be precisely predicted. The electrochemical permeation technique [2] allows highly-sensitive measurement of change in hydrogen flux as a function of time through cathodically-charged specimens. This technique is one of the major tools available for measurement of the permeability of metals to hydrogen and of the hydrogen diffusivity in the 20-125~
216
Hydrogen Permeation and Transport
temperature range in aqueous environments [3,4]. The versatility associated with electrochemical charging methods allows the spanning of a wide range of hydrogen equivalent pressures or activities. It also yields hydrogen concentrations in samples that are similar to those obtained by gas phase charging at high pressures. Moreover, the electrochemical permeation method is well suited for the study of the influence of microstructure on hydrogen diffusion and trapping, as it is well established that hydrogen transport and uptake are strongly dependent on hydrogen trapping [1] at microstructural defects. The aim of this paper is to discuss, for ferritic steels, the issue of quantifying the diffusible hydrogen concentration derived from electrochemical permeation and electrochemical degassing tests. For this purpose, numerical data are compared with the hydrogen concentration measured with the nitrogen-carrier-fusion thermal conductivity method.
2. Experimental 2.1. Materials Two bainitic steels - 2.25Cr- 1Mo (errs = 580 MPa, OUTS= 690 MPa) and 3Cr- 1MoV (errs = 660 MPa, OUTS= 740 MPa) [5] - and a normalized steel 4120 (errs = 310 MPa, OUTS= 500 MPa) were tested. Chemical compositions are provided in Table 1. Table 1 Chemical composition (wt.%) of the steels studied Steel 2.25Cr-lMo 3Cr-IMo-V 4120
C 0.13 0.12 0.19
Si 0.18 0.06 0.22
Mn 0.58 0.52 0.66
Cr 2.2 3.1 0.95
Mo 1.04 0.99 0.18
S 0.003 0.003 0.014
P 0.000 0.000 0.018
Ni 0.14 0.18 0.16
V w 0.22 --
Cu 0.07 0.046 0.20
2.2. Electrochemical permeation experiments Permeation tests were conducted with foils 0.1-0.13 cm thick in a Devanathan-type, double-electrolytic cell [2]. Temperature was maintained at 25 + 0.1~ by circulating water in the double jacket of the cell. Hydrogen was generated on the input surface of the samples by cathodic polarization in 0.1 N NaOH at-1350 mVscE and in 1 N H2SO4 at 5 mA/cm 2. Additional experimental detail can be found elsewhere [6]. The advantage of this technique over other methods is that the decay current corresponding to the hydrogen desorption from the material can be easily recorded by interrupting the cathodic polarization at steady-state. The boundary conditions usually considered for electrochemical permeation tests with thin foils of thickness L and a pure diffusion control [7] involve rapid hydrogen entry kinetics to achieve equilibrium at the charging side of the sample and constant sub-surface hydrogen concentration Co. Moreover, a negligible hydrogen concentration on the detecting side of the foil is assumed to be achieved by immediate oxidation of all exiting hydrogen atoms in an alkaline solution with a high pH (0.1 N NaOH, 0 mVscE). The flux measured at the output surface is related to Co through Fick's first law. Under cathodic polarization, the diffusible hydrogen concentration in the specimen's bulk
Volume 1: Chemistry, Mechanics and Mechanisms
217
increases with time and distance until a linear concentration profile corresponding to steady-state conditions is achieved. The steady-state output flux doo, expressed in cm3/cm2/s, is derived from the anodic current density at steady-state Ioo. The working surface area S. Joo can be expressed as: Joo = I ~ I S x F = D x C o / L
(1)
doo is related to the hydrogen pressure PH2 through Sieverts's law by: d~ = D x S ( T ) ~ p u ' / L
(2)
where S(T) is the solubility. The permeability, P, expressed in cm3xcm/cm2/s/bar1/2 (which is the product of the diffusivity and the solubility) is then related to the steady-state flux through: P = J~ x L / x/-Pn~
(3)
2.3. Hydrogen concentration measurements
Hydrogen concentration was measured with the nitrogen-carrier-fusion thermal conductivity method. For this purpose, plates 1 mm thick were charged from 1 N H2SO4 and 0.1 N NaOH under the same conditions as the permeation tests, but for 5 h to get a homogeneous hydrogen concentration profile throughout the thickness of the plates [8]. The samples were stored in liquid nitrogen after charging to avoid hydrogen degassing. They were repolished at low temperature before hydrogen analysis. To quantify the diffusible hydrogen concentration, several sets of samples were allowed to de-gas at room temperature for various aging times. 3. Results and discussion
3.1. Influence o f microstructure on the permeability to hydrogen
Data representative of the permeability of the steels to an unknownequivalent hydrogen pressure are presented in Table 2. Each value represents duplicate or triplicate experiments performed with separate samples. Cathodic charging from 0.1 N NaOH led to steady-state flux values smaller than those measured in 1 N H2SO4. This was due to the lower hydrogen activity associated with (i) the a smaller hydrogen coverage of the input surface at higher pH of alkaline solutions [9], and (ii) the value of the applied potential. For a given charging condition, the permeability of AISI 4120 pearlitic steel to hydrogen is higher than that of bainitic Cr-Mo steels which is consistant with previous results [ 10].
218
Hydrogen Permeation and Transport
Table 2 The average value of the steady-state hydrogen flux in Cr-Mo steels at 298 K under cathodic charging conditions 3| x L x 109 (cm 3 x cmlcm21s) I N H2SO4, 5 mA/cm 2 0.1 N NaOH,-1350 mVscE 6.2 + 0.50 38 + 4 9.7 + 0.60 33 + 2 15.0 + 0.02 55 + I
Steel 2.25Cr-IMo 3Cr-IMo-V 4120
Comparison of values ofJ~ x L (Table 2) with the permeability (Eq. (3)) of the same material [1 I] to a known pressure of H2 gas at the same temperature made it possible to evaluate the equivalent hydrogen pressure generated by cathodic charging. For the bainitic steels, P"2 yielded 50-100 bar and about 1500 bar for charging from 0. I N NaOH and I N H2SO4 respectively.
3.2. Apparent diffusion coefficient of hydrogen derived from build-up transients Typical normalized permeation and degassing curves are plotted in Fig. !. The dimensionless time variable r = Dt/L 2 was computed with the largest diffusion coefficient value reported for hydrogen diffusion in pure Fe at 298 K [ 12]. ~J)J ' i " - -- ~ '
0.8
'
~
'|
~ ' ~
'
~+'" % "
#
b
+
0.8
6'
~
o
_~
- - e - - 2.25Cr- 1Mo
+
- - + - - 3Cr- 1Mo-V
-~
~ O
I
- -o--4120
r r ~;
0.4
'
J/Jr
~5
0.6 ~
'
§
0.6
- -,--4120
0.4
- - e -- 2.25Cr-1Mo
#
o
-41
- -+-
- 3Cr-1 M o - V
(D
0.2
0.2
0
--
0
i
1] m~
50
,
,
I
100
,
i
i
i
I
150
,
,
,
, ,
200
oo-
01
e v
200
400
600
800
v
, j
1000 1200 1400
Fig. I. Typical normalized (a) permeation and (b) de-gassing curves for 4120, 2.25Cr-IMo and 3Cr-IMo-V steels. Cathodic polarization in I N H2SO4, 5 mA/cm2 at 298 K. The differences in the permeation curves corresponding to each steel reflected different trapping capacities [13-15]. The traps (e.g., dislocations, carbide and inclusion interfaces) can be generally classified as "reversible" or "irreversible" [I ], depending on the residence time of hydrogen at these sites (which act as energy wells for hydrogen atoms diffusing in the interstitial lattice sites). Strong or irreversible traps, where the residence time of hydrogen is "long" at room temperature, will lead to a greater decrease in hydrogen diffusivity than reversible traps [16].
Volume 1: Chemistry, Mechanics and Mechanisms
219
The apparent diffusion coefficients Defy(Table 3) were computed at the beginning of the rising transient with the time-to-breakthrough [17] and the time lag [2]. As expected [10,18], the hydrogen diffusivity was much larger in the normalized 4120 steel than in bainitic steels. This can be accounted for by a smaller density of trapping sites in pearlite than in bainite. Compared to 2.25Cr-lMo steel, the smaller diffusivity in 3CrI Mo-V steel could be ascribed both to the increase of the Cr-content and to the presence of V-containing fine precipitates [ 19-20], which were also shown to stabilize the dislocation networks [19]. Moreover, a large dependence of the hydrogen diffusivity on hydrogen activity was observed (Table 3), as previously reported for iron [21-24]. This dependence, rationalized with concentration-dependent diffusion plus trapping equations [13], was considered to result primarily from an increase of trap occupancy when the lattice hydrogen concentration CL was increased. This effect is due to the dependence on CL of the capture probability of hydrogen atoms by reversible traps [ 16,25]. The larger increase of Da with the external hydrogen activity observed for 4120 steel indicated a larger Occupancy of the saturable traps [25] in this material and/or a smaller trap density. Table 3 The average value of the apparent diffusion coefficients of hydrogen in Cr-Mo steels at 298 K derived from rising permeation transients under cathodic charging conditions Steel 2.25Cr-lMo 3Cr-1Mo-V 4120
De[iX 107 (cm2/s) 0.1 N NaOH,-1350 mVscE Db, Da 2.2 + 0.10 2.3 + 0.1 0.4 + 0.02 0.6 + 0.1 39.0 + 1 18.0 + 3.0
1 N H2SO4, 5 mAfcm 2 Db, Da 2.7 + 0.5 3.0 + 0.2 1.1 + 0.1 2.0 + 0.4 110.0 + 4.0 78.0 + 3.0
As expected [10,18], the hydrogen diffusivity was much larger in the normalized 4120 steel than in bainitic steels. This can be accounted for by a smaller density of trapping sites in pearlite than in bainite. The smaller diffusivity in 3Cr-IMo-V steel compared to 2.25Cr-lMo steel can be ascribed both to the increase of the Cr-content and to the presence of V-containing f'me precipitates [ 19-20], which were also shown to stabilize the dislocation networks [ 19]. Moreover, a strong dependence of the hydrogen diffusivity on hydrogen activity was observed (Table 3), as previously reported for iron [21-24]. This dependence, rationalized with concentration-dependent diffusion plus trapping equations [13], is considered to result mainly from an increase of the trap occupancy when the lattice hydrogen concentration CL is increased. The latter is due to the dependence on CL of the capture probability of hydrogen atoms by reversible traps [16,25]. The greater increase of Da with the external hydrogen activity observed for 4120 steel indicates a larger occupancy of the saturable traps [25] in this material and/or a smaller trap density.
3.3. Apparent diffusion coefficient of hydrogen derivedfrom decay transients In order to generate more information on the contribution of trapping to the permeation of hydrogen, the decay curve recorded by interrupting the cathodic polarization at steady-state can be compared with the first permeation curve. The decay
Hydrogen Permeationand Transport
220
rate is governed by the rate of hydrogen release from the reversible traps assumed to be in dynamic equilibrium with the diffusing hydrogen atoms, and by the diffusion coefficient [25]. As shown by Table 4, hydrogen de-gassing is slower from 3Cr-IMo-V steel than from 2.25Cr-1Mo steel. This suggests that the reversible trapping capacity is greater in the V-containing material. Furthermore, this indicates that the slower apparent diffusivity of hydrogen derived from rising transients (Table 3) for 3Cr-IMo-V steel cannot be attributed only to diffusion controlled by irreversible traps. The more rapid de-gassing of hydrogen from normalized AISI 4120 steel can be attributed to a smaller reversible trapping capacity, which is presumably due to a lower dislocation and interface density. The faster hydrogen extraction when the initial hydrogen activity is high may derive from the dependence of the diffusion coefficient on the concentration of diffusing hydrogen [23]. Moreover, for a given material and hydrogen activity, the laCgest value of the apparent diffusion coefficient derived from electrochemical permeation and decay curves is obtained from the time corresponding to the early stage !
of decay. The apparent diffusion coefficient Dbt calculated with the "time-to-decay" following cathodic charging at high hydrogen activities can therefore be considered less affected by trapping effects. Table 4 The average value of the apparent diffusion coefficients of hydrogen in Cr-Mo steels at 298 K derived from decay transients. Influence of previous cathodic charging conditions !
Steel
Def t
x 10 7 (cm2/s)
0.1 N N a O H , - 1350mVscE !
2.25Cr-lMo 3Cr-IMo-V 4120
!
1 N H2SO 4, 5 m A / c m 2 !
!
Dbt
Dtl
Dbt
Dtl
6.7 + 0.1 4.9 9 0.8 83.0 + 9.0
0.5 + 0.04 0.4 + 0.05 5.7 + 0.20
22 + 2.0 15 + 0.5 115 + 2.0
3.9 + 0.1 2.1 + 0.2 4.4 + 0.4
3.4. Diffusible hydrogen concentration measured with the nitrogen-carrier-fusion thermal conductivity method The change with aging time at room temperature of the total hydrogen concentration resulting from cathodic charging from 1 N H2SO4 is illustrated in Fig. 2. Each value corresponds to at least three experiments performed under identical conditions, without subtraction of the residual hydrogen content (0.5 to 0.6 wt.ppm) measured in uncharged samples. After cathodic charging from 1 N H2SO4, the total hydrogen content was larger in 3Cr-IMo-V steel than in 2.25Cr-lMo steel and 4120 steel. Aging at room temperature results in a strong decrease of hydrogen content in the three steels. De-gassing hydrogen from 2.25Cr- 1Mo steel and 4120 steel was achieved after about 15 days and 2 days respectively of aging at room temperature. At~er subtraction of the hydrogen concentration in the as-received specimens, the mean hydrogen content remaining in the samples after these time periods corresponded to about 0.1 wt.ppm of residual hydrogen from cathodic charging. Because 200 days are required to reach ~0.2 wt.ppm
Volume 1: Chemistry, Mechanics and Mechanisms
221
of residual hydrogen from cathodic charging, a much longer time was required for hydrogen degassing from 3Cr-IMo-V steel. As a consequence, most of the trap sites for hydrogen are reversible owing to the fact that almost all the initial hydrogen resulting from cathodic charging can be released at room temperature. However, the long time required for hydrogen de-gassing from 3Cr-I Mo-V at room temperature indicates the presence of a significant fraction of "strong reversible" traps compared to irreversible traps. These traps can account for the slower diffusivity of hydrogen in the Vcontaining material [26]. H 2 (wt ppm) 2.25Cr- 1 Mo 3Cr-IMo-V 4120
2
uncharged specimens 0 ~.:.:.:4 0
24
48
360
1632
2280
4800
aging time (hours)
Fig. 2. Change with aging time of the average hydrogen content absorbed in Cr-Mo steels during cathodic charging in I N H2SO4, 5 mA/cm2 at 293K.
meltextr.
The mean diffusible hydrogen concentration ,.. d~ can be assessed by subtracting, from the total initial hydrogen concentration, the hydrogen concentration remaining in the samples after aging at room temperature for 15 days and 2 days for 2.25Cr-IMo steel ,,-, meltextr.
and 4120 steel respectiveley. ' - d ~ yielded ~1.4 wt.ppm for 2.25Cr-IMo steel and 0.6 wt.ppm for 4120 steel, for the charging condition used. For 3Cr-IMo-V steel, c meltextr.
diff yielded 2.5 to 3 wt.ppm, depending on the criterion used for the definition of reversible trapping. 3.5. Diffusible hydrogen concentration assessed from permeation curves For a given external hydrogen pressure or activity, the sub-surface hydrogen concentration Co (Eq. (l)) in the interstitial lattice sites below the charging surface of the foils can be considered representative of the lattice hydrogen solubility. However, this holds true only in the case of ideal lattice diffusion in perfect materials. In this case, the amount of hydrogen diffusing through the foil at steady-state is equal to Co/2. it must be noted that the use of DI.~ for the calculation of Co for steels is questionable if the interstitial diffusion of hydrogen in steels is slower than that in pure iron due to interactions of hydrogen with substitutional alloy atoms.
Hydrogen Permeation and Transport
222
In Table 5, the lattice solubility of hydrogen in pure iron C~ e [10,27,28] computed for an equivalent hydrogen pressure of 50 bar and 2000 bar at 298 K is compared with
C~ Fe values computed from the steady-state permeation flux (Table 2) and with two values of DFe [ 12,28].
Table 5 Average diffusible hydrogen concentration in Cr-Mo steels derived from steady-state permeation flux and computed with different values of the diffusion coefficient. Comparison with the diffusible hydrogen concentration derived from melt extraction measurements and from graphical integration of decay transients Material
c?
C DFe
C D'bt
eff
C D tl
r" meltextr.
From Dtl
From melt From decay extractions integration
(wt.ppm)
(wt.ppm)
eff
"~diff
C deg
!
From solubility laws (wt.ppm)
From
x 10 2
x 10 2
From Dbt
DFe (wt.ppm) (wt.ppm)
(wt.ppm)
1 N H2SO4, 5 m A / c m 2
Pure Fe 4120 2.25Cr-lMo 3Cr-IMo-V
1.5-3.0 -
0.7-0.9 0.5-0.6
0.05 • 0.004 0.08 + 0.003 0.20 + 0.01 1.40 + 0.2
0.6 1.4
0.42 + 0.01 1.20 + 0.04
0.4-0.5
0.30 • 0.07
2.5-3.0
1.80 + 0.2
-
0.035 + 0.01 0.34 + 0.06 1.0 + 0.6
2.20 + 0.4
0.1 N NaOH, -1350 mVscE Pure Fe
0.2-0.5
.
.
.
.
.
4120 2.25Cr-lMo 3Cr-IMo-V
-
0.17-0.23 0.02 + 0.002 0.09 + 0.01 0.07-0.10 0.10 + 0.02 0.34 + 0.06 0.09--0.13 0.28 + 0.01 1.90 9 0.06
Table 5 shows that the use of DFe was not necessarily appropriate for estimating Co in the case of complex steel microstructures, since C DEe yielded values that could be smaller than the lattice solubility of hydrogen in iron C Fe . For the limiting case of a low fraction of occupied reversible traps, when irreversible trapping can be neglected, an apparent solubility Ce# (lattice + reversibly trapped subsurface hydrogen concentration) can be derived [16] from the expression:
Ceff = JooL / Deff
(4)
The problem is then to determine which apparent diffusion coefficient value must be used for the calculation of Ceyf. Computed with D'bt - the value of the diffusion coefficient that was less affected by !
t r a p p i n g - CDe~t - JooL/D'bt yielded 2 to l0 times the value of CcFe calcultated f o r d
Volume 1: Chemistry, Mechanics and Mechanisms
223 !
hydrogen pressure of 2000 bar. For this charging condition, ("eft ' D b t can thus be considered an upper limit of the apparent solubility of hydrogen in these steels. r, meltextr.
With the use of Da derived from the time-lag, Ceff values comparable to '--d/ff (Table 5) are found for bainitic steels. However, this is not the case for the normalized 4120 steel because for this material the large apparent diffusivity leads to small Ceff values. This situation clearly illustrates the difficulty of quantifying the trapping behaviour of hydrogen from built-up permeation transients. In agreement with a smaller trapping capacity of normalized microstructures, Ceff values were smaller for normalized AISI 4120 steel (Table 3) than for bainitic steels.
3.6. Diffusible hydrogen concentration assessed from decay curves It is also possible to assess the diffusible hydrogen content in the samples at steadystate by graphical integration of the decay curves (Fig. l(b)), owing to the fact that the area beneath the decay curve is a coulometric measure of hydrogen diffusing out through the output side of the sample, qd~. The corresponding value of the diffusible hydrogen concentration Cdeg = qdeg x 2 is presented in Table 5, and compared with (i) f , m e ltextr.
the diffusible hydrogen concentration measured from melt extractions ~d/ff
,
and
(ii) the diffusible hydrogen concentration derived from steady-state permeation c'Otl "-'elf " For bainitic steels, Cdeg and C eff ~ were almost identical, regardless of the charging condition. When these two values were derived from tests performed in 1 N H2SO4, ,,.-, m e ltextr.
they were also compared with the value of Wd/ff
,
which was derived from melt
extraction measurements. For normalized steel, the values of Cdeg and
r meltextr. ,.. d / f f
fairly similar, but very different from the diffusible hydrogen concentration c'~ ~"eff
"
were In the
limiting case of lattice diffusion through a perfect lattice, assuming that the hydrogen concentration at the input side instantaneously drops to zero upon arrest of cathodic charging, one third of the total diffusible hydrogen concentration at steady-state Qa~ was predicted to desorb through the output side of the foil [29]. The total diffusible hydrogen content Qdeg would then be equal to qdeg x 3, and the corresponding concentration equal to q d e g X 3 x 2. The values of qdeg and Qa~ = C e f f / 2 for the steels tested are compared in Table 6. The fraction of hydrogen degassed at the output side of the sample, qdeg/Qa~, was much larger than the value of 1/3 predicted for desorption controlled by diffusion in a perfect material [29]. Except for 4120 steel, qdeg/Qd~ff ~ 1 for permeation tests conducted in 1 N H2SO4. This suggests that the former input side of the sample was almost totally impermeable to hydrogen because, for this boundary condition, Qdeg= qdeg [29]. It would then be necessary to assume that the slight corrosion that can occur during cathodic charging of carbon steels impedes hydrogen de-gassing through the former input side, despite the polarization of the surface in 0.1 N NaOH (Section 2.2). However, this interpretation does not appear to be realistic. This is so because the ratio qdeg/Qd~ derived from tests performed in 0.1 N NaOH also yielded values larger than 1/3, even though the former input side was assumed to be free of corrosion in this medium.
224
Hydrogen Permeation and Transport
Table 6 Average hydrogen content degassed at the anodic side of the foil after interruption of cathodic charging from 1 N H2SO4 and 0.1 N NaOH. Comparison with the diffusible hydrogen content at steady state Steel 4120 2.25Cr-lMo 3Cr-IMo-V
1 N H2SO4, 5 mA/cm2 qdeg(wt.ppm) Qa~ (wt.ppm) 0.4 • 0.005 0.04 + 0.001 0.6+0.02 0.70+0.01 0.9 + 0.10 1.10 + 0.03
0.1 N NaOH, -1350 mVscE qdeg(wt.ppm) Qa~(wt.ppm) 0.035 + 0.01 0.045 + 0.005 0.170+0.03 0.170+0.03 0.510 + 0.02 0.940 + 0.30
The data in Table 5 also provide evidence of the dependence of hydrogen concentration extracted at the output side of the sample on the hydrogen activity of the previous charging solution, and of the larger diffusible hydrogen concentration in 3Cr1Mo-V steel in comparison to 2.25Cr-1Mo and 4120 steels.
4. Summary Electrochemical permeation and degassing tests were performed at room temperature with a pearlitic 4120 steel and with bainitic steels 2.25Cr-lMo and 3Cr1Mo-V. The diffusible (lattice + reversibly trapped) hydrogen concentration derived from melt extractions (nitrogen-carrier-fusion thermal conductivity method) performed after cathodic charging was compared, for each steel, to the hydrogen concentration calculated from permeation data and from the graphical integration of the desorption curves recorded upon the arrest of cathodic polarization. The hydrogen diffusivity and the total hydrogen concentration depended on microstructure. Hydrogen concentration measurements showed that the total hydrogen content was highest in 3Cr-IMo-V steel and lowest in 4120 steel, and that all three materials contained mostly reversible traps. However, a large fraction of "strong reversible" traps in 3Cr-IMo-V steel reinforced the trapping capacity of this steel. This accounts for the different diffusion behaviour of hydrogen in the two bainitic steels. A comparison of calculated and experimental data showed that, for ferritic steels with a high trap density, the use of the lattice diffusion coefficient of hydrogen in iron or the apparent diffusion coefficient derived from steady-state permeation measurements might not be appropriate for the calculation of the diffusible hydrogen concentration. However, the hydrogen concentration derived from the graphical integration of the decay curves allowed, with greater confidence, attaining values similar to those measured by melt extractions. Independent measurements of the hydrogen concentration were required for reliable quantification of permeation and degassing curves in terms of hydrogen concentration.
Acknowledgment This work was supported in part by the European Communities under Contract ECSC 7210-PR/110.
Volume 1: Chemistry, Mechanics and Mechanisms
225
References [ 1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [ 17] [ 18] [19] [20] [21 ] [22] [23] [24] [25] [26] [27] [28] [29]
G.M. Pressouyre, Acta Metall. 28 (1980) 895-911. M.A.V. Devanathan, Z. Stachurski, Proc. Roy. Soc. London A270 (1962) 90-101. A-M. Brass, A. Chanfreau, Acta Mater. 44 (1996) 3823-3831. A. Tumbull, in: A. Tumbull (Ed.), Hydrogen Transport in Metals, The Institute of Materials, London, 1995, pp. 129-141. Prediction of Pressure Vessel Integrity in Creep Hydrogen Service, E.C. BRITE/EURAM Programme, Final Report BE-1835, 2000. A-M Brass, F. Guillon, S. Vivet, Metall. Mater. Trans. A, 35A (2004) 1449-1464. C. Montella, J. Electroanal. Chem. 465 (1999) 37-50. J. Crank, The Mathematics of Diffusion, Oxford University Press, Oxford, 1975. E.G. Dat't, K. Bohnenkamp, H.J. Engell, Corros. Sci. 19 (1979) 591-612. W.C. Luu, J.K. Wu, Corros. Sci. 38 (1996) 239-245. P. Tison, Report of the Commissariat h l'Energie Atomique, Service de Documentation. CEN Saclay, 91191 Gif sur Yvette, CEA-R-5240 (1), 1984, p. 281. E. Riecke, K. Bohnenkamp, Z: Metallkunde 75 (1984) 76-81. G.R. Caskey, W.L. Pillinger, Metall. Trans. A, 6A (1975) 467-476. M. Iino, Acta Metall. 30 (1982) 367-375. J.B. Leblond, D. Dubois, Acta Metali. 31 (1983) 1471-1478. R.A. Oriani, Acta Metall. 18 (1970) 147-157. N. Boes, H. Ziichner, J. Less Com. Met. 49 (1976) 223-240. D.L. Dull, K. Nobe, Werkst. Korros. 33 (1982) 439-448. F. Gehrmann, H.J. Grabke, E. Riecke, in: A. Turnbull (Ed.), Hydrogen Transport in Metals, The Institute of Materials, London, 1995, pp. 216-226. J.C. Charbonnier, H. Margot-Marette, A-M Brass, M. Aucouturier, Metall. Trans. A, 16A (1985) 935-944. J. O'M Bockris, P.K. Subramanyan, Electrochim. Acta 16 (1971) 2169-2179. L. Nanis, T.K.G. Namboodhiri, Acta Metall. 18 (1970) 432-444. T.K. Govindan Namboodhiri, L. Nanis, Acta Metall. 21 (1973) 663-672. A.J. Griffiths, A. Tumbull, Corros. Sci. 11 (1995) 1879-1881. A. McNabb, P.K. Foster, Trans. TMS-AIME 227 (1963) 618-627. L. Coudreuse, P. Bocquet, in: A. Tumbull (Ed.), Hydrogen Transport in Metals, The Institute of Materials, London, 1995, pp. 227-239. H.H. Johnson, Metall. Trans. A, 9A (1988) 2371-2387. K. Kiuchi, R.B. McLellan, Acta Metall. 31 (1983) 961-984. L. Nanis, T.K.G. Namboodhiri, J. Electrochem. Soc. 119 (1972) 691-694.
227
Hydrogen diffusivity and straining effect at cathodic polarization of AI in NaOH solution E. Lunarska, O. Chemyayeva Institute of Physical Chemistry, Polish Academy of Sciences, ul. Kaprzaka 44/52, 01-224 Warsaw, Poland Abstract
Unusual runs of hydrogen permeation transients observed during electrochemical measurements of hydrogen permeation through aluminium in 0.01 N NaOH accounted for the complex transport of hydrogen within the membrane. Experimental study of the effects of membrane thickness, cathodic polarization and the mode of cathodic polarization application on the stress-strain relationship of aluminium, plus theoretical analysis, showed that the uphill diffusion of hydrogen caused by the membrane strain as a result of the application of cathodic polarization was involved. Taking into account the effect of strain-induced uphill diffusion on the hydrogen permeation transients, the hydrogen diffusivity was evaluated from permeation data. Results suggested that hydrogen uphill diffusion and hydrogen-induced stress relaxation should possible be taken into account when considering the stress and/or strain state at the tip of a growing corrosion crack in aluminium. I. Introduction
Hydrogen-induced degradation of metals is governed by hydrogen transport and behaviour. Especially, hydrogen diffusivity into a metal and thus the rate of reaching places where a crack(s) can initiate is of great importance. One of the methods for evaluating hydrogen diffusivity and hydrogen transport parameters involves electrochemical measurements of the hydrogen permeation rate through a metal membrane [ 1]. Data on hydrogen diffusivity in aluminium are scarce [2], with the hydrogen diffusion coefficient (at room temperature) ranging from 10-9 to 10-6 cm2/s [3,4]. Such a significant scatter can be explained by specific corrosion behaviour of aluminium alloys in aqueous solutions and the disturbance of hydrogen transport within membranes. In the latter case, the interaction of hydrogen with traps and the hydrogen uphill diffusion due to stress formation should be taken into consideration. In order to calculate hydrogen diffusivity from electrochemical permeation measurements, all phenomena that affect the experimental results should be identified. This would allow application
228
Hydrogen Permeation and Transport
of experimental procedures that provide undisturbed diffusion transport of hydrogen and the appropriate analysis of permeation data. In the present work, the effect of complex hydrogen transport through membranes exposed to an alkaline solution on hydrogen permeation data was evaluated. 2.
Experimental
procedure
Pure (99.99%) AI foils of thickness (L) from 25 to 2000 ~tm were used in this study. The double cell for electrochemical hydrogen permeation measurements was specially modified (i) to coat the egress side of a membrane with Pd, (ii) measure the potential and the near-electrode pH, and (iii) inject the corrosion and hydrogen promoters or inhibitors into the ingress cell (Fig. 1). In order to minimize contact with air and standardize the metal surface, special precautions were taken while removing the atrbome film from the membrane surface and pouring the electrolyte into both cells [5]. In the egress compartment containing 0.01 N NaOH, the anodic current as the measure of hydrogen permeation rate was recorded at the constant anodic potential of +150 mV (Hg/HgO) until the steady state value (Joo) was attained in the ingress cell. The membrane ingress side was subjected to either open circuit conditions or galvanostatic cathodic polarization (with current density from 20 to 120 mA/cm 2) in 0.01 N NaOH. The following modes of cathodic polarization were applied: Repeated application and interruption of the same cathodic current density (cycle mode), and Application of the step-by-step increased cathodic current density with its interruption atter each step (on-off mode). Tensile specimens (gauge length 30 mm, width 10 mm, and thickness 0.17 mm) were mounted into the electrochemical cell fixed in jaws of the Instron machine. Atter filling the cell with 0.01 N NaOH, each specimen was kept at the open circuit potential (Eocp) for 30 min before stretching the strain rate from 5 x 10-4 to 1 • 10-4 s-~. The effect of the application and interruption of cathodic polarization on the run of stress-strain curves was recorded.
Fig. 1. Schematic of the modified double cell for hydrogen permeation measurements.
229
Volume 1" Chemistry, Mechanics and Mechanisms 3. Results
3.1. Hydrogen permeation
The change of near-electrode pH and Eocp in the ingress cell, and the change of the hydrogen permeation rate (d) recorded in the egress cell atter the pouring the base electrolyte into the ingress cell, are shown in Fig. 2. With time, all measured parameters reached steady-state values. The steady-state hydrogen permeation (3~162 occurred under open-circuit conditions. Application of cathodic polarization caused immediate changes in the electrode potential and near-electrode pH in the ingress cell, which then reached appropriate steady-state values. However, the hydrogen permeation transient exhibited unusual behaviour. Initial decrease in the hydrogen permeation rate followed by an increase up to the steady-state value (foo) was recorded at the application of cathodic polarization (Fig. 3). As seen in Fig.3, the interruption of cathodic polarization resulted in a hydrogen permeation maximum. l
l
,
5
0
~
1
,
4
0
~
11,45 1 40
E3o
~soln "
'/~
-1450
] 0.1cm, 0.01N NaOH I]-1400
E
-1350~
20
.1300~.i~
0
100
200
300
400
500
-1200 700
600
time, s Fig. 2. Changes of hydrogen permeation (d) recorded in the egress cell as well as Eocpand nearelectrode pH at pouring electrolyte ($ soln.) into the ingress cell. AI, 1 mm, 0.01 N NaOH. 140
~ :t.
i
,
!
,
!
,
!
/ / !
,
i
'
120 100
80 60 40
,:r
20
t t;i
/
............ ,
189t'irnel,80/"6~'
" 800
Fig. 3. Hydrogen permeation current in the egress cell at the application ($ on) and interruption (1' off) of cathodic polarization in the ingress cell. AI, 1 mm, 0.01 N NaOH.
230
Hydrogen Permeation and Transport
|
|
4~ injecti~ 30
.
,
/NaOH
E 20
=" "~10
0t
.01N NaOH
I
0
,
I
60
9-x.: I
120
time,
180
240
s
Fig. 4. Effects of EDTA and NaOH injections into 0.01 N NAOH in the ingress cell on hydrogen permeation recorded in the egress cell under open-circuit conditions. AI, 1 mm. The application of cathodic polarization may result in decreasing near-electrode pH or suppressing corrosion processes, and thus may cause decreasing hydrogen evolution and hydrogen uptake. In order to determine whether the above effects could account for the change of near-electrode conditions due to cathodic polarization, a series of specially-designed experiments was performed. Specifically, ethylenediaminetetraacetic acid (EDTA) and NaOH were injected into the base solution (0.01 N NaOH) in the ingress cell under open-circuit conditions, and the response of hydrogen permeation in the egress cell was recorded. The above species were chosen due to their different effects [5]. EDTA decreases pH and inhibits hydrogen evolution, whereas NaOH increases pH and promotes the corrosion of aluminium. As seen in Fig. 4, the effect of above species was similar, despite their different influences on the nearelectrode conditions. Both species decreased the steady-state hydrogen permeation, but this decrease was preceded by the formation of the transition permeation maximum. Therefore, the effect of the species was similar to that produced by the cessation of cathodic polarization (Fig. 3). The cyclic application of cathodic polarization did not significantly change the appearance of the resulting hydrogen permeation minimum (Fig. 5(a)). In the case of the on-off mode of application of cathodic polarization, the minimum became less pronounced with increasing polarization (Fig. 5(b)). The application of cathodic polarization to the thin membrane resulted in the formation of the small maximum, followed by several permeation minima (Fig. 6). The permeation minimum observed at application of cathodic polarization might be described by the values of its depth (AJ) and by the time to attain the minimum (r~n), as shown in Fig. 5(a). The effects of polarization and membrane thickness on the parameters of transition permeation minimum are summarised in Fig. 7. Some of the scattering of results, especially for thin membranes, might be caused by the formation of complex minima in the case of such membranes, as shown in Fig. 6. However, the apparent trends could be traced. The time to attain the minimum increased with increasing thickness and decreasing polarization (Fig. 7(a)).
231
Volume 1: Chemistry, Mechanics and Mechanisms
0
2ol
. . . i .0 0 i'
!0 r
e
IV,,
-20
0 ...~ ~.~N,
| .
9
!; a."
\\
-.,0
f'
;80
i 9
. . . . 6 0j "s1 J
. s
;.. ,
sii
-30
-30
-40
i
on. o f f l
-40 I
0
,
i
20
,
i
40
,
i
,
60
i
9
80
i
,
i
9
100 120
"~-
i i
0
140
20
40
60
80
100
120
140
tim e, s
til~, s (a)
(b)
Fig. 5. Effect of the mode of cathodic polarization application on the transition permeation minimum. (a) Permeation minimum at the subsequent application of the cathodic current of 40 mA/cm2 and its interruption - Roman numerals. (b) Numbers near the curves denote the applied cathodic current density in mA/cm2. |
5
~
|
!
|
!
|
180
240
300
ir
o ,,, ~
-10 -15 '
60
'
120
'
360
time, s Fig. 6. The appearance of build-up hydrogen permeation transients recorded at the application of the cathodic current of 40 mA/cm2 (,1, ir on). 25-1am-thick membrane.
The degree of the effect, i.e., the depth of the permeation minimum, decreased with increasing polarization and thickness of the membrane (Fig. 7(b)). With the application of higher polarization (especially in thick membranes), the effect disappeared entirely, and no permeation minimum was observed. Similar effects of membrane thickness and polarization on the transition permeation maximum (including its disappearing for thick membrane) were detected at the interruption of cathodic polarization.
232
Hydrogen Permeation and Transport
100~
90
'
d
100j....---~
-
80 -
O
40
o
70 -
-10
/
-15
o
60
60 -
::I.
50
-20
.m t~
x, 1 ID
.=_. E O
o E
,
C "O tl--
,>))>))>))>))>))>)>>))>))>))>i~:x,X
.05
0 C .0_
E a..
J 0
1200
2400
3600
Time (s)
Fig. 5. Hydrogen-permeation curves of the steels tested.
Table 2 Diffusion coefficients of the steels tested Diffusion coefficient (m2/s) 1.1 x 10 -9 5.4 x 10-1~ 2.3 x 10-t~
Steel Extremely low-carbon steel Hypoeutectoid steel Tempered martensitic steel
(a)
--'-1 lain
(b)
Fig. 6. High sensitivity HMT results for extremely low-carbon steel after (a) 5 min and (b) 40 min of charging.
1 i~m
Hydrogen Permeation and Transport
246
(a)
1 ~tm
(b)
1 pm
Fig. 7. High sensitivity HMT results for hypoeutectoid steel after (a) 5 rain and (b) 40 rain of charging.
4.3. High sensitivity HMT results for hypoeutectoid steel As can be seen from Fig. 7(a), silver particles were localized at ferrite-carbide interfaces in hypoeutectoid steel after 5 min of charging, at which time the hydrogenpermeation current did not reach a steady state. In addition, under this charging condition, a few Ag particles were observed in proeutectoid ferrite. In contrast, 40-min charging with steady-state hydrogen diffusion revealed many Ag particles localized at carbide-ferrite imerfaces and also in ferrite distributed within a pearlite matrix. In addition, proeutectoid ferrite was covered by silver particles (Fig. 7(b)). It should be noted that an SEM image of carbide in pearlite that looked white under a 5-min charging condition (Fig. 7(a)) was black under a 40-min charging condition (Fig. 7(b)). The reversal of the contrast of SEM images of carbide was probably due to increasing the secondary electrons yield, which in turn was due to a large accumulation of Ag particles in ferrite distributed within pearlite at 40-min charging. Our results obtained under 5-min charging compared well with those obtained by Luu and Wu [8], who used conventional HMT. They also demonstrated that the carbide-ferrite interface was a preferential diffusion path for hydrogen and concluded that the main diffusion path in pearlitic steel was the carbide-ferrite interface. However, our results obtained under a 40-min charging condition disagreed with results of Luu and Wu [8]. The results indicated that the hydrogen diffusion path at steady-state diffusion in hypoeutectoid steel was ferrite distributed within pearlite and proeutectoid ferrite as well. On the other hand, our finding for a 40-min charging was supported by Hagi's results [9] for hypoeutectoid steels obtained by using the electrochemical hydrogen-permeation technique. According to Hagi, the rate of hydrogen permeation and the coefficient of hydrogen diffusion in hypoeutectoid steels decrease with increasing carbon content, i.e., the volume fraction of carbide. This trend indicates that ferrite is the main diffusion path and the carbide-ferrite interface in pearlite acts as a trapping site that delays hydrogen diffusion in steels. A question remains: Why were Ag particles localized at carbide-ferrite interfaces at 5-min charging? The explanation may be as follows" The decay curve of the hydrogenpermeation test for 5-min charging (Fig. 5) shows that the permeation current density
Volume 1: Chemistry, Mechanics and Mechanisms
247
for hypoeutectoid steel decreased slower than that for low-carbon steel. This suggests that hydrogen atoms were trapped reversibly at the carbide-ferrite interface. As a result, hydrogen atoms perhaps diffused along these interfaces at a non-steady state. 4. 4. High sensitivity HMT results for tempered martensitic steel
As Fig. 4(c) shows, the microstructure of tempered martensitic steel was covered by numerous cementite (Fe3C) particles. Consequently, it was difficult to investigate the distribution of Ag particles (an indicator of the distribution of hydrogen atoms) on the surface of this steel. To make it easier, we used not only SE images but BSE images as well. When comparing SE and BSE images for no hydrogen-charging conditions, we revealed a rather large difference: Fe3C particles gave the SE image more white colour, while almost no white regions were observed in the BSE image (Fig. 8). That was because under the condition of the accelerated voltage of 5 kV, the BSE image was not affected by an edge effect. At~er that, we employed the BSE image to investigate specimens aider 10 and 40 min of charging. We could not easily identify Ag particles by using the SE image because there were many white spots and regions in the image (Fig. 9(a)). However, at~er hydrogen charging, white spots and rows of white spots were visible along the interface between martensitic laths in the BSE image (Fig. 9(b)); these were identified as Ag particles by EDXS. This suggested that the interfaces of martensitic laths served as hydrogen diffusion paths in tempered martensitic steel. It is important to note that the high sensitivity HMT was useful to investigate the steel having a lot of Fe3C particles.
(a)
(b) l0 lam 10 lam Fig. 8. SEM images of tempered martensitic steel: (a) SE image and (b BSE image. No hydrogen charging.
5. Summary Hydrogen diffusion paths were examined in various steels using high sensitivity HMT. Major f'mdings are summarized below: 9 The crystal lattice served as the main diffusion path (diffusion within grains) in extremely low-carbon steel. No accelerated diffusion of hydrogen was observed along grain boundaries.
248
Hydrogen Permeation and Transport
(a)
10 lam
(b)
10 gm
Fig. 9. SEM images of tempered martensitic steel after 40 min of charging: (a) SE image and (b) BSE image.
9 In hypoeutectoid steel, proeutectoid ferrite and ferrite distributed within pearlite served as main diffusion paths for hydrogen. Under non-steady-state charging, hydrogen diffused along the carbide-ferrite interface that favored the trapping effect. 9 The interface of martensitic laths served as a hydrogen diffusion path in tempered martensitic steel. 9 High sensitivity HMT employed in this study had a spatial resolution of the order of 0.1 gm. The use of this technique with BSE image made it easier to investigate a steel surface with a lot of participations on it.
References
[ll [21 [31 [41 [51 [61 [71 [81 [91
A.R. Troiano, Trans. ASM 52 (1960) 54-80. G.M. Pressouyre, Metall. Trans. A, 10A (1979) 1571-1573. T.E. P6rez, J. Ovejero Garcia, Scripta Metall. 16 (1982) 161-164. J. Ovejero Garcia, J. Mater. Sci. 20 (1985) 2623-2629. K. Ichitani, S. Kuramoto, M. Kanno, Corros. Sci. 45 (2003) 1227-1241. H. Tsubakino, T. Mizuno, K. Yamakawa, Trans. ISIJ 26 (1986) 732-736. S. Kuramoto, K. lchitani, A. Nagao, M. Kanno, Tetsu-to-Hagane 86 (2000) 17-23. W.C. Luu, J.K. Wu, Corros. Sci. 38 (1996) 239-245. H. Hagi, J. Inst. Metals 57 (1993) 864-869.
249
Effect of deformation type on the hydrogen behavior in high-strength low-alloy steel E. Lunarska, K. Nikiforow Institute of Physical Chemistry of the Polish Academy of Sciences, ul. Kaprzaka 44/52, 01-224 Warsaw, Poland Abstract
The effect of cold work (applied during production or expected during exploitation) on hydrogen transport and susceptibility to hydrogen-induced cracking was studied for 0.25C-1CrIMn-lSi-lNi steel that has been recommended for the replacement of agricultural airplane parts in Poland. The tensile and compression types of macroscopic loading (including shot peening, tension-compression fatigue, and tensile stressing) were considered. Hydrogen lattice diffusivity in bainite steel was found to be independent of the kind of applied cold work that influenced the hydrogen trapping parameters and the mode of hydrogen-induced cracking. The hydrogeninduced cracks followed local plastic deformation paths of pre-strained materials. The difference in the hydrogen behaviour should be taken into account when considering stress corrosion and corrosion fatigue crack propagation. I. Introduction
Susceptibility of high-strength steels to hydrogen-induced cracking is affected by hydrogen diffusivity, which determines the rate at which hydrogen reaches the hydrogen traps. It is also affected by the hydrogen entry flux, which determines the build-up of local concentrations of trapped hydrogen to some critical level necessary to initiate and propagate crack growth [1,2]. A cold work affects susceptibility to hydrogen-induced degradation due to the creation of hydrogen traps that influence the rate of hydrogen transport. Hydrogen-induced degradation is also due to the creation of local stress sites and dislocation concentrations that may serve as the nuclei of hydrogen-induced cracks. The dislocation structure and the distribution of stresses within the plastically-deformed material differ depending on the kind of applied loading (tensile, torsion, or bending) and on the mode of its application (static or cyclic). This should also modify hydrogen trapping and transport. Although the effect of the cold work on hydrogen behaviour is well known (see, for example, Ref. [3]), the effect of different types of metal deformation has not been explored completely [4].
250
Hydrogen Permeation and Transport
After production, airplane parts are commonly subjected to the shot peening treatment in order to decrease their susceptibility to fatigue failures. It has been shown that shot peening applied to low-alloy steel (0.25C-1Cr-lMn-lSi-lNi) to ensure mechanical strength also increases its resistance to hydrogen degradation and stresscorrosion cracking (SCC) in CI-containing environments [5]. In the course of exploiting highly-loaded airplane parts, local plastic deformation can occur due to either a stress concentration or fatigue. As a result, even the local change of the metal structure and the stress and strain distribution alters hydrogen transport and the susceptibility to hydrogen embrittlement compared to material that was not cold worked. Therefore, a change in the resistance of steel to hydrogen embrittlement and SCC may be expected in the course of exploitation. In the present study, the hydrogen diffusivity, trapping efficiency and hydrogeninduced cracking of shot peened, pre-fatigued and tensile pre-strained steel were examined and correlated with the change of the steel structure produced by those types of deformation.
2. Experimental 2.1. Materials
Specimens were cut from slabs of low-carbon low-alloy steel (0.253% C, 1.13% Cr, 1.04% Mn, 1.12% Ni, 1.65% Si, 0.003% S, 0.017% P) specially designed for the replacement of airplane parts [6]. After machining, specimens were normalized at 900~ (heat treatment "N"). Certain normalized specimens were heated at 910~ cooled in oil at 80~ and then aged at 200~ for 3 h (heat treatment 'T'). All heattreated specimens were polished and then aged additionally at 120~ in oil for 3 h. Selected normalized (N) and heat-treated (I) specimens were subjected to shot peening. The codes of the peened specimens were "Ns" and "Is." Cast iron shots (0.43 mm mesh and 470 HV of hardness) were injected pneumatically for 60 s at an air pressure of 0.5 MPa, which provided coverage higher than 100%. Shot-peened specimens were then polished mechanically to obtain a smooth surface and remove surface defects that may have been produced by peening. The mechanical properties of the heat-treated and shot-peened specimens are presented in Table 1. For hydrogenpermeation measurements, the shot-peened membranes (2 mm thick) were polished mechanically from one side to a thickness of 1 mm. Table 1 Mechanical properties of the differently heat-treated and shot-peened specimens Specimen code N Ns I Is
UTS, MPa 570 1090 1590 1700
ef 0.22 0.19 0.12 0.11
The heat-treated specimens were subjected to axial tension fatigue tests at room temperature at a loading frequency of 20 Hz (Polish Standard PN-76/H-04325). The load-controlled fatigue tests were carried out at a stress ratio (R) of 0.3 and two levels
Volume 1: Chemistry, Mechanics and Mechanisms
251
of maximum stress: 1100 MPa (IF-I) and 1300 MPa (IF-2). The number of loading cycles were 2 x 105 for IF-1 specimens and 9.25 x 104 for IF-2 specimens. Pre-fatigued membranes (0.5 mm thickness and 6 mm diameter) were cut from the fatigue specimens perpendicular to the axis from the middle of the specimen gauge. Pre-strained (IP) membranes (0.5 mm thickness and 5 mm diameter) were cut from the uniformly-deformed (c = 0.075) gauge parts of heat-treated specimens tensile-tested at a strain rate of 2 x 10-4 S-1 in air. 2.2. Experimental procedure 2.2.1. Hydrogen permeation measurements Electrochemical hydrogen-permeation measurements were performed using the double cell [8] divided by the membrane studied. The egress side of the membrane coated with Pd was exposed to a 0.1 N NaOH solution and polarized at +150 mV. (Note: All potentials are given vs. Hg/HgO reference electrode.) Anodic current, the measure of the hydrogen permeation rate, was recorded in the egress side of the double cell. The aerated solution containing 30 g/l NaCI + 71 g/l NazSO4 with the small addition of H2SO4 (pH 3) simulating "acid rain" [7] was used as the test solution in the ingress compartment of the double cell. Using the step-by-step application of cathodic potential (from -750 to -2500 mV) on the ingress side of the membrane exposed to the test solution, the buildup hydrogen-permeation transients were recorded for each step in the egress cell until steady-state values were achieved. At the switching-off of polarization on the ingress side, the decay permeation transients were recorded in the egress cell. During all tests, the shot-peened surface served as the ingress side of the membrane. From hydrogen-permeation tests, three parameters were estimated [8,9]. These included the steady-state hydrogen permeation current ( f ) at given cathodic polarization, the apparent hydrogen diffusivity (D*) calculated from the buildup transient as:
D*=~,
L:
(1)
6./'0.63
and the apparent hydrogen diffusivity (Dtb) calculated from the decay transient, as O.05.L ~
D~b = ~
(2)
rib where L is the membrane thickness, /'0.63 is time needed to achieve 0.63 of d~ on the build-up transient, and rtb is break-through time established on the decay transient. 2. 2.2. Vacuum extraction Vacuum-extraction measurements were made after completing hydrogenpermeation tests. The mean content of hydrogen remaining inside the membrane (V) was measured at 400~ for 2 h.
252
Hydrogen Permeation and Transport
2.2.3. Characterizing of the structure Microstructure of steel and the thickness of the shot-peened layer were evaluated by optical microscopy at magnifications up to 1600x. The ingress surface of the membranes, following hydrogen permeation tests, was observed by scanning electron microscopy (SEM) at magnifications up to 10000x. The ferrite lattice parameters and the width of diffraction lines were estimated by X-ray spectral analysis. 3. Results
3.1. Hydrogen measurements
Fig. 1 presents the steady-state hydrogen permeation recorded in the egress cell at the application of cathodic polarization in the ingress cell. The highest level of permeability was exhibited by normalized steel. Shot peening decreased permeability in normalized steel and in heat-treated steel. A similar effect was observed after prefatigue and pre-stress treatments. It should be noted that in cold-worked, heat-treated steels IP, IF-I and IF-2, the permeation of hydrogen was detected at more negative cathodic potentials than in steel I that was not cold worked. Apparent hydrogen diffusion coefficients - including D* calculated from the first recorded build-up permeation transient (Eq. (l)), Dtb calculated from the decay permeation transient (Eq. (2)), and the amount of residual hydrogen V- are presented in Table 2. 3.2. Material character&ation
Normalized steel exhibited a ferrite structure with globular cementite particles. Isothermally quenched steel had the acicular bainite structure with he globular
14 12
----....
10 8
E O
4 20 -2 -2600
-2400 -2200 -2000 -1800 -1600 -1400 -1200 -1000
E, mV Hg/HgO
Fig. 1. Steady-state hydrogen permeation recorded in the egress cell at the application of cathodic potentials in the ingress cell.
253
Volume 1: Chemistry, Mechanics and Mechanisms
Table 2 Hydrogen diffusion coefficients (Otb and D*) calculated from the permeation measurements and the content of residual hydrogen (V) measured by vacuum extraction Material code N Ns Ns, layer#
Dtb, cm2/s
D*, cm2/s
3.5 x 10-6 1.2 x 10- 6 -
1.63 x 10 6.8 x 10-7 2.6 x 10 -7
I
4 . 7 x 10 -7
3.4 x 10 -7
Is Is, layer# IP IF- 1 IF-2
3.4 2.4 6.9 3.5
1.7 x 2.1 x 6.0 x 4.2 x 2.3 x
x 10 -7
x 10-7 x 10-7 x 10-7
-6
10 -7
10-8 10-8 10-s 10-9
"
V, ppm 0.32 0.63 1.4 0.6 1.3 5.7 3.6 2.2 3.5
# Values characterizing the deformed layer of the shot-peened materials: D*x is hydrogen apparent diffusivity in layer recalculated according to Eq. (3), and Vx is residual hydrogen content in the layer recalculated according to Eq. (4).
cementite particles, mostly situated on the boundaries of the bainite laths and the parent austenite grains. The ferrite structure of normalized steel was deformed plastically by shot peening, as evident in the slip pattern in the cross-section (Fig. 2). In the heattreated bainite steel, shot peening caused the refining of bainite laths [5]. The thickness of the shot-peened layer corresponding to the metal depth indicated that the change of the microstructure in comparison to the bulk one was 0.18 mm and 0.12 mm in the case of Ns and Is steels respectively. The ferrite (bainite) lattice parameter, a, and the width of diffraction line { 110},/5, calculated for the materials studied are given in Table 3. Heat treatment increased the lattice parameter and the width of the diffraction line of ferrite. The cold work of the heat-treated specimen decreased the lattice parameter a.
Fig. 2. Microstructure of the shot-peened layer of steel Ns, seen in the cross section.
Fig. 3. Appearance of the peeled-off ingress surface of Ns steel alter high polarization.
Hydrogen Permeation and Transport
254
Table 3 The ferrite (bainite) lattice parameters, a, and the width of the {110} line, ,8, for the studied steels Code a, A fl, de~
N 2.86488 0.359
Ns I 2.87115 2.87659 0.347 0.511
Is 2.87427 0.400
IP 2.87188 0.468
IF-I 2.85703 0.588
IF-2 2.86551 0.706
3.3. Hydrogen-induced blistering cracking Hydrogen-permeation transients recorded at polarization higher than -2000 mV exhibited maximum of hydrogen permeation, which revealed the formation of microdefects [10]. After prolonged cathodic polarization at highly negative potentials, parts of the ingress surface of membranes peeled off (Figs. 3-5). The only exception was the membrane of steel N. In the case of steel Ns, the ductile peeled-off surface revealed separation along the slip bands, cf. Fig. 3. Microvoids in membrane I (Fig. 4(a)) and microcracks in membrane Is (Fig. 4(b)) resembled the geometrical markings of the bainite microstructure. They were both situated along the boundaries of parent austenite grains and bainite laths.
(a)
(b)
Fig. 4. Peeled-off ingress surface of (a) I and (b) Is membranes subjected to prolonged high cathodic polarization.
In the case of IP specimens, the fracture surface had an intergranular appearance, with cracking occurring along the boundaries of parent austenitic grains (Fig. 5(a)). The peeled-off surface of the pre-fatigued steel exhibited the appearance of ladder-like strips, cf. Fig. 5(b).
255
Volume 1: Chemistry, Mechanics and Mechanisms
(a)
(b)
Fig. 5. Peeled-off ingress surface of (a) IP and (b) IF-2 membranes subjected to prolonged high cathodic polarization.
4. Discussion The shot-peening treatment affected only the surface layer. Therefore, the values of fo and D* estimated during the hydrogen-permeation tests of shot-peened steel membranes reflected the hydrogen transport through the plastically-deformed peened layer and through the strainless core. To estimate the parameters of hydrogen transport in the deformed layer, the solution of diffusion through the two-layer material given in [ 12] was adopted. In terms of the present study, it was written as:
D*x
6D*x
2D*c
1"0.63 --
I
X D*x
D*c~,6D* c
2D*x
l}
(3)
Lc D*c
where D*x is the apparent diffusion coefficient of hydrogen in the shot-peened layer, D*c is the apparent diffusion coefficient of hydrogen in the core (corresponding to the D* value for unpeened steel), X is the thickness of the peened layer, and Lc = L - X is the thickness of the core. Vacuum extraction conducted after the completion of hydrogen-permeation measurements and desorption of movable hydrogen at recording the decay permeation transient represented the amount of residual, irreversibly-trapped hydrogen. In the case of shot-peened specimens, it was the mean value of irreversible hydrogen in the layer and in the core. Hydrogen distribution between the shot-peened layer and the core can be estimated using a scheme from Ref. [13] and Eq. (4): VX ~
t.v-O-x)Vr
(4)
where V is the mean content of residual hydrogen measured in the shot-peened membrane, Vx is the mean content of residual hydrogen in the layer, and Vr is the mean content of residual hydrogen in the core (corresponding to the content of residual hydrogen measured in unpeened specimens).
256
Hydrogen Permeation and Transport
The re-calculated values of the apparent diffusion coefficients of hydrogen and the residual hydrogen content within the shot-peened layers are presented in Table 2. Fig. 1 shows that the steady-state hydrogen permeation recorded for cold-worked steel was lower than that for untreated steel. This could not be due to the effect of decreasing hydrogen absorption during the cold work, since the cold-worked steels exhibited a much higher content of residual hydrogen (Table 2). However, the cold work did not change electrochemical parameters during the cathodic polarization of steel [11 ]. Therefore, the decrease in hydrogen permeation due to cold work (Fig. 1) was caused by the retardation of hydrogen transport through metal and its more intensive irreversible accumulation in the plastically-deformed steel in accordance with Ref. [14]. Foe the types of cold work used in this study, the compressive (shot peening), tensile (tensile stretching) and the tensile-compressive (pre-fatigue) loads were applied macroscopically. However, real stresses and strains introduced into the metal, and especially the local stress-strain state, are quite complex and cannot be calculated precisely. Because the cold work has been known to introduce internal stresses and cause structural refinement, an attempt was made to describe the effect of cold work by X-ray data. The lattice distortion reflecting introduced internal stresses and the width of diffraction line affected by the lattice distortion and by the size of coherently reflected areas [ 15] were used to describe the state of bainite ferrite subjected to the cold work. From obtained interplanar distances, the internal stress, e, was calculated according to the equation: d, -do do
(5)
where d, is the { 112 } interplanar distance of the cold-worked ferrite lattice and do is the { 112 } interplanar distance of specimen I that was not cold worked. The calculated values of internal stresses and the width of the diffraction line were compared with two calculated hydrogen diffusion coefficients (Dtb and D*) and the measured amount of residual hydrogen content (V). Despite the high simplification, some trends could be detected in Figs. 6 and 7. The hydrogen diffusion coefficient Dtb was evaluated from decay permeation transients after completing the permeation test. This corresponded to the desorption of hydrogen from the steel by the filled traps. Therefore, the values of Dtb might be attributed to the hydrogen lattice diffusivity in the bainite ferrite. As seen in Figs. 6 and 7, the lattice diffusivity did not differ substantially for the differently cold-worked heat treated steels. The values of D* calculated from the first recorded build-up permeation transient resembled the hydrogen transport affected by the irreversible hydrogen trapping. For all steels, the values of D* were lower than those for Dtb. This difference was attributed to the hydrogen trapping efficiency of the steels [16]. With increasing internal stress, the efficiency of the irreversible trapping generally increased (Fig. 6(a)). Fig. 6(b) shows that the amount of residual hydrogen increased with increasing internal stresses. It might be concluded, therefore, that the trapping efficiency of the cold-worked bainite increased with introduced stresses regardless of the kind of cold work employed.
257
Volume 1: Chemistry, Mechanics and Mechanisms
I
'
I
'
I
t#}
E to o"
'
0
I
'
I
o
E4
I
'
I
9 Ip
>~
#
3
0 IF1
1E-7
F2
1E-8
'
9 o*
I o',
,
QIF1
2
, x,,
UI I
0,000 0,001 0,002 0,003 0,004
, ,
I
,
I,
=
I
I
0,000 0,001 0,002 0,003 0,004
(a)
(b)
Fig 6. Relationships between (a) hydrogen diffusion coefficients and internal stresses and (b) the residual hydrogen content and internal stresses for bainite steels.
I
~
'
o_ ~
I
'
'
"
,,,.I
w
I
-
"
6
L)
4
I
'
I
ca.
I
9 Ip
>T
'
I
IF-2 0
IF-2
1E-7
2
1E-8
'
Q D* I
,o 0,4
o,~I,
,
,
/IF-1
~.
1 O /
, ~
/ I
o,~
0,o 13
(a)
o,7
0,4
,
I
",
o,s
9
,
I
0,6
=
i
13 0,7
(b)
Fig. 7. Relationships between (a) hydrogen diffusion coefficients and the width of diffraction line { 110} and (b) the residual content of hydrogen and the width of diffraction line { 110} for bainite steels.
Shot peening and stretching decreased the width of the diffraction line, whereas prefatigue increased it. The effects were more pronounced for specimens subjected to a higher level of stress (IF-2), cf. Table 2. This suggested that, depending on the type of cold work applied, different effects either on the size of coherently reflected zones (increase in the shot-peened and stretched specimens and decrease in the pre-fatigued specimens) or on the stress state of deformed specimens were produced. However, in all cases, the level of reversible (Fig. 7(a)) and irreversible (Fig. 7(b)) hydrogen trapping efficiency increased.
258
Hydrogen Permeation and Transport
Under experimental conditions in the present study, the formation of cracks and the peeling off of membrane ingress surface occurred without the application of external stresses. Therefore, cracking was caused by local hydrogen segregation up to the critical value. In the case of the bainite steel (I) that was not cold worked, the formation of voids at the carbides situated at the boundaries between bainite laths and parent austenite grains (Fig. 4(a)) revealed hydrogen segregation around those particles. However, in the case of cold-worked steels, the effect of carbides on the hydrogen crack formation was not so obvious. The shot-peened normalized Ns steel peeled off along the deformation bands formed during its shot-peening treatment (cf. Figs. 2 and 3). In the shot-peened and tensilestretched bainite steel, the hydrogen-induced cracks had an intergranular appearance. In shot-peened (Is) steel, the cracks were located along the boundaries of bainite laths (Fig. 4(b)), whereas in tensile stretched (IP) steel, cracking occurred along the boundaries of parent austenite grains (Fig. 5(a)). This difference might suggest that the local increase of dislocation density and stresses, which promoted hydrogen segregation and cracking, was formed at the boundaries of bainite laths in the first case (Is steel) and at the boundaries of parent austenite grains in the second (IP steel). The appearance of the peeled-off surface of pre-fatigued steel (Fig. 5(b)) resembled the specific appearance of the deformation structure formed during fatigue. Such a structure consisted of persistent slip bands, within which dislocations collected in the strips and boundaries defining the ladder-like structure [ 17]. The appearance of the peeled-off fracture surface of cold-worked specimens resembled the pattern of the local increase of plastic deformation formed during prestraining. Therefore, the fracture mode of cold-worked steel depended on the kind of cold-work treatment applied. 5. Conclusions
1. The cold work of high-strength low-alloy bainite steel did not produce a pronounced effect on hydrogen lattice diffusivity. 2. All types of applied cold work (shot peening, uniform stretching, tensilecompressive fatigue) increased hydrogen trapping by steel. 3. Some correlation between hydrogen trapping efficiency and parameters evaluated by X-ray spectral analysis, describing the distortion of ferrite bainite lattice subjected to the different kinds of cold work, was evident. However, more detailed studies are necessary. 4. Hydrogen-induced cracking of cold-worked steel followed local deformation paths. Therefore, the kind of applied cold work determined the mode of the hydrogeninduced fracture of steel and the appearance of the fracture surface. References
[ 1] A. Turnbull, Mater. Sci. Forum 63 (1995) 192. [2] Z. Szklarska-Smialowska, Susceptibility of steels to hydrogen trapping evaluated by potentiostatic double pulse technique, in: J. Flis (Ed.), Proc. Polish-Japan Symposium on Environmental Effects in High Technology Materials, Warsaw-Chiba, 1997, p. 131. [3] B. Marandent, Effect of cold work on the dissolution and the diffusion coefficient of hydrogen in unalloyed carbon steel, in: R.W. Staehle, I. Hochmann, R.D. McCright, J.E.
Volume 1: Chemistry, Mechanics and Mechanisms
[4]
[5] [6] [7] [8] [9] [ l 0] [11]
[12] [13] [14] [15] [16] [17]
259
Slater (Eds.), Stress Corrosion Cracking and Hydrogen Embrittlement of Iron Base Alloys, NACE, Houston, 1977, p. 774. C.L. Ho, S.L.I. Chan, J.Y. Lin, Evaluation of three different surface modification techniques for resisting hydrogen embrittlement in steel, in: Corrosion Control for LowCost Reliability, Proc. 12th International Corrosion Congress, vol. 4, NACE International, Houston, 1993, p. 2367. E. Lunarska, K. Nikiforow, E. Sitko, Werkst. Korros. 55 (2004) 373. E. Lunarska, K. Nikiforow, E. Sitko, Adv. Mater. Sci. 6 (2003) 35. Acid Rain'95: Abstract Book, Kluwier Academic Publishing, 1996. M.A.V. Devanathan, Z. Stachurski, J. Electrochem. Soc. 111 (1964) 619. R.M. Barrer, Diffusion Trough and In Solids, Cambridge University Press, Cambridge, 1941. J.O'M. Bockris, P.K. Subramanyan, J. Electrochem. Soc. 118 (197 l) I I 14. K. Nikiforow, Hydrogen Embrittlement and Stress Corrosion Cracking of Bainite HighStrength Steels in C1- Containing Environment, Ph.D. Dissertation, Institute of Physical Chemistry, Warsaw, 2004. R. Ash, R.M. Barter, D.G. Palmer, Brit. J. Appl. Phys. 16 (1965) 873. E. Lunarska, J. Michalski, Mater. Corros. 51 (2000) I. M. lino, Acta Metall. 30 (1982) 367. C.A. Huber, X-ray diffraction characterization of nanophase materials, in: A.N. Goldstein (Ed.), Handbook of Nanophase Materials, Marcel Dekker, New York, 1997, p. 317. G.M. Pressouyre, I.M. Bemstein, Metall. Trans. A, 9A (1978) 1571. S. Kocafida, Fatigue Failure of Metals, Translated by E. Lepa, Sijthoff & Noordhoff International Publishers, Alphen aan den Rijn, 1978.
261
Strain-assisted transport of hydrogen and related effects on the intergranular stress corrosion cracking of alloy 600 J. Chine Laboratoire de Physico-Chimie de l 'Etat Solide, UMR CNRS n ~ 8648, Bat. 41 O, Universitd Paris-Sud, 91405 Orsay, France
Abstract
In order to characterize the interactions of hydrogen with moving dislocations in Ni-based alloys, the strain-assisted transport and redistribution of tritium in tritiated tensile specimens were investigated by fl counting at room temperature. A predictive analysis of the existence of these interactions as a function of the most pertinent parameters (temperature, hydrogen activity, strain level and strain rate) was performed and compared with the results of experiments designed to characterize the mechanism of hydrogen-induced intergranular fracture. Tensile tests conducted on hydrogen precharged tensile specimens of alloy 600 at various temperatures and strain rates showed that hydrogen-induced intergranular cracking required hydrogen segregation to grain boundaries during plastic deformation. These experimental data were used to identify a temperature/strain rate domain at which hydrogen-induced intergranular rupture of alloy 600 was observed. Its concordance with the domain of hydrogen transport by dislocations provided support for a major influence of a mechanism of hydrogen-accelerated transport and supersaturation in the vicinity of grain boundaries. This influence was apparent in the intergranular rupture observed in the 180-500 K temperature range. Additional tests confirmed the existence of a local supersaturation of hydrogen associated with dislocation transport, of the transient character of the action of hydrogen and of its effects on embrittlement. The possible contribution of this mechanism to the intergranular stress corrosion cracking of alloy 600 is discussed. I. Introduction
Evidence of hydrogen-induced intergranular fracture of nickel and nickel-based alloys has long been observed and discussed [ 1-3]. The determinant role of hydrogen (H) plasticity interactions in the environment-induced cracking mechanisms of metals, including stress corrosion cracking (SCC), has been addressed more recently [4,5]. H was shown to affect the deformation mechanisms (enhancement of dislocation motion and stress relaxation) [6,7]. Conversely, deformation played a role on H absorption and
262
Hydrogen Permeation and Transport
redistribution in the material (stress- and strain-assisted trapping and transport) [8,9]. H-induced intergranular cracking was considered to result from a local enhancement of H concentration in the vicinity of grain boundaries. The primary purpose of this study was to obtain deeper insight into the mechanism responsible for the increase of H concentration locally in H precharged tensile specimens of alloy 600.
2. Experimental The composition of alloy 600 was as follows (wt.%): 0.067 C, 0.003 S, 0.006 P, 0.33 Si, 0.28 Mn, 16.2 Cr, 8.2 Fe, 0.23 Ti, 0.03 Cu, 0.19 AI, bal. Ni. Small, fiat tensile specimens (8 x 2 x 0.5 mm) were machined from the alloy under mill-annealed conditions (MA 1093 K). Several specimens were given an additional solutionizing treatment for 1 h at 1423 K and then quenched. The mill-annealed microstructure exhibits a 20-30 lam grain size, with numerous intergranular and intragranular chromium carbides. After quenching, the solutionized alloy was almost free of chromium carbides, and grain size range was 200--400 gm. Different experimental procedures were used to investigate the strain-assisted transport of H and its effects on the intergranular cracking of alloy 600.
2.1. Strain-assisted transport of H A highly sensitive technique is required to measure the ultra-low H quantities involved in a possible transport by moving dislocation. This measurement can be achieved by fl counting of tritiated specimens strained in a liquid scintillation counter. The technique, described in detail elsewhere [ 10], allowed monitoring of tritium release from a specimen during tensile deformation. For this purpose, tritium was first introduced into the sample by cathodic charging at 423 K for 5 h in a tritiated molten salt. The stress-strain curve was correlated with the curve representative of the cumulative amount of tritium released during the tensile test. This was done to investigate the possible occurrence of a mechanism of dislocation transport and furthermore to obtain quantitative data on the amount of tritium involved in this phenomenon.
2.2. Correlation of H transport~H-induced intergranular cracking In order to study the role of internal H on the intergranular rupture of alloy 600, small, flat tensile specimens were hydrogenated by cathodic charging at 423 K for 4 h in molten salts [9]. These charging conditions ensure a concentration of H of about 1850 at.ppm homogeneously distributed within the specimen. The hydrogen concentration in the samples was measured before and aider the tensile test with the fusion thermal conductivity method. Tensile tests were conducted at various temperatures and strain rates ranging from 77-550 K and 10-5-10 -1 s-1 respectively in order to determine whether a correlation exists between (i) the temperature/strain rate domain of existence of H transport by moving dislocations, and (ii) the domain where H-induced intergranular fracture occurs. A comparison of the stress-strain curves recorded with H-charged and H-free specimens with the corresponding fracture mode analyzed by scanning electron
Volume 1: Chemistry, Mechanics and Mechanisms
263
microscopy (SEM) allowed characterization of the hydrogen embrittlement. The relatively homogeneous elongation loss (If) was used as an embrittlement index representative of the extent of H-induced intergranular cracking: If = (Eo- EH)/Eo
where E0 and EH are the homogeneous elongation of H-free and H-charged tensile specimens respectively. In addition to conventional tensile tests, tensile straining procedures were used to test the dependence of the intergranular cracking on the build-up and on the dissipation of a strain-induced local transient concentration of H: "Interrupted tensile tests" performed at 293 K with a 2% increase of the plastic deformation were conducted step-by-step after aging at 293 K for 2 h between each successive deformation, and - Various combinations of tests were used to characterize the respective influence of the strain-induced redistribution of H and of the strain-induced increase of the trapping sites density. These tests involved, respectively, a pre-straining at 77 or 293 K and fracture at 293 or 77 K of specimens charged with H before or after pre-straining. 3.
Results and discussion
3.1. Experimental evidence of a strain-assisted transport of H in alloy 600
The existence of a strain-assisted desorption of tritium in alloy 600 has been evidenced by fl counting of tritiated tensile specimens strained in a liquid scintillation counter (Section 2.1). This effect is illustrated in Fig. 1(a). A comparison of the stressstrain curve (dots) with the time dependence of the cumulated amount of tritium released reveals a tritium desorption enhancement at the onset of plastic deformation. The desorption rate tended to decrease with increasing deformation levels up to 23%. However, a significant accelerated desorption was still observed when the straining was stopped and during the first stage of the stress relaxation. This accelerated desorption of tritium observed during the plastic deformation of large-grained recrystallized samples probably results from a dislocation transport mechanism, in agreement with similar observations made on single crystals [10]. Similar tests performed with pre-deformed tensile specimens (28%) with small grains (mill-annealed microstructure) showed no significant accelerated transport of tritium during the plastic deformation (Fig. l(b)). The capability of moving dislocations for long-range transport was presumably reduced in a small-grained microstructure and when the moving dislocations encountered a large density of dislocations of the forest. Moreover, the barrier effect of grain boundaries to dislocation motion in a smallgrained microstructure facilitated a strain-induced enrichment of H in the vicinity of grain boundaries.
264
Hydrogen Permeation and Transport
Fig. 1. Strain-induced desorption of tritium during tensile straining of tritiated samples of alloy 600: (a) solutionized for 1 h, 1150~ + quenched, 4 x 10-~ s-1, 23% plastic strain; and (b) mill annealed + 28% elongation, 5 x 10-5 s-1, 4.5% plastic strain.
3.2. Experimental evidence of the role of a strain-assisted transport and redistribution of H on the intergranular fracture 3.2.1. The need for a strain-induced redistribution of H The effect of ultra-low temperatures on the behaviour of H precharged tensile specimens of alloy 600 points to a need for a strain-induced redistribution of H to achieve intergranular cracking [9]. This effect is illustrated in Fig. 2. H trapping on grain boundaries or intergranular chromium carbides was expected to be maximal at low temperatures. Consequently, the fully-ductile fracture of hydrogenated samples pulled to rupture at 77 K (Fig. 2(a)) indicated that the local concentration associated with the equilibrium segregation and trapping of H at grain boundaries could not fully account for the strong intergranular fracture near room temperature (Fig. 2(d)). Such intergranular cracking implies the segregation of H to grain boundaries during plastic deformation. This strain-induced H redistribution was hypothesized to result from a dislocation transport mechanism.
Volume 1: Chemistry, Mechanics and Mechanisms
(a)
(b)
77 K, 10-3 s-~, If = 0%
293 K,
1 0 -3 s-i , If =
(d)
82%
265
(c)
193 K, 10-5 s-~, If = 47%
193 K, 5.2x10 -2 s-~,If= 5%
473 K, 10-3 s-~, If = 7%
473 K, 10-' s-l, If = 21%
(e)
(f)
Fig. 2. Effects of temperature and strain rate on the fracture mode and the ductility loss (If) of Hprecharged tensile specimens of mill-annealed alloy 600.
The effect of strain rate on the intergranular cracking at a given temperature (comparing Fig. 2(b) with 2(c) and Fig. 2(e) with 2(f)) is discussed in Section 3.2.2.
3.2.2. Correlation between the transport of H by dislocations and H-induced intergranular rupture of alloy 600 Predictive analysis: In order to determine the possible role of H-deformation interactions in the intergranular fracture of Ni-based alloys, an attempt was recently made [9] to determine the dependence of a mechanism of H transport by moving dislocations in alloy 600 on temperature, strain rate, and H activity. This simplified approach was based on previous studies [11,12]. However, it was also based on the assumption that the transport of H by dislocations is controlled by the H mobility in the material and by the carrying capacity of edge dislocations. Let us assume that for the H atmosphere to accompany a moving dislocation, the segregation velocity of H to the dislocation must be larger than the dislocation mean velocity. Given this assumption, a critical strain rate for H transport can be computed as a function of temperature, H-dislocation binding energy and the density of moving dislocations [9]. It is thus possible to define, as a function of temperature and strain rate, the limit between the two domains where dislocation transport can or cannot occur. As shown in Fig. 3(a), the transport was not expected to occur at low temperature because H diffusivity was too low.
Hydrogen Permeation and Transport
266
(critical strain rate) (s 1 ) 102
(a)
-~
...i
/
100
10 "2
no transport
~~~'~
10 -4 i
j/ li' /~,,[,
10-6
10 .8
--
15 kJ 29 kJ
"
II 108 cm -2
-
10 "10
0
100
200
300
400 500 temperature (K)
600
Fig. 3. Evaluations of the role of temperature, strain rate, and H activity on the mechanism of H dislocation transport in alloy 600: (a) temperature dependence of the computed value of the critical strain rate for H transport; and (b) temperature dependence of the computed value of the equilibrium occupancy fraction of H on an edge dislocation (closed system).
Assuming the existence of a local equilibrium between H and a moving dislocation, the carrying capacity of a dislocation can be expressed by the equilibrium occupancy fraction (0) of H on the dislocation. The latter can be computed as a function of the
267
Volume 1" Chemistry, Mechanics and Mechanisms
temperature, the binding energy, and the experimental conditions that control the activity of diffusible H [9]. Fig. 3(b) shows that the carrying capacity and, therefore, the effective transport of H were expected to decrease with increasing temperature. Experimental results: The temperature/strain rate domain of H-induced intergranular cracking assessed with precharged tensile specimens of mill-annealed alloy 600 is shown in Fig. 4. The ductility loss (If) was taken as an index of the extent of intergranular cracking, which is in agreement with SEM observations (Fig. 2). The absence of any ductility loss at low temperature (below 180 K) is consistent with the predicted absence of any strain-assisted redistribution of H by dislocation transport (Fig. 3(a)). 100
,(%)
C (at.ppm)
I
t~,. " "
80
2000
10"5s-1 10% 1 4.2 1 0 % 1
60
A
5.2 1 0 % 1
"-
101s "~
k
40
L
500
20
0 100
.
.
.
.
i
200
.
.
.
.
i
300
.
.
.
.
.
400
.
.
.
500
.
0 600
temperature (K)
Fig. 4. Effects of temperature and strain rate on H-induced intergranular fracture of H-precharged tensile specimens of mill-annealed alloy 600.
The gradual drop of the ductility loss with increasing test temperature was not a direct consequence of H desorption from the samples during the test. This effect is clearly evidenced by the H concentration measured in the tested samples (Fig. 4). In contrast, the temperature dependence of the ductility loss and of the carrying capacity of dislocations exhibited a strong correlation (Fig. 3(b)). Moreover, the results presented in Figs. 2 and 4 demonstrate that H-induced intergranular cracking was facilitated by lowering the strain rate at low temperature and by increasing it at high temperature. At low temperature, the increase with the strain rate of the critical temperature, below which the intergranular cracking disappeared, was consistent with the rise in strain rate of the critical temperature for H transport shown in Fig. 3(a). At high temperature, the significant enhancement of intergranular cracking with increasing strain rate (Figs. 2 and 4) was probably a consequence of the dissipation that may have occurred by thermal diffusion of the transient supersaturation raised by dislocation transport in the grain boundaries vicinity. These correlations, together with the concordance between the temperature/strain rate domain of H-induced intergranular cracking shown in Fig. 4 and the domain of H-induced serrated flow in Ni [13,14],
268
Hydrogen Permeation and Transport
strongly support a dominant role of H transport by moving dislocations on the intergranular rupture of alloy 600. 3.3. Investigation on the strain-induced transient supersaturation of H in the grain boundaries vicinity
The dynamic character of H deformation interactions in metals makes it very difficult to obtain direct evidence of a local supersaturation of H associated with dislocation transport and redistribution. However, specific tensile testing conditions provided evidence of the existence of a concentration built-up, its dependence on testing conditions, and the consequences on intergranular cracking. 3.3.1. Interrupted tensile tests Results of interrupted tensile tests performed at room temperature (Section 2.2) are illustrated in Fig. 5(a). When the specimen was given successive 2% deformations at 293 K with 2 h of aging between deformations, the H-induced ductility loss decreased by a factor of two. As shown by H concentration measurements, H loss during the test could not account for this reduced intergranular cracking. The reduced intergranular cracking was probably a consequence of dissipation, by thermal diffusion during the aging periods, of the concentration build-up near grain boundaries. A drop in the concentration of diffusible hydrogen leading to a decrease in the carrying capacity of moving dislocation might also contribute to the lowering of the embrittlement, as discussed below.
(a)
(b)
Fig. 5. Effect of the tensile testing procedure on H-induced ductility loss and intergranularcracking: (a) comparison of the ductility loss obtained with a conventional test and "interrupted tests;" and (b) effect of 4% pre-deformation at 77 K or 293 K.
3.3.2. Role of a pre-deformation At high temperature, when the H diffusivity was large, the dissipation by thermal diffusion of the concentration build-up might play a major role. At low temperature, antagonistic effects of H transport by moving defects and by H trapping on the
Volume 1: Chemistry, Mechanics and Mechanisms
269
dislocations of the forest must be considered. Such effects were investigated by comparing the ductility loss and the fracture mode of hydrogenated samples predeformed in such a way that the strain-assisted transport of H could occur either during the pre-straining step (pre-straining at 293 K followed by fracture at 77 K) or when the specimen was pulled to fracture (pre-straining at 77 K and fracture at 293 K). The results of these tests, where the hydrogenated samples were given a 4% pre-strain at 77 K or 293 K, are illustrated in Fig. 5(b). Comparison of the tensile curves of the predeformed alloy with those obtained without pre-straining clearly shows that prestraining at low temperature reduced the intergranular fracture at 293 K, whereas prestraining at 293 K reduced the ductility at 77 K. Similar observations were obtained with a 25% pre-strain (Fig. 6(a)). For this condition, the hydrogen embrittlement at 293 K disappeared fully after pre-straining at 77 K. Two major conclusions could be drawn from these results: When the pre-strain temperature is large enough for the transport and redistribution of H in the specimen to occur (293 K), a significant intergranular cracking results from the H-deformation interactions, In contrast, when the pre-strain temperature is sufficiently low to allow the density of strain-induced defects to increase without a concomitant H redistribution (77 K), the ductility loss measured at 293 K is reduced. -
-
Fig. 6. Influence of an increase of the density of strain-induced defects on the H-deformation interactions: (a) role of a 25% pre-strain at 77 K on the tensile behaviour at 293 K of H precharged alloy 600; and (b) temperature dependence of the computed value of the occupancy fraction of H on an edge dislocation in a closed system with a low H concentration and a high density of traps.
These results confirm that (i) H redistribution associated with dislocation transport is required to initiate intergranular cracking, and (ii) an increase of the forest dislocation density affects the transport of H by moving dislocations. The reduced ductility loss and intergranular cracking observed in this case could be explained by a lowering of the carrying capacity of moving dislocations, as illustrated in Fig. 6(b). During tensile straining of H-precharged samples, the H-metal system behaved as a closed system. That is, for a given H concentration, the occupancy fraction of dislocations could
Hydrogen Permeation and Transport
270
decrease when the dislocation density increased. As shown in Fig. 6(b), this decreased the concentration of diffusible H at or below room temperature and, as a result, reduced the carrying capacity of the moving traps.
3.4. Role of the strain-assisted transport of H on the intergranular stress corrosion cracking of alloy 600 The experimental data presented herein indicate that the strain-assisted transport of H controls the intergranular cracking of H-precharged tensile specimens of alloy 600 strained in the 180-500 K temperature range. These results suggest that this transport mechanism might play a similar role in the intergranular SCC (IGSCC) of alloy 600 if (i) a local plastic deformation is involved in the SCC mechanism, and (ii) the carrying capacity of moving dislocations is large for the considered SCC conditions. Under SCC conditions, the H-metal system at the crack tip in contact with the environment behaved as an open system. To evaluate the possible role of H transport in IGSCC, the equilibrium occupancy fraction of H on an edge dislocation was computed for alloy 600 exposed to various values of hydrogen pressure [9]. The temperature dependence of the trap occupancy fraction for different values of H-trap binding energy is illustrated in Fig. 7. m m -.,ynm,
m m
m m
m ~
~ ~ .
^ ~
^ v
^ v
-v
~
0.8
,.
0.6 0
0.4 O 60KJ
0.2
I I 29 KJ
m
o
....
, ....
1oo
_:
m
".
"' 1 0 0 0 a t m " $ . . \ """'0.. I I . -. "Ik'"W','L',~L" " ', ""O-..~ , . :~. " : ' ~ ~-,m,-,.~...~...~._.~_._, v 200
300 400 temperature (K)
500
600
Fig. 7. Temperature dependence of the equilibrium occupancy fraction of H on an edge dislocation in alloy 600 (open system).
As shown in Fig. 7, the trap occupancy fraction and, as a consequence, the ability of dislocations to transport H in SCC conditions depend strongly on temperature, H activity and the binding energy with the moving trap. One can therefore infer that below 400 K, H transport by dislocation can occur in most hydrogen-rich environments, providing the temperature is not too low to prevent H segregation toward the dislocation. At higher temperatures, H transport requires either a high H activity on the surface or a
Volume 1: Chemistry, Mechanics and Mechanisms
271
large binding energy. The role of H on IGSCC in alloy 600 in a pressurized water reactor (PWR) environment remains an open question [4,5]. The contribution of a mechanism based on the transport of H by dislocations would imply a large trap occupancy fraction at 600 K and thus a large H activity at the crack tip (of the order of 1000 atm). This effect is not confirmed by the available experimental data [13]. Alternatively, H transport at 600 K would imply a high binding energy (--60 kJ) between H and the moving trap. This value is comparable to the binding energy reported for H-vacancies interactions [ 14]. Because H-vacancies interactions may occur at the crack tip in primary water, additional studies are required to better characterize the mechanisms of H-induced enhancement of the vacancy concentration and of H transport associated with the migration of vacancies. Moreover, the possible contribution of these mechanisms to the propagation of stress corrosion cracks at high temperature must be elucidated if a thorough understanding of the phenomenon is to be attained.
4. Summary 9 The strain-assisted desorption of tritium in alloy 600 evidenced by fl counting during tensile straining of tritiated specimens can be attributed to H transport by dislocations. 9 The intergranular cracking of H-precharged tensile specimens required a straininduced redistribution of H in the alloy. 9 The temperature/strain rate domain, where H can be transported by dislocations, correlated with the temperature/strain rate domain of H-induced intergranular rupture of alloy 600. 9 Specific tensile testing conditions provided evidence of a strain-induced concentration build-up of H in the grain boundaries vicinity and its dependence on temperature. 9 The strain-assisted transport of H controlled the intergranular cracking of alloy 600 that was strained in the 180-500 K temperature range. Further investigations are required to identify the role of the strain-assisted transport of H on the IGSCC of alloy 600 at high temperature (PWR environment).
References [1] T. Boniszewski, G.C. Smith, Acta Metall. 11 (1963) 165-178. [2] P. Menut, Y. Shehu, J. Chine, M. Aucouturier, R61e de la structure et de la composition chimique des joints de grains sur la rupture intergranulaire du nickel en pr6sence d'hydrog/me cathodique, in: P. Azou (Ed.), Hydrog~ne et Mat6riaux: Proc. 3rd International Congress on Hydrogen and Materials, vol. 2, Paris, 1982, pp. 857-862. [3] R.H. Jones, S.M. Bruemmer, Environment-induced crack growth processes in nickel-base alloys, in: R.P. Gangloff, M.B. Ives (Eds.), Environment-Induced Cracking of Metals, NACE, Houston, 1990, pp. 287-310. [4] T. Magnin (Ed.), Corrosion-Deformation Interactions, CDI'96, The Institute of Materials, London, 1997. [5] N.R. Moody, A.W. Thompson, R.E. Ricker, G.W. Was, R.H. Jones, (Eds.), Hydrogen Effects on Materials Behavior and Corrosion Deformation Interactions, TMS, Warrendale, 2003. [6] P. Sofronis, H.K. Birnbaum, J. Mech. Phys. Solids 43 (1) (1996) 49-90.
272 [7] [8] [9] [10] [11]
Hydrogen Permeation and Transport
J.P. Chateau, D. Delafosse, T. Magnin, Acta Mater. 50 (2002) 1507-1522. A-M. Brass, J. Chine, Mater. Sci. Eng. A242 (1998) 210-221. J. Chine, A-M. Brass, Metall. Mater. Trans. A, 35A (2004) 457-464. J. Chine, A-M. Brass, Scripta Mater. 40 (1999) 537-542. J.K. Tien, A.W. Thompson, I.M. Bemstein, R.J. Richards, Metall. Trans. A, 7A (1976) 821-829. [12] J.K. Tien, S.V. Nair, R.R. Jensen, Dislocation sweeping of hydrogen and hydrogen embrittlement, in: I.M. Bemstein, A.W. Thompson (Eds.), Hydrogen Effects in Metals, TMS, Warrendale, 1981, pp. 37-56. [13] D.S Morton, S.A. Attanasio, G.A. Young, P.L. Andresen, T.M. Angeliu, The influence of dissolved hydrogen on nickel alloy SCC: a window to fundamental insight, CORROSION/2001, NACE International, Houston, 2001, paper no. 1117. [14] F. Besenbacher, S.M. Myers, J.K. Norskov, Nucl. lnstrum. Meth. 7B-8B (1985) 55-66.
273
Hydrogen in trapping states harmful and resistant to environmental degradation of high-strength steels Kenichi Takai Department of Mechanical Engineering, Sophia University, Tokyo 102-8554, Japan Abstract
Hydrogen in trapping states that are resistant to environmental degradation of the mechanical properties of high-strength steels has been separated and extracted using thermal desorption analysis (TDA) and slow strain rate tensile tests. High-strength steel occluding only hydrogen desorbed at a low temperature (peak 1), as determined by TDA, decreases in maximum stress and plastic elongation with increasing occlusion time of peak 1 hydrogen. Thus the trapping state of peak 1 hydrogen is directly associated with environmental degradation. The trap activation energy for peak 1 hydrogen is 23.4 kJ/mol. Therefore, the peak 1 hydrogen corresponds to weaker binding states and diffusible states at room temperature. In contrast, the high-strength steel occluding only hydrogen desorbed at high temperature (peak 2), by TDA, maintains the maximum stress and plastic elongation in spite of an increasing content of peak 2 hydrogen. This indicates that the peak 2 hydrogen trapping state is resistant to environmental degradation, even though the steel occludes a larger amount of peak 2 hydrogen. The trap activation energy for peak 2 hydrogen is 65.0 kJ/mol, which indicates a stronger binding state and nondiffusibility at room temperature. I. Introduction
In order to clarify the mechanism of environmental degradation of the mechanical properties, thermal desorption analysis (TDA) of hydrogen in high-strength steels [ 1-6], titanium alloys, ceramics and vitreous silica glass [7] has been studied extensively over the past few years. TDA enables us to separate the hydrogen trapping states in materials based on the peak temperatures of hydrogen desorbed from materials during heating. These peak temperatures are dependent on the metallurgical microstructures of steels, as shown by Takai et al. [8-10]. Martensitic steel desorbs hydrogen at temperatures below 200~ (peak 1), whereas cold-drawn pearlitic steel desorbs not only peak 1 hydrogen but also hydrogen at temperatures between 200 and 500~ (peak 2). Total hydrogen content in high-strength steels was measured and correlated with environmental degradation of the mechanical properties. Thereatter, it was reported that the peak 1 hydrogen trapping state would affect environmental degradation as shown by
274
Hydrogen Permeation and Transport
Suzuki et al. [ 11 ], whereas the peak 2 hydrogen trapping state would not. However, experimental verification of these results is lacking. In addition, it is not yet established whether the peak 2 hydrogen trapping state does not affect environmental degradation, even though a large amount of peak 2 hydrogen is occluded in high-strength steels. Given such a background, high-strength steels occluding only peak 1 hydrogen and those occluding only peak 2 hydrogen have been prepared to investigate the relationship between each trapping state of hydrogen and environmental degradation of the mechanical properties. Based on the anticipated results, we intend to separate and extract the harmful and resistant trapping states of hydrogen in high-strength steels to the degradation, and then to clarify the trapping sites corresponding to the two hydrogen trapping states. Furthermore, the mechanism and model for harmful and resistant states of hydrogen to the degradation will be discussed.
2. Experimental 2.1. Materials
The high-strength steel employed in the present study was eutectoid steel (JIS SWRS 82B) with a chemical composition as follows: 0.84% C, 0.19% Si, 0.76% Mn, 0.008% P, 0.008% S, and 0.01% Cu. A specimen occluding only peak 1 hydrogen was fabricated as follows. The specimen was cold-drawn from 13 to 5 mm in diameter, then, a~er austenitizing at 950~ for 15 min, it was transformed isothermally in a lead bath at 350~ for 30 min. This specimen was designated as the "350~ '' The specimen occluding only peak 2 hydrogen was fabricated as described by Takai et al. [12]. An isothermal treatment consisting of austenitizing at 950~ for 15 min and transforming in a lead bath at 550~ for 30 min, followed by cold-drawing with 85% reduction (true strain 1.91), was applied to the specimen at 13 mm diameter to a final diameter of 5 mm. Because this cold-drawn eutectoid steel occludes both peak 1 and peak 2 hydrogen, the steel was heated to 200~ to remove peak 1 hydrogen. This specimen occluding only peak 2 hydrogen was designated as the "550~ specimen." Specimen preparation conditions and tensile strengths are shown in Table 1. The microstructures of these specimens were observed with transmission electron microscopy (TEM). Based on TEM observations, the metallurgical microstructure of the 350~ was bainite. The morphology of precipitated Fe3C in the 350~ specimen was similar to that in tempered martensite. The metallurgical microstructure 0 of 550 o C-85'~-specimen was cold-drawn pearlite. In addition, dark-field images obtained by TEM indicated that Fe3C in the 350~ was single crystal, while Fe3C in the 550~ was a fine-grained nano-polycrystal caused by colddrawing (the images thereof are omitted in this paper. 2.2. Hydrogen occlusion
Specimens were immersed in an aqueous solution of 20 mass % NH4SCN kept at a temperature of 50~ under various immersion times to occlude hydrogen. Hydrogen is easily occluded in steels in such a solution. The conditions for hydrogen occlusion were the same as those standardized by F6d6ration Intemationale de la Pr6contrainte (FIP) Report [ 13] for the delayed fracture tests of prestressed steels.
Volume 1: Chemistry, Mechanics and Mechanisms
275
Table 1 Specimen preparation conditions and tensile strength Specimen preparation Identity
350~ 550~
Isothermal transformation temperature (~ 350 550
Reduction in area (%) 0 85
Tensile strength (MPa)
1517 1918
2.3. Thermal desorption analysis The TDA apparatus consisted of a gas chromatograph, a thermal conductivity detector, a heating unit, and a recorder. The gas chromatograph was calibrated with standard-mixture gas (Ar + 50 ppm H2) to quantify the hydrogen content in the specimens. TDA was conducted in the temperature range 25-800~ Sampling of the carrier gas to the gas chromatograph was at 5 min intervals to obtain hydrogen desorption profiles. To prevent desorbtion, the hydrogen content in 5 mm diameter specimens was measured immediately after occluding of hydrogen. In order to calculate the trap activation energy of hydrogen, the specimens were measured at various heating rates (200, 300, 400 and 500~ immediately after immersion in NH4SCN solution for24 h.
2. 4. Environmental degradation test The degree of the environmental degradation of mechanical properties was measured as a decrease in maximum stress and plastic elongation during the application of tensile stress by using the slow strain rate tensile (SSRT) technique. The 350~ specimen and 550~ for SSRT tests were processed to 3 mm in diameter and 20 mm in gauge length. The 350~ was mounted on the SSRT test apparatus immediately after occluding peak 1 hydrogen. The 550~ was mounted on the SSRT test apparatus after occluding both peak 1 and peak 2 hydrogen, and then removing only peak 1 hydrogen by heating to 200~ SSRT tests were conducted at a strain rate of 5 x 10 -7 s -l i n ambient air. At this strain rate, the specimens failed within approximately 48 h. A negligible amount of hydrogen was released within the test time of 48 h, as will be discussed later in this paper. 3. Results
3.1. Separation of hydrogen trapping states in high-strength steels Fig. 1 shows the hydrogen desorption profiles during continuous heating of the specimens measured by using TDA at the heating rate of 200~ immediately after immersion in NH4SCN solution for 24 h. The 350~ occluded only the peak 1 hydrogen desorbed at the temperature of 120~ The 550 o C-85 O~-specimen occluded both peak 1 and peak 2 hydrogen desorbed at 120 and 370~ respectively. This
Hydrogen Permeation and Transport
276
difference in hydrogen desorption peaks was believed to be due to the trapping states of peak 1 and peak 2 hydrogen. ~'0.20
I (a) 350~
.,..~
&0.15 .~ 0.10 o >
= 0.05 O
~
0
i
0
~. 0.20 E
200 4()0 Temperature (~
600
(b) 550~
.,.~
~0.15 200o C =0.10 >
~0.05 o
0 0
.......Y ' ' -''.~ ..... 200 400 600 Temperature (~
Fig. 1. Hydrogen desorption profiles of high-strength steels by TDA.
3.2. Evaluation of environmental degradation Stress-strain curves of the 350~ occluded only peak 1 hydrogen, while the 550~ occluded only peak 2 hydrogen under various immersion times in NHaSCN solution, obtained by SSRT tests (Fig. 2). The maximum stress and strain of the 350~ decreased with increasing immersion time, significantly, increased content of peak 1 hydrogen. Decrease in the maximum stress and strain ceased after specimen immersion for 3 h. In contrast, the stress-strain curves of the 550 o C-85 O~-specimen were identical regardless of immersion time, that is, regardless of the content of peak 2 hydrogen, as shown in Fig. 2(b). However, the 550~ specimen, which occluded both peak 1 and peak 2 hydrogen without heating to 200~ decreased in maximum stress and strain with increasing immersion time. These results suggest that trapping states of peak 1 hydrogen were harmful, while those of peak 2 hydrogen were resistant to the environmental degradation.
277
Volume 1: Chemistry, Mechanics and Mechanisms
(a) 350~
2000
[ 3h
2200
Oh
lh
(a)
6h
M 2000 ] U
1500
(peak 2:2.9 ppm) M Fq
/,
550~
9
1800
1000
M
E 350oC_specimen
1600
500
~ ~ . i
0 2000
|
i
5 10 Strain (%)
b) 550~
15
~ ....
1400
0
(b) 4h
/
1000
/
...a
500
.o
lh
3h
Oh 6h
(peak 2:2.9 ppm)
M
6
@
/]'
T
550~
8 I
I
~, 1500
I_
I0
18h
/~
(pe~k 1:3.1 ppm) A 9 6 12 18 24 Immersion time (h) ,
4
[3
[3
350~ (peak 1" 3.1 ppm)
r~
2 w
9
0
,
,
.
,
.
i
.
.
,
|
,
i
a
5 10 Strain (%)
i
|
,
|
15
Fig. 2. Stress-strain curves of: (a) 350~ specimen occluding only peak 1 hydrogen and (b) 550~ occluding only peak 2 hydrogen.
Immersion time (h)
Fig. 3. Effects of immersion time in NH4SCN solution on (a) maximum stress and (b) plastic elongation.
Fig. 3 shows the maximum stress and plastic elongation as a function of occlusion times of peak 1 and peak 2 hydrogen. Concentrations in parenthesis show hydrogen content of peak 1 and peak 2. The hydrogen concentrations of only 18 and 24 h are shown in Fig. 3, since hydrogen distribution in the specimen is presumably uniform at immersion times of 18 and 24 h. The maximum stress and plastic elongation initially decreased with increasing occlusion time of peak 1 hydrogen, and then this decrease ceased. Note that the maximum stress and plastic elongation did not change even though the content of peak 2 hydrogen increased to as high as 2.9 mass ppm. Fig. 4 shows scanning electron microscopy (SEM) fractographs of specimens occluding only peak 1 and only peak 2 hydrogen. The 350~ occluding 0.8 mass ppm of only peak 1 hydrogen underwent brittle fracture (Fig. 4(a)). In contrast, the 550~ occluding as much as 2.9 mass ppm of only peak 2 hydrogen underwent ductile fracture with elongation and necking (Fig. 4(b)). Fig. 4(b) shows a fracture mode identical to that of the initial specimen without hydrogen occluded. These results indicate that the trapping state of peak 1 hydrogen impeded the
278
Hydrogen Permeation and Transport
environmental degradation of the mechanical properties of high-strength steel, even though the steel occluded a small amount of peak 1 hydrogen. Indeed, the trapping state of peak 2 hydrogen was resistant to degradation, even though the steel occluded a larger amount of peak 2 hydrogen.
Fig. 4. SEM fractographs of (a) 350~ specimen occluding only peak 1 hydrogen, and (b) 550~ occluding only peak 2 hydrogen.
Fig. 5. Hydrogen release in ambient atmosphere under various holding times for (a) 350~ specimen occluding only peak 1 hydrogen, and (b) 550~ occluding both peak 1 and peak 2 hydrogen.
4. Discussion
4.1. Trapping states and sites corresponding to peak 1 and peak 2 hydrogen In this study, the mobility of peak 1 and peak 2 hydrogen in high-strength steels was measured by using TDA. Fig. 5 shows the desorption process of peak 1 and peak 2 hydrogen for specimens kept in ambient conditions for various holding times at room temperature after immersion in NH4SCN solution for 24 h. The content of peak 1 hydrogen gradually decreased with holding time (Fig. 5(a)). Hence, peak 1 hydrogen corresponded to weaker binding states and diffusible states at room temperature. In
Volume 1: Chemistry, Mechanics and Mechanisms
279
contrast, the content of peak 2 hydrogen initially decreased with holding time, but then remained at 2.7 mass ppm atter 200 h (Fig. 5(b)). The content of peak 2 hydrogen of 3.9 mass ppm in the 550~ immediately atter occluding hydrogen also decreased to 2.7 mass ppm after heating to 200~ to remove only peak 1 hydrogen, but remained at 2.7 mass ppm at room temperature over 1200 h (the data were omitted in this paper). In addition, the content of peak 2 hydrogen below 2.7 mass ppm in the 550~ initially did not decrease with holding time; it remained relatively constant. From these results it is evident that peak 2 hydrogen above 2.7 mass ppm consisted primarily of nondiffusible hydrogen but also contained a small amount of diffusible hydrogen. However, the peak 2 hydrogen in the specimen used in SSRT tests corresponded to nondiffusible states at room temperature. Additionally, these hydrogen-desorption processes show that a minute amount of hydrogen was released within the SSRT time of 48 h. In order to evaluate the hydrogen-binding states qualitatively, the trap activation energy of hydrogen was calculated from the dependence of the peak temperatures on heating rates. The equation for trap activation energy (Eo) was shown by Choo et al. [14]: c31n(~/Tp2) / 0 (1/Tp ) = -EJR
(1)
where # is the heating rate (K/s), Tp is the peak temperature (K), and R is the gas constant (8.13(J.K-l.mol-1)). The trap activation energy for peak 1 hydrogen was 23.4 kJ/mol and that for the peak 2 hydrogen 65.0 kJ/mol. These values confirmed that the trapping state of peak 1 hydrogen was unstable, whereas that of peak 2 hydrogen was stable. The authors recently investigated hydrogen trapping sites corresponding to each desorption peak on the basis of TDA and of secondary ion mass spectrometry (SIMS) results for cold-drawn pure iron [8], eutectoid steel [12], and spheroidal cast iron [15]. The content of only peak 1 hydrogen increased with the reduced degree of the colddrawn area in pure iron, whereas the contents of both peak 1 and peak 2 hydrogen increased with the reduced degree of the cold-drawn area in eutectoid steel. These results indicate that peak 2 hydrogen was occluded into the specimen when the pearlite microstrueture consisting of lamellar Fe3C was cold-drawn. The trapping sites of peak 1 hydrogen should correspond to complex sites such as vacancies, vacancy clusters, strain field of dislocations, dislocation cores, grain boundaries, and ferrite/Fe3C interfaces. Accordingly, the trap activation energy obtained from Eq. (1) is the average value of desorption from these sites and from interstitial hydrogen in lattice without trapping. On the other hand, the trapping sites of peak 2 hydrogen should correspond to the strained interfaces between ferrite and Fe3C and/or interface dislocations enclosed between Fe3C lamellae due to cold-drawing. 4.2. Relationship between hydrogen trapping states and the environmental degradation
of mechanical properties The relationship between hydrogen trapping states and the degree of environmental degradation is discussed on the basis of the fact that the trapping state of peak 2 hydrogen is innocuous while that of peak 1 hydrogen is harmful to the environmental
280
Hydrogen Permeation and Transport
degradation of mechanical properties. Fig. 6 shows a schematic illustration of stressinduced hydrogen diffusion in front of a crack tip in high-strength steel with and without cold-drawing. It is widely recognized that the maximum stress region occurs around the elastic-plastic boundary in front of the crack tip in the case of the plane strain state. The stress-induced diffusion of hydrogen is depicted by arrows in Fig. 6. The gray scale in front of the crack tip indicates the stress level, while the smallest, black region corresponds to a high stress level. In the triaxial stress field in front of the crack tip, hydrogen accumulates in the region of the elongated lattice by means of elastic interaction. Furthermore, because dislocation and microcracks occur in the maximum stress region, hydrogen is locally accumulated by stress-induced diffusion.
(a) High-strengthsteel occludingpeak 1 hydrogen (b) High-strengthsteel occludingpeak 2 hydrogen Fig. 6. Schematic illustration of stress-induced hydrogen diffusion in front of the crack tip of high-strength steel with/without cold-drawing: (a) 350~ occluding only peak 1 hydrogen, and (b) 550~ occluding only peak 2 hydrogen.
In the case of the 350~ hydrogen trapped weakly in ferrite and along the interfaces of Fe3C is desorbed as peak 1 hydrogen (Fig. l(a)). When tensile stress is applied to the specimen during SSRT tests, a hydrogen potential gradient corresponding to the stress gradient appears at the front of the crack tip as the initiation of a microcrack. A much higher local hydrogen content than the average hydrogen content then arises around the elastic-plastic boundary. Thus, the diffusible peak 1 hydrogen in the 350~ is believed to accumulate, causing environmental degradation such as decrease of maximum stress and plastic elongation in spite of hydrogen of 0.8 mass ppm, as shown in Fig. 4(a). The trap activation energy of 23.4 kJ/mol for peak 1 hydrogen suggests that both the driving force energy required for stress-induced diffusion during elastic and plastic deformation, and that required for hydrogen dragging by dislocation mobility during plastic deformation, were higher than binding energy between hydrogen and trapping sites. The trapping state of peak 1 hydrogen was, therefore, associated directly with environmental degradation; that is, this was a harmful trapping state.
Volume 1: Chemistry, Mechanics and Mechanisms
281
In the case of the 550~ hydrogen trapped strongly along the strained interfaces of Fe3C was desorbed as peak 2 hydrogen (Fig. l(b)). Peak 2 hydrogen in the specimen immediately atter occluding hydrogen above 2.7 mass ppm consisted primarily of nondiffusible hydrogen but also contained trace amount of diffusible hydrogen. The content of peak 2 hydrogen in the specimen after heating to 200~ did not decrease with holding time: it remained at 2.7 mass ppm at room temperature over 1200 h. That is, peak 2 hydrogen in the 550~ for environmental degradation test was only in the irreversibly trapped state at room temperature. This suggests that the nondiffusible peak 2 hydrogen in the 550~ specimen did not accumulate in front of the crack tip and did not cause environmental degradation in spite of hydrogen content as high as 2.9 mass ppm (Fig. 4(b)). The trap activation energy of 65.0 kJ/mol for peak 2 hydrogen suggests that the driving force energy required for stress-induced hydrogen diffusion during elastic and plastic deformation, and that required for hydrogen dragging by dislocation mobility during plastic deformation, were lower than binding energy between hydrogen and trapping sites. Therefore, the trapping state of peak 2 hydrogen is not associated with environmental degradation. That is, this is an innocuous trapping state even though the steel occluded as much as 2.9 mass ppm of peak 2 hydrogen.
5. Summary Hydrogen in trapping states innocuous to environmental degradation of highstrength steels has been separated and extracted using TDA and SSRT techniques. The results of the present study can be summarized as follows: 9 The high-strength steel occluding hydrogen desorbed at low temperature (peak 1), as determined by TDA, decreased the maximum stress and plastic elongation with increasing occlusion time of peak 1 hydrogen. Fractographic examination showed that peak 1 hydrogen caused the brittle fracture of high-strength steel. Thus the trapping state of peak 1 hydrogen was associated directly with the environmental degradation of mechanical properties. 9 The high-strength steel occluding hydrogen desorbed at higher temperature (peak 2), as determined by TDA, retained the maximum stress and plastic elongation in spite of increasing content of peak 2 hydrogen. Fractographic examination showed that peak 2 hydrogen as high as 2.9 mass ppm caused the ductile fracture of highstrength steel. Thus the trapping state of peak 2 hydrogen was resistant to environmental degradation of mechanical properties even though the steel occluded a large amount of peak 2 hydrogen. 9 The peak 1 hydrogen corresponded to weaker binding states and diffusible states at room temperature. The trap activation energy for peak 1 hydrogen was 23.4 kJ/mol. The trapping sites of peak 1 hydrogen should correspond to complex sites such as vacancies, vacancy clusters, strain field of dislocations, dislocation cores, grain boundaries, ferrite/Fe3C interfaces, and interstitial hydrogen in lattices without trapping. 9 The peak 2 hydrogen corresponded to stronger binding states and nondiffusible states at room temperature. The trap activation energy for peak 2 hydrogen was 65.0 kJ/mol. The trapping sites of peak 2 hydrogen should correspond to the strained
282
Hydrogen Permeation and Transport
interfaces between ferrite and Fe3C and/or interface dislocations enclosed between Fe3C lamellae due to cold-drawing. 9 Because the peak 2 hydrogen was nondiffusible at room temperature and its trap activation energy was 65.0 kJ/mol, this hydrogen apparently does not accumulate in front of a crack tip nor does it cause environmental degradation in spite of being present in amounts as high as 2.9 mass ppm. The trap activation energy of 65.0 kJ/mol for peak 2 hydrogen suggests that the driving force energy required for stress-induced hydrogen diffusion during elastic and plastic deformation and that required for hydrogen dragging by dislocation mobility during plastic deformation were higher than binding energy between hydrogen and trapping sites. References [ 1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [ 12] [ 13] [14] [15]
M. Nagumo, ISIJ Intem. 41 (2001) 590-598. M. Nagumo, K. Takai, N. Okuda, J. Alloys Comp. 293-295 (1999) 310-316. M. Nagumo, M. Nakamura, K. Takai, Metall. Mater. Trans. A, 32A (2001) 339-347. D.G. Enos, J.R. Scully, Metall. Mater. Trans. A, 33A (2002) 1151-1166. K. Takai, Bull. Jap. Inst. Met. 39 (2000) 239-242 (in Japanese). K. Takai, Boshoku Gijutsu (Corros. Eng.) 49 (2000) 395-405. K. Takai, Report on the Project "Function of Hydrogen on Environmental Degradation of Structural Materials-IV," ISIJ, Tokyo, 2002 (in Japanese). K. Takai, G. Yamauchi, M. Nakamura, M. Nagumo, J. Jap. Inst. Met. 62 (1998) 267-275 (in Japanese). K. Takai, J. Seki, Y. Homma, Tetsu-to-Hagan6 81 (1995) 1025-1030 (in Japanese). K. Takai, Y. Homma, K. Izutsu, M. Nagumo, J. Jap. Inst. Met. 60 (1996) 1155-1162 (in Japanese). N. Suzuki, N. lshii, T. Miyagawa, Tetsu-to-Hagan6 82 (1996) 170-175 (in Japanese). K. Takai, A. Nozue, J. Jap. Inst. Met. 62 (2000) 669-676. F6d6ration lntemationale de la Pr6contrainte, Report on Prestressing Steel 5, FIP, Sept. 1980, pp. 1-56. W.Y. Choo, J.Y. Lee, Metall. Mater. Trans. A, 13A (1982) 135-140. K. Takai, Y. Chiba, K. Noguchi, A. Nozue, Metall. Mater. Trans. A, 33A (2002) 26592665.
285
Ductile crack initiation and growth promoted by hydrogen in steel Y. S himomura a, M. Nagumo b Department of Materials Science and Engineering, Waseda University, Tokyo, Japan a Present address: Steering System Engineering Group, Vehicle Component Engineering Department no. 2, NISSAN MOTOR Co. Ltd., Okazukoku, Atugi, Kanagawa 243-0192, Japan b Present address: Konodai 1-8-10, Ichikawa, Chiba 272-0827, Japan
Abstract The effect of hydrogen in successive stages leading to fracture was examined in a three-point bending test of low-carbon steel with the objective of elucidating the mechanism of hydrogenrelated failure. Measurement of the crack-opening displacement, detection of ductile crack initiation by means of an electric potential method, and R-curve analysis of stable ductile crack growth were conducted. Hydrogen (i) suppressed the blunting of the pre-notch even prior to the onset of a ductile crack, (ii) promoted crack initiation, and (iii) reduced the resistance to ductile crack growth. Fractographic features suggested that the effect of hydrogen was to reduce plasticity associated with crack growth. Enhanced strain localization by hydrogen associated with the creation of defects in front of the notch and crack is consistent with these results. Hydrogen likely enhances the creation of vacancies during plastic deformation, which would be consistent with the vacancy agglomeration model for the mechanism of hydrogen-related failure.
I. Introduction Researchers have discussed the mechanism of hydrogen-related failure of steels from various perspectives. Most previously-proposed mechanisms have focused on the function of hydrogen in elementary processes inherent to fracture. These processes include dislocation mobility [ 1], lattice cohesive strength [2], and the surface energy of a crack [3]. A model must be examined for conformity to characteristic features of the actual failure before it can be validated. Strong dependence of the susceptibility of steel to degradation on its microstructure is another subject that a model must explain. The degradation of mechanical properties results from successive stages that lead to the final fracture. Accordingly, the role of hydrogen in each stage of the overall fracture process must be examined. The origin of the microstructure dependence of the susceptibility to failure is also expected to be revealed in the effect of hydrogen in each stage.
286
Hydrogen-Assisted Cracking and Embrittlement
Striations that appear on the fracture surface of hydrogen-charged iron [4] and Fe3% Si steel [5,6] are characteristic features of hydrogen-related failure. These striations are observed along traces of slip planes coincident with the extension of highlydefective slip bands that develop within the specimens [7]. Another feature of hydrogen-related failure is strong shear localization that appears as slip bands associated with crack formation along characteristic slip lines in notched specimens subjected to bending tests [8,9]. Hirth [10] proposed an autocatalytic effect of void formation at hard particles and associated promotion of shear localization. Two possible roles of hydrogen have been posited: The first is that hydrogen causes separation of the interface between particles and the matrix by precipitating as molecular hydrogen. The other role is that hydrogen enhances localized plasticity caused by hydrogen in solution. Shear localization associated with defect creation is expected to affect ductile crack initiation and growth stages. A useful method of analyzing the fracture process is the resistance curve (R-curve) that denotes the J-integral value vs. advanced crack length [11]. Except in highly brittle cases, the initiation and stable growth of a ductile crack followed by the onset of unstable brittle fracture can be separated on an R-curve. Because the J-integral is a measure of the energy dissipated during straining, R-curve analysis of hydrogen degradation is capable of identifying the effect of hydrogen on degradation at each stage. Ductile fracture is a process that involves defect creation. Concerning the defects associated with plasticity, the creation of a substantial number of vacancy-type defects during straining has been revealed in ferritic and martensitic steels, using hydrogen as a tracer of defects [12,13]. It has been shown with ferritic steels that vacancy creation is associated with the constraint of slip extension across grain boundaries [14]. Ductile crack growth resistance correlates well with the increase in the amount of straininduced vacancies, the extent of which depends on the microstructure [ 15]. This implies that the ductile fracture process is strongly affected not only by the presence of secondphase particles as the void source, but also by the dynamic creation of defects during straining. Hydrogen enhances the strain-induced creation of vacancies [16]. Because the features of hydrogen-related failure suggest a role is played by plasticity, it has been assumed that hydrogen functions to enhance the creation of vacancies and their clusters in strain-concentrated areas such as slip bands [16,17], connecting the fractographic feature and the formation of defective slip bands. Correlations between strain-induced vacancies and the susceptibility to hydrogen-related failure have been demonstrated for martensitic steels with various microstructures [ 17,18]. Ductile crack growth is a process of both void initiation and linking for which plasticity is the critical factor. If the primary function of hydrogen is the creation of vacancies rather than the enhancement of plasticity, hydrogen clearly promotes ductile crack initiation. The J-integral at the onset of a ductile crack, J~c, is normally identified as the intersection of the blunting line and the stable crack growth line on an R-curve. However, the identification is seldom exact. The crack initiation stage preceding the growth of a stable ductile crack has been examined in our project based on the results of a study that demonstrated the effect of hydrogen on ductile crack growth resistance [19].
287
Volume 1: Chemistry, Mechanics and Mechanisms
2. Experimental procedures The chemical composition of the steel used in our study was 0.08 C, 0.52 Si, 0.005 Mn, 1.99 Ni, and 0.008 N (wt.%) with banded ferrite and peadite structures. The yield and tensile strength were 315 and 455 MPa respectively, and were within • of the specimens with/without hydrogen charging. A three-point bending test was conducted using specimens of the geometry shown in Fig. 1. The notch was 0.1 and 3 mm in width and depth respectively, at the tip of a machined notch of 1.5 and 6 mm in width and depth respectively. Crack propagation was in the transverse direction on the plane normal to the rolling direction.
I_ r
I_..,.
..J
50 100
III
--]
/
Fig. 1. Geometry of three-point bending test specimen. Tests were conducted in accordance with standard procedures for evaluating jr_ integral (JSME S001-1981) [20] and crack-opening displacement (COD) (BS 57621979) [21 ] at a bending deformation rate of 1 mm/min. Both quantities were measured simultaneously once per second, and the data stored in a computer. R-curves were obtained from J-integral values at various loads and corresponding crack extension lengths. The latter were observed by scanning electron microscopy of the fracture surface of specimens subjected to unloading and fracture at liquid nitrogen temperature. Details of the bending test procedure are described in Ref. [ 19]. The growth stage of a stable ductile crack was discerned from the blunting stage of the pre-notch by an inflection of the gradient of the R-curve. The onset of the stable ductile crack was simultaneously detected by using a potential method [22] that measured the electric potential between electrodes fixed across the ligament in front of the pre-notch under a constant current of 6 A, utilizing the increase in electric resistance associated with the reduction of the ligament length. Hydrogen charging of specimens was conducted prior to the tests by cathodic electrolysis in a 3% NaCI solution containing 3g/l of NH4SCN at a current density of 5 A/m2 for 24 h. The amount of hydrogen absorbed was 0.82 wt.ppm for a nondeformed specimen, and was thermally desorbed at temperatures lower than 200~ The specimen surface was cleaned after hydrogen charging, and then the clip gauge and wiring for potential measurement were fixed to the specimen. Bending tests started -15 min after hydro~gen charging. Preliminary tests were conducted with current densities of 10 and 20 A/m" to confirm that hydrogen charging did not cause a further decrease in the J-integral value.
Hydrogen-Assisted Cracking and Embrittlement
288
3. Experimental results The R-curves of specimens with/without hydrogen charging are shown in Fig. 2. Ductile crack growth resistance in terms of the slope of the R-curve was noticeably reduced in the hydrogen-charged specimen. Stable ductile crack growth originates as a deviation of the slope from the initial blunting line. The effect of hydrogen on J~c, however, was not definite because the deviation point was not distinctly determined. The increase in COD, 6, during loading is shown in Fig. 3, in which the abscissa denotes J/cry. The crack opening was almost identical for specimens with/without hydrogen charging in the initial stage of loading, but it tended to be smaller in the hydrogen-charged specimen. This implies that hydrogen suppresses the blunting of the pre-notch and ductile crack. 2.5 O Non-charged
]
I
A Hydrogen-charged]
2.0
1.5
Z
1.0 .=. 0.5
0.0
p
0
t
t
I
I
I
I
I
I
1
2
3
4
5
6
7
8
9
z~ a (mm)
Fig. 2. The R-curves of specimens with/without hydrogen precharging.
0.8
~,0.6
0.4 /
0.2
I/ 0
~"
0.5
1
'
Hydrogen-charged
1.5
2
2.5
J/~ ys (mm) Fig. 3.
The
increase in COD during loading for specimens with/without hydrogen precharging.
Volume 1: Chemistry, Mechanics and Mechanisms
289
Fig. 4. Increase in the electric potential difference associated with loading of specimens.
The results of the electric potential measurement are shown in Fig. 4. For the convenience of making a comparison with the R-curves described later, the abscissa again denotes d/cry. The deviation from the linear increase is ascribed to the onset of a ductile crack that preceded the blunting stage of the pre-notch. The deviation took place at a lower J/cry value in the hydrogen-charged specimen. Because the yield stress was virtually unaffected by hydrogen precharging, the promotion of ductile crack initiation by hydrogen is implied. An alternative method for determining the onset of ductile crack initiation on an Rcurve has been devised by Fujii et al. [23]. Their method is based on the finite element analysis procedure of Needleman and Tvergaard [24] for examining the crack-tip stress and strain fields in a ductile porous material. Assuming inclusion distributions of two types, i.e., small and large, Needleman and Tvergaard [24] predicted a small "jog" or uprise would appear in the COD vs. d-integral curves when larger voids coalesced with the crack tip. The predicted jog, however, is usually too small to be detected experimentally. The method devised by Fujii et al. [23] depended on the gradient of a line from the origin intersecting d vs. d/cry curve. The maximum of ~(d/cry) was assumed to appear at the end of the jog, and was successfully identified for a highstrength, low-carbon steel. In the present case, the tT(J/cry) vs. J/cry curve is shown in Fig. 5 together with the electric potential difference from Fig. 4 for specimens with/without hydrogen-charging. A peak appeared on the 67(J/cry) vs. J/cry curve for the non-charged specimen, and the J/cry value for the maximum of tT(J/cry) almost coincided with that obtained by the potential method in accord with the previous results reported for a high-strength, lowcarbon steel [23]. In the case of the hydrogen-charged specimen, no peak appeared. COD was reduced in the hydrogen-charged steel compared with that of the non-charged steel at the same J-integral value even before initiation of the ductile crack, as shown in Fig. 3.
Hydrogen-Assisted Cracking and Embrittlement
290
0.445
7
(a)
Initiation .......
0.42
10
(b)
, o O- 6.5
0.44
8
5.5
o ..,o~ o/
0.435
5
/
0.43
... 4.5
~
0.39
~
~
9,-
~
~
~"
/
ooiio
~~9
::t.
........ . . .......... . . . . . . . . .6. .
/ /
...
o
4
7
o ...... oO ..................o0 ......... ......
5
0.36
4
~
o
3.5
0.425
0.42 0.2
0.3
!
i
i
0.4
0.5
0.6
J/(Yys ( m m )
2.5 0.7
0.33 ~
J
~
~
~
0.2
0.3
0.4
0.5
0.6
1 0.7
J/Gys ( m m )
Fig. 5. The slope of COD vs. J/cry curve under increasing J/ay. Concurrently-observed electric potential difference is also shown (a) without and (b) with hydrogen precharging.
Fig. 6. Fracture surface of specimens (a) without and (b) with hydrogen precharging.
The fractographic features observed by means of scanning electron microscopy are presented in Fig. 6. The fracture surface of the non-charged sample had a dimple pattern consisting of relatively large primary dimples and small secondary ones. In the hydrogen-charged specimen, the periphery of the dimples tended to be blurred, and irregular striations that characterize quasi-cleavage were occasionally observed. The hydrogen-charged specimen was overall characterized by a shallower fracture surface,
Volume 1: Chemistry, Mechanics and Mechanisms
291
suggesting that the linking of primary dimples proceeded with reduced plasticity in this specimen. This feature is consistent with the previous result that showed shallowing of primary dimples in the hydrogen-charged specimen [19]. 4. Discussion
A substantial decrease in ductile crack growth resistance due to hydrogen was shown in Fig. 2. An R-curve is theoretically derived by calculating the J-integral from the stress and strain fields in front of a crack. The finite element method (FEM) is useful for calculating the stress and strain fields, but deterioration of the elements must be taken into account when defects are created during straining. We have previously found a high level of consistency between observed and calculated R-curves by taking into account the actual presence of voids in the yield function of the elements [25]. The observed decrease in ductile crack growth resistance caused by hydrogen was assumed to be attributable to an increase in the void volume fraction in the simulation [ 19]. Void nucleation in ductile fractures has been attributed primarily to the interface separation or cracking of second phase particles. However, fine dimples without particles and far exceeding the density of particles have been observed occasionally [26-28]. Microstructural heterogeneities such as twin and grain boundaries [26] or dislocation cell walls [29] were the sites of void nucleation. The mechanism for this proposed by Lyles and Wilsdorf [30] was the condensation of vacancies, the feasibility of which was theoretically demonstrated by Cuttifio and Ortiz [31 ], who considered vacancies generated primarily by the dragging of intersecting jogs. The estimated vacancy concentration was of the order of-10 -5 to 10-4 in atomic ratio. Generation of a high density of vacancies of the order of 10-3 in dislocation-rich walls was estimated by Essmann and Mughrabi [32] by annihilating dislocations of the opposite sign. Experimentally, vacancy clusters leading to nanovoid formation in the dislocation-free zone ahead of a crack tip were observed by Chen et al. [28] in an austenitic stainless steel by means of in situ transmission electron microscopy. Our findings concerning the role of hydrogen in enhancing the strain-induced creation of vacancies, and correlation between the vacancy concentration and ductile crack growth resistance, are in accord with the studies cited above. As for the interaction of hydrogen with vacancies, creation of a high density of vacancies far exceeding the thermal equilibrium value has been demonstrated experimentally and theoretically to be a consequence of the reduced formation energy of vacancies through combination with hydrogen [33-36]. A similar effect of hydrogen can be expected to enhance the high density of strain-induced vacancies in strain-concentrated areas. Promotion of void formation therein by hydrogen can thus be a consequence of vacancy agglomeration in shear bands that have been observed under a notch in bending tests [8,9]. As an effect of increasing the void volume fraction, the FEM calculations reported in our previous study [19] showed a prominent localization of tensile strain and void volume fraction in the crack tip region. The fractographic features in Fig. 6, which indicate reduced plasticity by hydrogen in forming the fracture surface, are consistent with the localization of strain and voids plus the decrease in crack growth resistance. They are also consistent with the small COD values in the hydrogen-charged specimen. The reduced COD accompanying the growth of an environmentally-induced crack in low-carbon, high-strength steel has been observed recently by Minoshima et al. [37] by
292
Hydrogen-Assisted Cracking and Embrittlement
using in situ atomic force microscopy. However, the fractographic features did not directly indicate the formation of nanoscale voids as a consequence of vacancy agglomeration. We have recently observed amorphization in a ductile crack front and beneath the fracture surface of hydrogen-charged ferritic steel [38]. The amorphization was preceded by fragmentation of the matrix and a reduction in dislocation density. The amorphization associated with a high density of vacancies is plausible according to the expected lattice instability caused by a high density of vacancies exceeding l0 -4 in atomic ratio, as pointed out by Cahn [39]. Plastic instability leading to strain localization is a characteristic of the mechanical properties of the amorphous phase, and ductility is much reduced in the presence of hydrogen [40]. In the fracture process, the formation of the amorphous phase in the strain-concentrated area in front of the crack is expected to reduce resistance to crack growth. The elastic modulus and yield strength are generally lower atter amorphization. Although in the presem case it was not clear whether nanovoid formation or amorphization operated in crack growth, both might reduce the capacity of the matrix to retain high stress states. The promotion of ductile crack initiation by hydrogen was detected by means of the electric potential method (Fig. 4). The coincidence of the initiation point detected by the potential method and the R-curve analysis in Fig. 5 for the specimen without hydrogen-charging supports the validity of the potential method. However, t~(d/cry) continued increasing with d/Cry, without peaking in the hydrogen-charged specimen. Successive formation of jogs in 8 vs. d-integral curves in the presence of hydrogen may have been the reason for the absence of a peak. The jog is a consequence of enlargement of the crack-tip opening associated with the coalescence of a large void with the crack tip. The calculation made by Needleman and Tvergaard [24] with their model assumed inclusions as the void source. The formation of a jog was ascribed to the loss of stress-carrying capacity of the elements between the crack tip and an island of large inclusions when the void volume fraction increased to a critical value due to straining. Deterioration of the elements between the crack tip and voids was not studied. According to Needleman and Tvergaard's model [24], the effect of hydrogen on crack initiation would be ascribed to the increase in the void volume fraction. However, the loss of stress-carrying capacity of the intermediate elements was the essential factor for jog formation, in which defect creation in the elements might also play a role. The disappearance of a peak in the fi/(d/Cry) vs. J/cry curve with a continued increase in 8 in the hydrogen-charged specimen (Fig. 5), has been attributed here to successive jog formation. This attribution may be justified, provided that strain concentration in front of the crack, which induces vacancy creation, proceeds further once the initial notch connects with a microvoid. The effect of hydrogen on enhancing strain localization was apparent even before the onset of a ductile crack, as implied in Figs. 3 and 5 by 8 values in the hydrogen-charged steel smaller than in the non-charged one. Once a ductile crack is initiated, the reduced crack tip radius and extended crack length may further promote strain localization concomitant with defect creation, leading to a further loss of stress-carrying capacity. The present findings indicate that hydrogen affects ductile fracture from the early stages being incorporated with plasticity. Hydrogen likely enhances the creation of
Volume 1: Chemistry, Mechanics and Mechanisms
293
vacancies during plastic deformation, which would be consistent with the vacancy agglomeration model for the mechanism of hydrogen-related failure.
5. Conclusions The effect of hydrogen in successive stages leading to fracture was examined in a three-point bending test of low-carbon steel with ferrite and pearlite microstructures. Measurement of COD showed that hydrogen suppressed the blunting of the pre-notch even prior to the onset of a stable ductile crack that was detected with an electric potential method. Hydrogen promoted the initiation of the ductile crack and reduced ductile crack growth resistance that was evaluated by means of R-curve analysis. Shallowing and features of quasi-cleavage characterized the fracture surface of the hydrogen-charged steel. Enhanced strain localization by hydrogen associated with the creation of defects in front of the notch and crack, reported by other studies, is in accord with our f'mdings. Hydrogen presumably enhances the creation of vacancies associated with plastic deformation, while the agglomeration of vacancies would promote the initiation of ductile crack and reduce successive crack growth resistance.
Acknowledgement The authors thank Dr. Y. Matsubara (former name, Fujii) for helpful discussions concerning experimental procedures.
References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [ 14] [15] [16] [17] [18]
R.A. Oriani, P.H. Josephic, Acta Metall. 22 (1974) 1065-1074. H.K. Birnbaum, P. Sofronis, Mater. Sci. Eng. A176 (1994) 191-202. H. Vehoff, W. Rothe, Acta Metall. 31 (1983) 1781-1793. F. Terasaki, T. Kawakami, A. Yoshikawa, N. Takano, Rev. M6tall.- CIT/Sci. G6nie Mat6r. (1998) 1519-1529. W.W. Gerberich, R.A. Oriani, M.-J. Lii, X. Chen, T. Foecke, Phil. Mag. A, 63 (1991) 363-376. T.J. Marrow, M. Aindow, P. Prangnell, M. Strangwood, J.F. Knott, Acta Mater. 44 (1996) 3125-3140. M. Nagumo, K. Miyamoto, J. Jap. Inst. Met. 45 (1981) 1309-1317 (in Japanese). I-G. Park, A.W. Thompson, Metall. Trans. A, 22A (1991) 1615-1626. O.A. Onyewuenyi, J.P. Hirth, Metall. Trans. A, 14A (1983) 259-269. J.P. Hirth, The role of hydrogen in enhancing plastic instability and degradading fracture toghness in steels, in: A.W. Thompson, N.R. Moody (Eds,), Hydrogen Effects in Materials, TMS, Warrendale, 1996, pp. 507-522. P.C. Paris, H. Tada, A. Zahoor, H. Ernst, The theory of instability of the tearing mode of elastic-plastic crack growth, in: J.D. Landes, J.A. Begley, G.A. Clarke (Eds.), ElasticPlastic Fracture, ASTM STP 668, ASTM, Philadelphia, 1979, pp. 5-36. M. Nagumo, K. Ohta, K. Takai, Scripta Mater. 40 (1999) 313-319. N. Suzuki, N. Ishii, Y. Tsuchida, T6tsu-to-Hagan6 80 (1994) 855-859 (in Japanese). M. Nagumo, T. Yagi, H. Saitoh, Acta Mater. 48 (2000) 943-951. T. Yagi, A. Itoh, M. Nagumo, T6tsu-to-Hagan6 81 (1995) 225-230 (in Japanese). M. Nagumo, M. Nakamura, K. Takai, Metall. Mater. Trans. A, 32A (2001) 339-347. M. Nagumo, Mater. Sci. Tech. 40 (2004) 940-950. M. Nagumo, H. Matsuda, Phil. Mag. A, 82 (2002) 3415-3425.
294
Hydrogen-Assisted Cracking and Embrittlement
[ 19] M. Nagumo, H. Yoshida, Y. Shimomura, T. Kadokura, Mater. Trans. 42 (2001) 132-137. [20] JSME S 001-1981: Standard Method of Test for Elastic-Plastic Fracture Toughness J~c, JSME-Japanese Society of Mechanical Engineers, 1981. [21] BS 5762-1979: Methods for Crack Opening Displacement Testing, BSI-British Standards Institution, 1979. [22] G.M. Wilkowski, J.O. Wambaugh, K. Prabhat, Single-specimen J-resistance curve evaluations using the direct-current electric potential method and a computerized data acquisition system, in: R.J. Sanford (Ed.), Fracture Mechanics: Fifteenth Symposium, ASTM STP 833, ASTM, Philadelphia, 1984, pp. 553-576. [23] Y. Fujii, A. Kikuchi, M. Nagumo, Metall. Mater. Trans. A, 27A (1996) 469-471. [24] A. Needleman, V. Tvergaard, J. Mech. Phys. Solids 35 (1987) 151-183. [25] H. Yoshida, M. Nagumo, ISIJ Int. 38 (1998) 196-202. [26] R.H. Van Stone, T.B. Cox, J.R. Low Jr., J.A. Psioda, Int. Met. Rev. 30 (1985) 157-179. [27] H.G. Wilsdorf, Acta Metall. 30 (1982) 1247-1257. [28] Q.-Z. Chen, W.-Y. Chu, Y.-B. Wang, C.-M. Hsiao, Acta Metall. Mater. 43 (1995) 43714376. [29] R.N. Gardner, H.G.F. Wilsdorf, Metall. Trans. A, 11A (1980) 659-669. [30] R.L. Lyles, Jr., H.G.F. Wilsdorf, Acta Metall. 23 (1996) 269-277. [31] A.M. Cutti~o, M. Ortiz, Acta Mater. 44 (1996) 427-436. [32] U. Essmann, H. Mughrabi, Phil. Mag. A, 40 (1979) 731-756. [33] Y. Fukai, N. Okuma, Phys. Rev. Lett. 73 (1994) 1640-1647. [34] V.G. Gavriljuk, V.N. Bugaev, Yu.N. Petrov, A.V. Tarasenko, B.Z. Yanchitski, Scripta Mater. 54 (1996) 903-907. [35] R.B. McLellan, Z.R. Xu, Scripta Mater. 36 (1997) 1201-1205. [36] H.K. Birnbaum, C. Buckley, F. Zeides, E. Sirois, P. Rozenak, S. Spooner, J.S. Lin, J. Alloys C0mp. 253-254 (1997) 260-264. [37] K. Minoshima, Y. Oie, K. Komai, ISIJ Int. 43 (2003) 579-588. [38] M. Nagumo, T. lshikawa, T. Endoh, Y. Inoue, Scripta Mater. 49 (2003) 837-842. [39] R.W. Cahn, Nature 273 (1978) 491-492. [40] M. Nagumo, T. Takahashi, Mater. Sci. Eng. 23 (1976) 257-259.
295
Hydrogen assisted stress-cracking behaviour of supermartensitic stainless steel weldments W. Dietzel
a, p. Bala
Srinivasan b, S.W. Sharkawy c
a GKSS-Forschungszentrum Geesthacht GmbH, Institute for Materials Research, D-21502 Geesthacht, Germany b Department of Metallurgical Engineering, National Institute of Technology, Tiruchirappalli 620 O15, India c Department of Metallurgy, Nuclear Research Centre, Atomic Energy Authority, Cairo 13 759, Egypt
Abstract
Supermartensitic stainless steel (SMSS) grades are gaining popularity as an altemate material to duplex and superduplex stainless steels for applications in the oil and gas production industry. This study addresses the stress-cracking behaviour of weldments of a high-grade SMSS in the presence of hydrogen. Welds were produced with matching consumables using electron beam and submerged arc welding processes. Weldments were subjected to slow strain rate tests in a 0.1 M NaOH solution with introduction of hydrogen into the specimens by means of potentiostatic cathodic polarisation at a potential of-1200 mV vs. Ag/AgCI electrode. Reference tests were performed in air for comparison. Results suggest that both the SMSS base material and the weld metals are susceptible to embrittlement under hydrogen charging conditions. Submerged arc welds show a marginally-better performance than electron beam welds. 1. Introduction
In the oil and gas production sector, conventional carbon and Cr-Mn steels as well as low-alloy steels are widely employed. The use of 12% Cr martensitic stainless steels with better mechanical and corrosion properties (compared to carbon steels) has considered for such applications, but these steels also have problems with weldability. Modified 13% Cr steels with better property combinations are gaining popularity as alternate materials for aggressive environmental service conditions [1,2]. In recent years, owing to their excellent combinations of properties duplex stainless steels and superduplex stainless steels have been preferred as alternatives for both offshore and land-based oil and gas applications [3-5]. However, due to cost factors, at present the focus is on the development of supermartensitic stainless steels (SMSSs) for the manufacture of oil and gas pipelines, especially for transport of mildly-sour gas. In the past five years, researchers around the world have been studying SMSSs in an effort to
296
Hydrogen-Assisted Cracking and Embrittlement
understand their behaviour and thereby assess their suitability for aforementioned applications under aggressive environmental conditions. Much research has addressed the structure-property relationship and especially the stress corrosion cracking (SCC) behaviour of SMSSs in sweet/sour environments [6,7]. The performance of weldments produced by using non-matching consumables has also been studied [8,9,10]. This study addresses the SCC behaviour of a high-grade SMSS base material and of weldments of this material produced by electron beam welding (EBW) and submerged arc welding (SAW) processes.
2. Experimental A high-grade supermartensitic stainless steel plate of 20 mm thickness was tested in this study. Joining was accomplished by SAW and EBW processes using a solid wire matching consumable, the chemical composition of which is also given in Table 1. The filler addition in the case of the EBW process was used to handle chemistry-structureproperty of the resultant weld metal.
Table 1 Chemical composition of SMSS base material and filler wire (wt.%), balance Fe Grade Base material Filler
C 0.006 0.013
Mn 1.87 0.67
Si 0.294 0.500
Cr 11.65 12.37
Ni 6.498 6.370
Mo 2.33 2.65
Cu 0.475 -
N 0.009 0.002
A double V-groove edge preparation was employed for SAW, and the root pass was laid by plasma arc welding. In the case of EBW, butt joints were welded in a single pass from one side. Solid wire 3 mm in diameter was used in both cases. Subsequently, cosmetic capping passes were laid on both the face and root sides by the gas tungsten arc (GTA) welding process, using the matching filler wire. Both the submerged arc and electron beam weldments were given a post-weld heat treatment (PWHT) at 630 + 10~ for 30 min. Microhardness measurements were made in a Shimadzu Vickers hardness tester under a load of 4.91 N to obtain the hardness profile across the weldment. The stress cracking susceptibility of SMSS base material and welds, resulting from hydrogen embrittlement, was assessed by slow strain rate tensile (SSRT) tests. Cylindrical specimens with the dimensions given in Fig. 1 were used for SSRT tests. Weld specimens were machined perpendicular to the fusion line, such that the thinner diameter gauge section of each specimen was sampling over the entire region of the weldment, including portions of unaffected base material at both ends. To achieve this, the gauge sections of the weld specimens were extended to 22 mm while the gauge sections of base material specimens were 10 mm. The overall length of specimens and gauge diameter dimensions were identical. Tests were performed in a 0.1 M NaOH solution at room temperature. Hydrogen was introduced into specimens by the cathodic polarisation of specimens at-1200 mV vs. Ag/AgCI electrode using a Wenking LT 78 potentiostat. Prior to the test, the specimens were pre-charged with hydrogen over a period of~24 h while kept under a
297
Volume 1: Chemistry, Mechanics and Mechanisms
3
R2
.__._.
_
it_
~
R1
&
_-.-_
m....i
_.
all dimensions in "ram"
~-- 1,5 x 45 ~ Fig. 1. Schematic diagram of a tensile specimen for SSRT tests made from weldments.
small pre-load of approximately 0.5 kN (~25 MPa -"pre-soaking"). During the subsequent monotonic loading of specimens with a constant strain rate of 10-6 s-l, their hydrogen charging was maintained. Reference tests were performed in ambient air. All tests were evaluated according to the ISO standard 7539, Part 7 "Slow Strain Rate Tests" [11 ]. After failure, the diameter of each specimen at its fracture area was determined with an optical microscope at equally-distributed circumferential positions. Percent reduction in area (%RA) was determined from the average of these values. Finally, the fracture surfaces of SSRT failed specimens were examined with a scanning electron microscope (SEM). 3. Results and discussion 3.1. Microstructure
The microstructure of high-grade SMSS base material in the as-received condition is shown in Fig. 2. A tempered martensitic structure characteristic of the hardened and tempered condition is apparent. Prior austenite boundaries are also visible in Fig. 2. The macrostructure of electron-beam and submerged arc weldments are presented in Fig. 3. In the electron beam weldment, apart from the narrow-bead geometry, the GTA cosmetic capping pass laid from either side of the plate clearly visible. These capping GTA passes were intended to free the welds from a serious defect-undercut. The weld produced by SAW was also found to be a complete penetration joint, free from defects. The microstructures of the high-grade SMSS electron beam weldment are shown in Fig. 4. The composite region comprising the weld metal, the heat affected zones (HAZ), and the parent material is depicted on the let~ of this figure. The as-cast weld metal structure characteristic of solidification and rapid cooling, with a dendritic structure and small amounts of ferrite stringers, is shown on the right. The microstructure looks predominantly like a tempered martensitic structure.
298
Hydrogen-Assisted Cracking and Embrittlement
Fig. 2. Optical micrograph of the SMSS base material.
(a)
(b) Fig. 3. Macrographs of (a) electron beam and (b) submerged arc SMSS weldments.
Representative micrographs of the submerged arc weldments are presented in Fig. 5. Coarse-grained (CG) and fine-grained (FG) regions were observed in these weldments. Moreover, as in the case of electron beam welds, these welds exhibited a tempered martensitic structure with small amounts of ferrite.
Volume 1: Chemistry, Mechanics and Mechanisms
299
Fig. 4. Optical micrographs showing the composite region (weld metal, WM) and HAZ of electron beam weldment at lower (left) and higher (right) magnifications.
Fig. 5. Optical micrographs of a submerged arc SMSS weldment with the composite region showing HAZ and weld metal at lower (left) and higher (right) magnifications.
3.2. Hardness
The base material of the high-grade SMSS material has registered hardness values of 295 _ 10 HV0.5 (0.5 kp load). The hardnesses of the weld metal and HAZ regions of the electron beam weldment were observed to be higher than those in the submerged arc weldment, even though both weldments received the same PWHT (Fig. 6). This difference is owing to the faster cooling rates associated with EBW, resulting in harder microconstituents/structure both in the weld metal and in adjoining HAZs. However, hardness values in different regions of this weldment were within the range prescribed by established codes and standards, and are thus acceptable for application. In the case of submerged arc weldments, the weld metal and HAZ regions registered hardness values of 300 _ 15 HV05 (atter PWHT), confirming that this weldment was too overmatched in terms of strength.
300
Hydrogen-Assisted Cracking and Embrittlement
Fig. 6. Hardness traverse profile of the electron beam SMSS weldment.
Hardness variations were small across weldment regions. This was possibly due to the combined effect of tempering during the multiple pass welding and the subsequent PWHT. The influence of carbon on hardness has been reported to be higher significant under welding conditions, and its level must be kept low in order to avoid problems of cold cracking in weldments [12]. In this case, because the carbon content was C2, C2 < Co, where Co is the average concentration of hydrogen atoms). Physically this means that a hydride boundary instantly captures atomic hydrogen from the solution and supplies it to the new phase with a higher hydrogen concentration. The influence of a stress field is that, in addition to the concentration gradient, the hydrogen atoms migrate due to the stress field gradient of triple grain boundaries. Because of this, the velocity of the hydride phase boundary increases. The relationship defining the hydride growth kinetics around triple grain boundaries can be written as [3]
1 c3C
(~2C
D Ot
Or 2
C(Rl,t) = C2,
1 + a OC
r
g3r
C(r,O) = Co(r > Ro),
(c,-c=)--~-
C(m,t) = Co,
(9)
Ll-&rI+ - 7 r=Rl
where Ro is the radius of hydride phase nucleus, and R1 is the current hydride radius. The distance between the lines of triple grain boundaries is assumed to exceed the typical size of hydride phase considerably. This allows for consideration of the new phase growth in the unlimited matrix. Changing the hydride size obeys the law Rl(t) = fl~Dt, where fl is a non-dimensional parameter, the value of which is taken from the mass balance equation on an interphase boundary. In relation to "a stationary interphase boundary," a quadratic equation was obtained for determining the parameter fl /3 2
2fl[C2-CO c,
2C2 -
c,
(lO) - ~
When a = 0 in Eq. (9), under invariable initial and boundary conditions we obtained 10C
D Ot
(~2C
10C
Or 2
r
Or
C(Rl,t) = C2, C(r,0) = Co(r >-Ro), C(r
= Co,
(11)
dRl = D( dC ] (C, -62)--d7~-~r jr=R' In this case, the growth of hydride results from the hydrogen atom concentration gradient. A transcendental equation can be obtained from a mass balance equation on the interphase boundary for determining parameter ill, according to the expression fll~/Dt
320
Hydrogen-Assisted Cracking and Embrittlement
2 C2-C o
(12)
c,-c
where Ko(X) and KI(X) are modified Bessel functions of the second series of zeroth and first orders respectively. Assume Co ~ 2.0 x 10-4, C2 = 10-4, Cl = 3.0 x 10-4 (at.%) without limiting community. This gives fl = 1.3 and fll = 0.8 for the non-dimensial parameters. The stress field near triple grain boundaries clearly accelerates the growth of the hydride phase. The essential change of volume hydride wedges off the grain boundaries, forming microcracks. The hydrogen embrittlement of metal is observed at macroscopic scales as a result of the formation and growth of hydrides. 3. Conclusion
Internal stresses, caused by a non-homogeneous deformation within the solid, occur in metals even without applied loads. Internal stresses have an essential effect on the character of diffusion processes and change the strength of the metals. The kinetics of the processes is described by a parabolic equation under corresponding initial and boundary conditions. The formation of microcracks along the grain boundaries occurs when hydrogen atoms interact with the metal. A possible mechanism of hydrogen embrittlement is as follows. Under some conditions (for example, under braking of grain-boundary slippage), cylindrically-shaped cavities are formed near triple grain boundaries. Molecular hydrogen creates an essential pressure within the cavities, followed by formation of microcracks along grain boundaries. In some metals, hydride phases are formed near triple grain boundaries. The change of the hydride volume also results in the formation of microcracks along grain boundaries. Because of this, structural defects like triple grain boundaries, when interacting with hydrogen contents, cause the formation of microcracks along grain boundaries. The embrittlement of metal is observed macroscopically when interacting with hydrogen. References
[ 1] N.M. Vlasov, V.A. Zaznoba, Fiz. Tverd. Tela 41 (1999) 64-67 [Phys. Solid State 41 (1999) 55-58.] [2] B.A. Kolachev, Hydrogen Brittleness of Metals, Metallurgy, Moscow, 1985 (in Russian). [3] N.M. Vlasov, I.I. Fedik, Int. J. Hydrogen Energy 27 (2002) 921-926.
323
Stress corrosion cracking of magnesium alloy with the slow strain rate technique H. Uchida, M. Yamashita, S. Hanaki, T. Nozaki Division of Mechanical Engineering, Universityof Hyogo, Himeji, 671-2201, Japan Abstract Stress corrosion cracking (SCC) tests of AZ31B magnesium alloy in distilled water and 2-8% NaCI solutions at 298 K were performed using a slow strain rate technique. The alloy showed susceptibility to SCC in distilled water, and the susceptibility increased as the strain rate decreased. The alloy was also highly susceptible to SCC in NaCI solutions under open-circuit condition regardless of the level of strain rate and the concentration of NaCI. It was found that SCC of the alloy in the 4% NaC1 solution occurred at potentials close to the corrosion potential. The latter lay in the potential range of hydrogen evolution. Examination of fracture surfaces revealed a transgranular crack growth mode of the alloy with a quasi-cleavage appearance. Results suggest that hydrogen embrittlement was the mechanism of SCC in AZ31B alloy exposed to chloride solutions. Anodic dissolution of the alloy due to the presence of chloride ions also played some role in SCC. I. Introduction
Due to advantages such as high strength-to-weight ratios, specific castability and recycling efficiency, magnesium alloys are used in a wide variety of industrial applications. These include the automotive, aerospace, electronics, and consumer product industries [ 1]. However, magnesium and its alloys are chemically highly active and electrochemically less noble than other engineering alloys. Moreover, the control of susceptibility to stress corrosion cracking (SCC) is becoming a major issue in industrial applications [2]. Many cast magnesium alloys have been studied to determine their susceptibility to SCC. In general, alloys with >1.5% AI tend to experience SCC, and this tendency becomes marked at higher AI contents [2,3]. It was also reported that magnesium forgings increase the susceptibility to SCC compared to magnesium castings [4]. However, there are few published reports on the SCC of magnesium and its alloys, and the investigation of the mechanism of SCC in magnesium alloys is still in its infancy [5-8]. In this study, SCC tests of AZ31B magnesium alloy in distilled water and sodium chloride (NaCI) solutions were performed using a slow strain rate tensile (SSRT) technique under various potential conditions. The purpose of the tests was to
324
Nonferrous Alloys
investigate the determining factor(s) of SCC susceptibility to establish a SCC mechanism(s).
2. Experimental procedures 2.1. Material
A commercial AZ3 I B magnesium alloy (2.5-3.5% AI, 0.7-1.3% Zn, >0.2% Mn, d. .I IE -o.6o
I-
- -
i 151 L~ 15 I lll
:5
I"
=i
I~1
.
I ~
Ii
E
o
E E m -0.65 i r
O
Q. .~ -0.70
Range of l Experimental/
ObservationIs t
r
o -0.75
Fig. 4. Crack potential at r = l mm for model results and experimental data. Model conditions: EApp = -0.495 VSCE,gTip -- -0.90 VscE, CTOD = 250 nm,/wall = 5 lam/cm2 and to= 0.0614 l/f~cm except for noted tip films.
Another means of examining the feasibility of crack tip film thickness and conductivity is to compare experimental and model crack potentials at a constant position within the crack. It is clear from Fig. 4 that if a very resistive film (to = 10-4 1/f~-cm) were present at the tip, it would necessarily be very thin. On the other hand, a substantially less resistive film ~10 lam thick cannot be excluded based on the 1-D modeling results presented here. 4. Discussion
4.1. Crack potential measurements and modeling The crack electrode potential during EAC of AA 7050 in aqueous chromate-chloride solution varied from the external applied value (Fig. l) due to ohmic drop, particularly during high rate da/dt. The steep crack potential gradient measured with the in-situ reference electrode indicates ion movement within the crack: cations are driven out of the crack while anions are attracted to the tip. In addition, the fact that there is a net current (and hence IR drop) within the crack indicates that the anodic reaction at the tip is not balanced by local cathodic reactions. Anode and cathode are separated in this system. The variability in crack wake potential profile during steady state da/dt shown in Fig. 1 is likely related to: (i) metastable localized corrosion events near the in-situ electrode; (ii) changes in crack resistance due to HE gas bubble formation and emission; and/or (iii) corrosion product accumulation and possible re-dissolution.
340
Nonferrous Alloys
4.1.1. Solution conductivity and crack wall current Potential gradients facilitate steady-state concentration gradients. The crack solution chemistry within AI alloys becomes acidic (pH ,~ 3) and concentrated in alloy dissolution products and anions from the bulk solution [6,12,13-17]. However, accounting for the increase in solution conductivity did not have a strong effect on the modeled crack tip current. This results because the bulk test solution is a strong supporting electrolyte. The bulk electrolyte and relatively concentrated crack solution have similar conductivity (to--0.061 and 0.08 to 0.16 1/f2-cm, respectively). Because /R-drop scales with ITiJ/(~ for a constant ETip, the increase in Ivip is at most 2.7x assuming the crack were uniformly filled with solution of tr = 0.16 vs. 0.061 l/f2-cm. The exception may the presence of a resistive film at the tip. Crack wall passive dissolution currents contribute approximately 10% of the IRdrop observed in the corrosion-inhibited system (Fig. 3). Although not insignificant, passive dissolution of the crack walls was not a major contributor to potential drop within the crack. The potential drop appears to be dominated by the crack tip geometry.
4.1.2. Significance of crack potential modeling One-dimensional modeling revealed that the parameters that most influence the crack potential distribution are: (i) crack tip geometry; and (ii) the presence of a resistive crack tip film. Comparison of the crack potential profile for a blunt crack-tip geometry (Fig. 3) with the experimental data shown in Fig. 1 indicates that a blunt crack does not closely reproduce the experimental results. These results suggest that IG EAC cracks in AA 7050 are likely sharper than predicted by elastic-plastic analysis. This interpretation is consistent with the observations of Somerday et al. [9] who examined the crack surface opening displacement (CSOD) of a ,fl-Ti alloy while under load in a scanning electron microscope. They reported that whereas transgranular fatigue cracks attained blunt-like crack tip morphology similar to that predicted via continuum elastic-plastic analysis with small scale yielding, intergranular EA cracks were much sharper. For an IG crack, the CSOD was ~300 nm at r = 3 mm [9]. In addition, Deshais and Newcomb [10] reported that IG EAC cracks of AA 7010 had ~50 nm wide corrosion fronts. The corrosion was restricted to the grain boundary precipitate free zone. The results of the I-D crack potential modeling suggest that IG EA cracks in AA 7050 may also be significantly sharper than predicted via elastic-plastic analysis. The choice of the crack tip corrosion area deserves discussion. Consider the case of a sharp crack with CTOD = 250 nm. For a 1 ~tm crack tip corrosion front (multiplied by the crack thickness to get the area), the closest 375 nm of wall on either side of the tip would also have to be considered to be corroding at the rate predicted by the Ecrack model results (i.e., /Tip ~ 0.1 A/cm2). Recall that observed crack growth rates were 2 to 8 x 10-5 mm/s at -0.495 VSCE [6]. Assuming a crack growth rate = 5 x 10-5 mm/s, the amount of penetration or blunting of the near-tip wall would be 250 nm in the time it took the crack to advance 375 nm after which the wall is assumed to repassivate. Unclear is whether ~250 nm of near-tip wall corrosion would be sufficient to cause crack arrest due to blunting. If the dissolution zone is approximately 1 ~tm high, then the Ecr~r modeling work suggests that anodic dissolution dominates crack growth. The 1-D model results presented in Fig. 3 indicate that there are a host of crack geometry-environment combinations that generally provide good agreement with the
Volume 1: Chemistry, Mechanics and Mechanisms
341
observed potential distribution. One such set of conditions is illustrated in Fig. 2 where the experimental Ecraek profiles shown in Fig. 1 are bounded if a 2-mm region adjacent to a sharp tip (CTOD = 250 nm) is presumed to contain a solution or precipitate of tc = 0.03 or 0.1 1/t2-cm. Such a scenario may represent the situation in which precipitated aluminum oxychloride and/or hydroxide forms over a significant distance back from the crack tip. Chloride and Cr6+-bearing aluminum hydroxides at pH ~ 3 had conductivity's ranging from 0.07 to 0.16 1/t2-cm. Given the very restricted crack geometry and sharp tip suggested by the steep potential gradient, it is possible that the near-tip solution is saturated in AI reaction products. As noted above, AI pits at high potentials become saturated in dissolution products [18,19]. Evidence suggests that the crack tip solution is concentrated and may also become saturated in dissolution products. Cooper et al. observed ~1 mole/L A13§ near the tip of AA 7050-T651 undergoing high-rate EAC in chromate-chloride solution [20]. Nguyen and coworkers also reported a concentrated crack tip environment for a similar alloy cracking at anodic potentials in NaCI solution [13]. Newman [21 ] developed a 1D model similar to the one described here for assessing the crack potential and derived an analytical equation for the concentration of metal ions at the tip (CTip) of a smoothsided, wedge-shaped crack. For a sharp crack (CTOD = 250 nm), /Tip = 0.1 A/cm 2, /wall = 5 ~tA/cm2, the crack tip AI 3+ concentration would be thus estimated to be 1.7 mole/L, consistent with reported values [ 13,20]. Based on reported and calculated crack tip AI 3+ concentrations, the metal ion solubility [22] is expected to be exceeded near the tip and precipitation of gelatinous hydroxide to occur. Scenarios involving conductive hydroxide and/or oxychloride precipitate such as depicted in Fig. 2 are thus predicted based on experimental and model results. Modeling cannot completely discern the crack conditions. However, it can provide insight into the likely nature of the crack tip geometry and allow the exclusion of some film thickness-resistance combinations on the basis that they do not replicate measured potential profiles. 4.2. Hydrogen embrittlement vs. anodic dissolution
Controversy over the crack tip damage mechanism of AI-Zn-Mg-Cu alloys persists despite extensive investigation [1,23]. Hydrogen is likely responsible for cracking in humid air where dissolution can be excluded based on the absence of a condensed water layer necessary for electrochemical reactions [24]. Substantial evidence exists implicating HE as the dominant process in the aqueous EAC of these alloys [ 1,23]. The Stage II crack path is similar under humid air and aqueous conditions [25]. In addition, crack growth kinetics in hot, humid air (60~ 90% relative humidity) and chromatechloride solution at EApp - -0.495 VSCE (at room temperature) are comparable for AA 7050-T651 [24]. For the measured crack tip potential and pH, the overpotential for the hydrogen evolution reaction is approximately -0.40 V. Obviously, there is a significant driving force for the production of atomic hydrogen. H-uptake during EAC of AA 7050-T651 in chromate-chloride solution was reported, including a positive correlation between crack wake H concentration and crack growth rate [6,25]. However, dissolution cannot be ruled out as playing a significant role in the EAC process. Of course, dissolution is responsible for generation of the critical crack tip chemistry. Furthermore, the crack tip current appears to be substantial. Even with very
342
Nonferrous Alloys
sharp tip geometry, crack potential modeling suggests that the tip dissolution rate is ~0.1 A/cm 2, equivalent to da/dtdiss ~ 3 x 10-5 mm/s. Thus, the observed crack growth rates for AA 7050-T651 (2 to 8 • 10-5 mm/s at Egpp = -0.495 VSCE) could be justified on the basis of dissolution alone. Nonetheless, the evidence for hydrogen effects from companion [6,24,25] and other studies [2,26] is compelling. It would thus appear that for AA 7050 in chromate-chloride solution, anodic dissolution and hydrogen embrittlement both contribute to crack propagation, although the relative contributions cannot be unequivocally determined. Based upon this work, the contributions of AD and HE appear to be on the same order of magnitude.
5. Conclusions A large potential gradient (~1 V/cm) exists near the tip of AA 7050 environmentassisted cracks. Tip potentials were -0.90 to -0.80 VSCE when EApp was -0.495 VSCE. Intergranular environment-assisted cracks in AI-Zn-Mg-Cu alloys are likely very sharp (CTOD ~ 250 nm); blunt cracks of the form predicted via continuum mechanics with small scale yielding do not generate sufficient /R-drop nor reproduce the observed Ecrack distribution, particularly the steep potential gradient near the tip. A highly resistive crack tip film is not necessarily required to achieve experimentally observed /R-drop; a tight, sharp crack can generate a steep potential gradient and significant IRdrop. However, unless a large crack tip corrosion area is assumed to exist, calculated crack growth rates based on Al dissolution significantly overestimated observed da/dt. The effective crack resistance may be greater than suggested by a smooth-sided crack filled with conductive electrolyte. Very thin (~20 nm), highly resistive films (ic ~ 10-4 1/f2-cm) or thicker (1 to l0 ~tm) moderately resistive films (ic~ l0 -2 1/~-cm) cannot be excluded as possibly existing at the crack tip of this alloy. Anodic dissolution contributes significantly to the Stage II crack growth of AA 7050 in aqueous chloride solution.
Acknowledgments This research was supported by the Alcoa Foundation, the technical assistance of J.P. Moran, E.L. Colvin, and J.T. Staley is acknowledged. Mobil Exploration and Production Technology Corp provided additional support. L.M. Young and R.P. Gangloff are acknowledged for their substantial assistance in the crack growth testing and analyses.
References [ 1] N.J.H. Holroyd, Environment-induced cracking of high-strength aluminum alloys, in: R.P. Gangloff, M.B. lves (Eds.), Environment-Induced Cracking of Metals, NACE, Houston, TX, 1990, pp. 311-345. [2] R. Gibala, R.F. Hehemann (Eds.), Hydrogen Embrittlement and Stress Corrosion Cracking, ASM, Metals Park, OH, 1984, pp. 271-296. [3] F.P. Ford, A quantitative examination of the slip-dissolution and hydrogen embrittlement theories of cracking in aluminum alloys, in: P.R. Swann (Ed.), Mechanisms of Environment Sensitive Cracking of Materials, TMS, Warrendale, PA, 1977, pp. 125-147. [4] T.R. Beck, Electrochem. Acta 29 (1984) 485-491. [5] T.R. Beck, Electrochem. Acta 30 (1985) 725-730.
Volume 1: Chemistry, Mechanics and Mechanisms
343
[6] K.R. Cooper, L. M. Young, R. P. Gangloff, R. G. Kelly, The electrode potential dependence of environment-assisted cracking of AA 7050, in: E.A. Starke (Ed.), Proc. 7th International Conference on Aluminum Alloys: Their Physical and Mechanical Properties, Trans Tech Publications, Enfield, NH, 2000, pp. 1625-1634. [7] R.H. Dean, J.W. Hutchinson, in: Fracture Mechanics: Twelfth Conference, ASTM STP 700, ASTM, Philadelphia, PA, 1980, pp. 383-405. [8] J.R. Rice, W.J. Drugan, T.-L. Sham, in: Fracture Mechanics: Twelfth Conference, ASTM STP 700, ASTM, Philadelphia, PA, 1979, pp. 189-221. [9] B.P. Somerday, L.M. Young, R.P. Gangloff, Fatigue Fract. Eng. Mater. Struct. 23 (2000) 39-58. [10] G. Deshais, S.B. Newcomb, The influence of microstructure on the formation of stress corrosion cracks in 7XXX series aluminum alloys, in: E.A. Starke (Ed.), Proc. 7th International Conference on Aluminum Alloys: Their Physical and Mechanical Properties, Trans Tech Publications, Enfield, NH, 2000, pp. 1635-1640. [ 11] E.P. Simonen, R.H. Jones, J. Windisch, C.F., A transport model for characterizing crack tip chemistry and mechanics during stress corrosion cracking, in: R.H. Jones, D.R. Baer (Eds.), New Techniques for Characterizing Corrosion and Stress Corrosion, TMS/AIME, Warrendale, PA, 1996, pp. 141-160. [12] K.R. Cooper, R.G. Kelly, E.L. Colvin, CORROSION/99, NACE, Houston, TX, 1999, paper no. 153. [13] T.H. Nguyen, B.F. Brown, R.T. Foley, Corrosion 38 (1982) 319-326. [ 14] A.H. Le, R.T. Foley, Corrosion 40 (1984) 195-197. [15] N.J.H. Holroyd, G.M. Scammans, R. Hermann, Environmental conditions within crevices and stress-corrosion cracks in aluminum alloys, in: A. Turnbull (Ed.), Corrosion Chemistry Within Pits, Crevices and Cracks, HMSO, London, 1984, pp. 495-510. [ 16] A.J. Sedriks, J.A.S. Green, D.L. Novak, Corrosion 27 (1971) 198-202. [17] A. Turnbull, Chemistry within localized corrosion cavities, in: H.S. Isaacs (Ed.), Proceedings of the Second International Conference on Localized Corrosion, NACE, Houston, TX, 1990, pp. 359-373. [ 18] D.W. Buzza, R.C. Alkire, J. Electrochem. Soc. 142 (1995) 1104-1111. [19] R.C. Alkire, M. Feldman, J. Electroch. Soc. 135 (1988) 1850-1851. [20] K.R. Cooper, R.G. Kelly, J. Chromatogr. A, 850 (1999) 381-389. [21 ] R.C. Newman, Corrosion 50 (1994) 682--686. [22] K. Wefers, C. Misra, Report No. 19, Aluminum Company of America, 1987. [23] T.D. Burleigh, Corrosion 47 (1991) 89-98. [24] G.A. Young, Hydrogen Environment Assisted Cracking of an AI-Zn-Mg-(Cu) Alloy, Ph.D. Dissertation, University of Virginia, Charlottesville, VA, 1999. [25] L.M. Young, Microstructural Dependence of Aqueous-Environment Assisted Crack Growth and Hydrogen Uptake in AA7050, Ph.D. Dissertation, University of Virginia, Charlottesville, VA, 1999. [26] R. Ambat, E.S. Dwarakadasa, Bull. Mater. Sci. 19 (1996) 103-114.
345
Influence of composite materials on the stress corrosion cracking of aluminum alloys Feng Lu a Wenjun Chang b Guoqiang Zhu c Xiaoyun Zhang a, Zhihui Tang a a Institute of Aeronautical Materials, Beij'ing 100095, China b Commission of Science Technology and Industry for National Defense, Beijing 100038, China c China Aviation Industry Corporation I, Beijing 100022, China
Abstract
A C-ring type of specimen was used to conduct stress-corrosion cracking (SCC) tests in a 3.5% NaC1 solution. A graphite epoxy composite material (GECM) was coupled with aluminum alloys LYI2CZ and LC4CS in order to determine the effect of galvanic corrosion between GECM/LY12CZ and GECM/LC4CS couples on the SCC behavior of LYI2CZ and LC4CS alloys. The fracture surfaces were examined by scanning electron microscopy. Results of the study showed that galvanic corrosion promotes the initiation and propagation of stress-corrosion cracks in LYI2CZ alloy, whereas SCC in LC4CS alloy is less affected by GECM. The effect of galvanic corrosion between GECM and LY12CZ alloy on SCC correlated with the level of applied stress. The effect of galvanic corrosion on the features of a fracture surface was insignificant. I. I n t r o d u c t i o n
The aluminum alloys LY 12CZ and LC4CS are widely used in aircratt structures due to their high strength-to-weight ratio. However, high-strength aluminum alloys are susceptible to environmental embrittlement. The mechanism of environmental effects on the mechanical behavior of aluminum alloys is discussed extensively in the literature. Two p r o c e s s e s - anodic dissolution and hydrogen embrittlement- were proposed to explain the environmental effects. According to the anodic dissolution approach, stress corrosion cracking (SCC) occurs due to preferential dissolution of precipitates formed near grain boundaries. The hydrogen embrittlement approach propose that hydrogen accumulating during dislocation movement results in the loss of alloy ductility [ 1-4]. Graphite epoxy composite materials (GECMs) are composed of high-strength, highmodulus graphite fibers combined with an amine-cured polymer epoxy resin. Graphite fiber behaves electrochemically like a noble metal (such as Au or Pt). Increasing the
Nonferrous Alloys
346
galvanic corrosion rate of metals coupled electrically to a "noble" GECM is one of the most common problems in this field. Such a problem can exist when structural aircraft alloys are coupled with GECM and exposed to a humid corrosive environment. For example, aluminum alloys as less noble metals will corrode due to galvanic action [5]. Generally, galvanic corrosion accelerates the dissolution rate, but does not increase susceptibility to SCC of aluminum alloys. Once cracks have initiated, the crack growth rate is influenced by the local electrochemical kinetics and by the nature and stability of surface films at the crack tip [6]. In order to evaluate the extent of galvanic corrosion and its effect on the SCC resistance of LY12CZ and LC4CS alloys, the alternate-immersion tests of GECM/LY 12CZ and GECM/LC4CS couples were performed in a 3.5% NaCI solution. The purpose of this study was to determine the susceptibility to SCC of LY 12CZ and LC4CS alloys coupled to GECM. In addition, the fracture surfaces were examined by scanning electron microscopy (SEM).
2. Experimental method 2.1. Materials A composite material T300/5222 containing graphite fiber T300 and epoxy resin 5222 was used in this study. All GECM samples were supplied by the Beijing Institute of Aeronautical Materials (BIAM). The geometry of aluminum alIoy/GECM couple specimens was about 20 x 50 mm and 2 mm in thickness. Chemical compositions and mechanical properties of aluminum alloys are given in Tables 1-3. In order to estimate the SCC performance, C-ring stress-corrosion test specimens were machined from a 30mm diameter rod of LY12CZ alloy and a 40-mm diameter rod of LC4CS alloy in the short-transverse (ST) direction relative to the grain structure in accordance with the ASTM G38 Standard. The applied stress level varied from 60 to 85% of yield strength (YS). Table 1 Chemical composition (wt.%) of LY 12CZ aluminum alloy Alloying elements Cu Mg Mn AI 3.8-4.8 1.2-1.8 0.3-0.9 Bal.
Impurities (max) Fe Si Zn 0.5 0.5 0.3
Ni 0.1
Ti 0.15
Fe + Ni 0.5
Others 0.1
Table 2 Chemical composition (wt.%) of LC4CS aluminum alloy Alloyin8 elements Cu M8 1.4-2.0 1.8-2.8
Mn 0.2-0.6
Zn 5.0-7.0
Cr 0.10-0.25
AI Bal.
Impurities (max) Fe Si Others 0.5 0.5 0.1
Table 3 Mechanical properties of the aluminum alloys tested Alloy
Identity
LYI2CZ LC4CS
030 mm rod 640 mm rod
UTS (MPa) 549 587
YS (MPa) 416 587
E (GPa) 72 68
El. (%) 15.5 13.2
347
Volume 1: Chemistry, Mechanics and Mechanisms 2.2. Test methods
C-ring type specimens - coupled or not coupled with T300/5222 - were tested by alternate immersion in 3.5% NaC1 solution in accordance with the ASTM G44 Standard. The galvanic corrosion between aluminum alloys and GECM was evaluated. The effect of galvanic corrosion on SCC performance of the alloys was studied. SEM was used for studying the micro-mechanism of crack propagation and the fracture surface of T300/5222. 3. R e s u l t s a n d d i s c u s s i o n
3.1. SCC results of aluminum alloys exposed to alternate immersion test
SCC is a delayed failure process wherein a crack initiates and propagates slowly until the stress in the remaining section of the test specimen or engineering component exceeds fracture strength. C-ring type specimens - coupled or not coupled with GECM - were inspected for cracks at certain (predetermined) time intervals. The time required for cracks to appear after the exposure of stressed specimens to a test solution and the threshold stress above which cracks appeared were used as a measure of SCC resistance. When GECM and aluminum alloys were in physical or electrical contact (through the conductive electrolyte), galvanic corrosion occurred. During galvanic corrosion, LYI2CZ and LC4CS alloys (more active metals) served as anodes and corroded, whereas GECM served as a cathode and was protected from corrosion by aluminum alloys. SCC results for LY 12CZ and LC4CS alloys are presented in Tables 4 and 5. It is evident that the LY12CZ alloy has higher susceptibility to SCC than the LC4CS alloy. Table 4 shows that the LY 12CZ alloy coupled with GECM failed atter a brief period of exposure. Galvanic corrosion promoted the initiation and propagation of stress-corrosion cracks in the LY12CZ alloy. However, SCC of the LC4CS alloy was less affected by GECM. The effect of galvanic corrosion between the LC4CS alloy and GECM on SCC correlated with the level of stress: the higher the stress, the higher was the effect of galvanic corrosion on SCC. In addition, it was found that when the LC4CS alloy was coupled with GECM and stressed at 60% YS, the general corrosion was too intensive to allow observation of SCC. Results of our studies showed that C-ring stress-corrosion tests are multi-purpose and efficient for the qualitative analysis of susceptibility to SCC of aluminum alloys in a wide variety of test conditions. Also, the results suggested that anodic dissolution was the dominant mechanism of SCC in the LY12CZ alloy, while HIC was the dominant mechanism of SCC in the LC4CS alloy [1 ]. Table 4 The SCC resistance of LYI2CZ alloy Alloy
LYI2CZ
Applied stress (%YS) 60 75 85
Not coupled with composites F/N* Time (hrs) 0/5 1/5 96 2/5 66.7, 66.7
Coupled with T300/5222 composites F/N* 5/5 5/5 5/5
Time (hrs) 78, 78, 78, 78, 78, 78 66.8, 66.8, 66.8, 66.8, 66.8 66.8, 66.8, 66.8, 96.3, 66.8
* F is the number of fractured specimens and N is the total number of specimens.
Nonferrous Alloys
348 Table 5
The SCC resistance of LC4CZ alloy Alloy LC4CS
Applied stress (%YS) 60 70 80
Not coupled with composites
F/N* 4/5 3/5 2/5
Time (hrs) 51; 51; 10; 10 51; 51; 51 78; 78
Coupled with T300/5222 composites F/N* Time (days) 0/5 3/5 51; 50; 50 2/5 7; 4
* F is the number of fractured specimens and N is the total number of specimens.
3.2. The fracture morphology of aluminum alloys SEM fi'actographs of the LY 12CZ alloy - coupled and not coupled with T300/5222 - are presented in Figs. 1 and 2 at two different levels of applied stress. The character of crack growth was intergranular, with many secondary cracks along grain boundaries. Metallographic examination of cross-sections through the secondary cracks also showed an intergranular character of stress-corrosion cracks in the alloy. Intergranular attack was more serious, and affected both crack initiation and crack propagation in the LY 12CZ alloy more strongly compared to the LC4CZ alloy. It is well known that only aluminum alloys containing an appreciable amount of soluble alloying elements - primarily Cu, Mg, Si, and Z n - are susceptible to SCC, and that SCC in aluminum alloys is characteristically intergranular. According to the electrochemical theory of corrosion, the intergranular mode of crack growth requires that grain boundaries be anodic in comparison to the rest of the alloy. In this case, corrosion will propagate selectively along the grain boundaries. Susceptibility to intergranular corrosion is one of the prerequisites for intergranular SCC. The heat treatment of aluminum alloys that improves resistance to SCC will also improve their resistance to intergranular corrosion [7-8].
Fig. 1. SEM fractographs of LY12CZ alloy at 75% YS. (a) LYI2CZ alloy, and (b) LYI2CZ alloy coupled with T300/5222.
Volume 1: Chemistry, Mechanics and Mechanisms
349
Fig. 2. SEM fractographs of LY12Cz alloy at 85% YS. (a) LYI2CZ alloy, and (b) LYI2CZ alloy coupled with T300/5222. The mechanism of SCC is specific for certain alloys or alloy systems in specific environments. In our tests, galvanic corrosion between the LY12CZ alloy and GECM promoted the local grain-boundary dissolution process. Due to the galvanic corrosion, the rate of crack growth in aluminum alloys varied greatly (Tables 4 and 5). When coupling aluminum alloys with GECM, the cathodic area was much bigger than that for uncoupled alloys. 3.3. The surface morphology ofT300/5222 composite material SEM fractographs of T300/5222 composite material- coupled and not coupled with LY12CZ alloy- during alternate-immersion tests in 3.5% NaCI are presented in Fig. 3. Compared to the fracture surfaces of the T300/5222 composite material that was not coupled with the LYI2CZ alloy, the fracture surface of GECM coupled with the alloy did not change significantly. The corrosion reactions that took place on the surface of the LY 12CZ alloy coupled with GECM are given as [5,7,8]: AI ~ AI3+ + 3e-
(1)
02 + 2H20 + 4e- ~ 4OH-
(2)
4. Conclusions
9 SCC in the LY12CZ alloy was predominantly intergranular. Coupling with the T300/5222 composite increased the amount of secondary intergranular cracks on the fracture surface of the LY 12CZ alloy. 9 The T300/5222 composite coupled with aluminum alloys showed no corrosion damage on its surface atter alternate-immersion corrosion tests. 9 Galvanic corrosion between the LY12CZ alloy and the T300/5222 composite increased the probability of crack initiation and crack propagation in the LY12CZ alloy, but had a lesser effect on SCC in the LC4CS alloy.
350
Nonferrous Alloys
Fig. 3. SEM fractographs of T300/5222 composite material. (a) T300/5222, (b) T300/5222 coupled with LYI2CZ at 75% YS, and (c) T300/5222 coupled with LYI2CZ at 85% YS.
Acknowledgments The authors are grateful for the financial support provided from the National Key Basic Research and Development Project (Grant No. G1999065004) and the National Defense Science Foundation (51412030301HK5104).
References [ 1] [2] [3] [4] [5] [6] [7]
T.C. Tsai, T.H. Chuang, Metall. Mater. Trans. A, 27A (1996) 2617-2627. A. Conde, B.J. Fernandez, J.J. de Damborenea, Corros. Sci. 40 (1997) 91-102. C.P. Ferrer, M.G. Koul, B.J. Connolly, A.L. Moran, Corrosion 59 (2003) 520-528. M. Bobby-Kannan, V.S. Raja, R. Raman, A.K. Mukhopadhyay, Corrosion 59 (2003) 881-889. Feng Lu, Qunpeng Zhong, Chunxiao Cao, Acta Metall. Sinica 16 (2003) 41-45. A. Tumbull, Corrosion 57 (2001) 175-189. R.H. Jones (Ed.), Stress-Corrosion Cracking: Materials Performance and Evaluation, ASM International, Materials Park, 1992. [8] P.A. Schweitzer (Ed.), Corrosion Engineering Handbook, Marcel Dekker, New York, 1996.
351
Study on stress corrosion cracking of aluminum alloys in marine atmosphere Xiaoyun Zhang, Zhihua Sun, Zhihui Tang, Minghui Liu, Bin Li Institute of Aeronautical Materials, Beijing 100095, China
Abstract
An investigation of the stress-corrosion cracking (SCC) resistance of high-strength aluminum alloys LYI2CZ and LC4CS in a marine atmosphere is presented. Results are compared with those obtained in laboratory accelerated tests. C-ring stress-corrosion test specimens machined from 30-mm (LY 12CZ) and 40-mm (LC4CS) bars in the short-transverse direction relative to the grain structure were used. Stress levels were 60, 70, 80 and 90% of yield strength. All specimens were exposed to the coastal atmosphere of Qingdao City, Shandong Province, China. In addition, alternate-immersion and soak-immersion tests in salt water were conducted in the laboratory. Results showed that SCC performance of aluminum alloys could be evaluated within one year for LC4CS alloy and 3 months for LYI2CZ alloy; otherwise, only exfoliation corrosion was observed. Alternate-immersion tests showed that SCC in aluminum alloys was consistent with SCC of alloys exposed to an outdoor marine atmosphere at the stress of 70% yield strength. I. Introduction
High-strength AI-Cu-Mg and AI-Zn-Mg-Cu alloys are widely used in airframe structures due to their relatively high strength and stiffness. However, the alloys containing appreciable amounts of soluble alloying elements such as Cu, Mg, Si and Zn are susceptible to stress-corrosion cracking (SCC) [ 1-2]. Most SCC data of the alloys available in the literature were obtained in a 3.5% NaCI solution using alternateimmersion and direct-tension stress-corrosion tests. Although the test results can be obtained in a short time by using such tests, the 3.5% NaCI solution does not reflect the condition of environment experienced by airplane service. Because of this, it was necessary to conduct atmospheric SCC tests in seacoast atmosphere that would allow for evaluation of the susceptibility of these alloys to SCC and comparison of the results with those obtained from laboratory accelerated tests [3-4]. The objectives of this study were to: (i) compare the results obtained from marine atmospheric tests with those obtained from laboratory accelerated tests; and (ii) find a correlation between the results of outdoor tests and indoor accelerated tests.
Nonferrous Alloys
352
2. Experimental procedures 2.1. Materials Materials tested were high-strength aluminum alloys LYI2CZ and LC4CS. Chemical compositions and mechanical properties of the alloys are given in Tables 1-3. C-ring type stress-corrosion test specimens was used. Specimens were machined from a 30-mm diameter rod of LY 12 alloy and 40-mm diameter rod of LC4 alloy in the shorttransverse (ST) direction relative to the grain structure in accordance with the ASTM G38 Standard. Tests in a marine atmosphere as well as laboratory accelerated tests were conducted. The applied stresses were equal to 60, 70, 80, or 90% of yield strength (YS). In addition, bent-beam specimens machined from 1.5-mm thick plates in accordance with the ASTM G39 Standard were used in order to compare the results and accuracy o f various types of accelerated tests. The level of applied stress for bent-beam specimens was 70% YS.
Table 1 Chemical composition (wt.%) of LY 12CZ alloy Alloyin8 elements Cu M8 Mn AI 3.8-4.8 1.2-1.8 0.3-0.9 Bal.
Impurities (max.) Fe Si Zn 0.5 0.5 0.3
Table 2 Chemical composition (wt.%) of LC4CS alloy Alloying elements Cu M~ Mn Zn Cr 1.4-2.0 1.8-2.8 0.2-0.6 5.0-7.0 0.10-0.25
Ni 0.1
A1 Bal.
Ti 0.15
Fe + Ni 0.5
Others 0.1
Impurities (max.) Fe Si Others 0.5 0.5 0.1
Table 3 Mechanical properties of the aluminum alloys Alloy
Identity
LYl2 LC4 LY12 LC4
~30 mm rod, CZ ~40 mm rod, CS 1.5-mmthick plate, CZ 1.5-mm thick plate, CS
YS (MPa) 416 587 377 482
UTS (MPa) 549 587 529 538
El (%) 15.5 13.2 -
E (GPa) 72 68 -
CA)YS, yield strength; UTS, ultimate tensile strength; El, elongation; E, modulus of elasticity.
2.2. Test methods C-ring stressed specimens were exposed to the seacoast atmosphere of Qingdao City, Shandong Province, East of China and removed at time of failure. The atmospheric parameters at the Tuandao Island Station are given in Tables 4 and 5, showing typical seacoast and industrial city environments in which NaCl and SO2 are the major components of atmospheric pollutant affecting metallic corrosion.
Volume 1: Chemistry, Mechanics and Mechanisms
353
In addition, other specimens with the same level of applied stress were tested by alternate immersion in the 3.5% sodium chloride (NaCI) solution in accordance with the ASTM G44 Standard. Then, seacoast results were compared with those obtained during laboratory accelerated tests. After testing, the fracture surfaces were examined by scanning electron microscopy (SEM) to supplement the SCC data. 3. Results 3.1. L YI2 and LC4 alloys exposed to seacoast environment
The time-to-failure of stress-corrosion specimens tested in the seacoast environment involved periods of crack initiation and crack propagation. SCC results of LY12 and LC4 alloys exposed to the seacoast atmosphere at the Tuandao Island Station are given in Table 6. The LY12 alloy was more susceptible to SCC than the LC4 alloy. Specimens made of LY12 failed after 7 to 51 days of exposure. For LC4 alloy, the earliest failure was observed at 10 days of exposure. Most failures took place after between one month and one year of exposure, and some specimens did not crack after one year of exposure. However, exfoliation corrosion was typical for the unfailing specimens of LC4 alloy. The results showed that the SCC performance of high-strength aluminum alloys in a seacoast atmosphere could be evaluated within one year for the LC4 alloy and within three months for the LY12 alloy. Fractographic analysis of fracture surfaces that was performed to obtain additional information on SCC behavior indicated the intergranular mode of SCC and some evidence of secondary cracks (Fig. 1). Although the amount of small-angle grain boundaries clearly increased with increasing stress level, the number of secondary cracks on fracture surfaces in our tests did not change significantly at various stress levels (from 60-90% YS) applied.
Table 4 Environmental parameters at the Tuandao Island Station Temperature Relative humidity Rain (~ (%) (days) Maximum Minimum Average Average Minimum 22.2 6.2 13.2 73.7 57.7 68
Rain (mm) 55.3
Wind speed (m/s) 4.75
Table 5 Atmospheric corrosion pollutant concentrations at the Tuandao Island Station Rain analysis (mg/m3) Ammonia Sea salt SO4 2- CI-
Continuous method (mg/(100 cm2.day))
NO2
H2S
SO4 2-
0.1142 0.0725 0.04169 0.0189
1.2182 70800 19998
pH
6.8
Nature dust (~(cm2-month)) Water No water soluble soluble 3 . 7 0 2 0 2.1962
Nonferrous Alloys
354 Table 6 SCC resistance of aluminum alloys exposed to seacoast atmosphere Alloy
Identity
LY 12
~30 mm rod, CZ
LC4
~40 mm rod, CS
Applied stress (%YS) 60 70 80 90 60 70 80 90
F/N (A)
3/6 7/8 8/8 6/8 2/5 5/7 5/7 3/7
Time (days) 21,7,51 14, 21, 31, 7, 18, 18, 23 14, 14, 31, 7, 7, 18, 51, 7 14, 31, 14, 7, 18, 7 320, 42 103, 159, 18, 42, 18 103, 272, 298, 51, 51 51, 10, 51
(A)F is the number of fractured specimens and N is the total number of specimens.
Fig. 1. SEM fractographs of LYl2 alloy exposed to the atmosphere at the Tuandao Island Station with the applied stress levels of: (a) 60, (b) 70, (c) 80, and (d) 90% YS.
3.2. Alternate-immersion SCC tests in 3.5% NaCI Results of the alternate-immersion tests of C-ring specimens in the 3.5% NaCI solution are given in Table 7, which shows that the SCC resistance of the LC4 alloy was higher than that of the LY12 alloy. Most failures in the LY12 specimens were observed during the first week. Several failures in LC4 specimens were observed during the first month. Those LC4 specimens that did not crack showed severe pitting corrosion on their surfaces.
355
Volume 1" Chemistry, Mechanics and Mechanisms
The results suggested that the SCC performance of high-strength aluminum alloys can be evaluated within one week for LY12 alloy and within two months for LC4 alloy by using alternate-immersion SCC tests in the 3.5% NaC1 solution. SEM analysis indicated intergranular mode of SCC (Fig. 2) similar to the cracking mode obtained during atmospheric SCC (Fig. 1). Cracks grew mainly through highand low-angle grain boundaries. The number of low-angle grain boundaries and susceptibility to cracking increased with increasing level of applied stress.
Table 7 SCC resistance of aluminum alloys during alternative-immersion tests in 3.5% NaCI Alloy
Identity
LY 12
t330 mm rod, CZ
LC4
040 mm rod, CS
(A) F
Applied stress ..... (%YS) 60 70 80 90 60 70 80 90
F/N (xF
1/10 4/10 3/10 3/10 5/10 4/10 4/10 5/10
Time (days) 3.8 3.8, 3.8, 3.8, 54 3.8, 3.8, 43 3.8,3.8,3.8 8,41,41,51,51 51,51,51 4,8,8,35 50, 50, 50, 43
is the number of fractured specimens and N is the total number of specimens.
Fig. 2. SEM fractographs of LY 12 alloy during altemative-immersion tests in 3.5% NaCI with the applied stress levels of: (a) 60, (b) 70, (c) 80, and (d) 90% YS.
Nonferrous Alloys
356
3.3. The comparison of different accelerated tests The bent-beam specimens were used for the comparison of different accelerated test methods conducted in the laboratory and seacoast atmosphere at the Tuandao Island Station. All specimens were tested at the applied stress level of 70% YS. Three methods were used: (i) alternative immersion in the 3% NaCI solution; (ii) soak immersion in the 3% NaCI + 0.5% H202 solution; and (iii) soak immersion in the 3% NaCI solution. The results of the comparative tests, given in Table 8, show that soak immersion in 3% NaCI + 0.5% H202 resulted in earliest SCC failures. During alternate-immersion tests in 3% NaCI, SCC was observed in all specimens but the time-to-failure was longer than during soak-immersion tests in the 3% NaCI + 0.5% H202 solution. No failure was observed during soak-immersion tests in 3% NaCI although the tests lasted 239 days. The results of soak-immersion tests in 3% NaCI + 0.5% H202 and alternate-immersion tests in 3% NaCI had good correlation with those obtained during tests in a marine atmosphere. The SCC resistance of the LC4 alloy was higher than that of the LYI2 alloy in all types of stress-corrosion tests. In addition, the LY12 alloy showed susceptibility to intergranular corrosion.
Table 8 SCC resistance of bent-beam specimens exposed to seacoast atmosphere and aqueous solutions(A) Alloy
Identity
Seacoast
F/N LY12 LC4
1.5-mm 5/5 thick plate, CZ 1.5-mm 0/5 thick plate, CS
Alternate immersion 3% NaC1 Time F/N Time (days) (days) "
1
|
03
....
102L
o ~-~ 'E -3 < oo ~ o 0
1
, ....
101 10o1..
O
, .... O
, ....
, ....
O
O
1031
o10~4'
'
''
103 I '
1021 - ~ 1011 _
O
'''
. .0. .
'''0
' ....
~ 10Ol ,'-- E 10.11._
0
I0-IF 1021 --
~ ~ (~ "0 ~"
''
0
o= < 10-21._ 1031"
o .~dic
-~~
''
0
' ....
0
O anodic
'
''
''
'
0 m 950 13_
0 0 in air
m 950 13.
O
"0-- " 0
o
O
o
O
o
--O-. . . . . . . . .
O
-O----"
O
in air~ .
~, 9oo
0~ 9 0 0 (!) (/)
0
O*
!-
O*
(/)
E = 850 E
~ 850 E
X
x
800
O,o, O
:~ 800
750 .... ' .... ' .... ' .... 2000 1500 -1000 500
Ti-6AI-4V pH 3 ' .... 0 500
Applied potential/mVsH
E
Fig. 2. Effects of applied potential on the maximum stress and cathodic current density in the solution with pH 3.
-
Ti-6AI-4V~ pH 1 750 2000
i
,
,
,
I
1500
,
,
.
,
I
,
-1000
a
,
,
I
,
,
,
-500
Applied potential/mVsH
,
,
0
,
,
,
500 E
Fig. 3. Effects of applied potential on the maximum stress and cathodic current density in the solution with pH 1. Asterisks denote conditions under which cracks were observed.
362
Nonferrous Alloys
(Fig. 3), a unique behavior in maximum stress against potential was observed. At a potential a b o v e - 5 5 0 mV, the maximum stress was almost identical to that in air, indicating no EAC. However, the maximum stress decreased when the potential was less than-550 mV. The lowest level of the maximum stress was observed a t - 1 0 0 0 mV. The maximum stress then increased again at potentials higher than -1000 mV. Also, the density of the cathodic current increased monotonically with decreasing potential. Therefore, unlike the hydrogen embrittlement mechanism, the cathodic reaction rate did not strongly affect the EAC susceptibility of the alloy. A few cracks were found on the side surfaces of specimens fractured at pH 1 at potentials near-1000 mV (see asterisks in Fig. 3). Fig. 4 demonstrates typical examples of cracks initiated under cracking conditions. The cracks propagated in the direction of the tensile axis. The direction of crack propagation was independent of the microstructure of the alloy not only along but also normal to the interface of a/13 lafnellar Structure. Lattice structural analyses of side surfaces of specimens fractured under all EAC conditions were conducted by using XRD method. Fig. 5 shows XRD patterns for specimens fractured at various applied potentials at pH 1 as well as before EAC tests. The XRD patterns of specimens fractured at pH 7 and 3 were similar to those obtained before EAC tests. That is, only peaks corresponding to substrate ot and 13 phases of Ti appeared. At pH 1, only peaks identical to the substrate material were detected in the specimen fractured at potentials a b o v e - 5 5 0 mV. In contrast, a t - 1 0 5 0 mV peaks different from those of the substrate were detected, and these were identical to ,/-Till.
Fig. 4. Typical ex.amPles of cracks on the side surfaces of Ti-6AI-4V fractured in the solution with pH 1 at applied potentials of (a)-800 and (b)-1050 mVsrm.
Fig. 5. X-ray diffraction patterns for side surfaces of Ti-6A1-4V before and after the EAC test in the solution with pH 1 at various applied potentials.
Volume 1: Chemistry, Mechanics and Mechanisms
363
As a potential lower than -1300 mV was applied, the peaks of hydride disappeared while the peaks of substrate remained. Therefore, it was inferred from Figs. 4 and 5 that the formation of y-Till and the initiation of cracks occurred readily at approximately -1000 mV at pH 1. The most severe EAC was observed under those conditions. 4. Discussion
Maximum stresses obtained during EAC tests are summarized as functions of potential and pH (Fig. 6). Fig. 6 can be termed a "potential-pH map" of EAC, and provides useful information when attempting to avoid EAC. The four symbols in Fig. 6 specify the groups classified by the level of maximum stress. An asterisk beside a symbol denotes the condition under which cracks were observed. It is clear from the summarized data that (i) the severe EAC evaluated by the maximum stress 933 933> or re=x>915 915> or max>900 900> Orm=x
r I r ..... '= S t r e s s / M P a . . . . /I / JJ J /I /
o
I0
i" ~
Gmax>933
I
1
1
J @ 933> o" max>g15 I
-
l e 915>o-,.,.,,~900 !.. "1 02
......
.",,,
/ -J ~ /
..........
~'/I ~ / 0*
o
0*
...... .~.~
H*
-
H2
VOH . .....
~-
9
@
' 0
'
o "
i
",
'
"-.:, . . . . ',
T ........
i
02
;.,!
@
,
~.1
H§ H2 -4
.6
,
/
9
'
,,
0 0 .2-4 -6
-2
~--. ',
s
"..
o
9
=
",2
. vo 2"
1
='
' " r'-,~Jl,
2
4
6
8
=, 10
I, 12
,/, 14 16
pH
Fig. 7. Summary of the effects of potential and pH on the EAC susceptibility of Ti-6AI-4V represented by maximum stress plotted on the potential-pH diagram of Al/water system.
0
0
V
""
0 O
-2 -2
0
2
4
6
8
10
12
14
16
pH
Fig. 8. Summary of the effects of potential and pH on the EAC susceptibility of Ti6AI-4V represented by maximum stress plotted on the potential-pH diagram of V/water system.
these two factors render the EAC negligible. The severest EAC susceptibility with formation of cracks was found in the soluble ions (Ti 2+ and Ti 3+) stable region. This EAC behavior can result in the two following mechanisms. When the pre-oxide film breaks down during the EAC test, Ti dissolves as ion species [18] and no oxide is formed at any breakdown site. The localized dissolution of Ti produces a notch on the material surface, a n d then causes material fracture at very low stress. This may be termed the "anodic dissolution mechanism." On the other hand, hydrogen easily enters into the material from the breakdown site because of the absence of an oxide film as
Volume 1: Chemistry, Mechanics and Mechanisms
365
well as large overpotential for hydrogen generation reaction. Therefore, hydrogen content in the material surface becomes large enough to form hydride, thereby causing material fracture. This may be termed the "hydrogen embrittlement mechanism." The main mechanism of EAC is not clear. In the hydride stable region, the hydride should be formed at the breakdown site of the pre-oxide film. Since hydride has a brittle nature [ 19,20], the hydride growing during the EAC test is considered to crack at low stress. However, almost no EAC susceptibility was found, and no hydride was detected in this hydride-stable region. The reason for this may be that the hydride resists entry of hydrogen, in which case the hydrogen content in the material becomes relatively low. The EAC map of Ti-6AI-4V was the same as that of the TiAI intermetallic compound shown in Fig. 9, which had been presented in Refs. [12,13]. A similar analysis can be applied to the potential-pH diagrams of A1/water and V/water [14] systems given in Figs. 7 and 8 respectively. The condition with severest EAC ( p H i and ~--1000 mV) occurs in the ion stable region of AI and V. Accordingly, we believe that hydrogen easily enters the substrate, and then hydride is formed for the same reason as that for Ti. The condition of pH 1 and -2000 mV is in the metal stable region of A1 and V, and should also generate hydrogen much more than that which enters the substrate at-1000 mV. However, the results indicate no EAC susceptibility with no hydride formation under the above condition. The findings strongly suggest that the very thin Ti hydride layer formed at -2000 mV prevents hydrogen entry, as mentioned earlier. The results suggest that the EAC behavior of Ti-6AI-4V could be explained from the thermodynamically stable chemical species in the potential-pH diagram of Ti/water system as a first approximation. The suggestion is quite similar to that made earlier for TiA1 [12,13].
Fig. 9. Effects of potential and pH on EAC susceptibility of T-TiAI plotted on the potential-pH diagram of Ti/water system [ 12,13].
366
Nonferrous Alloys
5. Conclusions The effects of electrochemical potential and pH on EAC in Ti-6A1-4V have been investigated in a 0.05 kmol/m 3 Na2SO4 solution at pH from 1-7 at room temperature by the SSRT technique. The following conclusions were drawn: 9 In the solutions with pH 3 and 7, the mechanical properties of Ti-6AI-4V were almost completely independent of applied potential; that is, there was no obvious susceptibility to EAC in the potential-pH region. 9 At pH 1, both maximum stress and fracture strain changed with applied potentials and showed their minimum values at ~--1000 mV. In addition, at this potential, several cracks were found and Ti hydride was detected from the side surfaces of specimens. Therefore, it was concluded that Ti-6AI-4V was susceptible to EAC at the potential of~-1000 mV and pH < 1. References [1] H.M. Burte, E.F. Erbin, G.T. Hahn, R.J. Kotfila, J.W. Seeger, D.A. Wruck, Met. Prog. 67 (1955) 115-120. [2] N.R. Moody, W.W. Gerberich, Metall. Trans. A, 11A (1980) 973-981. [3] N.R. Moody, W.W. Gerberich, Metall. Trans. A, 13A (1982) 1055-1061. [4] D. Hardie, S. Ouyang, Corros. Sci. 41 (1999) 155-177. [5] R.J. Kotfila, E F. Erbin, Met. Prog. 66 (1954) 128-131. [6] G.A. Lenning, C.M. Craighead, R.I. Jaffee, J. Metals 6 (1954) 367-376. [7] T. Takasugi, O. Izumi, Scripta Metall. 19 (1985) 903-907. [8] T. Takasugi, S. Hanada, J. Mater. Res. 7 (1992) 2739-2746. [9] A. Nozue, T. Uchimura, T. Okubo, Zairyo-to-Kankyo (Corros. Eng.) 41 (1992) 728-733. [ 10] A. Nozue, A. Sano, T. Okubo, Zairyo-to-Kankyo (Corros. Eng.) 44 (1995) 287-292. [11] J. Kuminick, H.H. Johnson, Metall. Trans. 5 (1974) 1199-1206. [12] T. Haruna, T. Shibata, T. Iwata, T. Sundararajan, Intermetallics 8 (2000) 929-935. [13] T. Haruna, T. Sundararajan, S. Fujimoto, T. Shibata, in: Proc. 15th International Corrosion Congress, Granada, 2002, paper no. 493 (CD-ROM). [14] M. Pourbaix (Ed.), Atlas of Electrochemical Equilibria in Aqueous Solutions, Pergamon Press, New York, 1966. [I 5] M. Pourbaix, Rapports Techniques CEBELCOR, 107, RT. 146, 1968. [16] M. Pourbaix, in: J.C. Scully (Ed.), The Theory of Stress Corrosion Cracking in Alloys, NATO Scientific Affairs Division, Brussels, 1971, pp. 17-63. [17] R.W. Schutz, L. C. Covington, Corrosion 37 (1981) 585-591. [18] A. Pourbaix, M. Marek, R.F. Hochman, in: J.C. Scully, R.F. Hochman (Eds.), Stress Corrosion Mechanisms in Titanium Alloys, NACE, Houston, 1971, pp. 35-44. [19] D.N. Williams, J. Inst. Metals 91 (1962/1963) 147-152. [20] G. Sanderson, J.C. Scully, Corros. Sci. 6 (1966) 541-542.
367
On the competitive effects of water vapor and oxygen on fatigue crack propagation at 550~ in a Ti6242 alloy C. Sarrazin-Baudoux
a
F Loubat b S Potiron
a
Laboratoire de Mdcanique et de Physique des Mat~riaux, UMR CNRS n ~ 6617, ENSMA B.P. 109, Chasseneuil de Poitou, 86960 Futuroscope Cedex, France b Laboratoire Mdcanique Physique, CNRS/Universitd Bordeaux I, 351, cours de la Libdration, F-33405 Talence Cedex, France a
Abstract
The influence of various gaseous atmospheres on the fatigue crack propagation behavior of a Ti6242 alloy was studied at 550~ Tests were conducted under selected environmental conditions (ambient air, high vacuum and humidified gaseous atmospheres (80% RH) including pure argon, 80% argon + 20% oxygen, and 80% nitrogen + 20% oxygen). The test frequency was 0.05 Hz with additional data collected at 5 Hz. The crack propagation rate was highly sensitive to environments, with a strongly detrimental effect of water vapor. A model is proposed to account for the influence of partial pressures of water vapor and oxygen, of test frequency and of effective stress intensity factor.
I. I n t r o d u c t i o n
Cracking and failure of structural materials in various environments remains a safety and economic problem despite the enormous effort that has been directed understanding the phenomena of fatigue, stress corrosion and creep. Because of their resistance to corrosion, high specific strength and low density, titanium alloys are of interest to the chemical processing industry. This is particularly the case in view of a new process of hydrothermal oxidation for organic effluent treatment. Such a process involves a supercritical water medium at a pressure of 280 bar and a temperature of 550~ Knowledge of the resistance of titanium alloys under such critical conditions is essential, but the tests involved are expensive and time-consuming. The objective of this study is to evaluate the corrosion fatigue resistance of the Ti6242 alloy under supercritical conditions based on tests performed in dry corrosion conditions at atmospheric pressure and a temperature of 550~
Nonferrous Alloys
368
2. Experimental The Ti6242 alloy (6% AI, 2% Sn, 4% Zr, 2% Mo) provided by Timet was received in the form of a 250-mm diameter disk 20-mm thick, fl-forged at 950~ Specimens were cut by sparkling machining, heat treated in a vacuum at 970~ for 1 h, argon quenched, maintained in a vacuum at 595~ for an aging time of 8 h and finally argon quenched. The resulting microstructure (Fig. 1) is composed of large globular primary a in a transformed fl matrix with fine secondary alpha precipitates.
Fig. 1. Microstructure of the Ti6242 alloy.
Fatigue crack growth experiments [1] were conducted on compact tension C(T) specimens 10 mm thick and 40 mm wide in accordance with ASTM Test recommendations for the Measurements of Fatigue Crack Growth Rates (E 647-88). A servo-hydraulic machine equipped with an environmental chamber and a furnace allowed testing in ambient air, high vacuum (3 x 10-4 Pa) and in humidified gaseous atmospheres (80% RH). These gaseous atmospheres included pure argon, 80% argon + 20% oxygen, and 80% nitrogen + 20% oxygen (synthetic air) at temperatures ranging up to 550~ Crack lengths were tracked using a direct current (DC) potential drop method. Specimens were submitted to sinusoidal loading at frequencies ranging from 10-3 to 5 Hz with a stress ratio (R) of 0.1. Microfractographic observations of cracked surfaces were performed by means of a scanning electron microscope (SEM) and an optical microscope.
3. Corrosion fatigue crack propagation 3.1. Intrinsic fatigue crack propagation To analyze the influence of the atmospheric environment on the fatigue crack propagation behavior of the Ti6242 alloy at temperatures ranging from 500 to 600~ reference tests were first performed in high vacuum to identify the intrinsic fatigue
Volume 1" Chemistry, Mechanics and Mechanisms
369
crack propagation regime. The fatigue crack growth rate was measured on a specimen tested at 5 Hz and 0.05 Hz in a vacuum of at least 10-4 Pa, as illustrated in Fig. 2. In the explored AK range, the crack growth rate da/dN progressively accelerated as AK increased, da/dN appeared independent of the test frequency.
vacuum, 5500C
10 -s
9
5Hz
9
0.05 Hz
s
A
o
mm
o 1.1
I
E Z "o t~ "o
10-7
w' II
/
104
. 20
.
f
ID le
4)
. 30
. 40
50
AK (MPaVm) Fig. 2. Intrinsic fatigue crack propagation at 550~ in high vacuum.
The stabilized regime (dashed line in Fig. 2) may well result from a pure fatigue mechanism, without contribution from time-dependant mechanisms such as creep or corrosion. That regime can be described using the relation derived by Petit et al. [2] for the intrinsic stage II propagation in the following form: (da/dN)o = A/Do (AKeff/E)4
(1)
where ,4 is a dimensionless parameter, Do is the cumulative plastic displacement at the crack tip, E the Young modulus, and AKeff the effective stress intensity factor range. As a first approximation, the crack closure contribution can be neglected in the explored rate range. Hence Eq. (1) can be used with AK ~ AKeff, as illustrated by the dashed line in Fig. 2. The ductile fracture surface morphology illustrated in Fig. 3 is consistent with a stage II crack propagation mechanism resulting from a plastic deformation accumulated at a crack tip during load cycling [2]. As classically shown in vacuum in metallic alloys, striations corresponding to a cycle-by-cycle crack advance are not observed. The coarse microscopic markings of the fracture surface may be more related to a step-bystep crack advance with each step corresponding to several loading cycles [3].
Nonferrous Alloys
370
Fig. 3. Microfractography of the fracture surface in high vacuum at 550~ at zkK= 20 MPa~/m.
=P
10a O vacuum
,.,/" 9
9 air
A
~
m m
o o E 10~
Z
"o "o
10"7
,
,
,
10
,
,
,
,
,
I
100 AK (MPaVm)
Fig. 4. Comparison of crack growth rates in air and in high vacuum at 550~
3.2. Fatigue crack propagation in air at high temperature Data generated by tests performed at 550~ in air at 0.05 Hz are plotted in Fig. 4 in comparison with the intrinsic data at high vacuum. Crack growth rates in air were substantially higher than in high vacuum at AK < 35 MPa~/m. This is consistent with previous studies on the intrinsic fatigue crack propagation of metallic alloys supported with a well-marked effect of environment, particularly in the low rate range and near the threshold [4]. An illustration of the cracked surface in air (Fig. 5) shows transgranular propagation with typical striations (Fig. 3) corresponding to a cycle-per-cycle crack advance. This is in contrast with results obtained in a vacuum. This detrimental effect of air agrees with the literature on tests performed at comparable temperature (520 and 590~ on the same alloy [5] and
Volume 1: Chemistry, Mechanics and Mechanisms
371
on other titanium alloys, including Ti6246 [2,6]. The detrimental effect of oxidation is more common explanation given for such a rapid propagation in air at elevated temperature in titanium alloys and other metallic alloys [5-8]. However, recent studies on various titanium alloys, including Ti6242, Ti6246 and IMI834, demonstrated a detrimental effect of atmospheric water vapor. Water vapor can have an effect even under very low partial pressure when operating at low frequency [2] and in the low crack growth rates domain.
Fig. 5. Microfractographic aspects of the fracture surface in air at 550~ and AK ~ 26 MPax/m. The effect of environmental factors is typically characterized by a change in the slope of the curves, switching from 2:1 in air to 4:1 in vacuum for AK < 35 MPa~/m (Fig. 4). For higher AK, the environmental assistance disappears, with the curves all having a slope of 4:1. To acquire a deeper insight into the mechanism governing the Ti6242 alloy, tests were performed in atmospheres containing various partial pressures of oxygen and water vapor at various frequencies. 3.3. Influence of partial pressure of water vapor and oxygen
The data resulting from tests performed at a frequency of 0.05 Hz in various humidified gaseous atmospheres (80% RH at room temperature) that included purified argon ( 1) and JITF was twice as high as Jic.
402
Iron and Nickel based Alloys
CLA steel (400~
air, 800x
CLA steel (400~
3.5% NaCI, 400x
CLA steel (200~
H2S solution, 400x
HSLA steel, air, 1100x
CLA steel (600~
3.5% NaCI, 700x
HSLA steel, H2S solution, 600x
Fig. 2. SEM fractographs showing the mechanisms of fracture in CLA and HSLA steels.
Volume 1: Chemistry, Mechanics and Mechanisms
403
4. Conclusions The susceptibility to SCC of CLA and HSLA steels exposed to NaCI solutions and a HES-Containing solution was studied. The following two conclusions were reached: 1. Two threshold parameters that determined the susceptibility of CLA and HSLA steels to SCC were measured at the onset of crack initiation (K/J)IIF and at the onset of terminal fracture (K/J)ITF. The ratios between these threshold parameters and the corresponding values of Kic determined the degree of susceptibility to environmentinduced cracking. The ratios ranged from 0.34 to 0.65 in NaCI solutions and from 0.28 to 0.42 in a H2S-containing solution. The H2S-saturated acidic solution was more aggressive than NaCI solutions. For HSLA steel tested in NaCI solutions, the ratios (K/J)~w and (K/J)m were both more than 1, indicating that the steel had an increased resistance to crack propagation. 2. The dominant micromechanism of crack initiation in CLA steels exposed to air and aqueous solutions was IG. However, during subcritical crack growth, features such as scattered pits, slip marks, secondary cracks, etc. appeared on the fracture surface when the steels were tested in aqueous solutions. For HSLA steel, a completely microvoid-coalescence mechanism was typical, and resulted in small equiaxed dimples on the plane of fracture in air. This mechanism changed when HSLA steel was tested in aqueous solutions: features such as elongated voids, fine cracks, joining blistering, deformation marks, and pits were observed. Hydrogen embrittlement appeared to be the dominant mechanism of crack growth in a H2Scontaining solution, while in NaCI solutions the dominant mechanism was apparently SCC. References [1] [2] [3] [4] [5] [6] [7] [8] [9]
[10] [11] [12] [13]
[14]
W.W. Gerberich, Y.T. Chen, Metall. Trans. A, 6A (1975) 271. H.W. Liu, J. Basic Eng., Trans. ASME 92 (1970) 633. T. Taira, K. Tsukada, Y. Kobayashi, H. Inagaki, T. Watanabe, Corrosion 37 (1981) 5. A.W. Thompson, Mater. Sci. Eng. 43 (1980) 41. I.M. Bemstein and A.W. Thompson, Resisting hydrogen embrittlement, in: J.K. Tien, G.S. Ansell (Eds.), Alloy and Microstructural Design, Academic Press, New York, 1976, p. 303. J. Gonzalez, F. Guti6rrez-Solana, J.M. Varona, MetaU. Mater. Trans. A, 27A (1996) 28. F. Guti6rrez-Solana, A. Valiente, J. Gonzales, J.M. Varona, Metall. Mater. Trans. A, 27A (1996) 291. B. Pawlowski, A. Mazur, S. Gorczyca, Corros. Sci. 32 (1991) 685. A.W. Thompson, Effect of metallurgical variables on environmental fracture of engineering materials, in: Z.A. Foroulis (Ed.), Environment-Sensitive Fracture of Engineering Materials, TMS-AIME, Warrendale, 1977, p. 379. B. London, D.V. Nelson, J.C. Shyne, Metall. Trans. A, 19A (1988) 2497. N. Eliaz, A. Shachar, B. Tal, D. Elieze, Eng. Failure Analys. 9 (2002) 167. A.N. Kumar, Ambient and Environmental Fracture of Low Alloy and HSLA Steel, Ph.D. Thesis, The Indian Institute of Technology, New Delhi, 1983. A.N. Kumar, R.K. Pandey, Subcritical cracking behaviour under environments in a low strength-high toughness steel, in: S R Valluri et al. (Eds.), Advances in Fracture Research: Proc. 6th International Conference on Fracture, vol. 4, Pergamon Press, New York, 1986, p. 2427. J.A. Begley, J.D. Landes, The d integral as a fracture criterion, in: Fracture Toughness: Proc. 1971 National Symposium on Fracture Mechanics, part 2, ASTM STP 514, ASTM, Philadelphia, 1972, p. 1.
404
Iron and Nickel based Alloys
[ 15] ASTM Standard, part 31, ASTM, Philadelphia, 1973. [16] A.N. Kumar, R.K. Pandey, Micromechanics of fracture in microalloyed steel, in: J. Carlssont, N.G Ohlson (Eds.), Mechanical Behavior of Materials-IV: Proc. 4th International Conference, vol. 2, Pergamon Press, New York, 1984, p. 1109. [17] A.N. Kumar, R.K. Pandey, Trans. Indian Inst. Met. 36 (1983) 436.
405
Corrosion-fatigue properties of surface-treated surgical implant stainless steel X2CrNiMo 18-15-3 G. Mori a
a
H. Wieser b H. Zitter
a
Department for General Analytical and Physical Chemistry, University of Leoben, Franz-Josef-Str. 18, A-8700 Leoben, Austria b Siemens AG A&D ET, Siemensstrafle 10, D-93055 Regensburg, Germany
Abstract
Certain applications of steels in surgical implants require a rough surface. The jagged surface structure facilitates the adhesion of bone tissue or surface coatings. For this purpose, certain steel implants are peened with coarse and sharp-edged peening mediums. Spherical peening balls are used to avoid a significant loss of fatigue strength. The present paper shows that the effect of surface roughening for stainless steel X2CrNiMol8-15-3 can be counteracted by the effect of compressive residual stresses in the surface layer. However, the effect of surface flaws cannot be ignored. In this study, the shape and depth of surface flaws in shot-peened surfaces was examined and a classification of surface flaws presented. Fatigue tests in laboratory air and in the standard physiological 0.9% NaC1 solution demonstrated that the depth and shape of surface flaws have a distinct influence on fatigue strength. The use of coarser peening balls and a subsequent electrolytic polishing treatment increased fatigue strength. Shot peening can have a positive and negative effect on corrosion-fatigue strength. I. Introduction
It is well known that shot peening has a positive influence on the fatigue properties of metallic components [1,2]. The two reasons for this effect are: (i) the formation of residual compressive stresses at the surface or at layers near the surface [3-5], and (ii) the cold working effect of the surface area yields a higher strength [6-9]. For a variety of applications in the automotive and the aeronautical industry, shot peening has become a standard treatment for cyclic loaded parts. Surgical implants are frequently shot-blasted with glass or ceramic particles. In such cases, it is otten not the improvement of the fatigue strength that is the primary focus of such a treatment. Rather, the purpose is a roughening of the surface that ensures a better adhesion of the implant to the bone tissue. Clearly, that surface flaws may have a negative influence on fatigue strength. Because shot peening with sharp-edged peening media generates only low residual
406
Iron and Nickel based Alloys
compressive stresses, the risk of fatigue or corrosion-fatigue cracking increases significantly. Therefore, it is advantageous to use spherical balls for the shot-peening treatment of surgical implant materials. Roughening of the surface by shot peening and its influence on fatigue or corrosionfatigue properties is rarely described in the literature. The commonly-used roughness parameters Ra (averages of all peaks and valleys of the roughness profile, see Fig. 1) or Rz (arithmetic mean of the single roughness depths, see Fig. 2) are not sufficient for the description of the influence of surface conditions of shot-peened implants. In order to estimate notching effects, it is necessary to have statistically-based information on the real shape and depth of the different types of surface flaws. This information can hardly be obtained by the use of commonly-used measuring methods for surface characterization.
Fig. 1. Definition of Ra.
Fig. 2. Definition of Rz (equal to arithmetic mean of single roughness depths Rzi).
2. Experimental All investigations were conducted with austenitic stainless steel X2CrNiMol 8-15-3 (material no. 1.4441) machined from bars. The chemical composition of this steel is given in Table 1. The values of ultimate tensile strength, yield strength and elongation at failure of cold-worked steel were 1220 MPa, 950 MPa and 15% respectively. The pitting resistance equivalent number (PREN -- % Cr + 3.3% Mo + 16% N) of the steel was 26.7, which represents a low probability for pitting in body fluids and in a physiological 0.9% NaCI solution [10].
407
Volume 1" Chemistry, Mechanics and Mechanisms
Table 1 Chemical composition of X2CrNiMol 8-15-3 steel (wt.%) C 0.018
Si 0.41
Mn 1.88
P 0.018
S 0.001
Cr 17.31
Ni 13.85
Mo 2.84
Table 2 Test program for differently-treated specimens Surface treatment
Fatigue tests (no. of load levels) Air 0.9% NaCI
Corrosion tests Metallographic Potentio- Corrosion examination of dynamic potential surface
Electrolytically polished 3 4 Yes Yes Yes Shot peened Z600 3 3 Yes Yes Yes Shot peened Z600, polished 3 3 Yes Yes Yes Shot peened Z300 7 Yes Polished, annealed* 2 . . . . Shot peened Z600, an.nealed* 2 --: * Annealing at 580~ for 1 h under helium atmosphere followed by furnace cooling.
Experimental procedures are described in Table 2. The following types of experiments were conducted: 1. Characterization of shot-peened surfaces and estimation of maximum surface flaws; 2. Investigation of pitting behaviour by potentiodynamic measurements in a physiological 0.9% NaCl solution; 3. Corrosion potential measurements in a 0.9% NaCl solution under unloaded, static and dynamic loading conditions; 4. Fatigue tests in laboratory air and a physiological 0.9% NaCl solution; and 5. Tensile tests with low stress rates in a physiological 0.9% NaCl solution. Shot peening was performed with ceramic peening balls having average diameters of 0.21 mm (Z210 balls), 0.3 mm (Z300), and 0.6 mm (Z600). The Almen-Intensity parameter was used as a measure of the deformation energy brought into specimen surfaces during peening. (Note: The bending height of thin strips of Ck70 carbon steel, hardened and tempered to the hardness from 44-50 HRC (A-strip: 67 x 19 x 1.3 ram, N-strip: 67 x 19 x 0.8 mm), due to the elongation of the surface during peening is called "Almen-Intensity" (in mm)). Almen-Intensities for various peening pressures are shown in Fig. 3. Surface coverage was 98%.
3. Shot-peened surface characterizations In most cases, shot-peened surfaces were characterized by using roughness measurements [ 11-18]. The determined characteristic values of Ra or Rz were used for quantitative description of surface roughness. It is not possible to characterize the surface flaws with the highest notching effect by using such measurements for the following reasons [ 19]:
408
Iron and Nickel based Alloys
1. The stylus tip cannot penetrate narrow gaps or surface flaws because of its rounded geometry; and 2. It is not possible to detect masked surface flaws such as overlappings or particles of the peening material that are pressed into the metal surface. Therefore, in the present study, the shot-peened surface profile was characterized by using sharp-edged metallographic sections and scanning electron microscopy (SEM). Fig. 4 shows four types of surface flaws that might be formed on shot-peened surfaces: 1. Shallow and rounded impressions formed by the impact of balls made of shotpeening material; 2. Sharp-edged notches can be formed by the broken parts of a peening medium. In some cases these broken parts are pressed deeply into the surface by the impact of other peening balls; 3. The local plastic deformation of the impact zone causes the formation of elevated areas of material around impact craters. If these are hit by other peening balls, overlappings will be formed; and 4. Cracks might be formed by local cyclic plastic deformations due to the impact of peening balls and maximum Hertzian pressures below the surface. Consequently, cracks on shot-peened surfaces may be perpendicular or parallel to the surface. It is reasonable to expect a small influence of shallow and rounded impressions in the surface on fatigue properties. The notching effect of shot-peened surfaces is determined by other types of surface flaws (Nos. 2 to 4 in the above list). Depth and shape of surface flaws were analysed statistically. Fig. 5 shows that more than 80% of all surface flaws were notches caused by broken peening parts sticking to the surface (see Fig. 4(c)). Most of them were very sharp, and thus they could be regarded as cracks rather than notches. The crucial parameter for the effect of cracks in addition to sharpness was their maximum depth. The estimation of the maximum depth of notches from surface-profile data (e.g., of sharp-edged sections) was challenging since it was very difficult to cut the notch exactly on the deepest point. This problem is discussed in detail elsewhere [20]. The best results could be obtained by using the statistical extreme values method [21 ]. ['i';::~'~"i "~- ~ ......................... ''-i' ......................
,,3 1
9 ~
~,,1 .~1
! o o H.s~,'im !!..A.,,~i~
~~ ' &.g 0,9 ..,v. ~ ~'I:~,~
.... ~ o.5 [ 0,3
0,45
o,
...
i I .7,.-""
< ...................................... "
.~ "r 0,35 ~ &
*
"'"
0.3 ~~ K ~ 0,25 ~
S o,~
0,2
.,...-
"
" -""
~'~
0.15 0,1 2
3
4
5
6
................. peem'ng pre,vsurr [bar] .........................
Fig. 3. Almen-Intensities for different shot-peening parameters.
Volume 1: Chemistry, Mechanics and Mechanisms
409
(d) Cracks formed by the repeated impact of peening balls Fig. 4. Metallographic (left) and SEM (right) surface views of various types of surface flaws.
410
Iron and Nickel based Alloys
Fig. 5. Fractions of various types of surface flaws of peened surfaces.
Fig. 6. Maximum depths of the surface flaws of differently peened surfaces.
The results of the analysis of extreme values for the parts that were shot peened with different parameters are presented in Fig. 6. It can be seen that the depth of surface flaws was influenced by the peening pressure more than by the size of peening balls. The larger peening balls caused higher Almen-lntensities and, therefore, higher depths of residual stresses. In order to achieve the high level and depth of residual stresses and the low degradation of the surface, the shot-peening parameters were fixed with the peening pressure at 3 bar and the diameter of peening balls at 0.6 mm (Z600). The surface profile of shot-peened surfaces after electrolytic polishing was irregular due to the local concentration of anodic current density in the elevated zones and edges of the rough surface. A similar effect was observed on non-metallic inclusions in electrolytically-polished smooth surfaces. The electrolytic polishing of a shot-peened surface smoothened sharp-edged surface flaws and resulted in a significant decrease of the notching effect. If the state of residual stresses was only slightly influenced by the
Volume 1: Chemistry, Mechanics and Mechanisms
411
electrochemical treatment, the electrolytic polishing treatment would have a positive influence on fatigue properties. In order to achieve this positive effect, it was necessary to remove a thin surface layer ~20 ~tm thick.
4. Fatigue and corrosion-fatigue tests The results of fatigue tests in laboratory air are shown as a probability-plot of the Weibull-distribution in Fig. 7. In these tests, electrolytically-polished specimens were used as a reference. The results could be summarized as follows: - The shot-peening treatment increased fatigue strength by ~15% in comparison to electrolytically-polished test specimens; - Subsequent electrolytic polishing of the shot-peened surface increased fatigue strength by a further 10% (thus the total increase in comparison to the reference state was ~25%); - The scatter of fatigue strengths for shot-peened and then electrolyticallysmoothened specimens was much less than that for other specimens; - The number of cycles to failure for electrolytically-polished specimens and shotpeened specimens varied between 105 and 1 0 6 , whereas shot-peened and electrolytically-smoothened specimens failed after 2 x 106 and 2 x 107 cycles. At the same Almen-Intensity, the use of smaller peening balls caused deeper surface flaws (Fig. 6). In order to investigate the influence of these deeper flaws on fatigue properties, the fatigue tests were conducted with specimens that were peened by smaller alumina balls with an average diameter of 300 ~tm and peening pressure of 7.5 bar. The fatigue strength of these specimens was ~10% lower than that of the specimens that were shot peened with larger balls (Fig. 8). In order to quantify the effect of surface flaws on fatigue properties, electrolyticallypolished and shot-peened specimens were annealed to the lower level of residual (compressive) stress. The results of fatigue tests in laboratory air are shown in Fig. 9. The annealing treatment of electropolished specimens caused only a small decrease in
Fig. 7. The effect of surface conditions on the Weibull plot for fatigue tests in laboratory air.
412
Iron and Nickel based Alloys
Fig. 8. The effect of various peening parameters on the Weibull plot for fatigue tests in laboratory air.
Fig. 9. The effect of surface flaws on the Weibull plot for fatigue tests in laboratory air. fatigue strength (5%). The annealing treatment atter shot peening reduced the fatigue strength by 40% compared to electropolished specimens. Surface flaws that still remained on the surface of specimens reduced their fatigue properties. The results of corrosion-fatigue tests are summarised in Fig. 10. The fatigue strength of electrolytically-polished specimens in the physiological NaCl solution was ~13% lower than that in laboratory air, although the specimens did not show any signs of localized corrosion during visual and SEM examination. The effect of a physiological solution on the fatigue strength was larger for shot-peened specimens (-30%). This can be explained by the roughening and cold working of the peened surface compared to the electropolished one:
Volume 1: Chemistry, Mechanics and Mechanisms
413
Fig. 10. Weibull-plot for fatigue tests in laboratory air and in 0.9% NaCI.
-
A rough surface has a higher specific surface area, and localized anodic attacks may occur in narrow depressions; and - The higher dislocation density on the surface of shot-peened specimens can cause the shift of corrosion potential to more negative values. This required additional electrochemical measurements, as described below. The lowest corrosion effect of the salt solution on the fatigue strength was attained with shot-peened and electrolyticaUy-smoothened specimens. A decrease of fatigue strength of only 4% occurred in the physiological solution. Obviously, the positive effect of electrolytic smoothening is larger in a corrosive environment than in air. 5. Measurement of corrosion potentials Reasons for the high loss of fatigue strength of shot-peened specimens in a corrosive environment were not obvious. Because of this and in order to check whether peening had an effect on corrosion potential, further studies were conducted to characterize the corrosion behaviour of variously treated specimens. Because pitting corrosion might occur in the physiological solution at a low probability [ 10], a pitting potential was determined by potentiodynamic measurements in this solution at 37~ The pitting potential was not influenced by various surface treatments. Its values were between 530 and 640 mVsHE (+59 mV). Furthermore, SEM investigations of failed specimens showed that pitting was not the cause of corrosionfatigue cracks. Therefore, the presence of pitting was not considered relevant to corrosion-fatigue properties under our testing conditions. In other experiments, the corrosion potential was measured during tensile tests at slowly-increasing stress rates (1.8 x 10-3 MPa s-l). Fig. 11 shows typical potential-time plots compared to potential-stress plots for different surface states. The initial potential of the reference state (electrolytic polishing) was ~300 mV higher than that of shotpeened specimens. Distinct decrease in the corrosion potential of electropolished
414
Iron and Nickel based Alloys
specimens was found at stresses far below yield strength. Such decrease can be explained by the microplastic deformation that takes place at low stress levels. This effect could not be observed with shot-peened specimens because the compressive residual stresses on their surface layer would preclude any plastic deformation at low stress levels.
Fig. 11. Corrosion potentials as functions of time and stress during slow strain rate tests.
Similar effects can be observed for corrosion potentials during cyclic loading. Fig. 12 illustrates the corrosion potentials before cyclic loading and after 5 x l0 5 cycles. Most electrolytically-polished specimens and specimens shot-peened and then smoothened showed significant decrease in corrosion potentials, whereas shot-peened specimens showed only small increase. The activation of specimens during shot peening was proved by their ignoble potential at the beginning of corrosion-fatigue tests. At this potential, higher corrosion rates were observed for pronounced decreases in the fatigue strength of shot-peened specimens when tested under corrosive conditions. 6. Conclusions
The literature clearly shows that both compressive residual stresses and surface hardening increase the fatigue strength of shot-peened metals. Moreover, this effect was counteracted by roughening of the surface. Because measurements of the depth and shape of surface flaws can be made only by using destructive test methods (e.g., investigation of sharp-edged metallographic sections), estimation of the degradation by surface flaws was very time-consuming. Furthermore, the influence of small surface flaws on fatigue properties can be calculated only with good difficulty, since small cracks or surface flaws have an irregular behaviour that still cannot be described exactly. However, the determination of the maximum depth of surface flaws might be helpful for the optimisation of shot-peening parameters.
Volume 1: Chemistry, Mechanics and Mechanisms
400
:~
415
..............
lO ~
-~
~O
~,~
., =: .~,, . . . . . . . . . . . . . . . . . . . -
!/
~~
J
/
-
/
"
0 shotJ~eened ," " l 0 shot ioeened ana /" " decl,otvticpolii~hed'
-,~ ~. W
.-"/
/
300 .......-l-O electl,olytic ~oli' ,hedO
,;.--
9
,, "
200 "
/
- ~It'~
~'
,,
ii
0
-" s~
!
100 .
410./
II
o
o "o"
"
.-"
-I00
~
. . . . . . . . . . . . . . . . . . . o.
-200
9
0 ....
..1
1
1
I ....
I
I
I
t
! I
ol
ii I
I
l
100 200 300 Corrosion potential before cyclic loading [m V.]
1
1
1_
400 "~
Fig. 12. Corrosion potentials measured before and during cyclic loading. The results of our fatigue tests are summarized in Fig. 13. As expected, the shotpeened and smoothened specimens showed the best results in both laboratory air and a corrosive environment. It was surprising that shot-peened specimens showed such large decrease in fatigue strength when tested in a corrosive environment. Visual and SEM examinations of tested specimens revealed no evidence of corrosion. Moreover, the results of corrosion potential measurements did not provide a satisfactory explanation for the lower resistance against corrosion fatigue aRer shot-peening treatment. The large decrease of fatigue strengths for shot-peened and annealed specimens (Fig. 13) indicate that the use of rough (e.g., sand-blasted) surfaces should be used for surgical implants only when specimens contain high compressive residual stresses at and near the surface. Therefore, the use of rounded peening balls for surface treatment is highly recommended. References
[ 1] B. Scholtes, E. Macherauch, Z. Metallkunde 77 (1986) 322-337 (in German). [2] H. Bomas, F. Hoffmann, P. Mayr, Fatigue behaviour of surface treated parts, in: D. Munz (Ed.), Proc. Conference on the Fatigue of Metallic Materials, DGM, Oberursel, 1984, pp. 321-367 (in German).
416
Iron and Nickel based Alloys
Fig. 13. Change of fatigue strength in comparison to the reference state (electrolytically- polished specimens).
[3] H. Wohlfahrt, Model for prediction of residual stresses in shot peened surfaces, in: E. Macherauch, V. Hauk (Eds.), Residual Stresses: Appearance, Measurement, Evaluation, DGM, Oberursel, 1983, pp. 301-319 (in German). [4] O. V0hringer, Changes in the state of the material by shot peening, in: H. Wohlfahrt, R. Kopp, O. V0hringer (Eds.), Shot Peening: Science, Technology, Application, DGM, Oberursel, 1987, pp. 185-204. [5] W. Cao, R. Fathallah, J. Barralis, L. Castex, Residual stresses in shot peened metal components. SHOTPEEN: An interactive prediction software, in: Proc. 4th International Conference on Residual Stresses, SEM, Bethel, 1984, pp. 589-597. [6] R. Clausen, Determination of influences on shot peening by single grain experiments, in: A. Niku-Lari (Ed.), Proc. First International Conference on Shot Peening, Pergamon Press, Paris, 1981, pp. 97-111 (in German). [7] L. Wagner, G. LtRjering, Influence of shot peening on the fatigue behaviour of titanium alloys, in: A. Niku-Lari (Ed.), Proc. First International Conference on Shot Peening, Pergamon Press, Paris, 1981, pp. 279-285. [8] H. Hanagath, O. V6hringer, E. Macherauch, Fatigue behaviour of shot peened TiAI6V4 in the temperatur range 20~ < T < 450~ in: H. Wohlfahrt, R. Kopp, O. VShringer (Eds.), Shot Peening: Science, Technology, Application, DGM, Oberursel, 1987, pp. 194-195. [9] L. Wagner, G. Ltitjering, Influence of the shot peening parameters on the surface layer properties and the fatigue life of TiAI6V4, in: H.O. Fuchs (Ed.), Proc. 2nd International Conference on Shot Peening, Pergamon Press, Chicago, 1984, pp. 194-200. [10] H. Zitter, D. Schaschl-Outschar, Mater. Corros. 32 (1981) 324-331 (in German). [11] H. Wohlfahrt, Shot peening and fatigue behaviour, in: A. Niku-Lari (Ed.), Proc. First International Conference on Shot Peening, Pergamon Press, Paris, 1981, pp. 675-693 (in German). [12] P. Starker, E. Macherauch, Z. Metallkunde 74 (1983), 109-115 (in German).
Volume 1: Chemistry, Mechanics and Mechanisms
417
[13] J. Broichhausen, M. Telfah, Metallurgy 35 (1981), 208-212 (in German). [14] P. Mille, E.M.M. Sutter, A. Comet, Relationship between surface treatment and fatigue damage of TiAI6V4 titanium alloy, in: G. Luetjering, U. Zwicker, W. Bunk (Eds.), Titanium Science and Technology, 1-4, DGM, Oberursel, 1984, pp. 2155-2162. [15] L. Wagner, C. Gerdes, G. Luetjering, Influence of surface treatment on fatigue strength of TiAI6V4, in: G. Luetjering, U. Zwicker, W. Bunk (Eds.), Titanium Science and Technology, 1-4, DGM, Oberursel, 1984, pp. 2147-2154. [16] O. Knotek, R. Elsling, Computer simulation of different surface topographies of metall produced by blasting processes, in: H. Wohlfahrt, R. Kopp, O. VShringer (Eds.), Shot Peening: Science, Technology, Application, DGM, Oberursel, 1987, pp. 361-368. [17] H.P. Lieurade, A. Bignonnet, Fundamental aspects of the effect of shot peening on the fatigue strength of metallic parts and structures, in: H. Wohlfahrt, R. Kopp, O. VOhringer (Eds.), Shot Peening: Science, Technology, Application, DGM, Oberursel, 1987, pp. 343359. [18.] M. Papakyriacou, H. Mayer, C. Pypen, H. Plenk, Jr, S. Stanzl-Tschegg, Int. J. Fatigue 22 (2000) 873-886. [ 19] H. Wieser, H. Zitter, Characterization of shot peened surfaces, in: D. Kirk (Ed.), Proc. Fifth International Conference on Shot Peening, Coventry University, Oxford, 1993, pp. 191198. [20] H. Wieser, Effect of Shot Peening on Fatigue and Corrosion Bahaviour of an Austenitic Surgical Implant Steel, Ph.D. Thesis, University ofLeoben, Leoben, 1997 (in German). [21 ] E.J. Gumpel, Statistics of Extremes, Columbia University Press, New York, 1957.
419
Stress corrosion cracking of austenitic stainless steel Type 3 16 in acid solutions and intergranular SCC mechanism" effects of anion species (CI-and SO42-) and sensitizing temperature R. Nishimura
a,
A. Sulaiman b, y. Maeda
a
a Department of Applied Materials Science, College of Engineering, Osaka Prefecture University, I-1 Gakuen-cho, SakaL Osaka 599- 8531, Japan b Research and Development Centrefor Metallurgy, Komleks P USPIPTEK, Serpong, Tangerang, 15314, Indonesia
Abstract The stress corrosion cracking (SCC) of a commercial austenitic stainless steel type 316 was investigated as a function of sensitizing temperature (800-1300 K) in 0.82 kmol/m3 sulphuric acid (H2SO4) and 0.82 k m o l / m 3 hydrochloric acid (HCI) solutions at 353 K by using a constant load method. The three parameters (lss, steady state elongation rate; tss, transition time; tf, time to failure) were obtained from corrosion elongation curves and were divided into three regions of applied stress, irrespective of sensitizing temperature and anion species, which are dominated by either stress, SCC or corrosion. In SCC-dominated region, lss became a relevant parameter for prediction of tf, although the slope depended on the sensitizing temperature. The maximum applied stress, the minimum applied stress and the value of tss/tf in the SCC-dominated region depended upon sensitizing temperature in both solutions. Specifically, at a sensitizing temperature of ~950 K the maximum applied stress was smaller in 0.82 kmol/m3 H2SO4 and larger in 0.82 kmol/m3 HCI than that of the solution annealed specimens. In addition, sulphate ions were found to become more aggressive than chloride ions for the SCC susceptibility of the specimens with the most severe sensitization. On the basis of the results obtained, the effects of sensitization on SCC, the role of anion species (SO42- and CI-) and an intergranular SCC mechanism were discussed.
1. Introduction For the solution annealed type 304 and type 316 specimens, it has been already reported [1,2] that specimens of these alloys have much less stress corrosion cracking (SCC) susceptibility in 0.82 kmol/m 3 sulphuric acid (H2SO4) solution than in 0.82 kmol/m 3 hydrochloric acid (HCI) solution. It was also found that a steady state elongation rate (1~, m s-~) obtained from a corrosion elongation curve (elongation vs.
420
Iron and Nickel based Alloys
time) became a relevant parameter for the prediction of time to failure (tf). On the other hand, for the sensitized type 316 specimens, sulphate ions were found to be more aggressive for SCC than chloride ions [3,4]. However, the steady state elongation rate was the parameter for the prediction of tf irrespective of sensitizing temperature. In the present work, the SCC behaviour of type 316 specimens as functions of applied stress and sensitizing temperature in 0.82 kmol/m 3 H2SO4 and 0.82 kmol/m 3 HCI solutions was reviewed (i) to confirm whether the steady state elongation rate becomes a parameter for predicting time to failure, (ii) to clarify the effect of anion species on the SCC behaviour of the sensitized specimens, and (iii) to propose qualitatively an intergranular SCC (IGSCC) mechanism in more details. 2. Experimental The specimens used were a commercial type 316 austenitic stainless steel (yield strength: 333 MPa and ultimate tensile strength: 647 MPa). The chemical composition (mass%) was as follows: C 0.054, Si 0.67, Mn 1.38, P 0.030, S 0.005, Ni 11.16, Cr 17.21, Mo 2.21. The specimen gauge for SCC experiments was 25.6 mm long, 5 mm wide and 1 mm thick. The specimens were solution annealed at 1373 K for 3.6 ks (1 h) under an argon atmosphere and then water quenched. Further the solution-annealed specimens were sensitized at a temperature in the range of 800-1300 K for 86.4 ks (24 h) under an argon atmosphere and then water quenched. Prior to the experiments, the sensitized specimens were polished to 1000 grit emery paper, degreased with acetone in an ultrasonic cleaner and washed with distilled water. The test solutions used were 0.82 kmol/m 3 H2SO4 and 0.82 kmol/m 3 HCI solutions. The test temperature was 353 + 0.5 K. All experiments were carried out under an open circuit condition. A lever-type constant load apparatus (lever ratio I:10) to which three specimens can be separately and simultaneously attached was used with a cooling system on the top to avoid evaporation of the solution during experiments. The specimens were insulated fi'om rod and grip by surface oxidized zirconium tube. A change in elongation of the specimens under the constant load condition was measured by an inductive linear transducer with an accuracy of +0.01 mm. 3. Results 3.1. Corrosion elongation curve
Fig. 1 shows a representative example of the corrosion-elongation curves for type 316 specimens sensitized at 923 K for 86.4 ks at various constant applied stresses (ty) in the 0.82 kmol/m 3 HC1 solution. Similar curves were obtained in the 0.82 kmol/m 3 H2SO4 solution. The corrosion elongation curves consisted of three regions up to failure with an initial sudden rise of elongation; primary, secondary and tertiary regions, which correspond to crack nucleation, steady crack propagation and terminal crack propagation periods, respectively, that is, a net SCC process until failure. From these curves, the following three parameters can be obtained: (i) the steady state elongation rate in the second region, lss; (ii) the transition time between the secondary and tertiary regions, tss; and (iii) the time to failure, tf. If SCC does not occur within a laboratory time scale, only the primary and secondary regions are observed. Therefore, we can estimate only the value of ls~, but not t~s and tf.
Volume 1" Chemistry, Mechanics and Mechanisms
421
I
I
I
type 316, sensitizing temperature 923 K for 86.4 ks 0.82 kmol/m 3 HC! solution at 353 K 403 MPa 351 MPa
297
tss
0 !
0
I
I
!
I
0.5
1.0
1.5
2.0
,
t/104"S
Fig. 1. The corrosion elongation curves of the type 316 specimens sensitized at a sensitizing temperature of 923 K for 86.4 ks in 0.82 kmol/m3 HCI solution at 353 K, where tf is the time to failure, tss the transition time between the secondary and tertiary regions, and lss the steady state elongation rate.
3.2. Applied stress dependence of three parameters (lss, tss, tf) Fig. 2 shows the representative examples of applied stress (or) vs. log tf relationships for the specimens sensitized at 923 K and 1123 K for 86.4 ks in 0.82 kmol/m3 H2SO4 solution and for specimens sensitized at 923 K for 86.4 ks in 0.82 kmol/m3 HCI solution, where the dashed lines are the results for the solution annealed specimens. The relationships for the sensitized specimens were divided into three regions as well as those for the solution annealed specimens in both solutiOns, which were designated by arabic numerals 1 to 3 in Fig. 2 and correspond to the stress-dominated region (region 1), SCC-dominated region (region 2) and corrosion-dominated region (region 3), respectively. In 0.82 kmol/m3 H2SO4 solution, the maximum applied stress (O'max) in the SCC-dominated region became larger for the specimens sensitized at 1123 K and smaller for the specimens sensitized at 923 K than for the solution annealed specimens. The minimum applied stress (Crn~n)in the SCC-dominated region became smaller for the sensitized specimens than for the solution annealed specimens and for the specimens sensitized at 923 K than for the specimens sensitized at 1123 K. In addition, the specimens sensitized at 923 K showed a much higher SCC susceptibility than the solution-annealed specimens. In 0.82 kmol/m3 HCI solution the SCC susceptibility for the sensitized specimens was not so different from that for the solution annealed specimens compared to that in 0.82 kmol/m3 H2804 solution. The O'max in the SCCdominated region became larger for the specimens sensitized at 923 K than for the solution-annealed specimens, while the Crmmremained the same as that for the solution annealed specimens.
Iron and Nickel based Alloys
422
600 :"
I
5 0 0 z~O
I
I
\o "--.
,
""-----.o.. Q 400Iz ;E
t
b 300--
3
'
e~
Type 316
200 -
sensitizotion
9
86.4 ks
923K 9 1123K o
I00
-
0.821~nol/m3 H a S 0 4 , 3 5 3 K 0.. z I0
I
I
I
i
I0 ~
I0 4
I0 5
I0
tf
6
/s
(a) .
!
1
.
.
.
!
.
.
!
600
I I
500
o 400 13.. b
Ty
300
~ " ~%~~,o~\
200
I00 l
0'
IO a
I,
IO S
IO4
IO 5
iO6
If /S (b)
Fig. 2. Relations between applied stress and the logarithm of time to failure for sensitized type 316 in (a) 0.82 kmol/m3 H2SOa and (b) 0.82 kmol/m 3 HCI solutions at 353 K. The dashed lines show the result for the solution annealed type 316 specimens.
Fig. 3 shows the relationships between cr and log l,s for the specimens sensitized at 923 K and 1123 K for 86.4 ks in 0.82 kmol/m 3 H2SO4 solution and for those sensitized at 923 K for 86.4 ks in 0.82 k m o l / m 3 HCI solution, where the dashed lines are the result for the solution annealed specimens. The relationships were also divided into three regions corresponding to those in Fig. 2. Fig. 4 shows the relationships between cr and the ratio of t~ to tf for the sensitized specimens in 0.82 kmol/m 3 H2SO4 and HCI
Volume 1: Chemistry, Mechanics and Mechanisms
423
solutions, where the dashed lines are the result for the solution-annealed specimens. In 0.82 kmol/m 3 H2SO4 solution, the values of t,s/tf for the sensitized specimens in the SCC-dominated region held constant independent of or and became larger for the specimens sensitized at 923 K and 1123 K than that for the solution annealed specimens. The values of tJtf in the stress- and corrosion-dominated regions for the sensitized and solution annealed specimens became larger than that in the SCCdominated region, although they depended upon or. Similar result was also obtained for the sensitized specimens in 0.82 kmol/m 3 HCI solution as shown in Fig. 4(b). 600
'
I'
'
.-
500 -
/
e"
'
-
'
_
/
/ S'~
I
/
/t
9
2
a. 400 b 30O
_
/@
/e
3
2
-
I
I
Type 516: sensitizotion,86.4ks 0 923K e 1125K _ O . 8 2 k m o l / m 3 HzS04
200 I00
3
,.
553K .... _, ..
io-9
fi-
,.
~
to-8
t6 r
..I
,
Io-6
io -"
is= / m.ff I
(a) I
I
I
i
'1'"
I
600 Type316 0"82km~
5OO
solution annealed
___ n13
~ -~--i
4OO
sensitized
~
/
"" ""
~
i
T
Ii
//
I
-
\ :300 b 200 I00 ~
Oc)i~
3 1
10-9
I
4-.
I(Ye
i
IO"r i.
/m.s
IO-6
..,
3 1
IO-5
I
-I
Fig. 3. Relations between applied stress and the logarithm of steady state elongation rate for sensitized type 316 in (a) 0.82 k m o l / m 3 H2SO4 and (b) 0.82 k m o l / m 3 HC1 solutions at 353 K.
Iron and Nickel based Alloys
424
,
i
60C
i
i
/,
t
l..i
I
I
i
i
600 Type316
500
0.82kmol/maHCI /
500
',! .~
~uj,.) 2 o
,I
400
I
400
n
=E
b
I I I
fl_ 300
\b
Type 3 i 6
200 -
I00
-
o0
300
200 3
"
sensitized
, 0.2
H2S04
, 0.4
, 0.6
, 0.8
I.O
tss / tf
(a)
%
j
j
I I
I00
~t~
2
I C I
9 2 3 K , O6.4ks
9 112:3 K
0"82km~
2 )
- - sotufion annealed I 1373K, 3.6ks
sensitization 86.4 ks 0923K
"j
I
353K
%
0 .' 2
' 0.4
0 '.6
o'. 8
I0
tss / tf
(b)
Fig. 4. Relations between applied stress and the ratio tss/tf for sensitized type 316 in (a) 0.82 kmol/m 3 H2SO4 and (b) 0.82 kmol/m3 HCI solutions at 353 K. The dashed lines show the result
for the solution annealed type 316 specimens.
The reduction in O'ma x at a sensitizing temperature of 923 K in 0.82 kmol/m 3 H2SO4 solution was unexpected, since trmax for the sensitized specimens in 0.82 kmol/m 3 HCI solution became larger than or equal to that for the solution annealed specimens. Therefore, we tried to conduct the same experiment using the specimens sensitized at 923 K for 86.4 ks, but a test temperature of 323 K was used to lower a corrosion rate of the sensitized specimens in 0.82 kmol/m 3 H2SO4 solution. From the relationships between cr and three parameters (tf, lss and tss), O'max became larger than that for the solution annealed specimens, but trmin was the same as that at the test temperature of 353 K. In addition, the value of tss/tf in the SCC-dominated region was independent of applied stress and was the same as that at the test temperature of 353 K. Thus, it is presumed that the reduction or increase in O'ma x is associated with the magnitude of corrosion current density as described later. With regard to the other sensitized specimens, the applied stress dependences of the three parameters were similar to those for the sensitized specimens in Figs. 2-4, although O'max,O'min and the value of tJtf were dependent on the sensitizing temperature as shown in Section 3.3.
3.3. Sensitizing temperature dependence Fig. 5 shows O'max and trim as a function of sensitizing temperature in 0.82 kmol/m 3 H2SO4 solution, where the solid circles and open triangles stand for those of the solution annealed specimens. At a test temperature of 353 K, the O'max in the range of 850-1050 K became smaller than for the solution annealed specimens and showed a minimum at a sensitizing temperature o f - 9 5 0 K, while the O'max in the range of 10501250 K became larger than for the solution annealed specimens. The O'min became smaller than for the solution annealed specimens in the range of 850-1250 K and
425
Volume 1" Chemistry, Mechanics and Mechanisms
600 500 400
no ", b
300
200 Type 316/ 0.82 kmol.m"~H2S04 ---
10O
323K
u 0
I
!
80O
9OO
.
I
I(X)O
353K
.
I100
sensitizing
j
i
1200
temp.,
1300
1400
T/K
(a) I
I
I
I
!
I
600 500 maximum stress in SCC,dominated region n~
400
I
300
I
I
I
I
I
Type316/O. 8 2 k m o l . nK aHCI, 3 5 3 K
b 200
minimum stress in s e e - d o m i n a t e d region I00
-
0 700
--o--o
i 800
I 900
I I000
I I100
I 1200
I 1300
I~K)O
sensitizing temp. , T / K
(b) Fig. 5. Maximum (Crmax)and minimum (crn~n) stresses in the SCC-dominated region as a function of sensitizing temperature for the sensitized specimens in (a) 0.82 kmol/m 3 H2SO4 and (b) in 0.82 kmol/m 3 HCI solutions.
showed a minimum at a sensitizing temperature of--950 K. However, when the test temperature was reduced from 353 to 323 K, the O'max for the specimens sensitized at 923 K became larger than that for the solution-annealed specimens as shown by the dashed line in Fig. 5, while the o~i, had almost the same value independent of test temperature. On the other hand, the O ' mxa in 0.82 kmol/m3 HC1 solution became larger than for the solution annealed specimens in the range of 800-1100 K and the tr~i. had almost the same value independent of sensitizing temperature. Fig. 6 shows the value of
Iron and Nickel based Alloys
426
0.7
Type316/0.82kmol.m3HzS04,353 K
0.6 "-0-
0.5
1
800
900
A
I000
=
,
~
I100 1200 sensitizing temp., T / K
t
1300
1400
Fig. 6. Relation between the value of qs/tf in the SCC-dominated region and sensitizing temperature for the sensitized specimens in 0.82 kmol/m3 H2SO4solution.
hs/tf as a function of sensitizing temperature for the specimens in 0.82 kmol/m 3 H2SO4 solution. The value of tss/tf was found to depend on sensitizing temperature. In this case, the value of tsJtf became larger than that for the solution annealed specimens in the sensitizing temperature range of 850-1250 K and showed the maximum at a sensitizing temperature o f - 9 5 0 K. These results indicate that the degree of sensitization becomes the maximum at a sensitizing temperature of---950 K. Similar results were obtained for the specimens in 0.82 kmol/m 3 HCI solution. As for the fracture appearances of the sensitized specimens in the SCC-dominated region, the fracture appearance of the specimens sensitized at 923 K was entirely intergranular as well as that of the specimens sensitized at 1023 K. The fracture appearance of the specimens sensitized at 1123 K appeared to be a mixture of an intergranular and a transgranular mode including somewhat an intergranular mode. On the other hand, the fracture appearances of the specimens sensitized at a sensitizing temperature of above 1200 K and below 850 K were predominantly transgranular. 4. Discussion
4.1. A parameter for prediction of time to failure Fig. 7 shows a representative example of the relationships between log/ss and log tf (o: variable) for the specimens sensitized at 923 and 1123 K with a test temperature of 353 K, and at 923 K with 323 K in 0.82 kmol/m 3 H2SO4solution. The relationships in the SCC-dominated region (arabic numeral 2 in the figure) became a good straight line with a slope smaller than that (-2) for the solution annealed specimens. However, the slope was found to depend upon sensitizing temperature. Similar results were also obtained in 0.82 kmol/m 3 HC1 solution. Hence we got the following empirical equation for the sensitized specimens,
427
Volume 1" Chemistry, Mechanics and Mechanisms
log lss = -A log tf + Ca
(1)
where the slope, -,4, depends on sensitizing temperature and Ca is a constant depending on sensitizing temperature. Fig. 8 shows the values of A in Eq. (1) as a function of sensitizing temperature in both solutions. The value of A became smaller than that (=2)
Type 316 sensitization, 86.4ks o 923 K _~ 9 1123K
2
15s
O.82kmol/m 3 H2S04
-4-
%
"4
2
& 323K o 9 353K
3
_
I
.--" 16'
t 16 8 e,
iC~~
I 103
,
I
,
,,
J
I
105
10 4
_
106
tf/s
Fig. 7. Relation between the logarithms of steady state elongation rate and time to failure for the sensitized type 316 specimens in 0.82 kmol/m 3 H2SO4 solution.
i
i
i
i
Type 316 / 0 . 8 2 ~ o l . m -~, 353K
2
e/
o HCl
IocJ Iss" - l l o c j t f ( ~ 9 variable)
9 H2S04 0
7'00
1
I
I
I
800
900
I000
I100
sensitizing
I
1200
4. Ca
I
,,,
1300
1400
temp., T / K
Fig. 8. Relation between the value of A of the linear Eq. (1) and sensitizing temperature for the sensitized specimens in 0.82 kmol/m 3 H2SO4 and HC1 solutions.
428
Iron and Nickel based Alloys
for the solution annealed specimens over the whole sensitizing temperature, showing the minimum at a sensitizing temperature of around 950 K. The linear Eq. (1) implies that 1~ becomes a relevant parameter for predicting tf even for the sensitized specimens as well as the solution annealed specimens, because l~s can be obtained at a time within 10 to 20% of tf from the corrosion elongation curve. It is confirmed from Fig. 8 that the severest sensitization takes place at a sensitizing temperature of around 950 K, where their slopes become maximum.
4.2. Effect of sensitization It is well known that the formation of Cr carbides and a Cr depletion zone takes place at or along grain boundaries by sensitization. The Cr carbides act as an obstacle to dislocation movement in the grains [8], by which the sensitized specimens would become harder than the solution annealed specimens. As a result, grain boundary sliding (GBS) becomes an important factor for the sensitized specimens instead of the dislocation movement in the grains for the solution-annealed specimens [3,4]. This corresponds to an increase in the ultimate tensile strength of the sensitized specimens to contribute to the increase in Crm~x. The increase in the ultimate tensile strength is associated with the increase in a true stress at tss (cr~e = o'/As~) under a constant applied stress condition by using an identical specimen geometry, where A~ is the cross sectional area at t~. For crt~ to increase, the distance of the crack propagation up to t~ (L~s) needs to increase for the reduction in As~, which means that ts~ becomes close to tf. This would be the reason why the value of t~/tf increases with increasing the degree of sensitization. Here the reduction in Crm~xat a test temperature of 353 K will be described in the next section. On the other hand, the Cr depletion zone serves as a preferential site for corrosion attack and/or crack nucleation. Hence an intergranular corrosion could predominantly take place to lead to the intergranular failure. However, the intergranular corrosion does not always lead to the intergranular failure as described in Section 4.3. In the case of the solution annealed specimens, in order that the dislocation movement in the grains can extend over several grains, GBS has little contribution to elongation or film rupture and therefore the preferential site is a slip step emerging on the surface. The difference in the preferential site would be a primary cause for the difference in the fracture mode.
4.3. Aggressiveness of sulphate ions to the sensitized specimens The present results show that sulphate ions become more aggressive to the specimens sensitized at the severest sensitizing temperature than chloride ions as shown in Fig. 2, but not to the solution annealed specimens, where it should be noted that sulphate ions themselves become much less aggressive for corrosion than chloride ions. It is reasonable to consider that a corrosion rate of the Cr depletion zone in 0.82 kmol/m 3 H2SO4 solution becomes larger than that of the other surface on which a more protective film is formed. This should lead to the formation of a local electrochemical cell between the Cr depletion zone and the other surface. The formation of the local electrochemical cell would accelerate the corrosion rate of the Cr depletion zone even in the absence of applied stress. In contrast, chloride ions themselves are more aggressive to the materials than sulphate ions, so that the formation of the protective film on the other surface becomes more difficult than that in 0.82 kmol/m 3 H2504
Volume 1: Chemistry, Mechanics and Mechanisms
429
solution. This suggests that the magnitude of the local electrochemical cell becomes smaller in 0.82 kmol/m 3 HCI solution than in 0.82 kmol/m 3 H2SO4 solution. The difference between the magnitude of the local cell in 0.82 kmol/m 3 HCI and in 0.82 kmol/m 3 H2SO4 solutions would cause the significant difference in the SCC susceptibility between the sensitized and the solution annealed specimens. In addition, the magnitude of the local electrochemical cell would increase by the application of applied stress. On the basis of the above consideration, a schematic representation of the difference in corrosion behaviour between grain and grain boundary in 0.82 kmol/m 3 HCI and in 0.82 kmol/m 3 H2SO4 solutions is shown in Fig. 9. With regard to the solution annealed specimens, such a local electrochemical cell could not be formed in the absence of applied stress, but when a stress is applied, the local electrochemical cell can be developed between the slip steps emerging on the surface and the other surface to show the SCC susceptibility. Sensitized Specimens HCi solution Js
J$
vL o"
solution
solution
j,
Js
H~S04 solution
(_Jsolution
solution
Fig. 9. Schematic representation of the difference in corrosion behaviour between grain and grain boundary in 0.82 kmol/m3 H2SO4and HCI solutions.
On the basis of the above consideration, the decrease or increase in trm,xcompared to that of the solution annealed specimens might be explained as follows. The corrosion rate of a Cr depletion zone becomes large by the combination of a high applied stress and a high test temperature (353 K). The high corrosion rate caused by the combination would be mainly associated with a high dissolution current density (jd) with less film formation current density (/f). Hence, the reduction in cross-sectional area by jd would be superior to that by crack propagation even if a crack will be formed. This could lead
430
Iron and Nickel based Alloys
to an intergranular corrosion and finally to a mechanical fracture, but not to IGSCC and is the cause of the reduction in Crmxobserved at a test temperature of 353 K. As applied stress is lowered, the corrosion rate by the combination is reduced and as a result the crack propagation would become predominant, resulting in the SCC susceptibility at an applied stress of less than Crm~x.In the case of the decrease in test temperature, which shows the increase in Crm~x,the corrosion rate of Cr depletion zone would decrease, so that the effect of the combination is lowered even at a high applied stress. Consequently, the SCC susceptibility is observed at the higher applied stress, that is, the increase in O'max. Next, consider the reduction in O'min with increasing degree of sensitization, where its reduction was not observed in 0.82 kmol/m 3 HCI solution [3]. In the case of the solution-annealed specimens in 0.82 k m o l / m 3 H 2 S O 4 solution, the applied stress range in the SCC-dominated region becomes very narrow due to a large rate of film formation (repassivation rate) at crack tips. This is supported from the fact that the specimens did not fail within a laboratory time scale (250~ concentrated caustic solutions and liquid metal. Commercial nickel is not especially susceptible to hydrogen embrittlement since the solubility and diffusivity of hydrogen in nickel are low and this material has low mechanical strength. 3.2. Ni-Cu alloys
As in the case of Ni 200, alloy 400 is not highly susceptible to SCC probably because it has low mechanical strength (Table 1). Alloy 400 was found to be susceptible to SCC in acidic solutions containing mercury salts as well as in liquid mercury, HF and fluosilicic acid (H2F6Si) [9]. In HF the cracking is transgranular and the highest susceptibility occurs in the vapor phase, especially in the presence of air [ 10]. Reduction of aeration reduces the susceptibility to cracking in hydrofluoric acid. Using U-bend specimens, it has been reported that the crack propagation rate in alloy 400 exposed to the vapor phase of 20% HF for 240 h decreased as the temperature increased from 66 to 93~ probably because less oxygen was available in the vapor phase as the temperature increased [ 11]. In the same study, U-bends of alloy 400 were found free from cracking while immersed in the liquid portion of 20% HF [6,11 ]. It has also been reported that highly stressed alloy 400 suffers SCC in ammonia vapors at 300~ [12]. Heat treatments that eliminate residual stresses and cold worked microstructures greatly reduce the susceptibility of alloy 400 to all types of environmentally induced cracking. 3.3. Ni-Mo alloys
Ni-Mo alloys are resistant to chloride-induced cracking in boiling magnesium chloride (MgC12) solutions [ 13]. When B-2 alloy and, to a lesser extent B-3 alloy, are exposed to temperatures in the range 550 to 850~ they loose ductility due to a solid phase transformation which forms ordered intermetallic phases such as NiaMo. The precipitation of these ordered phases changes the deformation mechanisms of the alloys making them susceptible to EAC such as hydrogen embrittlement [ 14,15]. In B-2 alloy, the precipitation of intermetallic phases can occur in the heat-affected zone (HAZ) during welding. It has been reported that B-2 alloy failed by intergranular SCC of the HAZ when exposed to organic solvents containing traces of H2SO4 at 120~ [ 16]. It has also been reported that B-2 alloy was prone to transgranular SCC in the presence of hydroiodic acid (HI) above 177~ [ 17]. SCC studies of B, B-2 and B-3 alloys in acidic solutions were carried out under laboratory and plant conditions [ 18]. The effects of the electrochemical potential, cold work produced by drilling, and two different aging processes (that would simulate welding and the subsequent cooling cycle) were investigated. At anodic potentials (200 mV above the free corrosion potential), Nakahara and Shoji found transgranular fissuring in all three alloys both for mill annealed and aged materials. At cathodic potentials (100 and 400 mV below the free corrosion potential) they found intergranular cracking only for the aged (sensitized) alloys. Since the amount of intergranular brittle
Volume 1: Chemistry, Mechanics and Mechanisms
441
cracking increased at the lower applied cathodic potential, this environmentally induced cracking was attributed to hydrogen embrittlement [ 18]. U-bend specimens of mill annealed B-3 (N 10675) alloy were found to suffer SCC in the presence of vapor and liquid phase of a 20% HF solution at 66, 79 and 93~ [6]. The cracking susceptibility of N10675 increased with the temperature and the liquid portion of 20% HF solution was more aggressive than the vapor phase. 3.4. Ni-Cr-Mo alloys
One of the major limitations of stainless steels is that these alloys are susceptible to chloride-induced localized attack such as crevice corrosion, pitting corrosion and SCC. Ni-Cr-Mo alloys are the most resistant Ni-based alloys to the classic chloride-induced localized corrosion that troubles the stainless steels. In some cases, SCC was reported in ~ high-strength materials; however, cracking only occurred in very aggressive conditions, such at temperatures higher than 200~ pH lower than 4 and presence of hydrogen sulfide (H2S) [ 19]. U-bend specimens of C-2000, C-22 and C-276 alloys were not susceptible to cracking in boiling (154~ 45% MgCI2 solution after 1008 h of testing [ 11]. C-276 and C-4 alloy were free from cracking in a 25% NaCI solution at 232~ however, these alloys were susceptible to cracking in an MgCI2 solution of same chloride content at the same temperature [ 14]. C-22 alloy was immune to SCC in 20.4% MgC12 solution up to 232~ even in the 50% cold worked condition and in the 50% cold worked plus aged at 500~ for 100 h condition [ 11 ]. Laboratory testing using U-bend specimens (ASTM G 30) had shown that Ni-Cr-Mo alloys such as C-276, C-22 and C-2000 alloy were susceptible to SCC in wet HF in both the liquid and vapor phases (Fig. 1) [6,11]. The most resistant of the Ni-Cr-Mo alloy to cracking in wet HF was C-2000 (N06200) probably because the beneficial effect of 1.6% Cu content. Just in opposite behavior to Ni-Cu alloy 400, Ni-Cr-Mo
Fig. 1. SCC of alloy C-276 (N 10276) U-bend specimen immersed in a 20% HF solution at 93~ for 240 h. 100x.
442
Iron and Nickel based Alloys
alloys were less susceptible to cracking in the vapor phase than in the liquid phase, suggesting that the presence of Cr is beneficial for HF vapor phase applications [6]. Ni-based alloys are known to be susceptible to caustic cracking. Under slow strain rate conditions, C-276 alloy was susceptible to transgranular cracking in 50% NaOH at 147~ [20]. On the other hand, mill annealed and aged at 677~ for 24 h C-shape specimens (ASTM G39) of C-22 alloy did not exhibit cracking after immersion in 50% NaOH solution at 147~ for 720 h [21 ]. When Ni-Cr-Mo alloys are aged at temperatures higher than 600~ for long time (e.g. 1000 h at 650~ long range ordering reactions and precipitation of tetrahedraUy close packed (TCP) phases (~t, P, a) may take place. The presence of the TCP phases produced by thermal aging may greatly reduce the ductility of Ni-Cr-Mo alloys. For example, for annealed C-276 alloy, the yield strength (YS) at room temperature is 360 Mpa, the ultimate tensile strength (UTS) is 807 MPa, the elongation to failure (ETF) is 63%. However, for a C-276 alloy that was aged at 760~ for 16000 h, the YS increases to 476 MPa, the UTS increases to 894 MPa and the ETF decreases to 10%. It has been reported that thermally aged C-276 alloy was susceptible to hydrogen-induced cracking in environments containing H2S [22,23]. Ni-Cr-Mo alloys were also found to suffer environmentally induced cracking in conditions associated to super critical water oxidation (SCWO). It has been reported that both C-276 (N10276) and alloy 625 (N06625) suffered intergranular cracking when exposed to various aqueous solutions in the vicinity of the critical point of water (374~ [24-26]. Because of its excellent resistance to SCC and other types of localized corrosion, C-22 (N06022) was selected by the U.S. Department of Energy to fabricate the outer shell of the high-level nuclear waste containers to be disposed permanently at the Yucca Mountain site [27-29]. C-22 has been extensively tested for its susceptibility to SCC in a variety of environments, mainly at GE Global Research, Southwest Research Institute and Lawrence Livermore National Laboratory (LLNL). This alloy was found extremely resistant to EAC in many different solutions at the corrosion potential, at all the tested temperatures from ambient to 110~ [30-33]. Tests were carried out using cyclic loading, constant load, constant deformation and slow strain rate tensile (SSRT) tests in solutions from 14 molal MgCl2, to simulated concentrated ground waters from pH 3 to 13. U-bend specimens of C-22 (N06022) and other nickel alloys such as C-4 (N06455), G-3 (N06985), 825 (N08825) and 625 (N06625) were being used to characterize their SCC susceptibility in a variety of environments [34]. Gas tungsten arc welded (GTAW) and non-welded U-bend specimens were exposed for more than 5 years at the corrosion potential to the vapor and liquid phases of three different solutions (pH 2.8 to 10) simulating up to 1000 times the concentration of ground water both at 60 and 90~ None of these alloys suffered any indication of environmentally induced cracking [34]. Alloy C-22 was found susceptible to EAC when SSRT test was performed on mill annealed specimens in hot simulated concentrated water (SCW) at anodic applied potentials [33,35,36]. SCW is a multi-ionic alkaline solution approximately 1000 times more concentrated than a Yucca Mountain ground water. It is likely that the small amount of fluoride ions presem in this solution (1400 ppm) contributed to the cracking of C-22 [36]. The susceptibility to cracking of C-22 was strongly dependent on the applied potential and the temperature of the solution. The highest susceptibility to EAC
443
Volume 1" Chemistry, Mechanics and Mechanisms
was found at ~90~ at +400 mV (vs. a saturated silver chloride electrode, SSC) (Fig. 2). At the corrosion potential, C-22 was free from EAC even at 90~ Similarly, at anodic applied potentials, C-22 was free from EAC at ambient temperatures and as the temperature increased the time to failure in the tests decreased (Fig. 3).
Fig. 2. Alloy N06022 strained in SCW solution at 86~
The applied potential of +400 mV
(SSC).
130
'
I
Alloy 2 2 ( N 0 6 0 2 2 ) 1 . 6 7 x 10 -6 s -1 S?W, +400 i V SSC
~O,~.
O
120 -
110 .I--1
1000 .~
90 -
80--
70
' 20
I 40
'
I
'
60
I 80
' 100
Temperature (~ Fig. 3. Effect of temperature on the SCC susceptibility of N06022 in SCW solution at the applied potential of +400 mV (SSC).
444
Iron and Nickel based Alloys
It has also been reported that alloy C-22 (N06022) may suffer some embrittlement when it is slow strained under cathodic applied potentials (or currents) [36-38]. The maximum susceptibility to cracking under cathodic conditions seemed to occur at ambient temperatures suggesting a hydrogen-related failure mechanism. 3.5. Ni-Cr-Fe alloys
This is one of the largest groups of Ni-based alloys since it covers Inconel 600 (N06600), Incoloy 825 (N08825) and 800 (N08800) and Hastelloy G-30 (N06030) type alloys. Since alloy 600 has been used to fabricate the tubes of steam generators in nuclear power plants, it has been by far the most studied nickel alloy regarding its SCC behavior, especially in hot water and in caustic solutions. Alloy 600 has been found to suffer SCC in high temperature pure water (>300~ both in service and in the laboratory. Due to its importance for the nuclear industry, the stress cracking of alloys 600 and 690 in pure water and in caustic solutions has been extensively researched in the last three decades [39,40] and more than one thousand technical papers have been published in this subject. The susceptibility to cracking of alloys 600 and 690 depend strongly on environmental factors such as temperature, level of tensile stresses, deformation rate, presence of hydrogen gas, solution pH and electrochemical potential, and metallurgical factors, such as presence of minor alloying elements (impurities), the amount of cold work and heat treatment (intragranular or intergranular carbides). Cracking in alloy 600 is predominantly intergranular (IGSCC) in nuclear service. In same cases (e.g. lead contamination from the secondary side), it could be transgranular (TGSCC). Alloy 600, like other Ni-based alloys, also suffers SCC in hot caustic solutions (150-200~ Alloy 690, which has double the amount of chromium in alloy 600, has been found to be more resistant than alloy 600 to high temperature cracking in pure water and in caustic solutions. Due to its high content of Ni (76%), alloy 600 is resistant to SCC in chloride-containing solutions; however, alloy 600 was susceptible to localized attack in HF-containing environments [ 10]. Alloy 800 is also used in the nuclear power generation. In steam generators, alloy 800 is generally more resistant than alloy 600 to cracking. It was shown that alloy 800 (N08800) was susceptible to caustic cracking [41] and even more susceptible than alloy 690, probably because of the higher Cr content of the latter [42]. Alloy 825 is more resistant to SCC in chloride solutions than 316 stainless steels ($31600) due to the higher content of Ni in alloy 825. SSRT and U-bend tests have shown that alloy 825 was susceptible to TGSCC in 45% MgCI2 solutions at temperatures above 146~ Alloy 825 is used extensively in the oil and gas production in sour wells, performance of other nickel alloys such as C-276 and G-50 (N06950) is still superior of that of alloy 825 [43]. These nickel alloys are used mainly in the cold worked condition for increased strength. NACE International provides guidelines (MR0175) on the maximum allowable hardness. For example, the maximum hardness for cold worked alloy 825 is 35 HRC (the annealed hardness is 85 HRB, Table 1). Environmental factors that may affect the stress cracking performance of alloy 825 (and other alloys) in oil and gas wells include temperature, amount of chloride and the presence of H2S gas [43]. Data on the SCC behavior of G-30 alloy is scarce. It has been reported that G-30 components used in the industrial production of HF suffered cracking [11]. U-bend
Volume 1: Chemistry, Mechanics and Mechanisms
445
specimens of G-30 alloy did not crack atter exposure for 500 h in 45% MgCI2 solution at 154~ It has been found that G-30 as well as other nickel alloys would suffer cracking in the aggressive conditions encountered in SCWO treatments.
4. Concluding remarks Nickel alloys are more resistant than stainless steels to EAC. Nickel alloys are practically immune to EAC in hot chloride-containing solutions at temperatures below 200~ Nickel alloys are also not especially susceptible to hydrogen-assisted cracking. The environments that induce cracking of nickel alloys are highly specific and therefore they may be avoidable by design. Nickel alloys may be prone to EAC in wet HF and in hot alkalis. Ni-Cu alloys (e.g. alloy 400) are more resistant to cracking in wet HF than Ni-Cr-Mo alloys (e.g. C-276). Similarly, commercially pure Ni such as Ni-200 is more resistant to cracking in hot alkalis than Ni-Cr-Mo alloys.
Acknowledgment This work was partially performed under the auspices of the U.S. Department of Energy (DOE) by the University of California Lawrence Livermore National Laboratory under contract No. W-7405-Eng-48.
References [1] J.R. Crum, E. Hibner, N.C. Farr, D. R. Munasinghe, in: CASTI Handbook of Stainless Steels and Nickel Alloys, vol. 2, CASTI Publishing, Edmonton, 2000, p. 287. [2] G.Y. Lai, High-Temperature Corrosion of Engineering Alloys, ASM International, Materials Park, OH, 1990. [3] B. Gleeson, in: M. SchUtze (Ed.), Corrosion and Environmental Degradation: A Comprehensive Treatment, vol. 2, Wiley-VCH, Weinheim, 2000, p. 173. [4] R.B. Rebak, in: M. SchtRze (Ed.), Corrosion and Environmental Degradation, vol. 2, WileyVCH, Weinheim, 2000, p. 69. [5] C.P. Dillon, Corrosion Control in the Chemical Process Industry, NACE International, Houston, TX, 1994. [6] R.B. Rebak, J.R. Dillman, P. Crook, C.V.V. Shawber, Mater. Corros. 52 (2001) 289. [7] R.B. Rebak, P. Crook, CORROSION/2000, NACE International, Houston, TX, 2000, paper no. 00499. [8] D.C. Agarwal, J. Kloewer, CORROSION/2001, NACE International, Houston, TX, 2001, paper no. 01325. [9] Corrosion Engineering Bulletin CEB-5, The International Nickel Company, Inco, New York, NY, 1968. [10] S.J. Pawel, Corrosion 50 (1994) 963. [11] R.B. Rebak, in: R.D. Kane (Ed.), Environmentally Assisted Cracking: Predictive Methods for Risk Assessment and Evaluation of Materials, Equipment, and Structures, ASTM STP 1401, ASTM, West Conshohocken, PA, 2000, p. 289. [12] G.J. Theus, R.H. Emanuelson, J. Russell, CORROSION/82, NACE, Houston, TX, 1982, paper no. 209. [13] J. Kolts, CORROSION/82, NACE, Houston, TX, 1982, paper no. 241. [14] D.C. Agarwal, U. Heubner, M. Kohler, W. Herda, Mater. Perform. 33(10) (1994) 64. [15] M.M. James, D.L. Klarstrom, B.J. Saldanha, CORROSION/96, NACE International, Houston, TX, 1996, paper no. 432. [16] Y. Takizawa, I. Sekine, CORROSION/85, NACE, Houston, TX, 1985, paper no. 355.
446
Iron and Nickel based Alloys
[17] N. Sridhar, G.A. Cragnolino, in: R.H. Jones (Ed.), Stress-Corrosion Cracking: Materials Performance and Evaluation, ASM International, Materials Park, OH, 1992, p. 131. [ 18] M. Nakahara, T. Shoji, Corrosion 52 (1996) 634. [19] J. Kolts, in: ASM Handbook, vol. 13: Corrosion, 9th ed., ASM International, Metals Park, OH, 1987, p. 647. [20] A.I. Asphahani, in: G.M. Ugiansky, J.H. Payer (Eds.), Stress Corrosion Cracking-The Slow Strain-Rate Technique, ASTM STP 665, ASTM, Philadelphia, PA, 1979, p. 279. [21] Haynes International Database, Kokomo, IN. [22] R.D. Kane, M. Watkins, D.F. Jacobs, G.L. Hancock, Corrosion 33 (1977) 309. [23] N. Sridhar, J.A. Kargol, N.F. Fiore, Scripta Metall. 14 (1980) 1257. [24] D.B. Mitton, S.-H. Zhang, M.S. Quintana, J.A. Cline, N. Caputy, P.A. Marrone, R.M. Latanision, CORROSION/98, NACE International, Houston, TX, 1998, paper no. 414. [25] P. Kritzer, N. Boukis, E. Dinjus, CORROSION/98, NACE International, Houston, TX, 1998, paper no. 415. [26] D.W. Alley, S.A. Bradley, CORROSION/2003, NACE International, Houston, TX, 2003, paper no. 03351. [27] U.S. Department of Energy, Office of Civilian Radioactive Waste Management, Yucca Mountain Science and Engineering Report, DOE/RW-0539, Las Vegas, NV, May 2001. [28] G.M. Gordon, Corrosion 58 (2002) 811. [29] G.A. Cragnolino, D.S. Dunn, Y.-M. Pan, in: Scientific Basis for Nuclear Waste Management XXV, vol. 713, Materials Research Society, Warrendale, PA, 2002, p. 53. [30] P.L. Andresen, P.W. Emigh, L.M. Young, G.M. Gordon, CORROSION/2003, NACE International, Houston, TX, 2003, paper no. 03683. [31] D.S. Dunn, Y.-M. Pan, G.A. Cragnolino, CORROSION/2002, NACE International, Houston, TX, 2002, paper no. 02425. [32] Y.-M. Pan, D.S. Dunn, G.A. Cragnolino, in: R.D. Kane (Ed.), Environmentally Assisted Cracking: Predictive Methods for Risk Assessment and Evaluation of Materials, Equipment and Structures, ASTM STP 1401, ASTM, West Conshohocken, PA, 2000, p. 273. [33] J.C. Estill, K.J. King, D.V. Fix, D.G. Spurlock, G.A. Hust, S.R. Gordon, R.D. McCright, G.M. Gordon, R.B. Rebak, CORROSION/2002, NACE International, Houston, TX, 2002, paper no. 02535. [34] D.V. Fix, J.C. Estill, G.A. Hust, L.L. Wong, R.B. Rebak, CORROSION/2004, NACE International, Houston, TX, 2004, paper no. 04549. [35] K.J. King, J.C. Estill, R.B. Rebak, in: Proc. 2002 ASME Pressure Vessels and Piping Conference (PVP 2002), vol. 449, ASME, New York, NY, 2002, p. 103. [36] K.J. King, L.L. Wong, J.C. Estill, R.B. Rebak, CORROSION/2004, NACE International, Houston, TX, 2004, paper no. 04548. [37] S. Kesavan, Ph.D. Dissertation, The Ohio State University, Columbus, OH, 1991. [38] K.M. Scammon, M.S. Thesis, University of Central Florida, Orlando, FL, 1994. [39] Z. Szklarska-Smialowska, R.B. Rebak, in: Control of Corrosion on the Secondary Side of Steam Generators, NACE International, Houston, TX, 1996, p. 223. [40] R.W. Staehle, J.A. Gorman, Corrosion 59 (2003) 931; Corrosion 60 (2004) 115. [41] A. Mignone, M.F. Maday, A. Borello, M. Vittori, Corrosion 46 (1990) 57. [42] W. Yang, Z. Lu, D. Juang, D. Kong, G. Zhao, J. Congleton, Corros. Sci. 43 (2001) 967. [43] E.L. Hibner, C.S. Tassen, CORROSION/2000, NACE International, Houston, TX, 2000, paper no. 00149.
449
Environment induced crack growth of ceramics and glasses R.H. Jones Pacific Northwest National Laboratory, Richland, WA 99352, USA
Abstract
Understanding the effects of the environment on the fracture and crack growth behavior of ceramics and glasses is important for applications where these materials are used as structural components but also where they function as barrier coatings, reinforcement phases in composites, or as inclusions within or oxide layers on metals. Silica glasses and oxides are known to exhibit subcritical growth in water and water vapor while carbides and nitrides are general more stable in water but experience high temperature corrosion and crack growth. In some cases, with SiC as one such case, a silica glass layer forms in oxidizing environments and serves as a "passive" layer protecting the SiC from accelerated corrosion. Ceramic composites made with a ceramic reinforcement and ceramic matrix such as Nextel fibers in mullite or SiC fibers in SiC exhibit an additional high temperature corrosion issue in that they otten contain an interphase between the fiber and matrix. This interphase is used to control mechanical properties but can be a factor in the high temperature environment induced crack growth behavior. This paper will review the behavior of glasses in water and water vapor and ceramics and ceramic composites in high temperature gas environments with the idea that this will be useful for broadening our understanding of these materials for a variety of uses as described above. I. Introduction
Ceramics and ceramic matrix composites have the potential to operate at hightemperatures and are, therefore being considered for a variety of advanced energy technologies such as combustor liners in land based gas turbo/generators, heat exchangers and advanced fission and fusion reactors. Ceramic matrix composites exhibit a range of crack growth mechanisms driven by a range of environmental and nuclear conditions. The crack growth mechanisms include: (i) fiber relaxation by thermal (FR) and irradiation (FIR) processes; (ii) fiber stress-ruptta'e (SR); (iii) interface removal (IR) by oxidation; and (iv) oxidation embrittlement (OE) resulting from glass formation including effects of glass viscosity. Silicon carbide composites are being considered for gas turbine combustion liners but high temperature moisture reacts with SiC to produce a volatile product and hence very accelerated corrosion rates.
450
Ceramics and Glasses
Environmental barrier coatings are being developed to protect them. Silicon carbide composites are also being considered for nuclear applications in advanced high temperature fission and fusion reactors. While the requirements for the fission and fusion applications vary somewhat, one common feature is the use of a He gas coolant. The impurities in the He gas are the primary corrosion issue in each case. Silicon carbide composites are also being evaluated for use as intermediate heat exchangers for the high temperature hydrogen production reactor. The environments could be He on the high temperature side along with its impurities and molten salt on the low temperature side. The He environment issues will parallel those for the use of these materials in the reactor core so there should be considerable data generated on their stability in impure He. However, there is currently very little known about the stability of these materials in molten salts. Glass and glass-ceramic matrix composites were among the earliest composites developed for high-temperature applications such as turbine engine components. Glass matrix composites do not have adequate high-temperature properties above about 600~ so glass-ceramic composites have received most of the attention for these applications. The matrices that have received attention are calcium-aluminosilicate (CAS), magnesium-aluminosilicate (MAS) and barium-magnesium-aluminosilicate (BMAS). Oxide matrix reinforced with oxide fibers have received attention more recently because of the chemical instability of SiC in high temperature turbine engine environments containing water vapor. While these composites are more chemically stable than SiC/SiC they lack high-temperature strength at temperatures of 1000~ or so. Examples of materials include alumina or alumina-YAG matices reinforced with Nextel 610 or 720 fibers.
2. Aqueous environments The subcritical crack growth of glass in water and water vapor has been evaluated by Wiederhorn et al. [ 1], and Michalske and Freiman [2] with resulting K - V curves not too dissimilar to that for the stress corrosion of metallic materials. Stress-intensity crack-velocity data for soda-lime-silica glass in water and water vapor shows a Stage I and strong K dependence, a Stage II with little or no K dependence and a Stage III with strong K dependence (Fig. 1). A Stage II crack velocity of 10-1 mm/s was observed in 100% relative humidity air. This is a fast crack velocity which requires fast crack-tip reaction kinetics. The crack velocities observed in glass exposed to water and water vapor are similar to those for liquid metal embrittlement of metallic material. In this latter case, a decrease in surface energy from absorption is thought to be the only chemical reaction needed to drive crack growth. Subcritical crack growth of glass has also been demonstrated in N2, CO, acetonitrile, nitrobenzene, ammonia, hydrazine and formamide by Michalske and Freiman [2]. A key difference in these environments was the absence of a Stage II plateau for ammonia, hydrazine, formamide and water. The plateau results from the rate-limiting process of water transport to the crack tip when water is a minor constituent in the solvent gas or liquid. A molecular mechanism for the stress-corrosion cracking of vitreous silica was presented by Michalske and Freiman. This mechanism involves a crack-tip reaction of water to the Si-O bond, simultaneous proton and electron transfer from the water to the Si-O bond and formation of a surface hydroxyl group which ultimately fractures. In
Volume 1" Chemistry, Mechanics and Mechanisms
451
10 -3 Relative humidity, %
~/,
lOO
10 -4
30 10 III
f
10 -s v
1.0
=>
~e 0.2
"~
o
H2C
10 -6
'1 D
9 o|
o
0.017
10 -7
10 -8 40
I
I
!
!
50
60
70
80.
Stress intensity factor, K~ (MNm -a~ x 10 s) Roman numerals identify regions of crack propagation
Fig. 1. Stress intensity-crack velocity data for soda-lime-silica glass in water and water vapor [ 1]. soda-lime glass, there is evidence for sodium ion leaching being involved in crack growth [3]. Crack velocities in vitreous silica and soda-lime silica are comparable at 0.5 MPa/m, which suggests the rate limiting step may be similar in these two materials while the crack advance process may differ. Michalske and Bunker [4] proposed that molecule size affects crack tip reactivity such that molecules greater than 0.5 nm do not produce stress corrosion of glass. This size comes from the atomic dimensions of the silica network such that chemically active species with sizes larger than the network opening do not induce subcritical crack growth. The exposure of glass and ceramics to water can produce subcritical crack growth, which can result in the development of a critical-sized flaw and catastrophic failure. Fiber optics, bone implants, dental porcelain and glass fiber reinforced composite materials are examples where the stress corrosion of glass and ceramic materials is technologically significant. Failures of fiberglass-reinforced plastics (RMP) have been reported in applications in which the material was stressed and in contact with an aqueous environment [5]. Rodriguez [6] showed that cracks develop in E-glass fibers immersed for 336 h in a 5% H2SO4 + water solution at 23~ Rodriguez [6] concluded that these cracks could clearly result in crack growth and failure of FRP materials in contact with acidic solutions. Stress corrosion of dental porcelain was demonstrated by Jones and Suton [7]. Because it is a feldspathic-type glass, porcelain is susceptible to
452
Ceramics and Glasses
aqueous stress corrosion, as are other glasses. Jones and Sulton [7] found that the fracture strength of porcelain varied with material composition and pH of the solution and was probably related to the solubility of potassium, sodium and silicon in water. There is some interest in the use of SiC in aqueous environments such as in nuclear reactors. The environments are either those for boiling water reactors (BWRs) or pressurized water reactors (PWRs). Both operate around 300~ but PWRs operate at a higher pH adjusted by the addition of OH-. In pure water at 360~ Kim et al. [8,9] found that the corrosion rate of SiC in water at 360~ depended on the nature of the SiC material. High purity chemical vapor deposited (CVD) material had the lowest corrosion rate at 1.0 mg/cm 2 over a 10 day period while reaction bonded silicon carbide (RBSC) had the highest corrosion rate at 8.0 mg/cm 2 and sintered silicon carbide (SSC) had an intermediate corrosion rate of 3.5 mg/cm 2. The difference in their response was related to the quantity of free Si in the structure with CVD material have very little free Si. ~ To date there is no crack growth rate data on these materials in aqueous environments but clearly the corrosion rates found for the RBSC and SSC material is too high for these materials to be considered for application in water at 360~
3. High temperature gas environments Ceramic matrix composites (CMCs) are being developed to take advantage of the high-temperature properties of ceramics while overcoming the low fracture toughness of monolithic ceramics. Toughening mechanisms, such as matrix cracking, crack deflection, interface debonding, crack-wake bridging, and fiber pullout, are being incorporated in CMCs to reduce the tendency for catastrophic failure found in monolithic ceramics. Ceramics reinforced with particulate, whiskers, and continuous fibers exhibit varying aspects of these toughening mechanisms; however, reinforcement with continuous fibers offers the greatest improvements in toughness. Composites with carbide, oxide, glass, and carbon matrices are being utilized in the development of CMCs. In the case of carbide, oxide, and glass matrix CMCs, the matrix exhibits excellent high-temperature corrosion resistance so that a goal of the composite development is to not detract from this preexisting property. This is not the case for carbon matrix composites, which frequently need coatings to provide adequate corrosion protection. The purpose of the chapter is to review the database and understanding of corrosion behavior of CMCs with the intent that this information will be useful in the development of materials with improved performance and reliability. Composite materials are chemically and microstructurally heterogeneous, consisting of matrix, matrix-reinforcement interface, and reinforcement constituents. The corrosion behavior of each of these constituents will likely not be equal whether evaluated individually or within the composite. The composite corrosion resistance may be more complex than each constituent because of inter-actions between corrosion reaction products and the composite constituents. An example of this interaction would be a reaction product from the corrosion of the matrix that protects a less corrosionresistant reinforcement or reinforcement-matrix interface. However, there is far more data on the corrosion behavior of the constituents as monolithic materials than there are on the composites made from these constituents. It is important to note that the constituents of a composite may differ slightly or substantially in chemical composition, crystal structure, and microstructure from their monolithic equivalent. Examples of these differences are the SiC fibers and matrix produced from
Volume 1: Chemistry, Mechanics and Mechanisms
453
polycarbosilane or its variants that may contain considerable oxygen or may be nanocrystals embedded in an amorphous matrix. There is considerable value in evaluating the corrosion resistance of the monolithic equivalent of the composite constituents because these generally represent the baseline corrosion behavior. An effort will be made to identify where significant differences are likely to occur between the composite constituents and the monolithic equivalent. High-temperature composites are being developed to operate in a variety of environments containing alkali elements, mildly oxidizing mixed gases such as He + 02, highly oxidizing environments, and H2 + 02, highly oxidizing environments, and H2. Not all ceramic composites are being considered for each of these environments, so the corrosion data are not available for each ceramic composite in every environment. 3.1. Corrosion reactions for silicon carbide
Silicon carbide will chemically react with 02, H2, and H20 according to the following reactions: SiC(s) + O2(g) = SiO(g) + CO(g): low pO2
(la)
SiC(s) + 3/202(g) = SiO2(s) + CO(g): high pO2
(lb)
SiC(s) + 2HE(g)= Si + CH4(g)
(2)
SiC(s) + 2HEO(g) = SiO(g) + CO(g) + 2HE(g)
(3)
Silicon carbide is thermodynamically unstable in 02, H2, and H20 environments under certain conditions. However, the kinetics of these reactions can be affected by the formation of a protective layer of SiO2 at high pO2 such that SiC is stable in many corrosive environments. The stability of the passive layer then becomes critical to the stability of SiC. The 02 pressure for the SiC active passive transition was determined by Gulbransen and Jansson [10] as shown in Fig. 2. At temperatures of 800=1000~ this transition occurs at 02 pressures of 10-8 atm. Therefore, the pertinent reactions then become those between SiO2 and specific gaseous or molten-salt environments, except in molten Li with a low O2 activity, where SiO2 is unstable. Some relevant reactions with SiO2 are below: SiO2(s) + H2(g) = SiO(g) + H20(g)
(4)
xSiO2(s) + Na2SO4(P) = Na20 - x(SiO2)(P) + SO3(g)
(5)
where M is an alkali element such as Na and Li. A phase diagram for the Na20-SiO2, system is shown in Fig. 3. Alkali elements such as Na and Li cause a break-down of the passive SiO2 film by the formation of low-melting alkali silicates such as those which occur at 800~ in the Na2-SiO2 system and 1024~ in the Li20-SiO2 system. The eutectic temperature in the Li20-SiO2 system is about 250~ higher than that in the Na20-SiO2 system and, therefore, Li is expected to have less effect on the passive film on SiC at temperatures below 1000~ than Na. A summary of the behavior of SiC in gas-molten-salt environments as presented by McKee and Chatterji [ 11] is shown in Fig. 4. Passivation occurs at high pOE, and active oxidation (formation of gaseous SiO) occurs at low pOE. A basic salt or salt melt with a low pO2 at the salt-SiC interface will cause active corrosion, as depicted by reaction
454
Ceramics and Glasses
(OC)
Temperature
850
950
0
1050
1150
1250
1350
1450
1550
=
i
Pasive Oxidation
/
Po2 > Ps~ *co A
-4
~
-6
E 4~ m
/ I
o
~
_1
Active Oxidation Po2 < Psio+co
-10
,"
-12 0.9
I
I
I
I
I
I
I
0.85
0.8
0.75
0.7
0.65
0.6
0.55
lIT
0.5
x 10 3
Fig. 2. Transition pressures for SiC active-passive oxidation versus temperature, according to Gulbransen and Jansson [ 10]. 1700 Cristobafite + Liquid
1600
1500
1400
Liquid
1300
1200
/ 't
Tridymite + Liquid
x = 1.33
1100
x = 3.55 lO00 2N~O 900
800
* SiO2
N~O 2SO~+ Quartz 700
N~O 9 600 0.44
0.69
1.03
1.55
2.40
4.12
9.27
x in Na:O 9 x(SiO2)
Fig. 3. Phase diagram for the system Na20-SiO2.
455
Volume 1" Chemistry, Mechanics and Mechanisms BEHAVIOR OF SiC IN GAS'MOLTEN SALT ENVIRONMENTS
////sic
HZ
/ / // /
NO REACTION
NO MATERIAL LOSS INERT
PASSIVATION
FORMS PROTECTIVESiOz SCALE, ALMOST INERT TO SALT MELT
HIGH Poz SiOz-'~) / / / S I C " / / / 7 ACID OR NEUTRAL HIGH
SALT MELT
POt
t_..~....!..:-.-.i::i::: ::i:;~ sioz-~
9
9/ / ;'
LOW POz t i t ,Sig_..CO
/ //'/SIC" / J' / BASIC HIGH SALT Pot
ii:: ii: iis'~
'ACTIVE' OXIDATION
ACCELERATED CORROSION
EVOLUTION OF SiO(g) AND CO(g} AT SiC SURFACE FORMATION OF SOLUBLE IONIC SPECIES
SiOz~-////Si C / / /
0 z DEPLETED HIGH POe
SMELTALT _.._~. _.':~:i ~"'~iO:r'LSiO 3Z" 9 : 't:'., :21.: /z/~SiC 9/ /
GASIFICATION AND ACTIVE CORROSION
FORMATION OF SiO, SiO2 SCALE ABSENT OR RUPTURED, SILICATE FORMATION IN MELT
Fig. 4. Possible modes of behavior of SiC in gas-molten-salt environments [ 11].
schemes Nos. 4 and 5. McKee and Chatterji [ 11 ] suggest that SiC will not react with H2; however, more recent analysis by Herbell et al. [12] indicates that the equilibrium partial pressure of CH4 (Eq. (2)) is 10-4 and at temperatures of 850-1400~ for a H2 pressure of 1 atm. McKee and Chatterji [ 11 ] measured the sample weight change in the test environment as a function of time in 1 atm of H2 at 900~ and observed no reaction between SiC and H2 (scheme 1, Fig. 4). However, their gas may have had sufficient 02 or H20 to cause passivation. Herbell et al. [12] calculated the SiO(g) partial pressure for Eq. (4) with 1 atm of H2 containing 1 ppm of H20 to be 10-7 atm. Therefore, it would appear that a very small amount of H20 mixed with 02 to promote SiO2 formation would be sufficient to cause a significant reduction in the reaction rate of SiC. Jacobson [ 13] evaluated the kinetics and mechanisms of the corrosion of sintered (zSiC in molten salts at 1000~ In the reaction of Na2SO4/O2 with SiC, the reaction occurred primarily in the first few hours with the formation of a protective SiO 2 layer. This observation was demonstrated with results showing the total weight of the corrosion products/unit area reaching 6 mg/cm 2 (SiO2 + Na20-x(SiO2) after a few hours and remaining constant up to 20 h. Jacobson and Smialek [14] also noted that SiC is subject to pitting corrosion in molten salts. Pits occurred at structural discontinuities and bubbles formed during the formation of SiO2. Pitting corrosion is detrimental
456
Ceramics and Glasses
because it demonstrated that the passive SiO2 layer has been degraded and because the pits act as flaws resulting in reduced fracture strength. 3.2. Corrosion of silicon carbide matrix composites: oxygen effects
Ceramic matrix composites have the potential to operate at high-temperatures and are, therefore, being considered for a variety of advanced energy technologies such as combustor liners in land based gas turbo/generators, heat exchangers and advanced fission and fusion reactors. Ceramic matrix composites exhibit a range of crack growth mechanisms driven by a range of environmental and nuclear conditions. The crack growth mechanisms include: (i) fiber relaxation by thermal (FR) and irradiation (FIR) processes; (ii) fiber stress-rupture (SR); (iii) interface removal (IR) by oxidation; and (iv) oxidation embrittlement (OE) resulting from glass formation including effects of glass viscosity. Creep of brittle matrix composites such as SiC/SiC occurs by matrix cracking and fiber creep; therefore, understanding crack growth of an individual crack is an important aspect in understanding bulk creep of these materials. Oxidation of the interphase materials, which are typically pyrocarbon or boron nitride in SiC/SiC composites, is the primary environment induced crack growth mode for these composite materials. In this crack growth mode, the interface is oxidized and thus removed such that the FR mechanism is accelerated. At high oxygen concentrations in the environment an oxidation embrittlement mechanism can operate where sufficient SiO2 formation occurs such that the fiber/matrix interface characteristics are altered and the fiber strength is degraded. Oxidation embrittlement results in the loss of composite properties and behavior more like that of monolithic ceramics. Another high temperature corrosion mechanism is that between SiO2 and H20 to form a volatile hydroxide as follows: SiC + 2H20 = SiO2 + CH4
(6)
SiO2 + H20 = SiO(OH)3- + H +
(7)
The presence of oxygen in the system can also produce SiO2, as per equation 1b, with subsequent reaction to form the volatile hydroxide; however, the oxygen will also reduce the H2 activity. Reaction (7) is a serious issue for the use of SiC/SiC composites for combustor liners and will require the use of environmental barrier coatings (EBCs) for them to survive in these environments. The environment-induced crack growth behavior of SiC/SiC composites has been analyzed by Henager et al. [15,16] by a combined experimental and modeling approach. A dynamic crack-growth model, developed to predict crack growth in ceramic composites containing creeping fibers in an elastic matrix, has been used to predict effects of temperature, time, PO2, and irradiation on crack growth of SiC/SiC composites. Mechanics for frictional bridging and both linear and nonlinear fiber-creep equations are used to compute the dynamic crack extension. Discrete, two-dimensional fiber bridges are employed, which allows separate bridge "clocks", to compute crackgrowth rates for composites containing fibers undergoing any variety of time-dependent processes. The approach to modeling time-dependent bridging controlled by fiber creep starts by expressing the crack-opening displacement as a function of time and position along the crack face for both applied and bridging tractions. The force or stress on each bridge as a function of time and position is then solved. The bridges conform to the
Volume 1: Chemistry, Mechanics and Mechanisms
457
mechanics of frictionally bonded fibers and follow an appropriate parametric fibercreep law, which can be either linear or nonlinear in stress and time. Thermal creep of polymer derived ceramic fibers is often nonlinear, or viscous-like, in time and stress, but oxygen causes the crack growth to be linear in time. Reaction rate equations were determined from experimental measurements for the reaction of 02 with the C interphase. The interphase recession rates were modeled from these reaction rates. The reaction rates for 02 with SiC to form SiO2 were obtained from the literature. Irradiation creep of these same fibers also appears to be linear in both time (dose) and applied stress. Accordingly, a standard thermal creep equation and a simple irradiation creep equation suitable for SiC-based fibers was used to generate some model results for two test temperatures, 1273 and 1373 K (1000 and l l00~ The thermal creep equation has an activation energy of about 600 kJ/mol. The irradiation creep equation assumes a temperature independent regime below 1173 K (900~ and an activation energy of 50 kJ/mol for temperatures greater than 1173 K. The creep rate is linear in dose rate and stress. We observe that irradiation creep of the fibers dominates the fiber deformation process for temperatures below 1273 K (1000~ but thermal creep dominates at higher test temperatures. The subcritical crack growth mechanism of SiC/SiCf composites is a function of temperature, stress, environment, loading mode and time as well as other secondary variables. The crack growth mechanisms include oxidation embrittlement (OE) in high oxygen containing environments, interphase removal (IRM) in intermediate oxygen containing environments, fiber relaxation (FR) and fiber irradiation creep relaxation (FIR) in low oxygen containing environments, stress-rupture (SR) of the fiber at high temperatures and viscous sliding at high temperatures and oxygen concentrations, Fig. 5. The transition between these mechanisms is also a function of temperature, stress and time. The slowest crack growth rate and hence the longest lifetime for SiC/SiC in a He + O: environment is in the FR dominant regime. Even with some 02 in the environment it could be possible for the crack velocity in SiC/SiC to be slow enough that a component will not fail for the life of the reactor. For example, the dynamic crack growth model predicts a lifetime of about 50 years in a He coolant containing 0.5 appm 02 for cracks growing in the IR dominant regime shown in Fig. 5. The effect of irradiation creep on the FR process (designated the FIR dominant regime) is uncertain since this effect is relatively new and additional data is needed to fully evaluate its effect on crack growth; current data clearly suggests that irradiation will produce fiber creep at temperatures substantially below the thermally activated creep regime. The slowest crack growth rate and hence the longest lifetime for Sic/sic in a He + 0 2 environment is in the FR dominant regime. Even with some 02 in the environment it could be possible for the crack velocity in SiC/SiC to be slow enough that a component will not fail for the life of the reactor. For example, the dynamic crack growth model predicts a lifetime of about 50 years in a He coolant containing 0.5 appm 02 for cracks growing in the IR dominant regime shown in Fig. 5. The effect of irradiation creep on the FR process (designated the FIR dominant regime) is uncertain since this effect is relatively new and additional data is needed to fully evaluate its effect on crack growth; current data clearly suggests that irradiation will produce fiber creep at temperatures substantially below the thermally activated creep regime.
458
Ceramics and Glasses
Fig. 5. Map showing regions of dominant crack growth mechanism as a function of oxygen concentration in the environment and temperature.
3.3. Corrosion of silicon carbide composites: effects of hydrogen Herbell et al. [ 12] have evaluated the thermodynamic stability of SiC in pure H2 at 1 atm, as shown in Fig. 6. The primary gaseous reaction product is CH4 as described in Eq. (2), whereas other reactions which produce Sill4 and Sill are also possible at temperatures as low as 900~ A small amount of H20 can alter the phase stability such that at 1400~ and about 1000 ppm of H20, the dominant gaseous reaction products become SiO and CO. Results for lower temperatures were not reported, but the reaction of H20 with SiC occurs at much lower temperatures, so similar reaction products would be expected at lower temperatures. No loss in the room temperature flexural strength of sintered SiC was noted by Herbell et al. [12] for samples exposed to H2 saturated with H20 for 100 h at temperatures from 800 to 1400~ In dry H2 (25 ppm H20), Hallum and Herbell [ 17] noted a 33% decrease in the fracture strength of sintered SiC after exposures of 500 h at 1100 and 1300~ A statistically significant decrease in flexural strength was also observed after 50 h at 1000~ The stability of SiC in an Ar-H20-5% H2 mixture was calculated by Jacobson et al. [ 18], in the same manner as the H2-H20 mixtures. Except for the lower gas pressures and the shift in the relative activities of SiO and CH4 in Region III, the results are essentially identical. At 1300~ Jacobson et al. [18] measured a weight loss of around 1 mg/cm 2 atter 24 h in a region II Ar-H2-H20 mixture.
459
Volume 1: Chemistry, Mechanics and Mechanisms
CH4
|
4"
i l= tl. -7
J
SiH4
J
f
g ..I
-10 850
i
,
J
I
I
950
1050
1150
1250
1350
1450
Temperature, *C
Fig. 6. Gases for equilibrium partial pressures of reaction products for reaction of SiC with pure H2 at 1 atm [12].
Hydrogen may also react with the carbon interphase to form CH compounds. This reaction would be in addition to the direct reaction with matrix and fiber as described by Herbel et al. [12]. Springer et al. [19] evaluated the reaction ofAr + H2 environments on the weight change of SiC-SiC composites which had a carbon fiber-matrix interphase. They used a thermogravimetric analyzer to study the weight change at 1000-1200~ with Ar + 0.1% H2. They found that the reactivity of the carbon interphase of H2 was substantially less than 02. For instance, Ar + 100 ppm 02 produced a weight loss 26 times greater at 1000~ relative to Ar + 0.1% Hz. Nightingale [20] found an activation energy of 65 kcal/mol for H2 reacting with bulk graphite, whereas Springer et al. [ 19] found activation energies of 18 and 34 kcal/mol for 0.1% H2 and 1% H2, respectively. The carbon interphase material is a mixture of amorphous carbon and graphite, so that the lower activation energies observed for the carbon interphase material could be the result of the lower stability of the interphase material relative to bulk graphite. A conclusion of the study by Springer et al. [19] is that H2 is much less of a concern than O2, but that for environments with low pO2, the reaction of both SiC and C with H2 could be a significant environmental stability issue, especially at temperatures above 1200~ 3.4. Glass matrix composites
Glass and glass-ceramic matrix composites are the most developed class of ceramic matrix composites. These composites are easier to prepare than SiC-SiC or oxide matrix composites and so have received further development and evaluation than other CMCs. The glass matrices employed in these composites include calciumaluminosilicate (CAS), lithium-aluminosilicate (LAS), magnesium-aluminosilicate (MAS) and barium-magnesium-aluminosilicate (BMAS). There have been a number of microstructural, mechanical property, and environmental effects studies of materials reinforced with Nicalon-type fibers.
460
Ceramics and Glasses
3.4.1. High-temperature air environments
Alteration of the fiber-matrix interface is one of the primary effects of oxidation on glass matrix-Nicalon composites. Daniel et al. [21] evaluated the oxidation of CASNicalon composites over the temperature range of 375-600~ in air for 100 h. They evaluated the change in the fiber-matrix interfacial properties with a nanoindentation push-down test and four-point bend tests. At exposure temperatures of 450~ and above, the composites exhibited brittle failure with minimal fiber pullout. The transition from tough behavior with fiber pullout for lower-temperature exposures to brittle fracture was associated with an increase in the fiber-matrix frictional shear stress. This increase in the frictional shear stress is accompanied by the loss of the fiber-matrix interfacial carbon layer and the resulting residual stress causing the matrix to apply a compressive stress to the fiber. This clamping stress on the fiber reduces fiber pullout and causes brittle-type behavior. Microstructural evaluation of the fiber-matrix interfacial region of CAS and LASNicalon fibers exposed to air at 600 or 900~ have been reported by Cooper and Chyung [22]. The oxidized foils were very fragile, consistent with the embrittlement noted by Daniel et al. [21]. The interface was found to have a silicate composition instead of the graphite composition. This is in contrast to the conclusion reached by Daniel et al. [21 ] that the loss of the graphite layer by oxidation resulted in a clamping stress on the fiber and the resulting brittle-"type" fracture. The formation of a silicate that forms a strong bond between the fiber and matrix will accomplish the same result and would also be consistent with the increased interfacial shear stress observed by Daniel et al. [21 ]. High-temperature mechanical property tests of BMAS-Nicalon composites in air by Sun et al. [23] showed only limited oxidation of near-surface fibers in tests where the stress was below the proportional limit. However, dynamically loaded samples loaded above the proportional limit, or matrix cracking stress, exhibited limited fiber-matrix interface oxidation. Oxygen diffusion along matrix microcracks created by stresses exceeding the proportional limit was thought to be the primary cause for the fibermatrix interfacial oxidation. This effect was most pronounced under cyclic loading compared to static or quasistatic loading. Embrittlement of a MAS-Nicalon composite during fatigue loading at 500~ in air was also reported by Heredia et al. [24]. A reduction in fatigue life was noted after only 1000 cycles at 500~ relative to roomtemperature tests. Heredia et al. [24] related this loss to a "pest" process where the Nicalon is embrittled and they suggest that the compressive matrix stress in the glassceramic matrix-Nicalon composites requires a cyclic stress to reveal this process. Sorensen et al. [25] studied the effect of environment and frequency on the fatigue properties of a CAS-Nicalon composite. They concluded that fiber-interfacial wear processes play a significant role in the loss of fatigue life of these composites and that the environment enhances this wear-induced loss of strength. 3.4.2. Hot corrosion environments
High-temperature salt environments will occur in engine components such as a Navy gas turbine engine and heat exchangers in coal-fired power plants. There is a strong emphasis on increasing the trust-to-weight ratio of Navy planes, and the low density and high-temperature performance of CMCs are needed to achieve these goals. CAS- Nicalon and LAS-Nicalon composites have been evaluated for this application by Wang et al. [26,27]. They examined the reaction of sodium sulfate with these
Volume 1: Chemistry, Mechanics and Mechanisms
461
composites by coming specimens and heating them to 900~ in either air or argon atmospheres for up to 100 h. The CAS-Nicalon composites exposed in air showed surface cracking and extensive reaction between the salt and the Nicalon fibers. The surface fibers were completely attacked and were totally removed. X-ray diffraction was used to identify the presence of CaSiO3 and NaA1SiO4. The unreinforced CAS glass exposed to the same conditions reacted to form NaAISiO4 but not CaSiO3. Therefore, the SiC fibers contributed to the reaction products and altered the corrosion reaction. The tensile strength and strain to failure of the CAS-Nicalon composite exposed to sodium sulfate in air was reduced relative to the as-received properties and those for material annealed at 900~ for 100 h but without the presence of the salt. However, the properties of material exposed to salt in an argon atmosphere showed no degradation in properties. The authors concluded that oxidation is the primary reaction responsible for the strength degradation of the composite. In contrast, the LAS-Nicalon composites did not form additional phases, although there were surface cracks and interdiffiasion of Na into the composite and Mg outward diffusion. A 30% strength reduction was noted for the LAS-Nicalon composite exposed to the salt, presumably a result of the surface cracking and Na and Mg interdiffusion. A thermodynamic evaluation was conducted by Kowalik et al. [27] for the CASNicalon composite exposed to sodium sulfate at 900~ as reported by Wang et al. [26]. This study suggested the following reaction path: (i) SiC oxidizes to form SiO2; (ii) the silica reacts with the Na20 in Na2SO4 (Na20"SO3); (iii) the result of reaction two may lead to a liquid oxide (or soda slag) phase which may attack the CAS matrix; (iv) the SO3 from reaction two may combine with the CaO in the matrix to form CaSO4; and (v) this lost CaO from the matrix is replaced by Na20 to yield NaA1Si3Os. These thermodynamic predictions closely match the experimemal results reported by Wang et al. [26]. Step one shows the significance of oxidation in the high-temperature corrosion of these materials. High-temperature corrosion studies of CAS-Nicalon and BMAS-Tyranno exposed to sodium and magnesium salts have also been conducted by Scott et al. [28]. The environments were 3.5% NaC1, 3.5% magnesium salts, and a mix of 3.5% of both sodium and magnesium salts. The samples were coated with these solutions by immersion and then heated to 600, 800 or 1100~ for up to 60 h. Reaction occurred primarily between the Ca and Mg ions, and the Nicalon in the CAS-Nicalon composite, but the Na ions penetrated the glassy phase and lowered its viscosity in the BMASTyranno composites. The authors concluded that both reactions were of concern for the stability of these composites in high-temperature salt environments. 3.5. Oxide matrix composites
The chemical instability of the SiC in the presence of alkali elements and the fibermatrix interphase in SiC-SiC composites in oxidizing environments are factors that encourage the development of oxide matrix composites. Much of the high-temperature corrosion data for oxide matrix composites exists for particulate-, whisker-, or plateletreinforced material, which have not been optimized for strength and toughness. Oxide fiber development has progressed to the state where continuous-fiber composites are being produced: however, there is no high-temperature oxidation data for these materials. Examples are alumina and alumina-YAG matrix composites reinforced with Nextel 610 and 720, as reported by Goettler [29]. Interphase layers of ErTaO4 or
462
Ceramics and Glasses
CaWO4 are being evaluated for producing fiber pullout and fracture resistance. However, the stability of these interphase materials in oxidizing, reducing, or salt environments has not been evaluated. Even though the matrix and fiber may exhibit excellent behavior in oxidizing environments, uncertainty about the composite chemical stability remains. Also, the high-temperature strength of oxide fibers is less than SiC fibers, such that further improvements in strength must occur before continuous-fiber oxide-oxide composites are attractive for high-temperature applications. Borom et al. [30] have examined the oxidation behavior of A1203 reinforced with SiC and MoSi2 particles and SiC whiskers. The particulate volume fractions ranged from 10 to 30% and tests were conducted at 1200-1500~ in air; the oxidation rate was determined by weight change and reaction layer thickness. Both SiC and MoSi2 form protective SiO2 layers when oxidized as bulk materials. Borom et al. [30] reported a 15fold increase in the oxidation rate of these phases when incorporated into an A1203 matrix. This increase was postulated as resulting from the volume change of the reaction product that forms on the composite and the thermal expansion mismatch of the reaction product with the composite. Both of these factors were less favorable for the composite as compared to bulk SiC and MoSi2. Larger volume fractions of these phases produced a large volume fraction of mullite in the reaction scale and this was favorable because the silica in the mullite will produce a more viscous scale that will allow more stress relaxation and accommodation for mismatch stresses. These authors suggested that a mullite matrix is preferred because the reaction product will contain aluminosilicate plus mullite, which will flow and relax thermal mismatch stresses. The bend strength of SiC-whisker-reinforced (28 vol.%) A1203 was found by Leaskey et al. [31] to increase by 33% when oxidized in air at 1600~ for 15 min. Composites with SiC particle reinforcement showed a 66% improvement in the bend strength following an oxidizing treatment of 2 h at 1600~ The authors suggested that the improved properties are the result of the oxidation of the SiC reinforcement to produce a compressive surface layer. The following conditions were necessary for this improvement: (a) a sufficiently large SiC content to produce a continuous oxide surface layer; (b) oxidation conditions that produce a low porosity layer with a critical thickness; and (c) elimination of large flaws in the bulk of the material. The tensile strength of Nicalon-fiber-reinforced A1203 following heat treatment in air at 750~ has been reported by Heredia et al. [24]. This material contained 00/90~ fiber orientation with a BN-SiC interphase. The room-temperature tensile strength was found to decrease from about 250 MPa to about 120 MPa following exposure to air at 750~ for 24 h. The formation of a glass phase on the Nicalon fiber was suggested as the cause of the observed oxidation embrittlement. Corrosion studies of SiC-reinforced A1203 have been conducted in coal combustion environments by Watne et al. [32] and Breder et al. [33]. In the study reported by Watne et al. [32], the materials produced by the Lanxide Corp. contained 50% SiC with 10% residual Si. Following a 100-h pilot-scale combustion test at about 1350~ in a radiant zone of the furnace, the composite, in the form of a tube, was intact but had a 0.85-mm reduction in the wall thickness. This loss was suggested as resulting from erosion from slag flow on the tube. The original wall thickness was 5.25 mm. A smaller amount of loss was found for a tube placed in the convective pass region of the combustor where the temperature was about 1200~ Breder et al. [33] exposed a similar tube made by Lanxide Corp. to coal slag obtained from two coal combustion
Volume 1: Chemistry, Mechanics and Mechanisms
463
plants. Exposures were conducted in a box furnace with the tube and coal slag at temperatures of 1090, 1260 and 1430~ Fracture tests were conducted on samples removed from the tubes following a 500-h exposure. The tube strengths were reduced by 20-45% at 1260~ depending on the type of slag. Although A1203 is the most commonly used matrix for oxide matrix composites, composites with other oxides such as MgO, ZrO2, and mullite have also been evaluated. The oxidation kinetics of SiC particulate-reinformed MgO has been examined by Hallum [34], and Camey and Readey [35]. Hallum [34] studied MgO reinforced with 5, 10, and 15 vol.% SiC particles or whiskers over the temperature range of 1100-1500~ The reaction-product thickness increased with the square root of time and was a function of the volume fraction of SiC in the composite. Mg cation diffusion was proposed as controlling the growth rate with a reaction layer formed by Mg cation diffusion through the reaction layer to the atmosphere where oxidation produced a columnar growth region. Camey and Readey [35] identified three oxidation-product layers unlike the single layer observed by Hallum [34]; however, they agreed with Hallum [34] regarding the growth rate being controlled by Mg cation diffusion through the product layer. Luthra and Park [36] evaluated the oxidation of SiC in mullite and alumina matrices and found parabolic rate constants that were three orders of magnitude larger than SiC. Xu et al. [37] measured the effects of adding ZrO2 to mullite on the oxidation of mullite-zirconia-SiC composites. They found that the addition of ZrO2 to mullite-SiC composites increased the reaction rate with oxygen. They rationalized this as being due to the increased diffusion rate of oxygen in the zirconia phase. A rapid "Mode II" type of oxidation, where oxygen can penetrate deep into the sample before the outer region is completely oxidized, occurred at 1200-1400~ and with the volume percent of ZrO2 greater than 20%. 4. Liquid metal and molten salt environments
There are select applications where ceramics and ceramic composites are considered for use in liquid metal and molten salt environments. One application is the use of SiC/SiC in contact with Pb-Li in a magnetic fusion blanket as coolant and tritium breeder. Liquid Pb-Li is considered as the coolant and breeding material for the TAURO and ARIES-AT reactor designs. Fenici and Scholz [38] reported that CVI SiC/SiC composite samples were stable in a static solution of Pb-17Li at 800~ for up to 1500 h. They concluded that SiC should be very stable in this environment because the free energy change for the following reaction is ~99 kJ/mol over the temperature range of 700 to 900~ 2 SiC + 2 Li = Li2 C2 + 2 Si
(8)
Terai et al. [39] also reported that SiC/SiC composite and monolithic SiC exhibited excellent stability in Pb-16Li at 300 and 500~ for 666 h exposure. The largest weight change reported was 1.5% for a SiC/SiC composite with Hi-Nicalon fibers, pyrolytic carbon fiber/matrix interface and matrix produced by the PIP process. Two other composite materials showed a factor of 10 less change in weight. In contrast, samples immersed in Li at 427~ for the same period of time were totally dissolved with the exception of a high-purity, monolithic CVD SiC. These results are consistent with the
464
Ceramics and Glasses
conclusion of Fenici and Scholz [38] regarding the small free energy change for the reaction given by Eq. (8). Reactions with other phases such as residual Si and C are a likely reason for the high reactivity of the other samples tested. In a review of temperature limits for fusion reactors, Zinkle and Ghoniem [40] concluded that the limits for SiC in Li, Pb-17Li and Sn-Pb-Bi are 800~ and >760~ respectively. The current corrosion results in Pb-Li are limited to static immersion tests but the blanket material will be in contact with the flowing coolant where combined erosion-corrosion effects could occur. Therefore, the effect of coolant flow rate must be evaluated. Silicon carbide composites are also being considered for intermediate heat exchangers for the high temperature hydrogen production reactor. These heat exchangers will be exposed to impure He from the reactor on the high temperature side and may be exposed to a molten salt on the low temperature side. Examples of molten salts being considered are Flibe and NaF-ZrF4 at temperatures up to 1000~ There is very little data on the stability of monolithic SiC let alone SiC/SiC composites in these molten salt environments. Petersen [41] calculated the solubility of SiC in Flibe and NaF-ZrF4 equilibrated with Be for Flibe and Zr for NaF-ZrF4. The results shown in Fig. 7 suggest that SiC is very stable in both salts but it is much more stable in Flibe than NaF-ZrF4 where the solubility for SiF4 is 10 -13 mole fraction in NaF-ZrF4 and about 10-17 mole fraction at 1000~ Static solubilities are only the first level of evaluation since, the kinetics of dissolution, transport between hot and cold legs and the effects of the molten metal or salt on the performance of the material must also be evaluated.
Fig. 7. Solubility of Si and SiC in Flibe and NaF-ZrF4 as a function of temperature. The salts were in equilibirum with their metal, Be for Flibe and Zr for NaF-ZrF4 to control the redox potential [41 ].
Volume 1: Chemistry, Mechanics and Mechanisms
465
Ceramic composites are being considered for a variety of high-temperature applications in which their corrosion properties will be important for their performance. Examples include combustor liners and blade shrouds for gas turbines, heat exchangers in coal-fired power plants, burner nozzles, gas injection lances, sensor shields, tundish nozzles for molten A1 and steel plants, and furnace/reformer tubes. Each of these applications involves some form of corrosion. The corrosion of ceramic composites is more complicated than that of a monolithic ceramic because composites are chemically and microstructurally heterogeneous. The high-temperature corrosion of CMCs is often affected by the fiber, fiber-matrix interphase, or the method used to produce the matrix of the composite. Hightemperature oxidation of the C or BN interphase in SiC-SiC composites is a clear example where the interphase causes the corrosion performance of the composite to be less than that of monolithic SiC. The presence of the SiC in mullite or alumina matrix composites were also found to increase the parabolic rate constants for oxidation by several orders of magnitude, whereas the presence of the SiC fiber resulted in a different reaction product in a CAS-Nicalon composite than in the unreinforced matrix when reacted with a high-temperature salt environment. Therefore, the corrosion performance of CMCs differs from the unreinforced monolithic ceramic and must, therefore, be carefully evaluated for each application.
Acknowledgements This research was supported by the United States Department of Energy, under Contract No. DE-AC06-76RLO 180 with Pacific Northwest National Laboratory. Pacific Northwest National Laboratory is operated by Battelle Memorial Institute. The assistance of Bev Wardlow with manuscript preparation is greatly appreciated.
References [ 1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [ 11] [12]
S.M. Wiederhom, E.R. Fuller, Jr., R. Thomson, Metal Sci. 14 (1980) 450. T.A. Michalske, S.W. Freiman, J. Amer. Cer. Soc. 66 (1983) 284. S.M. Wiederhom, Com. Amer. Cer. Soc. 65 (1982) C-202. T.A. Michalske, B.C. Bunker, J. Amer. Cer. Soc. 70 (1987) 780. F.R. Jones, J.W. Rock, J.E. Bailey, J. Mater. Sci. 18 (1983) 1059. E.L. Rodriguez, J. Mater. Sci. Lett. 6 (1987) 718. D.W. Jones, E.J. Suton, British Cer. Trans. J. 86 (1987) 40. W.-J. Kim, H.S. Hwang, J.Y. Park, J. Mater. Sci. Lett. 21 (2002) 733. W.-J. Kim, H.S. Hwang, J.Y. Park, W.-S. Ryu, J. Mater. Sci. Lett. 22 (2003) 581. E.A. Gulbransen, S.A. Jansson, Oxid. Met. 4 (1972) 181. D.W. McKee, D. Chatterji, J. Amer. Cer. Soc. 59 (1976) 441. T.P. Herbell, A.J. Eckel, D.R. Hull, A.K. Misra, in: R.H. Jones, R.E. Ricker (Eds.),
[13] [14] [ 15] [ 16] [17] [18]
Environmental Effects on Advanced Materials, TMS, Warrendale, PA, 1991, p. 159. N.S. Jacobson, J. Amer. Cer. Soc. 69 (1986) 74. N.S. Jacobson, J.L. Smialek, J. Electrochem Soc. 133 (1986) 2615. C.H. Henager, Jr., C.A. Lewinsohn, R.H. Jones, Acta Mater. 49 (2001) 3727. C.H. Henager, Jr., R.G. Hoagland, Acta Mater. 49 (2001) 3739. G.W. Hallum, T.P. Herbell, Adv. Cer. Mater. 3 (1988) 171. N.S. Jacobson, A.J. Eckel, A.K. Misra, D.L. Humphrey, J. Amer. Cer. Soc. 73 (1990) 2330.
466
Ceramics and Glasses
[19] G.D. Springer, C.F. Windisch, Jr., R.H. Jones, J. Nucl. Mater. 233-237 (1996) 1271. [20] R.E. Nightingale, Nuclear Graphite, Academic Press, New York, NY, 1962, p. 423. [21] A.M. Daniel, A. Martin-Meizoso, K.P. Plucknett, D.N. Braski, Cer. Eng. Sci. Proc. 17 (1996) 280. [22] R.F. Cooper, K. Chyung, J. Mater. Sci. 22 (1987) 3148. [23] E.Y. Sun, S.R. Nutt, J.J. Brennan, J. Amer. Cer. Soc. 79 (1996) 1521. [24] F.E. Heredia, J.C. McNulty, F.W. Zok, A.G. Evans, J. Amer. Cer. Soc. 78 (1995) 2097. [25] B.F. Sorensen, J.W. Holmes, P. Bmdsted, in: R. Naslain, J. Lamon, D. Dougmeingts (Eds.), Proc. Conference on High-Temperature Ceramic-Matrix Composites I, Woodhead Publishing Ltd., Cambridge, 1993, p. 343. [26] S.-W. Wang, R.W. Kowalik, R. Sands, Cer. Eng. Sci. Proc. 14 (1993) 385. [27] R.W. Kowalik, S.-W. Wang, P.D. Ownby, D.M. Thompson, W.T. Thompson, Cer. Eng. Sci. Proc. 16 (1995) 893. [28] V. Scott, S Bleay, R. Cooke, in: R. Naslain, J. Lamon, D. Dougmeingts (Eds.), Proc. Conference on High-Temperature Ceramic-Matrix Composites I, Woodhead Publishing Ltd., Cambridge, 1993, p. 691. [29] R. Goettler, Reported at the Continuous Fiber Ceramic Composites Working Group Meeting, Lake Tahoe, NV, 1997. [30] M.P. Borom, M.K. Brun, L.E. Szala, Cer. Eng. Sci. Proc. 8 (1987) 654. [31] L.A. Leaskey, R.O. Loutfy, J.C. Withers, in: N.P. Bansal, J.P. Singh (Eds.), Advances in Ceramic-Matrix Composites III, American Ceramic Society, Westerville, OH, 1994, p. 991. [32] T.M. Watne, J.P. Hurley, J.R. Gunderson, in: Proc. 20th Annual Conference on Composites, Advanced Ceramics, Materials, and Structures-A, Cer. Eng. Sci. Proc. 17(3) (1996) 462. [33] K. Breder, J.M. Canon, R.J. Patten, in: Proc. 20th Annual Conference on Composites, Advanced Ceramics, Materials, and Structures-A, Cer. Eng. Sci. Proc. 17(3) (1996) 479. [34] G.W. Hallum, High Temperature Effects of Oxidation of MgO-SiC Composite, Ph.D. Dissertation, Ohio State University, Columbus, OH, 1990. [35] M.E.F. Camey, D.W. Readey, Cer. Eng. Sci. Proc. 16 (1995) 863. [36] K.L. Luthra, H.D. Park, J. Amer. Cer. Soc. 73 (1990) 1014. [37] Y. Xu, G. Fu, A. Zangvil, Proc. 20th Annual Conference on Composites, Advanced Ceramics, Materials, and Structures-A, Cer. Eng. Sci. Proc. 17(3) (1996) 433. [38] P. Fenici, H.W. Scholz, J. Nucl. Mater. 212-215 (1994) 60. [39] T. Terai, T. Yoneoka, S. Tanak, Compatibility test of SiC with liquid metal breeders, International Town Meeting on SiC/SiC Design and Material Issues for Fusion Systems, Oak Ridge National Laboratory, Oak Ridge, TN, January 18-19, 2000. [40] S.J. Zinkle, N.M. Ghoniem, Fusion Eng. Design 51-52 (2000) 55. [41 ] P. Peterson, private communication.
467
Study of delayed fracture of PZT-5 ferroelectric ceramics K.W. Gao, Y. Wang, L.J. Qiao, W.Y. Chu Department of Materials Physics and Chemistry, University of Science and Technology Beo'ing, Beijing, 100083, P.R. China Abstract
The delayed fracture of lead-zirconate-titanate (PZT-5) ferroelectric ceramics under an applied constant load in a solution or during dynamic hydrogen charging was studied. Additionally, the effect of an applied sustained electric field on stress corrosion cracking (SCC) of PZT-5 ferroelectric ceramics was investigated. Experimental results indicated that, during dynamic hydrogen charging under an applied constant load, hydrogen can delay fracture of PZT5 ferroelectric ceramics. The threshold stress intensity factor decreased linearly with the logarithm of hydrogen concentration Co. The threshold stress intensity factor of hydrogeninduced cracking (HIC) revealed anisotropy for the specimens with different poling directions. However, the normalized threshold stress intensity factor of HIC did not reveal anisotropy. SCC of PZT-5 ferroelectric ceramics occurred in water, silicone oil, formamide and humid air. It also exhibited anisotropy in the threshold stress intensity factor of SCC. Positive and negative electric fields had identical effects on apparent fracture toughness, and both decreased the apparent threshold stress intensity factor of SCC in silicone oil. I. Introduction
Lead-zirconate-titanate ceramics (PZT), representative of ferroelectric ceramics, have been used extensively in sensors, actuators and transducers. PZT ceramics can become semiconductors atter reduction in a hydrogen atmosphere [1 ]. However, the electric properties of PZT capacitors deteriorated and leakage currents increased after the capacitor was exposed to hydrogen [2]. The physical properties of ferroelectric multilayer ceramic capacitors [3] and Ba(Pb)TiO3-based semiconducting ceramics [4] underwent degradation after cathodic charging, but could be restored after annealing in air at 650~ This indicates that atomic hydrogen can diffuse into ceramics during the charging process. Under an applied load, hydrogen can transport through stress-induced diffusion. For an alumina ceramic, hydrogen-induced delayed failure occurred during dynamic hydrogen charging under constant loading [5]. One objective of this project is
468
Ceramics and Glasses
to study the possibility of hydrogen-induced delayed fracture of the PZT-5 ferroelectric ceramics during dynamic hydrogen charging under an applied constant load. Stress corrosion cracking (SCC) of ceramics such as A1203, ZrO2, and YBa2Cu307.x has been investigated extensively. Results indicate that moist atmospheres and water can induce SCC of ceramics even if only residual stress existed [6-9]. A1203 with sintering additive of 0.4 wt.% MgO displayed corrosion fatigue in humid air or in water, and an acid environment can facilitate the corrosion fatigue [ 10]. Generally, it has been found that the stress generated by the difference in thermal expansion coefficient for different phases in ceramics and the internal stress generated during cooling will induce microcrack initiation [ 11], and the applied stress will concentrate at the microcrack tip. In solutions, the microcrack tip can absorb aqueous molecules, thereby causing a decrease in surface energy. At the same time, the stress concentration will facilitate hydrolyzation and induce Si-O bond transfer to Si-OH bond in SiO2. As a result, delayed fracture occurs under the combined effect of applied stress and environment. Do ferroelectric ceramics, such as PZT ceramics, undergo SCC? Poled ferroelectric ceramics exhibit anisotropy in fracture toughness [12-15]. Our experimental results showed that the threshold stress intensity factor Kic for the poling direction parallel to the crack plane and crack line was larger than that perpendicular to the crack plane. This effect was attributed to the stress-induced 90 ~ domain switching process. The stress-induced 90 ~ domain switching process also exists during SCC, and may cause anisotropy in the threshold stress intensity factor of SCC for poled ferroelectric ceramics. Therefore, the second objective of this project is to investigate SCC properties of PZT-5 ferroelectric ceramics. Experimental results indicate that an electric field may cause failure of ferroelectric ceramics [16]. An applied electric field along the poling direction decreases fracture toughness, resulting in crack propagation [13,17,18]. Fu's study [19] indicates that an electric field can reduce the bend strength of PZT-841. How does an applied electric field affect the delayed fracture of ferroelectric ceramics under an applied constant mechanical load? The third objective of this project is to investigate the effect of a sustained electric field on SCC behaviour of PZT-5 ferroelectric ceramics.
2. Experimental Sott lead-zirconate-titanate ceramics Pb (Zrl.xTix) 03 with a Zr/Ti ratio of 52:48 (PZT-5) was pressured into a bulk and sintered in the atmosphere at a temperature of 1200~ for 20 h. The average grain diameter was 2.5 lam, while the coercive field was 11 kV/cm. The specimens with a dimension of 0.9 x 8 x 40 mm cut from the bulk were Ni-electroplated for poling or charging with hydrogen. An edge notch of 4 mm depth with a root radius of 0.3 mm was created by diamond cutting at the centre of each specimen. Several specimens were poled along the longitudinal direction at temperatures higher than the Curie point, i.e., 350~ or along the thickness direction at room temperature in an electric field of 30 kV/cm. The tensile specimens were loaded in a constant load machine, and the fracture toughness measured in air. SCC tests were carried out in water, silicone oil, formamide and humid air. Hydrogen charging was performed in a 0.2 mol/l NaOH + 0.25 g/l As203 solution with various current densities. The concentration of diffusible hydrogen in the specimens after charging was evaluated using a glass tube filled with silicone oil at room temperature. In order to determine the effect of an electric field on delayed fracture, the specimens with sustained electric
Volume 1: Chemistry,Mechanicsand Mechanisms
469
fields first were loaded to failure in air to obtain the apparent fracture toughness under various sustained electric fields. Next the delayed fracture tests were carried out in silicone oil under a constant mechanical load and a sustained electric field. Each tensile test result represented an average value of six specimens. The fi'acture surfaces of specimens were examined with scanning electron microscopy. 3. R e s u l t a n d d i s c u s s i o n
3.1. Thefracture toughness of PZT-5 ceramics G
The average fracture toughness K~c, measured without an applied field for the specimen poled along the thickness direction (i.e., the poling direction parallel to the crack plane and crack line) was 1.34 + 0.25 MPa~/m. The average Kbc for the specimen poled along the longitudinal direction (i.e., poling direction perpendicular to the crack plane) was 0.92 + 0.10 MPa~/m. The fracture toughness of the poled PZT-5 ceramics a
revealed an anisotropy factor of Kzc / Kbc = 1.4.
3.2. Hydrogen-induced delayedfracture of PZT-5 ceramics Hydrogen charging was performed with various current densities. The diffusible hydrogen concentration increased with increasing charging current (Fig. 1). The threshold stress intensity factors of HIC Kin decreased with the logarithm of the hydrogen concentration, and revealed anisotropy (Fig. 2(a)). The relationship between the threshold stress intensity factor and the hydrogen concentration can be given as: Q
KbH = 0.366 -- 0.133 In Co
K1/4 = 0.53 8 - 0.208 In Co
lO 8
do4 2
0
0
50
100
150
200
250
300
350
i, m A / c m 2
Fig. 1. Hydrogen concentration vs. charging current density.
470
Ceramics and Glasses
0.6
(a)
,e-
E m
0.5
~
o.4 3:
~=-
0.3
,0
0.2
m
3:
0.1 0.0 -0.5
0.0
0.5
1.0
1.5
2.0
2.5
In C O
9 p~,k~
0.4 ,O
9
o_
perpendicular
(b)
o.3
"~ .O
,--
0
0.2
o_ "
0.1
I
o.o -0.5
0.0
0.5
1.0
1.5
2.0
2.5
InC o Fig. 2. (a) Threshold stress intensity factor of HIC and (b) normalized threshold stress intensity factor of HIC vs. logarithm of hydrogen concentration. The applied field was E = 0.
In contrast, the normalized threshold stress intensity factor did not reveal anisotropy,
and K~n/K~c - K~H/K~c (Fig. 2(b)). This indicates that the anisotropy factor of the threshold stress intensity factor of HIC is the same as that of K~c, and is independent of hydrogen concentration.
3.3. SCC of PZT-5 ceramics The stress intensity factor (K~) vs. time to failure (tF) during SCC in water, silicone oil, formamide and humid air (40% RH) are shown in Fig. 3. Because the specimens underwent delayed fracture in silicone oil and formamide that contained no water, it appears that SCC in PZT-5 ceramics was not related to hydrogen. This can be inferred
Volume 1" Chemistry, Mechanics and Mechanisms
471
because there was no hydrogen ion in the solution and no hydrogen atom diffused into the ceramics during SCC. 1.0 9
0.9
9 A 9
0,8
fom'mmide silicon oil air
0.7 0.6 0.5
0.4 0
20
40
60,
80
100
120
t ,h Fig. 3. Stress intensity factors (/(i) vs. the time to failure (tF) during SCC of PZT-5 in water, formamide, silicone oil and moist air. The applied field E = 0.
The value of stress intensity factor (Kt) for specimens with poling direction parallel or perpendicular to the crack plane vs. the time to failure during SCC in water and formamide are shown in Fig. 4. According to these results, the susceptibility of poled PZT-5 ceramics to SCC revealed anisotropy. The susceptibility to SCC for poling direction perpendicular to the crack plane was larger than that for poling direction parallel to the crack plan e . Threshold stress intensity factors for SCC (Kiscc) were obtained on the basis of Fig.
4. K~scc values for specimens with poling direction parallel to the crack plane were 0.88 and 1.0 MPa~/m in water and formamide respectively. These values were higher b than K~scc for specimens with poling direction perpendicular to the crack plane: b K~scc = 0.42 and 0.54 MPa~/m in water and formamide respectively. Anisotropy a
b
factors Kiscc/K~scc were 2.1 in water and 1.8 in formamide, indicating there was a (/
difference with a factor Kzc/Kbic that was equal to 1.4 (see Section 3.1). It was concluded that the anisotropy of threshold stress intensity factors for SCC was related not only to the 90 ~ domain switching process but also to the susceptibility of SCC to anisotropy.
3.4. The effect of sustained electric field on SCC of PZT-5 ceramics Specimens were loaded to failure in air under sustained electric fields of + 12.8, +6.4 and +3.2 kV/cm, and the apparent fracture toughness KIc(E) values obtained. Fig. 5
472
Ceramics and Glasses
shows that the apparent fracture toughness decreased as the applied electric field was increased. The positive electric field along the poling direction and the negative electric field had the same effect on the apparent fracture toughness.
1.5 II
(a) water
o
1.2 Parallel 0.9
~ " 0.6 0.3
0.01
0.1
h
tv,
1.5 ~|.
(b) formamide
--
.
~
Parallel
o
o
0.9 , , ~"
0.6 ,
~--'-~--..
0.3
0
. 0.01
0
~ 0.1
1
D 40
60
80 100120
t ,h Fig. 4. Stress intensity factors (KI) vs. time to failure (tr) during SCC of PZT-5 with the poling direction parallel or perpendicular to crack plane in (a) water and (b) formamide. The applied field E = 0.
Delayed fracture tests under various stress intensity factors KI were performed in silicone oil under sustained electric fields of +2.0 and +4.0 kV/cm, and the time to failure recorded. Apparent threshold stress intensity factors for SCC in silicone oil with the applied sustained electric field, K1scc(E), were also obtained. The variation of time to failure (tr) with Kt under various electric fields is shown in Fig. 6, while the relation between Klscc(E) and the applied electric field is shown in Fig. 7. The apparent
473
Volume 1: Chemistry, Mechanics and Mechanisms
1.0 0.8
S ~ tll
0.6 0.4 0.2 0.0 -20
" -15
' -10
' -5
0
E,
5
10
15
20
kV/cm
Fig. 5. The apparent fracture toughness in air as a function of the applied electric field. ,2
ii
i
i
(a) positive 9 E=O 9 E=2kVlcm 9 E=4kVlcm
I'
I' 1 0 I" 9 k
.E ~"
0.41-
0.0
0.1
~
~'-*-------~---
....................... 1
t~,h
10
100
1.2 (b) negative
,
1.0 E
"7
9 9 9 o
0.8
E:O E=-2kVIcm
E=-4kV/cm E=-6kVlcm
m 5
0.6
0.4
~
_
I~-..
0.2
~ ~
0.1
.
.
.
.
. . . J
.
1
.
.
.
.
.
.
.i
10
t~,h
.
.
.
.
.
.
.
J
.
100
Fig. 6. KI vs. tF during SCC in silicone oil with various (a) positive and (b) negative electric fields.
Ceramics and Glasses
474 0.6 silicon oil ..~
0.5 l
E
m o,. o.4.
- 0.3~
0 0
0.2-
~ - 0.1 0
-16
. -12
-8
0
~ -4 0 4 E, kVIcm
8
12
16
Fig. 7. The apparent threshold stress intensity factor Kxscc(E) as a function of the applied electric field E in silicone oil.
threshold stress intensity factor KIScc(E) decreased as the applied electric field was increased. The positive and negative electric fields had identical effects on Ktscc(E).
3.5. Fractography Examination by SEM revealed that the fracture surfaces of specimens tested under various conditions were intergranular (Fig. 8). Poling direction, hydrogen contents (HIC tests), environment (SCC tests) and electric fields on the fracture surface of PZT5 ceramics produced no obvious effects. 4. D i s c u s s i o n
In ferroelectric material applications, (i) an electric field, (ii) a stress (including the internal stress caused by the electric field), and (iii) an environment (such as air, oil, hydrogen, etc.) may exist simultaneously. Stress can cause mechanical fracture or fatigue of ceramic materials. The present study indicated that for ferroelectric ceramics, SCC or hydrogen-induced embrittlement occurs when the stress and the environment act simultaneously (region 1 in Fig. 9). The alternating electric field can induce the degradation of electric and mechanical properties as well as the initiation and propagation of microcrack(s) in ceramics [16]. Our recent study [20] showed that the combined effect of electric field and environment produced the delayed fracture of ferroelectric ceramics (region 2 in Fig. 9). In this study, the electric field acting together with stress (Fig. 5) resulted in the change of the fracture toughness of ferroelectric ceramics (region 3 in Fig. 9). Two types of interaction exist when an electric field, a stress and an environment act concurrently (region 4 in Fig. 9). These types are (i) the influence of electric field on SCC or hydrogen-induced embrittlement (Fig. 7), and (ii) the effect of stress on the delayed fracture of ferroelectric ceramics induced by the sustained electric field [21 ].
475
Volume 1: Chemistry, Mechanics and Mechanisms
(a)
(b)
(c)
(d)
(e)
(f)
Fig. 8. The fracture surfaces of PZT-5 ceramics: (a) specimen failure in air with poling direction perpendicular to crack plane; (b) specimen failure in air with poling direction parallel to crack plane; (c) specimen failure in formamide with poling direction perpendicular to crack plane; (d) HIC of specimen with poling direction perpendicular to crack plane; (e) specimen failure in air under an applied load and sustained electric field; and (f) specimen failure in silicone oil under a constant load and sustained electric field.
476
Ceramics and Glasses
Fig. 9. A schematic diagram of the combined effect of stress, electric field and environment.
5. Summary Our study of (i) SCC in water, silicone oil, formamide and moist air, (ii) HIC with various charging current densities, and (iii) the effect of a sustained electric field on SCC in PZT-5 ceramics led to the following conclusions: 1. SCC of PZT-5 ferroelectric ceramics occurred in water, silicone oil, formamide and moist air, and the threshold stress intensity factor Kzscc revealed anisotropy for poled specimens. 2. Hydrogen charging can result in delayed fracture of PZT-5 ceramics, and anisotropy of threshold stress intensity factors (Kin) for HIC exists. 3. Apparent fracture toughness K/c(E) decreased with increased applied electric field, and the positive and negative fields had identical effects on KIc(E). 4. The apparent threshold stress intensity factor for SCC Klscc(E) decreased as the applied electric field was increased and the positive and negative electric fields had identical effects on K~scc(E).
References [ 1] N.R. Rajopadhye, S.S.V. Bhoraskar, A.P.B. Sinha, J. Mater. Sci. 23 (1988) 2631. [2] J.T. Evans, L.L. Boyer, G. Velasquez, R. Ramesh, S. Aggarwal, V. Keramidas, Jap. J. Appl. Phys. 9B (1999) 5361. [3] W. Chen, L. Li, J. Qi, Z. Gui, J. Mater. Sci. Lett. 17 (1998) 899. [4] W. Chen, L. Li, Y. Wang, Z. Gui, J. Mater. Res. 13 (1998) 1110. [5] B. Shi, Y.J. Su, L.J. Qiao, W.Y. Chu, J. Mater. Sci. Lett. 20 (2001) 2047. [6] S.M. Spearing, F.W. Zok, A.G. Evans, J. Amer. Cer. Soc. 77 (1994) 562. [7] J.E. Ritter, J.N. Humenik, J. Mater. Sci. 14 (1979) 626. [8] C.P. Chen, W.J. Knapp, J. Amer. Cer. Soc. 60 (1977) 87. [9] T. Okabe, M. Kido, T. Miyahara., Eng. Fract. Mech. 48 (I 994) 1373. [10] S.M. Barinov, L.V. Fateeva, N.V. Ivanov, S.V. Orlov, V.J. Shevchenko, Scripta Mater. 38 (1998) 975. [ 11] J.H. Gong, Fracture Mechanics of Ceramics, Beijing, Tsinghua University Press, 2001. [12] G.G. Pisarenko, V.M. Chushko, S.P. Kovalev, J. Amer. Cer. Soc. 68 (1985) 259.
Volume 1: Chemistry, Mechanics and Mechanisms
477
[13] A.G. Tobin, E. Park, Proc. SPIE, Smart Structures and Materials 1993: Smart Mater. 1916 (1993) 78. [14] C.S. Lynch, Acta Mater. 46 (1998) 599. [ 15] F. Fang, W. Yang, Mater. Lett. 46 (2000) 131. [ 16] S.J. Kim, Q. Jiang, Smart Mater. Struct. 5 (1996) 321. [17] C.T. Sun, S.B. Park, Proc. SPIE, Smart Structures and Materials 1995: Smart Mater 2441 (1995)213. [18] S.B. Park, C.T. Sun, J. Amer. Cer. Soc. 78 (1995) 1475. [ 19] R. Fu, T.Y. Zhang, Acta Mater. 48 (2000) 1729. [20] Y. Wang, W.Y. Chu, K.W. Gao, Y.J. Su, L.J. Qiao, Appl. Phys. Lett. 82 (2003) 1583. [21 ] Y.J. Su, Y. Wang, W.Y. Chu, K.W. Gao, L.J. Qiao, Acta Mater. 52 (2004) 3752.
481
Liquid metal-induced embrittlement of a Fe9CrlMo martensitic steel J.-B. Vogt, I. Serre, A. Verleene, A. Legris Laboratoire de Mdtallurgie Physique et Gdnie des Matdriaux, Universitd de Lille, UMR CNRS 8517, 59655 Villeneuve d'Ascq, France
Abstract
This paper analyzes the influence of liquid metal on the mechanical behaviour of a Fe9Crl Mo martensitic steel. First, the occurrence of a transition from ductile to brittle behaviour induced by the adsorption of liquid metal is addressed with monotonic tensile tests. The results showed that a specific combination of several parameters (heat treatment of the steel, test temperature, notch effect) is required to observe cleavage fracture as opposed to ductile fracture. Atomic-scale numerical simulations indicate that the reduction of the surface energy induced by the adsorption of liquid metal atoms can be significant, which supports an adsorption-induced mechanism of liquid metal embrittlement. The fatigue behaviour of a Fe9CrlMo alloy at 300~ in air and in liquid lead bismuth eutectic was studied using low cycle fatigue tests performed under total strain control (0.4% < Ae~ < 2.4%). It is demonstrated that the stress amplitude and the cyclic accommodation depend only on strain level. Fatigue life is reduced by increasing the strain level and by a liquid lead bismuth environment instead of air. The reduction in fatigue resistance is explained by metallographic observations that suggest the liquid metal promotes the nucleation and/or growth of short cracks, skipping the microcracks coalescence step altogether. I. Introduction
Increasing attention is currently being directed world-wide at Accelerating Driving Systems (ADS), which involve a spallation reaction induced by energetic protons in high-Z targets. For Spallation Neutron Sources (SNS) in general, liquid metals are of interest because they do not experience radiation damage. Typical targets are mercury (Hg) for SNS [1] and lead-bismuth (Pb-Bi) for ADS [2,3]. In the content of European research programs, the Megapie-test R&D project employing partitioning and transmutation techniques focuses on ADS for typical applications such as nuclear waste transmutation or subcritical reactors [4]. Though a subcritical core is of great significance to safety, the reliability of some components must be considered to optimize the operation. One of the most critical aspects is the beam w i n d o w - the area where the proton beam enters the liquid metal target. The window is subject to cyclic
482
Liquid Metal Embrittlement
solicitations, irradiation and corrosion damage. It is widely expected that synergistic effects of stress and environment modify or accelerate damage mechanisms. In the present study, we investigate the risk of liquid metal embrittlement (LME) and the influence of LME on the fatigue resistance of the structural alloy candidate for the window. LME is difficult to predict because the concept of specificity based on the apparent immunity of certain liquid metal-solid metal pairs is questionable [5]. This paper presents the results we obtained for a Fe9CrlMo (also referred to T91) alloy. This alloy was selected for its high resistance to creep and swelling under irradiation. Particular attention is paid on the role and properties of the microstructure.
2. Experimental procedure 2.1. Materials Two heats of Fe9CrlMo steel supplied by Creusot Loire Industries (CLI) and Ascom6tal were investigated with the monotonic or cyclic tests. The chemical composition of the steel is given in Table 1. The recommended heat treatment for the T91 steel consists of an austenisation at 1050~ followed by air quenching and then tempering at 750~ for 1 h. The tempering temperature has a strong effect on hardness. While tempering at 750~ gives the softest microstructure (HV = 220), the hardest is obtained when the steel is tempered at 500~ (HV = 400). This point is pertinent because it is known that high-strength steels are more sensitive to LME than are lowstrength steels. The effect of tempering temperature (500 and 750~ was studied for the T91 from CLI, in relation to the transition from ductile to brittle behaviour induced by liquid metal. The steel from Ascom6tal was studied only with a tempering at 750~ for fatigue experiments. For both steels, the microstructure was fully martensitic, with an average grain size of 20 pm. Table 1 Chemical composition (wt.%) of the T91 steels Steel CLI Ascom6tal
Cr 8.80 8.50
Mo 1.00 0.95
V 0.25 0.21
Mn 0.38 0.47
Si 0.41 0.22
Ni 0.17 0.12
C 0.11 0.10
Nb 0.07 0.06
Fe Bal. Bal.
2.2. Specimen and mechanical tests Cylindrical tensile specimens with a diameter of 4 mm and a gauge length of 20 mm were machined from as-received ingots. The heat treatment was then applied (tempering at 500 and 750~ In some specimens, a notch (depth of 0.5 mm and curvature radius of 0.2 mm) was mechanically machined. The tensile tests were performed at a constant stroke speed corresponding to a strain rate of 10-4 s-~ using a Schenck RMC 100 servo-electric test machine. Tests were conducted in air, in liquid lead at 350, 375, 400 and 425~ and in liquid lead-bismuth eutectic (LBE, 56 at.% Pb-44 at.% Bi) at 260~ To study the effect of liquid metal on fatigue properties, low-cycle fatigue experiments were conducted using T91 steel submitted to standard heat treatment (i.e.,
Volume 1: Chemistry, Mechanics and Mechanisms
483
tempering at 750~ The specimens were smooth and cylindrical, with a gauge length of 13 mm and a gauge diameter of 10 mm. Their surfaces were carefully electropolished in order to avoid effects of surface roughness on fatigue life. Tests at 300~ were performed in air and in the liquid LBE alloy using a servo-hydraulic MTS machine with a load capacity of 100 kN. The fatigue tests were conducted in a fully push-pull mode (R~= -1) at imposed total strain variations At~ ranging from 0.4 to 2.4%. An extensometer for the strain control, a triangular waveform and a constant strain rate of 4 • 10-3 s-! (corresponding to loading frequencies ranging from 0.08 to 0.5 Hz) were used. During cyclic loading, hysteresis loops were recorded periodically, providing measurements of stress variation A~r for each cycle. A FEI Quanta 400 scanning electron microscope in the secondary electron mode was used to perform fractographic analysis. 3. Results
3.1. Occurrence o f a liquid metal embrittlement in the T91 steel
it was first confirmed that the T91 steel in its standard condition (tempering at 750~ is ductile with a yield stress o f - 2 6 0 MPa and an elongation to fracture of 22% in air and liquid Pb at 350~ Tempering at 500~ increased the yield stress up to 900 MPa, but the plasticity domain remained large up to 15% at fracture atter a test at 350~ in air and liquid Pb. Major differences occurred in the load-elongation curves obtained at 350~ in liquid Pb and air with notched samples tempered at 500~ (Fig. 1). A strong reduction in the fracture elongation for the test in liquid lead compared to that in air suggests a possible embrittlement of the steel by the liquid metal. Before material failure, the curves obtained in air and liquid Pb coincide, which means that the presence of the liquid Pb does not modify the bulk plastic properties.
Fig. 1. Tensile curves obtained at 350~ in air and liquid Pb for T91 notched specimens tempered at 500~
484
Liquid Metal Embrittlement
(a)
(b)
Fig. 2. Typical features of the fracture surface during the occurrence of ductile (a) and brittle (b) behaviour.
(a)
(b)
Fig. 3. The hard cold worked "coat" viewed from the fracture surface (a) and microcracked after loading (b).
The ~actographic analysis reveals ductile fracture with characteristic dimples for specimens tested in air (Fig. 2(a)), and cleavage-like transgranular brittle fracture for specimens tested in the liquid metals (Fig. 2(b)). Close inspection of a specimen's surface revealed that the machining of the notch causes the formation of a hard coldworked coat-10 tam thick (Fig. 3(a)) that breaks during tensile tests (Fig. 3(b)). In a liquid metal environment the microcracks generated in this manner have a brittle propagation, while this does not appear to be the case in air. A similar effect (i.e., a transition l~om ductile to brittle) has been observed in other liquid metal environments (LBE and Sn) at temperatures ranging from 250 to 400~ [6], but requires both notched and 500~ tempered specimens. In the present study, 750~ tempered specimens never exhibited brittle fracture.
Volume 1: Chemistry, Mechanics and Mechanisms
485
Fig. 4. Evolution of fracture energy with temperature atter fracture in air and in liquid Pb of notched specimens of T91 tempered at 500~
A ductility indicator is obtained by integrating the load-displacement curves. That indicator represents the mechanical energy required to break the samples. As the temperature is raised, the embrittlement effect for the T9 l-liquid Pb case progressively disappears, with the steel showing a fully ductile behaviour above 450~ (Fig. 4). This recovery in ductility is attributed to an enhanced plasticity and thus to a lower yield stress, while the surface energy remains unchanged. This transition is similar to that observed in standard brittle-to-ductile transitions.
3.2. Effect of liquid metal on low cycle fatigue behaviour The evolution of the stress amplitude Ao/2 vs. the number of cycles N for the tests conducted at 300~ in air and in the LBE is shown in Fig. 5. The T91 exhibits a cyclic sottening early in the fatigue life. A marked decrease of the stress amplitude that follows is related with the propagation of a macroscopic crack into the bulk before final failure. This is typical for martensitic or ferritobainitic steels even if tempering heat treatment has been adequate [7]. The same pattern was observed when cycling in the LBE environment. That is, the cyclic accommodation and the stress values do not differ in air and the LBE environment. Again, this represents a surface effect that occurs in corrosion fatigue [8] and thus does not affect the stress level that would be observed if a bulk effect existed. However, for a given strain amplitude, the fatigue life is much shorter in the LBE than in air. The fatigue resistance is decreased by at least a factor of 2 for tests involving large plastic deformation (Fig. 6).
486
Liquid Metal Embrittlement
600
LBE 9 A~=2.4%
Air
O []
500
,O
9
At;=1.6% t
9
Aa=0.6% t
4,..a om,
623 K. Following that, fractographic studies were performed to determine the fracture mode. Pb304-based lubricant was used to check the formation of metal Pb using a differential scanning calorimeter analysis at a temperature >613 K. It was found that 10.7CrMoVNbN and 12CrMoWV steels could be sensitive to liquid-metal embrittlement (LME) by Pb. In addition, for steels containing 10.7-12% Cr, it is thought that hardness may affect the sensitivity of LME more so than the chemical composition. At 613 K or above, the Pb3Oa-based lubricant covering the bolt thread can serve as a source of liquid Pb, which can give rise to a LME of the bolts. I. Introduction 12Cr steel bolts, currently employed in steam turbines, occasionally fractured in a brittle manner on their threaded portion. Elemental analysis maps revealed that Pb had adhered to cracks inside fractured bolts (Fig. 1). Normally the threaded portion of bolts was coated with solid lubricant, because these bolts performed at 623 K or higher. Fractured bolts had been coated with lead oxide (Pb304)-based lubricant. Fracture surface observation and chemical analysis suggest that the fracture may be caused by liquid Pb. The liquid-metal embrittlement (LME) phenomenon was first reported by Huntington in 1914 [1]. A report by Rostoker et al. [2] indicated that only certain combinations of liquid metal and solid metal were sensitive to LME, but a combination
492
Liquid Metal Embrittlement
of Pb and steel was not. However, according to recent reports, 91 steel (9Cr-lMo) is sensitive to LME [3,4].
Fig. 1. SEM image and map for Pb inside the fractured bolt.
In this study, to determine the reason for fracture of 12Cr steel bolts, tensile tests were performed in a liquid Pb environment at 623 K and higher temperatures. In addition, fractographic studies of the fractured surface of specimens were conducted. Formation behavior of metal Pb from Pb304-based lubricant were examined by differential scanning calorimeter (DSC) analysis at a temperature above 613 K.
2. Experimental procedures 2.1. Materials
Chemical compositions of 10.7Cr-Mo-Nb-N and 12Cr-Mo-W-V steels used in this study are shown in Table 1. These steels were supplied in the form of hot-rolled bars. Two heat treatments were performed on the 10.7Cr-Mo-Nb-N steel in order to get two levels of hardness. The steel was heated to 1363 K and oil quenched, then tempered at 903 K and air cooled or tempered at 803 K and air cooled, resulting in hardnesses of Hv400 and Hv320 respectively. In order to get a hardness of Hv320, 12Cr-Mo-W-V steel was heated at 1313 K and oil quenched, then tempered at 903 K and air cooled. It was confirmed that the microstructure of these steels was uniform tempered martensite.
Table 1 Chemical compositions of the steels tested Steel 10.7Cr-Mo-Nb-N 12Cr-Mo-W-V
C 0.18 0.22
Mn Si Ni Cr Mo Nb 0.66 0.49 0.59 10.93 0 . 9 2 0.39 0 . 5 1 0 . 3 2 0 . 6 8 11.43 0.98 -
V 0.22 0.28
N 0.05 -
W 0.96
Volume 1: Chemistry, Mechanics and Mechanisms
493
2.2. Observation of separation behavior of Pb from Pb304 Unused Pb304-based lubricant was heated at 613-763 K for 30 min in air and reducing atmosphere (N2). Analysis using DSC revealed that metal Pb had separated from Pb304.
2.3. Tensile tests Cylindrical tensile specimens with a 6 mm diameter and a gauge length of 38 mm were fabricated for tensile tests. To prevent the formation of a thin oxide layer on Pb and the specimen in a high-temperature environment, the tests were performed in a vacuum of 133 Pa (1 torr). Test temperatures ranged from 623 to 823 K, which is higher than the melting point of Pb (600 K). Because the heating system of the furnace was eleCtromagneticaUy controlled, the specimen's temperature reached 623 K in just a few minutes. A schematic view of the experimental set-up is shown in Fig. 2. Tests were performed in two environments: vacuum and liquid Pb. After reaching test temperature, the specimen was maintained at that temperature for about 10 min. The tensile tests were started when the test temperature was reached, and the pull rate was 0.01 mm/s. The Pb used for tensile tests was of 99.99% in pure. The oxide layers that formed on the surface of Pb were removed with a solution of composition HzO2 : CH3CH2OH : CHaCOOH = 1 : 1 : 1 immediately prior to the tensile tests. Liquid Pb filled the test holder where the specimen was installed (Fig. 2).
2.4. Fractography Pb was removed from fracture surface of specimens by dissolution in an appropriate solution of composition H202 " CH3CHzOH 9 CH3COOH = 1 91 91 before observation. Optical microscopy and scanning electron microscopy (SEM) were used for fractographic analysis.
Fig. 2. Schematic view of the experimental set-up.
Liquid Metal Embrittlement
494 3. Results and discussion
3.1. Pb formation One constituent of solid lubricant is oil. Because the lubricant was coated on the threaded portion of bolts, it can be assumed that the oil will evaporate gradually in a high-temperature environment. However, it may still remain on the threaded portion of bolts. To investigate the effect of oil and reducing atmosphere on Pb formation, heat treatment tests were performed on solid lubricant in two cases: as-received and with oil removed. DSC was used to determine the melting point of lead. The results of DSC analysis in Table 2 show that Pb formed at 613 K for 30 min when the solid lubricant was heated in a reducing atmosphere. In tests where the oil had been removed from solid lubricant, Pb formation was not observed when the solid lubricant was heated in air. For the actual fractured bolts, it is thought that the threaded portion was not exposed to air because the Pb304-based lubricant had coated the entire threaded portion.
Table 2 Results of high temperature tests of Pb304-based lubricant Pb304-based lubricant
Atmosphere
As-received Oil removed
N2 N2
Air
Temperature (K) 613 653 Pb Pb No No
673 Pb No -
693 Pb Pb -
723 Pb Pb No
763 Pb Pb No
P b - metal Pb was formed N o - metal Pb was not formed
3.2. Fractography Tensile tests of 10.7CrMoVNbN steel with hardnesses of Hv320 and Hv400, and 12CrMoWV steel with a hardness of Hv320 in a vacuum environment without liquid Pb showed that the specimens fractured smoothly (ductile failure), with cup and cone shapes. SEM observation revealed that ductility dimples existed in all areas of the fracture surface where tests were performed without liquid Pb (Fig. 3). For tests performed at 623 K with liquid Pb, the fractured specimens did not show cup and cone shapes, and optical micrographs revealed brittleness failure surfaces at the edge of the fracture surface (Fig. 4). SEM analysis revealed quasi-cleavage fractures on one part of fracture surface and dimples on the other part. Based on the above observations, it is possible that LME can occur on both 10.7CrMoVNbN steel and 12CrMoWV steel. For a hardened steel like 10.7CrMoVNbN steel with hardness of Hv400, the fracture surfaces showed cleavage fracture (Fig. 4), and the area of a brittle fracture part (Hv400) was larger than the material with hardness of Hv320. The above observation is concerned with the results obtained by Nieaise et al. [4] that LME occurred only on hardened material with hardness of Hv400, but occurred neither in base material with hardness of Hv250 nor in softened material with hardness of Hvl50. We believe that occurrence of LME by Pb of high Cr martensitic steel is related to the hardness more so than to the chemical composition. It was observed that
Volume 1: Chemistry, Mechanicsand Mechanisms
495
LME only occurred on specimens tested at 623 K. For specimens tested at 673 K and higher, the fracture surface showed full ductile mode. This is the case because with an increase of temperature, ductility of specimens gradually recovered. However, when the pull rate of tensile tests was
Anodic
6
Cathodic
Anodic
20
Anodic
Current density, mA/sq in
Fig. 18. The effect of applied current on time-to-failure for a 12% Cr martensitic stainless steel in oxygenated 3% NaCI solution at different pH values (atter Bhatt and Phelps [ 181]).
International Congress on Metallic Corrosion in Moscow, Bhatt and Phelps [235] reviewed the early applications of the method for studying the SCC behaviour of highstrength steels. It was pointed out that in polarization experiments the potential is a more fundamental parameter than current density and that for best results, polarization measurements should be made potentiostatically. In experiments in which the current density is controlled, it is possible that the potential might vary significantly to cause erratic results. It was also mentioned that the experiments could be conducted with practically any type of specimen that is suitable for SCC tests. In 1967, Leckie [231] adopted this approach to studying SCC in fracture mechanics precracked specimens loaded statically. In 1971, Barsom [219] and Gallagher [220] pioneered the application of the controlled-potential technique to distinguish between hydrogen embrittlement and active path dissolution during corrosion FCG. In addition, Barsom [219] and Meyn [221 ] first examined in greater detail the features of the fracture surfaces produced by corrosion FCG in high-strength steels and titanium alloys under applied potentials by TEM. These examinations were conducted to provide further information concerning the corrosion FCG mechanisms. In particular, the examinations showed that changes in cyclic frequency, electrode potential, pH, solution composition, or temperature produced clearly marked transitions on fracture surfaces as soon as the changes are made. Despite the advances in understanding some features of crack propagation, none of the above mentioned attempts, nor numerous attempts of other authors who employed the controlled-potential technique and fractographic evidence, allowed for the possibility of quantifying the role of any one mechanism in crack propagation.
Volume 1: Chemistry, Mechanics and Mechanisms
547
3. 7.2. Solution chemistry within crack
' One of the major obstacles to a better understanding of the mechanism of stresscorrosion and corrosion-fatigue cracking has been the almost total lack of information about the nature of the electrolyte within a growing crock, particularly near the advancing crack front. While it was generally conceded that the local chemistry within the corrosion crack differs from that of the bulk solution, especially in relation to pH, the demonstration of this difference was deterred by inaccessibility and the problems of obtaining representative samples of solution. It was only in 1969 that Brown and his coinvestigators [236] were able to measure the pH in the crack-tip region during SCC of a 7075 Al alloy, Ti-8AI-IMo-IV alloy, and 0.45% C steel tested in 3.5% NaCI solution with pH 6.5 by using an original method. The method required removing the specimen from the solution atter the stress-corrosion crack had propagated along a suitable distance (~10 mm) and immersing it in liquid nitrogen to freeze the corrodent in the crack in place. The specimen was subsequently removed from the liquid nitrogen and broken apart to expose the frozen solution on the SCC surface for analysis immediately on thawing. The 7075 AI alloy and 0.45% C steel specimens showed pH values of 3.5 and 3.8, respectively, near the advancing crack tip. The Ti-8AI-1Mo-1V alloy specimen indicated a pH as low as 1.7 at the crack tip. This low pH and its apparent constancy suggested not only that an anodic reaction was occurring at the crack tip region, but also that it must be buffered to some extent. In 1970, Brown and his co-investigators [237] fwst performed simultaneous measurements of the crack tip potential and pH values during SCC of AISI 4340 high-strength steel in 3.5% NaCI (pH 5.7). Wedgeopening-loading (WOL) specimens were used in conjunction with a small, rugged probe consisting of a pH dependent 1.6-mm solid cast Sb/Sb203 electrode together with a solid 1.6-mm Ag/AgCI reference electrode. During the experiment, the probe was positioned against the WOL specimen just ahead of an advancing crack front. The tests were conducted under freely corroding conditions and under conditions where the bulk pH was regulated between 2 and 10 by adding either HCI or NaOH. In all cases, the pH at the crack tip was found to be in the range of 3.5 to 3.9. It was immediately evident that the pH measured at the crack tip was virtually independent of the bulk pH of the corrodent, and was determined solely by the electrochemical reactions taking place at the advancing crack front. When plotting the combination of pH and potential observed during cracking on a Pourbaix diagram for iron [238] (Fig. 19), the points always fell in that region where hydrogen evolution is possible. This meant that in all cases (freely corroding, anodically polarized, and cathodically polarized), the electrochemical conditions at the crack front were such that hydrogen would be liberated. Because of this, it was concluded that "there does not appear to be any valid reason to require a mechanism other than hydrogen embrittlement to account for the cracking of highstrength steels in salt water" regardless of external potential or pH of the bulk solution. The conclusion contradicted that of Bhatt and Phelps [181] who assumed that the mechanism of any cracking which occurred under anodic polarization was active path dissolution. As an additional argument, Brown et al. [237] appealed to the recent work of Troiano et al. [239] on hydrogen permeability through thin (760 ~tm) foils made of martensitic HP 9-4-45 steel which unequivocally demonstrated that hydrogen was absorbed by steel under anodic polarization and this hydrogen pick-up was directly related to the steel delayed failure behaviour. Troiano et al. [239] concluded, "Thus, the argument that the decrease in time-to-failure under anodic polarization rules out a generalized hydrogen embrittlement concept of SCC is quite invalid." On the other
548
History of SCC Research S CE
Hydrogen
1500
STRESS CORRoSIoNOF 43'40 STEEL IN 0.6M NoCl SOLUTION S.PECIM.EN CONDITION
I000
_ _ ~
BULK pH-
A. Sb/Sb203 ELECTRODE DATA I FREELY CORRODING t. A 2. A t II ANODIC CONTROL
5 0 0 --
15OO
(INITIAL)
~
k
5.7 9.0
1. +
o
9
._>
4.
._
E
-- I 0 0 0
57 2
I0.0
IO.5
+3
--
500
1II CATHODIC CONTROL O --
...I
~
1. O
,,:I
~B.
\
I'Z I.l.I I--
~_
+ ~
o a. - 5 0 0
-
5.7
GLASS ELECTRODE DATA I FREELY CORRODING, ANODIC ~ 8 CATHODIC CONTROL
\
]
5.7
t.u
_-72"---
2.0 - 500
-lO00
~ -
OaD -15OO
- 2
~ 0
I 2
i .... 4
L 6
1 8
,
1 I0
1 12
l 14
-I000
16
pH
Fig. 19. Crack tip potential and solution pH values in the stress-corrosion crack of AISI 4340 steel in 3.5% NaCI for various bulk pH values and conditions of external electrochemical control (al~er Smith et al. [237]).
hand, when analyzing Brown's data (Fig. 19), Parkins [196] emphasized that "electrochemical conditions for hydrogen discharge are not exclusive of other processes, such as dissolution. Thus, the demonstration of acidification within the confines of a stress corrosion crack is not, per se, a demonstration of a hydrogen embrittlement mechanism, since acidification may also result in enhanced dissolution." 4. Recapitulation
9 In the last third of the 19th century, it was recognized that hydrogen can diffuse into and be absorbed by iron and steel at room temperature, and thus it was understood that such hydrogen absorption was in some way connected with a degree of embrittlement of the steel. The two sources of hydrogen which have usually been associated with this embrittlement are environments with which the metal naturally comes in contact during the course of its manufacture, and nascent hydrogen liberated at the surface of the steel either by electrolysis or by chemical action. 9 Among the first commercial alloys recognized as being susceptible to cracking under a combination of stress and environment were the common copper alloys. The considerable commercial importance of these alloys, and the fact that the elimination of the tendency to SCC was a matter of real concern to industries,
Volume 1: Chemistry, Mechanics and Mechanisms
549
including the rapidly growing electric power industry, led to an intensive investigation of the problem, beginning in 1906. In 1918, the first ever SCC symposium- a Topical Discussion on Season and Corrosion Cracking of B r a s s was arranged in Atlantic City, USA. The symposium was confined entirely to the season cracking of brass cartridge cases, a problem in war time. 9 The damaging effect to the strength of metals caused by corrosion and cycle stress was the subject of much investigation, and in the earlier years, particular attention was paid to the greatly accelerated damage which could take place when these influences acted simultaneously. In 1917, it was understood that this effect was not due to the reduction of cross-section caused merely by general corrosion, nor to the presence of deep holes formed by local attack, but was primarily due to sharp, deep cracks produced under the joint action of what is now termed corrosion fatigue. 9 In the mid-1930s, the first structural failures due to SCC of austenitic stainless steels Were observed. In the early 1940s, it was reported that the resistance of Fe-Cr-Ni alloys to SCC increased with Ni content. Also, in the early 1940s, the electrochemical nature of SCC was first recognized. The first major symposium on SCC was held in 1944. Although by that time SCC was recognized as a serious industrial problem, the second major symposium occurred only in 1954. These two symposia discussed a wide range of materials. However, their proceedings contain experimental data pertinent to a number of different areas and hence do not readily provide useful information about advances in SCC. The approach to research was very limited and did not draw on scientific methodologies available at the time as SCC researchers do today. In 1957 and 1961, the first comPonent failures due to SCC arose in pressurized and boiling water nuclear reactors, respectively. Since then, SCC and stress-induced localized corrosion in various components of watercooled reactors have come to represent a major generic problem in the nuclear power generation industry. 9 By the 1960s, fracture mechanics had become a serious consideration for studying SCC and corrosion FCG in high-strength, low-alloy steels and high-strength light alloys. The SSRT method was developed and first applied for SCC evaluation in 1966. It was shown that titanium alloys earlier thought to be immune to SCC on the basis of conventional smooth specimen tests were, in fact, highly susceptible to SCC when evaluations made use of precracked specimens. In the late 1960s, major efforts were made to identify the mechanism(s) of SCC and corrosion FCG, and the first measurements of the pH and electrochemical potential within a growing stresscorrosion crack were conducted.
Acknowledgments Firstly I wish to acknowledge my students in the Faculty of Materials Science at Moscow State University who, many years ago, prompted my investigations into the history of EAC research. Secondly, I especially appreciate the assistance of the library staff of the University of Calgary in securing reprints of 19th century documents relevant to this paper, and the assistance of Enoch Ng and Krishna Panchalingam of the University of Calgary with the preparation of figures. Special thanks to Dr. Stan Lynch of Monash University, Melbourne who also helped me to locate references. I would like to thank the Canadian Institute of Mining, Metallurgy and Petroleum (CIM) and Verlag Stahleisen GmbH, Germany for permission to reproduce copyrighted photographs in
History of SCC Research
550
this paper. Finally, I would like to acknowledge the financial support of the Natural Sciences and Engineering Research Council (NSERC) of Canada.
References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]
[15] [ 16] [ 17] [18] [ 19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31 ] [32] [33] [34] [35]
W.H. Johnson, Proc. Manchester Lit. Phil. Soc. 12 (1873) 42. W.H. Johnson, Proc. Manchester Lit. Phil. Soc. 12 (1873) 74-80. W.H. Johnson, Proc. Manchester Lit. Phil. Soc. 13 (1874) 60-61. O. Reynolds, Proc. Manchester Lit. Phil. Soc. 13 (1874) 93-96; also published in Chem. News 29 (1874) 118-119 and J. Chem. Soc. 12 (1874) 546 (abstract only). W.H. Johnson, Proc. Manchester Lit. Phil. Soc. 13 (1874) 100-111. W.H. Johnson, Proc. Manchester Lit. Phil. Soc. 13 (1874) 130-131. W.H. Johnson, Proc. Roy. Soc. London 23(158) (1875) 168-179. D.E. Hughes, J. Soc. Telegr. Eng. 9 (1880) 163-167; also published in Sci. Am. Suppl., no. 237, July 17, 1880. W.C. Roberts-Austen, Proc. Roy. Inst. Great Brit. 11 (1886) 395-412. Von A. Ledebur, Stahl Eisen 7 (1887) 681-694; 9 (1889) 745-755 (in German). M. Bellati, S. Lussana, Atti R. Ist. Ven. 7(6) (1889) 1321-1341 (in Italian). W.C. Roberts-Austen, Proc. Inst. Mech. Eng. (1899) 35-68. T.P. Hoar, Stress-corrosion cracking, in: Proc. 2nd International Congress on Metallic Corrosion, NACE, Houston, TX, 1966, pp. 14-22. B.F. Brown, A preface to the problem of stress corrosion cracking, in: Stress Corrosion Cracking of Metals - A State of the Art, ASTM STP 518, ASTM, Philadelphia, PA, 1972, pp. 3-15. C. Diegel, Verh. Ver. Bef. Gew. 85 (1906) 177-184; Metallurgie 3 (1906) 568-569 (in German). E.S. Sperry, Brass World 2 (1906) 39-44. F. Rogers, unpublished report, 1906 (quoted in [ 18]). H. Moore, S. Beckinsale, C.E. Mallinson, J. Inst. Met. 25 (1921) 35-124. C.E. Stromeyer, J. Iron Steel Inst. 81 (1910) 679-680; Report to the Manchester Steam Users' Association, 1910 (quoted in [20]). J.H. Andrew, Trans. Faraday Soc. 9 (1913) 316-329. S.W. Parr, Univ. Illinois Bull. 14 (Eng. Expt. Station, Bull. No. 94) (1917) 1-57. P.D. Merica, Met. Chem. Eng. 16 (1917) 496-503. B.P. Haigh, J. Inst. Met. 18 (1917) 55-77. F.H. Keating, Chemical manifestations of internal stress, in: Symposium on Internal Stresses in Metals and Alloys, The Institute of Metals, London, 1947, pp. 311-331. F.B. Allen, Trans. ASME 4 (1883) 142-149. A.M. Greene, Jr., History of the ASME Boiler Code, ASME, New York, NY, 1955. W.B. Le Van, Trans. ASME 4 (1881) 516-538. Steam-boiler explosions in England, Manuf. Build. 8 (1876) 107. J.M. Milton, Trans. Inst. Naval Arch. 47 (1905) 359-383. Topical Discussion on Season and Corrosion Cracking of Brass, Proc. ASTM 18(2) (1918) 147-219. H.L. Logan, The Stress Corrosion of Metals, John Wiley and Sons, New York, NY, 1966. Symposium on Stress-Corrosion Cracking of Metals, ASTM, Philadelphia, PA, 1945. Stress-corrosion cracking of metals: summary of symposium, Mining Met. 26 (1945) 40. Symposium on Internal Stresses in Metals and Alloys, The Institute of Metals, London, 1947. W.D. Robertson (Ed.), Stress Corrosion Cracking and Embrittlement, Wiley, New York, NY, 1956.
Volume 1: Chemistry, Mechanics and Mechanisms
551
[36] R.W. Staehle, A.J. Forty, D. Van Rooyen (Eds.), Fundamental Aspects of Stress Corrosion Cracking, NACE, Houston, TX, 1969. [37] P.R. Swann, F.P. Ford, A.R.C. Westwood (Eds.), Mechanisms of Environment Sensitive Cracking of Materials, The Metals Society, London, 1977. [38] A. Turnbull (Ed.), Corrosion Chemistry within Pits, Crevices and Cracks, HMSO Books, London, 1987. [39] S.A. Shipilov, Tech. Law Insur. 1 (1996) 131-142. [40] H. Lee Craig, Progress towards standardization of SCC test techniques by the American Society for Testing and Materials, in: Specialists Meeting on Stress Corrosion Testing Methods, AGARD-CP-98, NATO, Brussels, 1971, pp. 2.1-2.7. [41] R.W. Staehle, Comments on the history, engineering and science of stress corrosion cracking, in: R.W. Staehle, A.J. Forty, D. Van Rooyen (Eds.), Fundamental Aspects of Stress Corrosion Cracking, NACE, Houston, TX, 1969, pp. 3-14. [42] H. Nathorst, Weld. Res. Council Bull. 6 (1950) 1-18. [43] R.W. Staehle, A point of view concerning mechanisms of environment-sensitive cracking of engineering materials, in: P.R. Swann, F.P. Ford, A.R.C. Westwood (Eds.), Mechanisms of Environment Sensitive Cracking of Materials, The Metals Society, London, 1977, pp. 574-602. [44] R.W. Staehle, J. Hochmann, R.D. Mccright, J.E. Slater (Eds.), Stress Corrosion Cracking and Hydrogen Embrittlement of Iron Base Alloys, NACE, Houston, TX, 1977. [45] W.B. Lisagor, Environmental cracking- stress corrosion, in: R. Baboian (Ed.), Corrosion Tests and Standards: Application and Interpretation, ASTM, Philadelphia, PA, 1995, pp. 240-252. [46] T.W. Richards, R.C. Behr, Carnegie Inst. Washington, Publication no. 61, part 2, 1906. [47] A.R. Troiano, General keynote lecture, in: I.M. Bernstein, A.W. Thompson (Eds.), Hydrogen in Metals, ASM, Metals Park, OH, 1973, pp. 3-15. [48] S. Rask, Stress corrosion in copper and copper alloys. A bibliography, in: A. Bresle (Ed.), Recent Advances in Stress Corrosion, Bull. no. 25, IVA:s Korrosionsniimnd, Stockholm, 1961, pp. 81-107. [49] J.B. Kommers, Proc. Int. Soc. Test. Mater. 2, paper 5 (1912) 4a (quoted in [21,22,67]). [50] L. Lin, Y. Zhao, D. Cui, Y. Meng, Microstructural sensitivity of stress corrosion cracking in copper alloys due to dynamic recrystallization, in: S.A. Shipilov, R.H. Jones, J.M. Olive, R.B. Rebak (Eds.), Environment-Induced Cracking of Materials: Chemistry, Mechanics and Mechanisms, Elsevier, Oxford, 2006, pp. 387-392. [51] L. Lin, Y. Zhao, Stress corrosion cracking of aluminium brass induced by marine organism fouling, in: S.A. Shipilov, R.H. Jones, J.M. Olive, R.B. Rebak (Eds.), Environment-Induced Cracking of Materials: Prediction, Industrial Developments and Evaluation, Elsevier, Oxford, 2006, pp. 369-373. [52] G.H. Clamer, Proc. ASTM 15(2) (1915) 110-111 (discussion). [53] E.S. Sperry, Brass World 8 (1912) 345. [54] E. Heyn, O. Bauer, Int. Z. Metallogr. 1 (1911) 16-50 (in German). [55] E. Heyn, J. Inst. Met. 12 (1914) 3-37. [56] E. Jonson, Trans. Amer. Inst. Met. 8 (1914) 135; Proc. ASTM 15(2) (1915) 101-108. [57] P.D. Merica, Failure of brass. 2 - Effect of corrosion on the ductility and strength of brass, Technologic Paper of the Bureau of Standards No. 83, Washington, Government Printing Office, 1916. [58] W.H. Bassett, Proc. ASTM 18(2) (1918) 153-162. [59] W. Arthur, Trans. Amer. Inst. Met. 10 (1916) 173-177. [60] W. Rosenhain, Trans. Faraday Soc. 17 (1921) 2-16. [61] P.D. Merica, R.W. Woodward, Proc. ASTM 18(2) (1918) 165-178. [62] F.S. Spiers (Ed.), The Failure of Metals Under Internal and Prolonged Stress: A General Discussion, Trans. Faraday Soc. 17 (1921 ) 1-210.
552
History of SCC Research
[63] P.D. Merica, R.W. Woodward, Failure of brass. 1 -Microstructure, and initial stresses in wrought brasses of the type 60 per cent copper and 40 per cent zinc, Technologic Paper of the Bureau of Standards No. 82, Washington, Government Printing Office, 1917. [64] A.D. Flinn, E. Jonson, Proc. ASTM 18(2) (1918) 201-205 (discussion). [65] A.D. Flinn, Eng. Record 68 (1913) 527. [66] P.D. Merica, R.W. Woodward, Trans. Amer. Inst. Met. 9 (1915) 298. [67] C.H. Desch, J. Inst. Met. 11 (1914) 57-106. [68] W. Rosenhain, S.L. Archbutt, Proc. Roy. Soc. London A96 (1920) 55-68. [69] W.C. Hothersall, Trans. Faraday Soc. 17 (1921) 201-208. [70] A. Pinkerton, W.H. Tait, J. Inst. Met. 36 (1926) 233-236. [71] I.V. Williams, Bell Lab. Rec. 8 (1929) 77-79. [72] J. Czochralski, H. Schreiber, Korros. Metallsch. 13 (1937) 181-183 (in German). [73] G. Edmunds, Season cracking of brass, in: Proc. Symposium on Stress-Corrosion Cracking of Metals, ASTM, Philadelphia, PA, 1945, pp. 67-89. [74] H.P. Croft, Proc. ASTM 41 (1941) 905-926. [75] H.P. Croft, G. Sachs, The Iron Age 151(10) (1943) 47-50; 151(11) (1943) 62-67. [76] N.W. Mitchell, Petrol. Process. 1 (1946) 176. [77] L.E. Gibbs, Met. Prog. 45 (1944) 881-886. [78] M. Cook, The relation of composition to stress-corrosion cracking in copper alloys, in: Symposium on Internal Stresses in Metals and Alloys, The Institute of Metals, London, 1947, pp. 73-84. [79] E.H. Dix, Jr., Proc. ASTM 41 (1941) 928-929 (discussion). [80] L. Baker, Trans. Inst. Marine Eng. 65 (1953) 26-29. [81 ] C. Breckon, P.T. Gilbert, Met. Ind. 93 (1958) 89-92. [82] U.R. Evans, The Corrosion and Oxidation of Metals: Scientific Principles and Practical Applications, E. Arnold, London, 1961. [83] B. Irving, Weld. J. 71(12) (1992) 37-40. [84] D.J. Wulpi, Understanding How Components Fail, 2nd ed., ASM International, Materials Park, OH, 1999. [85] W.H. Hatfield, G.L. Thirkell, J. Inst. Met. 22 (1919) 67-91. [86] G. Masing, Elektrotechn. Z. 43 (1922) 152-155. [87] A.V. DeForest, Proc. ASTM 18(2) (1918) 205-209 (discussion). [88] W.B. Price, Proc. ASTM 18(2) (1918) 209-210 (discussion). [89] M.E. Whitaker, E. Voce, A.R. Bailey, Metallurgia 39 (1948) 21-29, 66-70. [90] D.H. Thompson, A.W. Tracy, Trans. AIME 185 (1949) 100-109. [91] T.A. Read, J.B. Reed, H. Rosenthal, The mechanism of season cracking of brass, in: Symposium on Stress-Corrosion Cracking of Metals, ASTM, Philadelphia, PA, 1945, pp. 90-110. [92] U.R. Evans, Metallic Corrosion, Passivity and Protection, E. Arnold, London, 1946. [93] C.L. Bulow, Stress-corrosion testing of copper-base alloys, in: Symposium on StressCorrosion Cracking of Metals, ASTM, Philadelphia, PA, 1945, pp. 19-35. [94] U.R. Evans, Influence of residual stress on chemical behaviour, in: Symposium on Internal Stresses in Metals and Alloys, The Institute of Metals, London, 1947, pp. 291-310. [95] Steam-boiler explosions in England, Manuf. Build. 7 (1875) 82, 106; 8 (1876) 33-34, 58. [96] W. Rosenhain, D. Hanson, J. Iron Steel Inst. 102 (1920) 23-30. [97] C.H. Hewison, Locomotive Boiler Explosions, David and Charles, Newton Abbot, 1983. [98] E.B. Wolf, J. Iron Steel Inst. 96 (1917) 137-158. [99] C.E. Stromeyer, J. Iron Steel Inst. 96 (1917) 159-163 (discussion). [100] C.H. Desch, J. Iron Steel Inst. 96 (1917) 172 (discussion). [101] J.A. Jones, Trans. Faraday Soc. 17 (1921) 102-109.
Volume 1: Chemistry, Mechanics and Mechanisms
553
[102] W. Andrews, Metallurgist 11 (1938) 160. [103] M. Smialowski, Korros. Metallschutz. 15 (1939) 81-82. [104] E. Houdremont, H. Bennek, H. Wentrup, Stahl Eisen 60 (1940) 757-763, 791-801 (in German). [105] S.L. Hoyt, M.A. Scheil, Trans. ASM 27 (1939) 191-226. [ 106] J.C. Hodge, J.L. Miller, Trans. ASM 28 (1940) 25-81. [107] H.J. Rocha, Tech. Mitt. Krupp, A: Forsch.-Ber. 5 (1942) 1-14 (in German). [ 108] M.A. Scheil, Some observations of stress-corrosion cracking in austenitic stainless alloys, in: Symposium on Stress-Corrosion Cracking of Metals, ASTM, Philadelphia, PA, 1945, pp. 395-410. [109] R. Franks, W.O. Binder, C.M. Brown, The susceptibility of austenitic stainless steels to stress-corrosion cracking, in: Symposium on Stress-Corrosion Cracking of Metals, ASTM, Philadelphia, PA, 1945, pp. 411-420. [110] O.B. Ellis, Some examples of stress-corrosion cracking of austenitic stainless steel, in: Symposium on Stress-Corrosion Cracking of Metals, ASTM, Philadelphia, PA, 1945, pp. 421-424. [ 111 ] S.W. Parr, F.G. Straub, Proc. ASTM 26 (1926) 52-79. [112] H. Nathorst, Weld. Res. Coun. Bull. 6 (1950) 1-18. [113] R.B. Johnson, SAE J. 61(12) (1953) 28-29. [114] W.L. Badger, Stress corrosion of 12% Cr stainless steel. Presented at the SAE Annual Meeting, held January 11-15, 1954, Detroit, MI. [115] F.K. Bloom, Corrosion 11 (1955) 351t-361t. [ 116] S.C. Langdon, Trans. Amer. Electrochem. Soc. 36 (1919) 131 (discussion). [ 117] D.J. McAdam, Jr., Proc. ASTM 26 (1926) 224-254. [118] H.J. Gough, J. Inst. Met. 49 (1932) 17-92. [119] H.J. Gough, D.G. Sopwith, J. Inst. Met. 49 (1932) 93-122. [ 120] G.D. Lehmann, Aeronaut. Res. Comm. Rep. Mem. (1926) 1054. [121] A.M. Binnie, Aeronaut. Res. Comm. Rep. Mem. (1929) 1244. [122] F.N. Speller, Proc. ASTM 26 (1926) 276-277 (discussion). [123] S.F. Dorey, Trans. Inst. Naval Arch. 75 (1933) 200-217. [124] W.E. Harvey, Met. Alloys 1 (1930) 458-461. [125] P.T. Gilbert, Met. Rev. 1 (1956) 379-417. [ 126] A.V. DeForest, J. Appl. Mech. 3 (1936) A23-A25. [127] A.J. Gould, U.R. Evans, Iron and Steel Institute Special Report no. 24, 1939. [128] U.R. Evans, M.T. Simnad, Proc. Roy. Soc. London A188 (1947) 372-392. [129] U.R. Evans, The electrochemistry of corrosion fatigue, in: Proc. Symposium on the Failure of Metals by Fatigue, University Press, Melbourne, 1947, pp. 84-90. [130] M.T. Simnad, U.R. Evans, Trans. Faraday Soc. 46 (1950) 175-186. [131] M.T. Simnad, J. Electrochem. Soc. 97 (1950) 31-44. [132] E.H. Dix, Trans. AIMI 137 (1940) 11-40. [ 133] R.B. Mears, R.H. Brown, E.H. Dix, A generalized theory of stress corrosion of alloys, in: Symposium on Stress-Corrosion Cracking of Metals, ASTM, Philadelphia, PA, 1945, pp. 323-344. [ 134] J.J. Harwood, Corrosion 6 (1950) 290-307. [135] P.T. Gilbert, S.E. Hadden, J. Inst. Met. 77 (1950) 237-261. [136] E.C.W. Perryman, S.E. Hadden, J. Inst. Met. 77 (1950) 207-235. [137] R.B. Mears, R.H. Brown, Ind. Eng. Chem. 33 (1941) 1001-1010. [138] Y. Druet, P.-A. Jacquet, Met. Corros. 22 (1947) 139-141 (in French). [139] H.L. Logan, J. Res. Nat. Bur. Stand. 25 (1940) 315-325. [140] J.C. Chaston, Discussion on "The origin, control, and removal of internal stresses," in: Symposium on Internal Stresses in Metals and Alloys, The Institute of Metals, London, 1947, pp. 416-417. [141] H.L. Logan, J. Res. Nat. Bur. Stand. 48 (1952) 99-105.
554
History of SCC Research
[142] J.T. Waber, H.J. McDonald, Stress Corrosion Cracking of Mild Steel, Corrosion Publishing Company, Pittsburgh, PA, 1947. [143] Korrosion IX: Zur D6utung der Spannungskorrosion der Nichteisenmetallegierungen, Verlag Chemie GmbH, Weinheim, 1958 (in German). [144] T.N. Rhodin (Ed.), Physical Metallurgy of Stress-Corrosion Fracture, Interscience Publishers, New York, NY, 1959. [145] Papers presented at the Symposium on Stress Corrosion Cracking, held March 11-15, 1963, in: Proc. 2nd International Congress on Metallic Corrosion, NACE, Houston, TX, 1966, pp. 33-235. [146] H.R. Copson, Effect of composition on stress corrosion cracking of some alloys containing nickel, in: T.N. Rhodin (Ed.), Physical Metallurgy of Stress Corrosion Fracture, Interscience Publishers, New York, NY, 1959, pp. 247-267. [147] C. Edeleanu, J. Iron Steel Inst. 173 (1953) 140-146; 175 (1953) 390-392. [148] W.L. Williams, J.F. Eckel, J. Amer. Soc. Naval Eng. 68 (1956) 93-104. [149] M.A. Streicher, A.J. Sweet, Corrosion 25 (1969) 1-6. [150] E.C. Bain, R.H. Abom, J.B. Rutherford, Trans. Amer. Soc. Steel Treat. 21 (1933) 481509. [151] J.A. Board, J. Inst. Met. 101 (1973) 241-247. [152] T. Kondo, H. Nakajima, R. Nagasaki, Nucl. Eng. Design 16 (1971) 205-222. [153] C.F. Cheng, Reactor Tech. 13 (1970) 310-319. [ 154] W.J. Singley, I.H. Welinsky, S.F. Whirl, H.A. Klein, Stress corrosion of stainless steel and boiler water treatment at Shippingport atomic power station, in: Proc. American Power Conference, vol. 21, Illinois Institute of Technology, Chicago, IL, 1959, pp. 748-766. [155] C.R. Cupp, Effect of neutron irradiation on post-irradiation stress corrosion cracking of stainless steel, in: T.N. Rhodin (Ed.), Physical Metallurgy of Stress Corrosion Fracture, Interscience Publishers, New York, NY, 1959, pp. 270-271. [156] G.M. Gordon, W.D. Miller, Nucl. Eng. Int. 33(402) (1988) 44-46. [157] H. Coriou, L. Grall, Y. Le Gall, S. Vettier, Corrosion sous contrainte de l'Inconel dans l'eau ~ haute temp6rature, in: 3~me Colloque de m~tallurgie sur la corrosion, North Holland Publishing Co., Amsterdam, 1959, pp. 161-169 (in French). [158] D. van Rooyen, Corrosion 31 (1975) 327-337. [ 159] B. GrSnwall, L. Ljungberg, W. Hiibner, W. Stuart, Nucl. Eng. Design 6 (1967) 383-390. [ 160] H. Coriou, L. Grall, C. Mahieu, M. Pelras, Corrosion 22 (1966) 280-290. [161 ] H. Coriou, L. Grall, P. Olivier, H. Willermoz, Influence of carbon and nickel content on stress corrosion cracking of austenitic stainless alloys in pure or chlorinated water at 350 C, in: R.W. Staehle, A.J. Forty, D. Van Rooyen (Eds.), Fundamental Aspects of Stress Corrosion Cracking, NACE, Houston, TX, 1969, pp. 352-356. [ 162] H.R. Copson, S.W. Dean, Corrosion 21 (1965) 1-8. [163] H.R. Copson, Discussion to "Influence of carbon and nickel content on stress corrosion cracking of austenitic stainless alloys in pure or chlorinated water at 350 C," in: R.W. Staehle, A.J. Forty, D. Van Rooyen (Eds.), Fundamental Aspects of Stress Corrosion Cracking, NACE, Houston, TX, 1969, p. 358. [164] D.J. DePaul, Corrosion 13 (1957) 75t-80t. [165] D.J. DePaul (Ed.), Corrosion and Wear Handbook for Water Cooled Reactors, McGrawHill Book Co., New York, NY, 1957, pp. 187-223. [ 166] J. Blanchet, H. Coriou, L. Grall, C. Mahieu, C. Otter, G. Turluer, Historical review of the principal research concerning the phenomena of cracking of nickel base austenitic alloys, in: R.W. Staehle, J. Hochmann, R.D. Mccright, J.E. Slater (Eds.), Stress Corrosion Cracking and Hydrogen Embrittlement of Iron Base Alloys, NACE, Houston, TX, 1977, pp. 1149-1160. [ 167] A.R. Troiano, Trans. ASM 52 (1960) 54-80. [168] R.P. Frohmberg, W.J. Barnett, A.R. Troiano, Trans. ASM 47 (1955) 892-925. [169] H.H. Johnson, J.G. Morlet, A.R. Troiano, Trans. TMS-AIME 212 (1958) 526-536.
Volume 1: Chemistry, Mechanics and Mechanisms
555
[170] R.D. Daniels, R.J. Quigg, A.R. Troiano, Trans. ASM 51 (1959) 843-860. [171] A.R. Troiano, Corrosion 15 (1959) 207t-212t. [ 172] J.R. Rice, Mechanics aspects of stress corrosion cracking and hydrogen embrittlement, in: R.W. Staehle, J. Hochmann, R.D. McCright, J.E. Slater (Eds.), Stress Corrosion Cracking and Hydrogen Embrittlement of Iron Base Alloys, NACE, Houston, TX, 1977, pp. 11-15. [173] A. Phillips, V. Kerlins, B.V. Whiteson, Report AFML-TR-64-416, Air Force Systems Command, Wright-Patterson AFB, OH, January 1965. [ 174] C.D. Beachem, Metall. Trans. 3 (1972) 437-451. [175] J.K. Tien, R.J. Richards, O. Buck, H.L. Marcus, Scripta Met. 9 (1975) 1097-1101. [176] J.R. Rellick, C.J. McMahon, Jr., H.L. Marcus, P.W. Palmberg, Metall. Trans. 2 (1971) 1492-1494. [177] C.J. McMahon, Jr., C.L. Briant, S.K. Banerji, The effects of hydrogen and impurities on brittle fracture in steel, in: D.M.R. Taplin (Ed.), Advances in Research on the Fracture of Materials: Fracture 77, Vol. 1, Pergamon Press, New York, NY, 1978, pp. 363-385. [178] H.H. Johnson, Hydrogen brittleness in hydrogen and hydrogen-oxygen gas mixtures, in." R.W. Staehle, J. Hochmann, R.D. McCright, J.E. Slater (Eds.), Stress Corrosion Cracking and Hydrogen Embrittlement of Iron Base Alloys, NACE, Houston, TX, 1977, pp. 382388. [179] M.A.V. Devanathan, Z. Stachurski, Proc. Roy. Soc. London A270 (1962) 90-102. [180] R.T. Davis, Jr., T.J. Butler, J. Electrochem. Soc. 105 (1958) 563-568. [ 181 ] H.J. Bhatt, E.H. Phelps, Corrosion 17 (1961) 430t-434t. [182] P. Bastien (Ed.), L'Hydrog/me dans les M6taux, Editions Science et lndustrie, Paris, 1972 (in French). [183] Hydrogen Embrittlement Testing, ASTM STP 543, ASTM, Philadelphia, PA, 1974. [184] I.M. Bernstein, A.W. Thompson (Eds.), Hydrogen in Metals, ASM, Metals Park, OH, 1973. [ 185] A.J. Forty, The initiation and propagation of cracks in the stress corrosion of ~t-brass and similar alloys, in: T.N. Rhodin (Ed.), Physical Metallurgy of Stress Corrosion Fracture, Interscience Publishers, New York, NY, 1959, pp. 99-114. [186] A.J. Forty, The metal physics of stress corrosion, in: A. Bresle (Ed.), Recent Advances in Stress Corrosion, Bull. no. 25, IVA:s Korrosionsn~.mnd, Stockholm, 1961, pp. 22-36. [187] W.D. Robertson, A.S. Tetelman, A unified structural mechanism for intergranular and transgranular corrosion cracking, in: Strengthening Mechanisms in Solids, ASM, Metals Park, OH, 1962, pp. 217-252. [ 188] P.R. Swann, Ph.D. dissertation, University of Cambridge, 1960. [189] T.P. Hoar, J.G. Hines, J. Iron Steel Inst. 182 (1956) 124--143. [ 190] R. Bakish, W.D. Robertson, Acta Met. 4 (1956) 342-351. [191] T.P. Hoar, J.M. West, Nature 181 (1958) 835. [192] P.R. Swann, J. Nutting, J. Inst. Met. 88 (1960) 478-480. [193] P.R. Swann, Corrosion 19 (1963) 102t-112t. [194] M. Henthorne, R.N. Parkins, Corros. Sci. 6 (1966) 357-369. [ 195] M. Henthome, R.N. Parkins, Brit. Corros. J. 2 (1967) 186--192. [ 196] R.N. Parkins, Brit. Corros. J. 7 (1972) 15-28. [197] G.P. Dean, Ph.D. dissertation, University of Newcastle upon Tyne, 1971. [198] J.C. Scully, D.T. Powell, Corros. Sci. 10 (1970) 719-733. [199] M.E. Indig, D.A. Vermilyea, The straining electrode around 300 C, in: D. de G. Jones, J. Slater, R.W. Staehle (Eds.), High Temperature High Pressure Electrochemistry in Aqueous Solutions, NACE, Houston, TX, 1976, pp. 558-567. [200] W.E. Andresen, Discussion to "The role of corrosion products in crack propagation in austenitic stainless steel. Electron microscopic studies," in: T.N. Rhodin (Ed.), Physical Metallurgy of Stress Corrosion Fracture, Interscience Publishers, New York, NY, 1959, pp. 147-148.
556
History of SCC Research
[201] N.A. Nielsen, The role of corrosion products in crack propagation in austenitic stainless steel. Electron microscopic studies," in: T.N. Rhodin (Ed.), Physical Metallurgy of Stress Corrosion Fracture, Interscience Publishers, New York, NY, 1959, pp. 121-143. [202] G.R. Irwin, J. Appl. Mech. 24 (1957) 361-364. [203] A.A. Wells, Naval Research Laboratory Report no. 4705, March 1956. [204] E.A. Steigerwald, Proc. ASTM 60 (1960) 750-760. [205] G.R. Irwin, J.A. Kies, H.L. Smith, Proc. ASTM 58 (1958) 640--657. [206] Report of a Special ASTM Committee: Fracture testing of high-strength sheet materials, ASTM Bull. 243, 1960, pp. 29-40. [207] H.H. Johnson, A.M. Willner, Appl. Mater. Res. 4 (1965) 34-40. [208] B.F. Brown, C.D. Beachem, Corros. Sci. 5 (1965) 745-750. [209] Stress Corrosion Testing, ASTM STP 425, ASTM, Philadelphia, PA, 1967. [210] C.D. Beachem, B.F. Brown, A comparison of three precracked specimens for evaluating the susceptibility of high-strength steels to stress corrosion cracking, in: Stress Corrosion Testing, ASTM STP 425, ASTM, Philadelphia, PA, 1967, pp. 31-40. [211 ] S.M. Wiederhom, J. Amer. Cer. Soc. 50 (1967) 407-4 17. [212] M.H. Peterson, B.F. Brown, R.L. New Begin, R.E. Groover, Corrosion 23 (1967) 142148. [213] W.G. Clark, Jr., J.D. Landes, An evaluation of rising load Kmcc testing, in: H.L. Craig, Jr. (Ed.), Stress Corrosion- New Approaches, ASTM STP 610, ASTM, Philadelphia, PA, 1976, pp. 108-127. [214] E.P. Dahlberg, D.B. Lytle, Fatigue crack propagation in high strength 4340 steel, NRL Memorandum Report no. 1471, Naval Research Laboratory, November 1963. [215] E.P. Dahlberg, Trans. ASM 58 (1965) 48-53. [216] W.A. Van Der Sluys, Trans. ASME, J. Basic Eng. 87 (1965) 363-373. [217] A. Hartman, Int. J. Fract. Mech. 1 (1965) 167-187. [218] R.P. Wei, J.D. Landes, Mater. Res. Stand. 9 (1969) 25-27. [219] J.M. Barsom, Int. J. Fract. Mech. 7 (1971) 163-182. [220] J.P. Gallagher, J. Mater. 6 (1971) 941-964. [221] D.A. Meyn, Metall. Trans. 2 (1971) 853-865. [222] O. Devereux, A.J. McEvily, R.W. Staehle (Eds.), Corrosion Fatigue: Chemistry, Mechanics and Microstructure, NACE, Houston, TX, 1972. [223] M.O. Speidel, M.J. Blackburn, T.R. Beck J.A. Feeney, Corrosion fatigue and stress corrosion crack growth in high strength aluminum alloys, magnesium alloys, and titanium alloys exposed to aqueous solutions, in: O. Devereux, A.J. McEvily, R.W. Staehle (Eds.), Corrosion Fatigue: Chemistry, Mechanics and Microstructure, NACE, Houston, TX, 1972, pp. 324-345. [224] A.J. McEvily, R.P. Wei, Fracture mechanics and corrosion fatigue, in: O. Devereux, A.J. McEvily, R.W. Staehle (Eds.), Corrosion Fatigue: Chemistry, Mechanics and Microstructure, NACE, Houston, TX, 1972, pp. 381-395. [225] I.M. Austen, E.F. Walker, Quantitative understanding of the effects of mechanical and environmental variables on corrosion fatigue crack growth behaviour, in: The Influence of Environment on Fatigue, IMechE, London, 1977, pp+ 1-10. [226] Symposium on Stress-Corrosion Cracking of Titanium, ASTM STP 397, ASTM, Philadelphia, PA, 1966. [227] H.H. Uhlig, Corrosion and Corrosion Control, John Wiley & Sons, New York, NY, 1963. [228] N.D. Tomashov, Theory of Corrosion and Protection of Metals, Macmillan, New York, NY, 1966 (English translation). [229] G.C. Kiefer, W.W. Harple, Met. Prog. 63 (1953) 74-76. [230] B.F. Brown, Mater. Res. Stand. 6 (1966) 129-133. [231 ] H.P. Leckie, Corrosion 23 (1967) 187-191. [232] H.H. Uhlig, Met. Prog. 57 (1950) 486-487.
Volume 1: Chemistry, Mechanics and Mechanisms
557
[233] R.B. Mears, Discussion to "New perspectives in the stress corrosion problem," in: T.N. Rhodin (Ed.), Physical Metallurgy of Stress Corrosion Fracture, Interscience Publishers, New York, NY, 1959, p. 22. [234] M.C. Bloom, Discussion to "Stress corrosion cracking of austenitic stainless steels in high temperature chloride waters," in: T.N. Rhodin (Ed.), Physical Metallurgy of Stress Corrosion Fracture, Interscience Publishers, New York, NY, 1959, pp. 238-239. [235] H.J. Bhatt, E.H. Phelps, The effect of electrochemical polarization on the stress-corrosion behavior of steels with high yield strength, in: Proc. 3rd International Congress on Metallic Corrosion, vol. 2, Mir Publisher, Moscow, 1969, pp. 285-291. [236] B.F. Brown, C.T. Fujii, E.P. Dahlberg, J. Electrochem. Soc. 116 (1969) 218-219. [237] J.A. Smith, M.H. Peterson, B.F. Brown, Corrosion 26 (1970) 539-542. [238] M. Pourbaix, Atlas of Electrochemical Equilibria in Aqueous Solutions, (Translated by J.A. Franklin), Pergamon Press, Oxford, 1966. [239] C.F. Barth, E.A. Steigerwald, A.R. Troiano, Corrosion 25 (1969) 353-358.
559
Author Index A Arimura, M. Ashkenazi, D. Aubert, I.
I
491 201 179
lchitani, K. Inagaki, S.
239 491
J
Jones, R.H.
B
Babout, L. Bala Srinivasan, P. Balch, D.K. Banks-Sills, L. Barnes, A. Boellinghaus, Th. Brass; A.-M. C Cadden, C.H. Chang, W. Chine, J. Chemyayeva, O. Chu, W.Y. Cooper, K.R. Cui, D.
69 295 305 201 47 125 179,215
305 345 261 227 467 333 387
449
K
Kanno, M. Kelly, R.G. Kumar, A.N. L Lebedev, B.D. Lee, J.S. Legris, A. Li, B. Lin, L. Liu, M. Loubat, F. Lu, F. Lunarska, E. Lynch, S.P.
239 333 395
497 81 481 351 387 351 367,377 345 227,249 167
D
Delafosse, D. Dietzel, W.
189 295
E
Eliasi, R. Eliaz, N. Engelberg, D.L. Evitts, R.W.
201 201 69 95
F Feaugeas, X. Fedik, I.I.
189 315
G Gangloff, R.P. Gao, K.W. Girardin, G.
141 467 189
M
Maeda, Y. Malkin, A.I. Marrow, T.J. Matsuda, S. McCartney, L.N. McFadden, S.X. M6nard, M. Meng, Y. Mori, G.
419 105,497 69 239 19 305 179 387 405
N
Nagumo, M. Nakaj ima, K. Newman, R.C. Nikiforow, K. Nishimura, R. Nozaki, T.
285 491 47,69 249 419 323
H
Hall, M.M. Hamasaki, M. Hanaki, S. Harlow, D.G. Haruna, T. Heppner, K.L.
59 359 323 3 359 95
O Olive, J.M.
179, 377
P
Petrova, I.V. Polukarova, Z.M.
497 497
560 Potiron, S. Puskar, J.D.
Author Index
367 305
Q Qiao, L.J.
435
S
Sarrazin-Baudoux, C. Scully, J.R. Senior, N. Serre, I. Sharkawy, S.W. Shchukin, E.D. Shibata, T. Shimomura, Y. Shipilov, S.A. Somerday, B.P. Sulaiman, A. Sun, Z. Switzer, M.A. T Taguchi, T. Takai, K. Tang, Z.
19
U Uchida, H.
323
V Vankeerberghen, M. Verleene, A. Viyanit, E. Vlasov, N.M. Vogt, J.-B.
.115 481 125 315 481
467
R
Rebak, R.B.
Tumbull, A.
367 81 47 481 295 497 359 285 507 305 419 351 81
491 273 345, 351
W
Wang, Y. Watanabe, O. Wei, R.P. Wieser, H.
467 491 3 405
Y Yamashita, M.
323
Z Zanozin, V.M. Zhang, X. Zhao, Y. Zhou, S. Zhu, G. Zitter, H.
497 345,351 387 19 345 405
505
Subject Index 08Ch 18N 10T stainless steel
431
A acoustic emission 449 alcohol environments 337 alloy 22 175 alloy 182 107 alloy 600 3, 95, 107, 123, 143, 153 alloy 690 3, 123 alloy EP-823 185 alloy X-750 123 aluminium alloys 375, 439, 471 aluminium brass 369 ammonia converter 411 analytical transmission electron microscopy 95 applied potential 123, 163, 211,243,267, 291,303, 313, 493 aspect ratio 199 austenitic stainless steels 3, 55, 95, 107, 163, 389, 401, 411, 421, 431,439, 449 B
backscattered electron images 401 boiling water reactors 3, 55, 95, 107, 153 buried pipelines 199, 323 C CANDU reactor 421 carbon steels 267, 337, 349, 401,459 case studies 337, 389, 401, 411,421,431 cathodic polarization 123 cathodic protection 211,233, 243, 267, 279, 291,303, 313, 323, 389, 493 caustic cracking 401 chemo-mechanical model 243 circumferential notch tensile specimen 459 chloride SCC 401 constant extension rate tests 3, 55 constant-load tests 185, 291 control rods drive mechanism nozzle 143 corrosion 175, 337, 363, 369, 389 corrosion-enhanced plasticity 123 corrosion fatigue 153, 221 corrosion potential 107, 123, 153, 175, 185 corrosion rate 175, 211,255, 313, 337, 363, 369 corrosion-based design approach 3 crack closure 199, 221,349 crack coalescence 199 crack colonies 199
crack growth mechanisms 221,233, 243 crack growth rate 3, 55, 107, 123, 243,255 crack initiation 55, 233, 243,267 crack mouth opening displacement 471 crack tip blunting 211 crack-tip plasticity 107,267 crack-tip strain rate 107,303 crack-tip structure 95 crack-wall structure 95 creep damage 123 crevice corrosion 175,185 C-ring sample 95 cyclic frequency effect 153,199,221 cyclic loading 153, 199, 211,221,243, 255, 267 cyclic potentiodynamic polarization 185 D
D20 255, 421 damage function analysis 55 de-alloying 411 degree of sensitization (DOS) 163, 421 deposits 3, 401 deuterium 255 Devanathan-Stachurski cell 243, 255 dissolution-enhanced plasticity 243 dissolved oxygen 163 double-loop electrochemical potentiokinetic reactivation (DL-EPR) 163 E
electrochemical corrosion potential 3, 55, 163 electrochemical impedance spectroscopy 255 electrochemical noise 449 electrode potential 337 energy-dispersive X-ray spectrometry (EDXS) 95, 123, 369, 389, 401,421,431 environmental scanning electron microscopy 303, 401 exfoliation corrosion 375 F
failure analysis 389, 401, 411, 421, 431 failure prediction 3, 55 failures 337, 349, 369, 389, 401 fatigue crack growth 199, 221,349 finite element analysis 107, 375, 483,493 flow assisted corrosion 3, 421 fracture mechanics 107, 153, 199, 211, 221, 255, 291,349, 459, 471
506 fracture toughness four-point bend tests fuel ethanol
Subject Index
459, 471 483 337
G gas tungstenarc welding 153 grain boundaries 3,123,143 grain boundary sliding 123 groundw~er 199,211,233,243,255,267, 303 H
hardening 411 heat-affected zone 291,303, 398, 401,483 heat treatment 411 high resolution X-ray tomography 439 high-temperature water 3, 55, 95, 107, 123, 143, 153, 163,431 HSLA steels 279, 493 hydrazine 3, 95 hydrocracker exchanger 401 hydrocracking reactor 349 hydrogen absorption 123,255, 349, 363 hydrogen content 123, 349, 363 hydrogen degradation 349 hydrogen diffusion coefficient 255, 349 hydrogen effects 123, 211 hydrogen embrittlement 123, 221,233, 279, 291, 313, 493 hydrogen-enhanced plasticity 243 hydrogen evaluation 255 hydrogen-induced cracking 243,267,389 hydrogen-induced SCC 267 hydrogen isotope effect 255 hydrogen permeation 243,255 hydrogen sulphide 221,323,349 hydrogen water chemistry (HWC) 55,107 hydrogenated primary water 123
linear polarization resistance 255 liquid metal embrittlement 389 localized corrosion 175, 185 localized plasticity 267 low alloy steels 349 low pH SCC 221,233, 255 low-potential SCC 3 low-temperature sensitization 163 M
marine atmosphere 363 marine organism fouling 369 metallography 313, 401, 411, 421, 431,449 methanol 267, 337 microbiological corrosion 323,389 microhardness 411,483 micro-notched specimens 267 microprocess sequence approach 3 microstructure 267, 279, 349 mild steels 267, 465 MnS precipitates 267 modeling crack growth 55, 221,243 modeling IGSCC mitigation 55 modeling PWSCC 143 N near-neutral pH SCC 199, 211, 221,233, 243, 255, 267, 291,303, 313 neutron flux 3 nickel-based alloys 3, 95, 107, 123, 143, 153, 175 nickel-based weld alloys 107, 153 nitriding 411 noble metal coating 55 non-destructive evaluation 389 non-destructive inspection 375 NOVA trapped water 211 NS4 solution 199, 211,243, 267, 279, 291, 303, 313
I
integrated (accumulated) damage 55 interdendritic cracks 153 intergranular carbides 3, 95, 123 intergranular corrosion 3, 123, 439 internal oxidation 123 irradiated-assisted SCC 95
O oil trunk line steel 349 once-through steam generator 3 optical microscopy 185, 199, 267, 279 oxygen effect 233 P
J
J-integral
123
L laser microscopy life-time assessment
153 55
parallel-detection electron-energy-loss spectrometry (PEELS) 95 passive films 421 PbSCC 3, 95 pH 185, 211,233, 323 photoelastic images 375
Subject Index
507
pipelines and line pipe steels 199, 211, 221, 243, 255, 267, 279, 291,303, 313, 323, 483, 493 pit nucleation rate 55 pit propagation rate 55 pit-to-crack transition 55, 243 pitting corrosion 185, 337, 389, 431 point defect model 55 polarization curves 243, 255, 313 polythionic acid SCC (PASCC) 401 post weld heat treatment 107, 153, 401 potential-pH diagrams 3, 143 pressure vessel steel 349 pressurized water reactors 3, 95, 107, 123, 143 protective films 3 PWR steam generators 3, 95 PWSCC 123, 143 PWSCC modeling 143 R
redox potential refinery equipment reinforcing steel residual stress residual water
313, 323 389, 401 363 221,483 349
S scanning electron microscopy (SEM) 153, 163, 185, 199, 211,255, 267, 279, 291, 313, 369, 401,421,431,439, 449, 459, 471,493 secondary ion mass spectroscopy 421 seawater 363, 369 sensitization 163, 421,449 shielded metal arc welding 107, 153 slip-step dissolution model 123 slow strain rate bending tests 493 soil environments 221. 303, 313, 323 special grain boundaries 123 spiral notch torsion specimen 471 SSRT tests 123, 143, 185, 233, 255, 267, 279, 291,303, 313, 337, 349 stacking fault energy 3, 163 steel composition 267, 279, 291 steel reinforced concrete 363 strain rate 233,267, 279, 291,303, 313, 411, 493 stress concentrators 221 stress intensity factor 107, 153, 175, 199, 459, 471 stress intensity factor range 199,211, 221, 255, 349
stress ratio 153, 199, 221 submerged arc welding 279 sulphate-reducing bacteria (SRB) 221, 313, 323,349 super duplex stainless steel weld metal 483 super martensitic stainless steel 483 superposition model 199, 221 T temperature effect 175, 185, 349 thermocycling in hydrogen 349 thermomechanical treatment 291 three-point bend tests 493 time-of-flight inspection 375 transdendritic cracks 153 transmission electron microscopy (TEM) 3, 95, 123, 431 Type 302 stainless steel 439 Type 304 stainless steel 3, 55, 95, 107, 163, 389, 401,421,449 Type 308 stainless steel 401 Type 321 stainless steel 389 Type 347 stainless steel 411 U U-bend tests
3, 163
W
waste disposal 175, 185 weight loss 211,363 welding and weldments 107, 153, 279, 291, 401,483 WOL specimens 291
X46 pipeline steel 313 X52 pipeline steel 267 X60 pipeline steel 199, 243, 313, 493 X65 pipeline steel 211,221,255 X70 pipeline steel 303, 493 X80 pipeline steel 493 X 100 pipeline steel 291 X-ray diffraction 401 X-ray fluorescence analysis 411 Y yield strength Yucca Mountain Z Zn-induced cracking
279, 291 175, 185
389
ix
Preface This volume is the second of two containing most of the papers presented at the Second International Conference on Environment-Induced Cracking of Metals (EICM-2), which was held at The Banff Centre, Banff, Alberta, Canada in September 2004. ~This conference was the fitth in a series of major international meetings sponsored or cosponsored by NACE in the area of environment-induced cracking. The previous four conferences, recognized as landmark events, were the: 9 International Conference on Fundamental Aspects of Stress Corrosion Cracking held at the Ohio State University, Columbus, OH, USA, 1967 9 International Corrosion Fatigue Conference held at the University of Connecticut, Storrs, CT, USA, 197 I 9 International Conference on Stress Corrosion Cracking and Hydrogen Embrittlement of Iron Base Alloys held in Unieux-Firminy, France, 1973 9 First International Conference on Environment-lnduced Cracking of Metals held in Kohler, WI, USA, 1988 EICM-2 was jointly sponsored by the Committee TEG 186X on Environmentally Assisted Cracking of NACE International, the Working Party WP5 on Environment Sensitive Fracture of the European Federation of Corrosion and the Joint ASM/TMS Corrosion and Environmental Effects Committee in association with the Technical Committee of Hydrogen Embrittlement and Stress Corrosion of the Chinese Society of Corrosion and Protection and the Technical Committee TC!0 on Environmentally Assisted Cracking of the European Structural Integrity Society. This conference was international in scope and attracted a significant number of researchers and professionals. There were 115 participants from countries that included Australia, Japan, China, India, Saudi Arabia, Israel, Russia, Ukraine, Poland, Slovak Republic, Austria, Slovenia, Italy, Sweden, Germany, Belgium, France, the United Kingdom, Brazil, the United States and Canada. The conference provided a forum for dialogue, assessment and critique among researchers committed to different methodologies and actively promoted discussion and cross-fertilization of ideas among established and emerging researchers working in different areas related to the problem of environment-induced cracking in materials. The technical program consisted of three plenary sessions, 12 technical sessions with 72 oral presentations, a poster session and a luncheon talk. In total, 87 papers that described both fundamental research studies and more practical engineering applications were presented. Topics included but were not limited to: SCC, hydrogen embrittlement, corrosion fatigue, liquid metal embrittlement, localized environmentinduced attack, modeling environmental effects, crack growth mechanisms, hydrogen permeation and transport, hydrogen-plasticity interaction, test methods and interpretation of test data, materials degradation in service, failure analysis, life prediction of corrodible structures, and the history of EICM research. Together the papers provided a comprehensive account of advances in research on environmentinduced cracking and constituted the most recent fundamental and industrial survey of Volume I is published under the title "Environment-lnduced Cracking of Materials: Chemistry, Mechanics and Mechanisms," S.A. Shipilov, R.H. Jones, J.M. Olive, R.B. Rebak (Editors), Elsevier, Oxford, 2007 564 pages.
x
Preface
the subject. Plenary lectures were presented by Richard Gangloff, Digby Macdonald, Roger Newman, Roger Staehle, Alan Turnbull and Robert Wei. A luncheon lecture entitled "Growing Old Gracefully? A Perspective on Pipeline Safety" was given by Alan Murray of the National Energy Board Canada. A highlight of the conference was the General Discussion which featured a panel comprising the plenary speakers with John Scully as a moderator. During the discussion, recent developments and a number of key questions in environment-induced cracking research were considered, including how to move from theory to validation to practice, needs and opportunities for mechanistic advances, progress towards life prediction/prognosis/damage evolution, the challenge to mitigate and control EICM, and needs and opportunities for technological and practical progress. As the potential audience for the proceedings represents a wide spectrum of professionals, including researchers, engineers, practitioners and consultants with different background and specific research and/or industrial interests, it was found feasible to divide the papers into two volumes, each with a specific focus- as indicated in their titles - which can then be sold and/or used together or separately. This necessitated a different arrangement of the papers in the books than the order in which they were presented at the conference, in each volume, the papers are organized to provide a reasonable reflection of research development in the subject areas. Either volume may be used as an independent reference source that reviews the current state of the fundamental and/or industry-oriented research and provides a comprehensive introduction to the field. Applied researchers and specialists in industry may find volume two highly relevant, especially if they are engaged in the areas related to nuclear power generation, oil and gas transportation, the disposal of radioactive waste, aircrat~ maintenance, chemical and marine applications, failure analysis in industrial equipment, and assessing the SCC resistance; while volume one may find a ready audience among researchers who focus upon investigating the phenomenological aspects of SCC, hydrogen embrittlement, liquid metal embrittlement and corrosion fatigue, including environmental, microstructural, electrochemical and mechanistic aspects, in high performance steels, nonferrous alloys, ceramics and glasses. Whatever the individual reader's preference, there are synergies and connections across the volumes which make their simultaneous publication an important occasion for researchers and practitioners alike. Of the 87 papers that were presented at the conference, 81 are offered to the reader across the two volumes. The papers were rigorously peer-reviewed, revised and edited extensively to meet the high standards for scientific publications intended for an international and discerning audience. The present volume contains 38 papers divided into eight sections and is the first publication in the field of environment-induced cracking that addresses a broad range of industrial applications, including nuclear power generation, the disposal of radioactive waste, oil and gas transportation, aircraft maintenance, chemical and marine applications, failure analysis, and assessing the SCC resistance. Section I titled "Prediction of stress corrosion cracking" composes of two plenary papers: "Predicting failures in light water nuclear reactors which have not yet been observed - microprocess sequence approach (MPSA)" by R.W. Staehle and "The electrochemistry of stress corrosion cracking- from theory to damage prediction in practical systems" by D.D. Macdonald et al., which are concerned with the prediction of in-service SCC failures in water-cooled nuclear reactors. New insights into the issues associated with the SCC of stainless steels and Ni alloys used as components of light
Preface
xi
water reactors, including boiling water reactors and pressurized water reactors, are provided in Section 2 titled "Stress corrosion cracking in L WR environments." Section 3 titled "Corrosion and cracking of waste package materials" is devoted to localized corrosion and SCC of Alloy 22 and Alloy EP-823, which are the materials recommended for containers used to dispose high-level radioactive waste and spent nuclear fuel in Yucca Mountain. Various aspects of corrosion fatigue and SCC in C-Mn and micro-alloyed steels used as components of oil and gas pipelines are covered in twelve papers that compose Sections 4 and 5 titled "Crack growth in pipeline steels under cyclic loading" and "SCC and hydrogen embrittlement of pipeline steels" respectively. The emphasis on hydrogen charging and embrittlement in most of these papers reflects the strong focus in recent years on the role of hydrogen in the service failures of buried pipelines having cathodic protection. Section 6 titled "Degradation of materials under in-service conditions" composes of six papers that investigate the "time-dependent" degradation of steels and alloys in different industrial applications where components and structures come into contact with natural or technological environments. Eleven cases of service failures caused by corrosion and EICM in industrial equipment are investigated in Section 7 under the heading "Stress corrosion cracking case studies" and further demonstrate the complexity of the EICM problem in industry. The final section titled "New test methods for SCC studies" describes several experimental techniques which were recently developed for assessing the susceptibility of materials to SCC. The total number of references to the literature in this volume is 664. Many individuals and organizations contributed to the success of EICM-2 and these volumes, and I would like to acknowledge their assistance. The conference and its proceedings would not have been possible without the generous support of my coeditors, Russ Jones, Jean-Marc Olive and RaO! Rebak. These dedicated individuals contributed greatly of their time and skills to arrange various aspects of the meeting. They were instrumental in helping to develop the program, with organizing sessions and assisting with reviewing the manuscripts. I also wish to thank Winston Revie for his assistance at the early stage of the planning of the conference. Special acknowledgments are extended to chairs of sponsoring committees for helping to organize international participation: Jorge Perdomo, Jean-Marc Olive, Ra~l Rebak, Lijie Qiao and Wolfgang Dietzel. Pierre Crevolin, President of NACE in 2003/2004 and Interim Executive Director in 2004, his staff and especially Gretchen Jacobson, Publishing Director, provided invaluable assistance in publicizing EICM-2. Special thanks go to my students, Don Boll, Vicky Carathanassis, Christie Millington, Enoch Ng, Krishna Panchalingam, Feng Wang and Hong Wang, who helped with registration and assisted me through all stages of the conference, including the preparation of several graphs for this publication. Sincere thanks also to Kim Allan for providing technical support for the conference website. 1 would like to extend my special thanks to the session chairs, Jenny Been, Anne-Marie Brass, Steve Bruemmer, Jacques ChEne, Wolfgang Dietzel, Noam Eliaz, Russ Jones, Rob Kelly, Fraser King, Amar Kumar, Graham Lobley, Stan Lynch, Jean-Marc Olive, Alan Plumtree, Rat~l Rebak, John Scully, Raman Singh, Brian Somerday, Bob Sutherby, Mirna Urquidi-Macdonald and Marc Vankeerberghen, who worked very effectively to keep the program organized and the floor discussions effective. ! greatly appreciate the work of the technical experts who joined in reviewing the manuscripts for these volumes, and who approached the manuscript reviews with the detail and exacting standards that mark the review process
xii
Preface
at technical journals. The names of these reviewers are listed on the next page. Without the generosity of our financial sponsors, and the individuals who secured sponsorship funds, the conference would not have happened and these papers would not have seen the light of day. Hence, my heartfelt thanks to NACE International (Pierre Crevolin), National Energy Board Canada (Alan Murray), Metallurgical Consulting Services Ltd. (lain Le May), the University of Calgary (Ron Bond), Canadian Energy Pipeline Association (Jake Abes), NOVA Chemicals Corporation (Fraser King), ASM Canada Council (Steve Yue), ASM Calgary Chapter (Sammy Tang), and Broadsword Corrosion Engineering Ltd. (Pat Teevens). During the preparation of the proceedings, ! was very ably assisted by Michael Aleksiuk who provided diligent and insightful proofreading of several of the manuscripts, and by Lucy Dickinson, Nicola Jones and Kristi Green of Elsevier Ltd. for their kind co-operation at all stages of the work. My sincere thanks go to the authors of the papers for making time in their busy lives to put their work in the public domain. To all those who have helped, ! express my sincere "Thanks!" Last, but certainly not least, i would like to acknowledge my thanks and appreciation to my wife, Hyacinth, whose support, encouragement and companionship sustained me through both the organizing of EICM-2 and the preparation of this publication.
Sergei Shipilov Editor and Conference Chair
xiii
List of Reviewers The quality of the papers that appear in this volume reflects not only the obvious efforts of the authors but also the unheralded, though essential, work of the reviewers who joined with the editors in reviewing the manuscripts for the proceedings. The editors sincerely acknowledge the following reviewers who provided comments and constructive suggestions for the revision of manuscripts.
K. Arioka, Institute of Nuclear Safety Systems, Fukui, Japan J. Been, NOVA Chemicals Corporation, Calgary, Canada J. Chine, Universit6 Paris-Sud, Orsay, France G.A. Cragnolino, Southwest Research Institute, San Antonio, USA J. Congleton, University of Newcastle upon Tyne, UK R. Eadie, University of Alberta, Edmonton, Canada F.P. Ford, General Electric Company, Schenectady, USA J.P. Frayret, Universit6 Bordeaux 1, Talence Cedex, France G.M. Gordon, Framatome ANP, Las Vegas, USA H.E. H~inninen, Helsinki University of Technology, Helsinki, Finland A.P. Jivkov, UMIST, Manchester, UK R.H. Jones, Exponent Failure Analysis Associates, Bellevue, USA J.-M. Olive, Universit6 Bordeaux I, Talence (;edex, France J.J. Perdomo, Smurfit-Stone Container Corporation, Carol Stream, USA M. Puiggali, Universit6 Bordeaux i, Talence Cedex, France R.B. Rebak, Lawrence Livermore National Laboratory, Livermore, USA R.E. Ricker, National Institute of Standards and Technology, Gaithersburg, USA P.R. Roberge, Royal Military College of Canada, Kingston, Canada S.A. Shipiiov, Metallurgical Consulting Services, Calgary, Canada B.P. Somerday, Sandia National Laboratories, Livermore, USA W.-T. Tsai, National Cheng Kung University, Tainan, Taiwan, China J.-B. Vogt, Universit6 de Lille, Villeneuve d'Ascq, France S. W~.stberg, Det Norske Veritas (DNV), Hovik, Norway W. Zheng, CANMET Materials Technology Laboratory, Ottawa, Canada
Predicting failures in light water nuclear reactors which have not yet been observed- microprocess sequence approach (MPSA) Roger W. Staehle Department of Chemical Engineering and Materials Science, Universityof Minnesota, Minneapolis, MN 55455, USA Abstract Stress corrosion cracking continues to provide the most serious challenge to the reliable performance of water cooled nuclear plants. As the lives of these plants are extended, the occurrence of new failures is inevitable. In the past, none of the modes of failure has been predicted, and experimental work on behalf of these failures had all been re-active. Past history has also shown that some of the corrosion failures have taken many years before they became evident. The purpose of this discussion is to describe an approach to predicting future failures. This approach is based first on identifying microprocesses that can occur in the chemistry of water, on the surfaces of components, associated with deposits, in crevices, in the bulk materials, and associated with nuclear radiation. Second, series and parallel sequences of these microprocesses are identified as they might lead to failures. Third, approaches to quantifying the critical microprocesses as they would affect the length of time and the likelihood are identified. Examples of sequential microprocesses that are associated with past failures are described as examples; examples of new sequential microprocesses are suggested.
1. Background The purpose of this discussion is to describe an approach to predicting corrosion failures that have not yet been observed but could occur atter long times, such as those associated with light water reactors (LWRs) that are expected to be re-licensed. This approach is in the category of "pro-active" prediction, where possibly future failures are intentionally sought out, and the credibility for producing failures is assessed. This approach also challenges conventional assumptions about the cause and nature of corrosion failures. In the past, failures have occurred first; and the nuclear materials community, then, has responded usually with excellent work aimed at explaining the observations. This is "re-active" research. We are concerned here, rather, with the mechanics of"pro-active" prediction. This discussion deals with predicting corrosion processes in LWRs, although the approaches described here would apply broadly to other industries as well as social
4
Prediction of Stress Corrosion Cracking
processes. This discussion is also mainly concerned with stress corrosion cracking (SCC) and corrosion fatigue (CF), as they are connected; and, some aspects of flowassisted corrosion (FAC) are included. These are the most likely modes of corrosion that can produce serious failures. Other topics of accelerated damage include wear at anti-vibration bars (AVB), but these are not discussed here, although such modes are within the scope of this approach. An essential assumption of this discussion is that very long times until failure are not related to a monotonic progression of SCC processes. Rather, the long times are most likely associated with other factors that produce specific local conditions that "open the gate" for SCC or other rapid processes at later times. These preliminary processes occur over a "precursor period." Schematically, such cases are shown in Fig. 1. Case I corresponds to SCC, starting upon initial exposure to an environmental condition that produces SCC, as described in Section 2.1. Case I corresponds to failure processes (e.g. to low-potential stress corrosion cracking, LPSCC) that initiate as soon as the surfaces are exposed to primary water at operating temperature. Case II corresponds to SCC that starts after necessary conditions for initiation are achieved in relatively short times (e.g. 89 to 20 years), for which there are already examples as described in Section 2.2. The approach described here for predicting failures, which have not yet occurred, is the "Microprocess Sequence Approach" (MPSA). This approach utilizes sets of elements from the environments and materials where these elements can be identified SCC starts after necessary initial condition achieved (e.g., SCC at TSPs, OTSG upper bundle, SCC after ( ~ denting)
Depth
SCC starts as soon as exposed (e.g., LPSCC at high-stress location) ( ~
•
Initiation v
SCC starts I No precursor period (Case I)
L
J_
-~/~ropaga~n ~/~ ~ ~
Initiation
Necessary initial condition for scc SC( not yet achieved
,
-~/~C'ropag~bn ~ . . l"
,qF
,qF
t
SCCstarts Relatively short time to achieve
initiation condition Short precursor period (Case II)
@
J_/
- ~7~ovag~g. ' r
t
SCCstarts Long time before initiation condition
(e.g., after license renewal)
precursor period (Case llI)
Time Since Startup Fig. 1. Schematic view of three cases for the time-dependence of SCC. Initiation and propagation times assumed to be the same. The three cases are differentiated by the length of the precursor periods. LPSCC = "low-potential SCC." TSPs = "tube supports." OTSG = "once-through steam generator."
Volume 2: Prediction, Industrial Developments and Evaluation
5
and quantified, and connected sequentially or in parallel, to provide a scenario leading to the initiation of failure. The overall procedure is described in Fig. 38 of Section 5. What is actually being predicted here is not the course of the SCC itself but rather the time to arrive at the conditions for SCC to start, as shown for shorter cases in Case II and much longer times in Case III of Fig. 1. It is assumed here that existing correlations, (e.g. from Staehle and Gorman [1]) for the occurrence of SCC will be activated once the precursor period has produced the necessary conditions for SCC to initiate. Such longer times should be compared with the range of time over which relicensing of LWRs in the U.S. is expected to occur as shown in Fig. 2. Relative to the schedule for re-licensing, it appears that instances of SCC in the nuclear industry have occurred in essentially three stages as suggested in Fig. 3. In Stage I, failures occurred in the early use of stainless steel tubes and then in Alloy 600 tubes. These failures were extensive. In Stage II, the present stage, SCC is occurring in the laboratory for Alloy 690TT and in the non-decorated grain boundaries of stainless steel. This present discussion about prediction applies to a Stage III, where a pro-active approach is required and where future failures are assessed by the reasonableness of scenarios as described in Section 5 and Fig. 38. In this Stage III, a relatively long time is consumed by the precursor period in which conditions for the occurrence of SCC must first develop before SCC can start, as shown for Case III in Figure 1. The main theme of this report is illustrated in Fig. 1 for Case III. Here, the conditions that are necessary for SCC to occur have not yet fully developed during the initial licensing period. The central question, then, is: what are the processes that could produce the necessary conditions to activate the SCC at this later time? My assertion here is that predicting SCC/CF that has not yet occurred, i.e. pro-active prediction, can be approached credibly by using information and understandings that are already available and linking them in sensible ways to predict the time required for completion of the precursor period, of Fig. 1 Case III, before the SCC can start. The challenge, then, is to identify this information and explore our present understandings of phenomena that would be considered in long precursor periods. Such an approach is described in Sections 3 and 4.
Fig. 2. Number of U.S. licenses issued and expired vs. time compared with a 20-year life extension after a presumed re-licensing.
6
Prediction of Stress Corrosion Cracking
Stages of incidence of SCC in nuclear industry Stage I
Stage II
Stage III
> 9 Alloy600MA 9 Austenitic stainless steel
9 SCC of 690TT arising from Pb, S
9 SCC has not yet occurred in practice
9 SCC of undecorated grain boundaries
9 Needmicroprocess path definitiorL~ 9 NeedSCC amelioration target
Fig. 3. Schematic illustration of the stages of SCC occurrences over time in the LWR industry.
There are, presently, no serious predictive methods for SCC that are not based on extrapolating from already existing failures. Current methods include: 1. Accelerated testing and carrying forward the data, usually the mean value, with experimentally determined dependencies such as the activation energy (e.g. Q = 40 kcal) or the stress exponent (e.g. n = 4). This is more or less how performance is predicted at the present. 2. Developing correlation equations and choosing limits for scatter (e.g. three sigma) for design, i.e. safety factors. 3. Enclosing scattered data with an enveloping curve and assuming that the maximum boundaries of this envelope gives conservative values. 4. Taking a Bayesian statistical approach where successive failures on a cumulative distribution give progressively more confidence to the shape factor and its extrapolation. 5. Using probabilistic fracture mechanics where the probability of occurrence of critically sized defects and the probabilistic evaluation of the critical stress intensity provide the basis for the probability of future failures [2-6]. 6. Applying a "fitness for service" approach where defects are assessed at some time during service, (e.g. an inspection period), and then assessing whether these defects can lead to potential failures. 7. Using the "corrosion-based design approach" (CBDA), as described by Staehle [7], involving the ten segments: environmental definition, material definition, mode definition, superposition, failure definition, statistical definition, accelerated testing, prediction, monitoring and inspection, and feedback. 8. Using the "Locations for Analysis" (LA~) approach as described by Staehle [7], where obvious locations containing multiple stressors of relatively intense values occur together. 9. Using physically based statistics for predicting the "First Failure" as described by Staehle [8,9]. Here, each of the statistical parameters is modeled using existing data for the seven primary variables as shown in Fig. 4 [ 1,9]. The final distribution then, includes the three, now physically dependent, statistical parameters with their respective dependencies.
Volume 2: Prediction, Industrial Developments and Evaluation
1. (E, potential variable) 9
I
4~l
I
I
2. and 3. (pH, x, pH and sjl~e!ies varia variables)
I
7
.
(C, metal composition variable) 1400
I
\
30(PC
@
"~~176
Alloy
1200
@ ~
10%NaOH, 325"C MA(II00"Cx 0.1 W.Q.) 0.~-x%Ni- 15-30%Cr %%@ o : x%Ni-15qbCr-Fe 9
2o
I
1o
\/sgorr
ii
9
Ecorr
{.,
~
I~
o
20o0.
"'~oo
Depth of SCC Penetration ~
\
looo-
\\\
[~o, ~oo~o
\
~
4
10 40 100 NaOH, g/l
~
x=A[lt+]n[x]p~me~
50O
0
f E-E~
1~
340
320
300
I
I
I
X 5 0 % NaOH
~ I/
~
|
AHoy 6OO MA
950
1000 1050 MA Temperature, *C
1100
!
~o
|
i
|
i
i
|
4O
3 1.4 kcal/mol
1o
r~
I
~o
7. (O, stress variable)
280
i=
r..) 0.0 /
i
,o
Q
Q = 69.0 kca]/mol
~ o.ioo.
A
e~tq
Temperature, *C
ioo
o t--.i 2.~r
N I C o n t e n t , w/o
(T, temperature variable)
.
(M, metal structure variable)
-- 15Cr
200
o
o
~'--'--"--~
"
I~I
~
~
PotenUal, Ecorr + mV
x~Ni-~-Feo
0.010-
_
0
Alloy 600 M A 10% NaOH at316"C
l & 2100 3~00 4 ~ 50100J00 70~0 Failure T h n e , hrs.
1.65
1.70 1.75 1000/T, *K
1.80
Fig. 4. Examples of dependencies on the seven primary variables for SCC as they apply to a correlation equation that might be used to model the statistical parameters for a statistical distribution. These examples are taken from data, which describe SCC in alkaline solutions. These data are discussed by Staehle and Gorman [ 1]. 9 2004 NACE International.
A practical question here, is how can failures that have not yet occurred be found? The essence of the approach in this discussion, the MPSA, for predicting the precursor conditions that must coalesce before SCC occurs, as indicated in Fig. 1, consists of six steps as follows: 1. The influences on materials can be divided into six domains and their respective microprocesses. The categories of the domains is somewhat arbitrary, but practical and convenient choices are: global environment, bulk environment, outside surface, protective surface layer, inside metal surface, and bulk metal. These domains, with examples of possible microprocesses, are identified in Fig. 5. These domains are discussed in Sections 3 and 4. 2. Identify the modes of failure of interest, (e.g. SCC, CF, FAC), and develop information bases for their dependencies of occurrence, as described in Section 6. These def'me targets for scenarios, as described in Section 5, which involve practical aggregations of domains and their microprocesses.
8
Prediction of Stress Corrosion Cracking
3. Identify the "microprocesses" of these domains that could affect the modes of failure. These microprocesses, as they apply for the six domains, are described in Section 4; examples are shown in Fig. 5. 4. Develop likely scenarios as suggested in Fig. 6 and Section 5, which connect microprocesses of the six domains, and which have a high likelihood for leading to failure. These scenarios would constitute the precursor period as identified in Fig. 1 for Cases II and III. 5. Quantify possibly critical microprocesses in terms of their dependence on the variables that lead to critical conditions for SCC to occur at the end of the precursor period. Note that the quantitative bases for most of these are available in the literature. 6. Develop critical experiments to assess whether the proposed scenarios and their component microprocesses are credible.
Fig. 5. Schematic view of six domains for quantifying microprocesses relating to the continuum from a global environment through to the bulk metal. Examples of microprocesses indicated.
Volume 2: Prediction, Industrial Developments and Evaluation
9
Predicting SCC in LWRs in the past has been hindered by overly restrictive and otten poorly informed assumptions on the microprocesses such as: 1. SCC occurs only in the presence of"specific ions." 2. SCC does not occur either in pure environments or in pure materials; i.e. SCC of Alloy 600 in pure deoxygenated water is not credible. 3. Boiling MgCI2, or a similar accelerated test, is a suitable environment to assess the dependence of SCC on alloy composition. 4. Pure water cannot produce SCC of sensitized stainless steel. 5. Water chemistry used for fossil boilers should be adequate for nuclear boilers. 6. Tube support crevices will not accentuate any chemistry that could lead to SCC. 7. There is not sufficient Pb in feed water to produce PbSCC even if it is concentrated. 8. SCC due to Pb in Alloy 600MA is transgranular. 9. Stainless steel without sensitization will not sustain SCC. 10. The high purity of water in secondary once-through steam generator (OTSG) water will not produce deposits on superheated upper bundle surfaces.
Fig. 6. Schematic view of a scenario that might be developed, which links microprocesses in successive domains as the scenario would apply to the precursor stage of Fig. 1.
10
Prediction of Stress Corrosion Cracking
There are more such assumptions. Rarely have engineers recognized and thought critically to question and test such assumptions. Similar assumptions may still hinder our capacity to predict performance. In Section 2, examples of failures, which have occurred and are already known, and which follow this microprocess route are considered and provide examples of important aspects of the MPSA. Section 3 describes domains and their inherent microprocesses; Section 4 describes physical details of microprocesses and their implications; Section 5 describes how scenarios are developed and applied; Section 6 describes the failure target (e.g. start of SCC after the precursor produces necessary conditions) for the development of scenarios; finally, in Section 7 some predictions for future and not yet observed damage are developed based on the procedures described in this discussion.
2. Examples of past sequential failures The purpose of this section is to describe examples of failures that correspond to Cases I and II of Fig. 1. These failures provide insights to how failures with longer precursor times might evolve. Section 2.1 describes failures that begin when the application starts and where the time-to-penetration is associated with the evolution of the SCC or FAC itself with no need to develop precursors over time. This corresponds to Case I of Fig. 1. Section 2.2 corresponds to Case II of Fig. 1 where a relatively short (e.g. 89to 20 years) precursor period precedes the evolution of the SCC itself. 2.1. Failures without time-dependent pre-conditions, no precursors, Case I 1. SCC on the primary side, L P S C C - no precursor: On the primary side of steam
generator (SG) tubes in pressurized water reactors (PWRs) there are no crevices, and the bulk chemistry is generally constant with time. The major stressors are residual stresses and temperature; thus, any SCC that occurs does not depend on accumulation processes in a precursor period. The highest residual stresses are initially present either at small diameter U-bends or at roll transitions at the top of the tubesheet. Fig. 7(a) shows data for the temperature dependence of LPSCC in the small diameter U-bends of SGs [ 10]. The earliest failures at the highest temperature, about 306~ occurred atter 20-30 months. Fig. 7(b) shows cumulative distribution functions (cdf) in Weibull coordinates for the LPSCC in US and French plants [11]. Here, the values of 0 (characteristic "space parameter" in the Weibull distribution) are in the range of 10 to 41 equivalent full power years (EFPY) and the values of,8 (slope or "shape factor" in the Weibull distribution) are in the range of 1.36 to 4.93 (noting that these data were analyzed with a two parameter Weibull fit). Such data suggest that the first failures occur in the range of about 0.1 of the mean. The data of Fig. 7, taken together and recognizing the differences in temperature, indicate that early LPSCC can occur in about a year. Details of dependencies of LPSCC are described by Staehle and Gorman [ 1]. 2. Local cold work on the secondary side, Case I - no precursor: Fig. 8 describes a situation in which relatively deep scratches, which were present at the start of the operation, produced local cold work that initiated extensive SCC. Staehle and Gorman [1] have summarized numerous such instances. Such SCC has penetrated the full thickness of the tubes five years after the start
Volume 2: Prediction, Industrial Developments and Evaluation
11
(a) Temperature, ~
lO3
315 T
310 !
_
305 !
300 !
295 !
290 !
285 I .......
_
Row 1 U-bends " Alloy 600MA
Q (apparent)~ 85 kcal/mol
102
l0 t
lOo
9 = No reported ECI or leakage 9 = Eddy current indications or leakage (first indication only) I I I I I 1 I
t
1.70
1.71
1.72
I
1.73 1 . 7 4 1.75 1 . 7 6 1 . 7 7 1 . 7 8 1.79 1000/T, K "I
0.90 0.700.50. 0.30. ~,
~- 0~0.
I
(b) I
!
I
I II _ At r
P r i m a r y side L P S C C
__
315-3 2 0 ~
0.01
0.001
0.oo01 0.01
,
, 0.1
,
121161
Service 1. F r e n c h plants, LTMA tubing 0 = 18.7 EFPY [3 = 4.48 r 2 = 0.82
_ _ ~ i "~__~'
~ 10
: 100
Time, EFPY
2. P a r t d e p t h 3. U . S . - m a d e r o w 2 roll U-bends, French plants 0 = 10.0 EFPY 0 = 41.1 EFPY ~ =4.29 ~ =3.07 r 2 = 0.95 r 2 = 0.95
4. U . S . . m a d e r o w 1 5. R o l l 6. P r i m a r y s i d e U-bends, French transitions denting plants for k i s s rolls 0 = 1.0.95 EFPY 0 = 7.13 EFPY 0 = 22.23 EFPY 1~= 4.93 = 4.29 l] = 1.36 r2= 1.0 r 2 = 0.77 r 2 = 0.98
F i g . 7. (a) T i m e - t o - f a i l u r e f o r R o w 1 U - b e n d s o f P W R S G s o n t h e p r i m a r y side. F r o m B e g l e y et al. [10]. S F E N , 1985. (b) P r o b a b i l i t y vs. s e r v i c e t i m e ( E F P Y ) f o r t h e L P S C C o c c u r r i n g o n t h e p r i m a r y s i d e o f t u b e s f r o m o p e r a t i n g S G s in P W R s . T e m p e r a t u r e s in t h e r a n g e o f 3 1 5 to 3 2 0 ~ F r o m S t a e h l e et al. [ 11 ]. 9 1 9 9 4 N A C E I n t e r n a t i o n a l .
12
Prediction of Stress Corrosion Cracking
of operation and was found on the surfaces of the cold leg of the secondary side. Hundreds of SCC events per inch occurred on some of the scratches. The fact that such SCC initiated and propagated on the free span of the cold side attests to the severity of the initial cold work. Fig. 8 shows first in (a) and then (b) the location of the failures [12]. Fig. 8(c) shows details of the scratch in plan view, and Fig. 8(d) shows the scratch in cross section from which the SCC proceeds [12]. Fig. 8(e) [13] and (f) [14,15] shows the accelerative effects of cold work, which supports the conclusion that local cold work started the cracks nucleating early. The scratches here seem to be deeper than ordinary scratches from processing. Fig. 8 shows also a scenario starting with M-1 ("M" identifies a "microprocess") through M-4, i.e. from the scratch through the cold work to the SCC, as accelerated by cold work and the SCC propagating as it moves beyond the cold work. However, here, this SCC certainly began at the start of life of the plant and the scratches accelerated the initiation. The relatively long time for perforation, compared with Fig. 7, results from the lower cold leg temperatures on which these SCC events occurred. 3. F A C at M i h a m a - no precursor: Fig. 9 shows a failed pipe in a section from Mihama 3 at the orifice between the #4 low pressure heater and the deaerator operating at approximately 142~ The pipe was 560 mm in diameter and the wall was 10 mm thick. The flow rate of water was 22 m/s at a pressure of 0.93 MPa and a pH of 8.6-9.3. The failure occurred after 185,700 hours of operation. The failure resulted from FAC at a location where there had been no inspections. The resulting failure killed five people and injured four others. The pipe was a thin wall large diameter pipe with high velocity water flowing at 22 m/s. The FAC involved in this accident most certainly started at the initial operation and did not involve any precursor period according to Case I of Fig. 1. Although the FAC proceeded at a relatively slow rate, no inspections were performed and no analysis of an acceptable operating period had been performed. 4. S C C in sensitized and non-sensitized stainless steels in B WR applications - no precursor: Fig. 10 shows three examples of data for SCC in stainless steels,
which are used in boiling water reactor (BWR) applications. These include sensitized Type 304, non-sensitized Type 304, and stabilized Types 321 and 347. Numerous instances of SCC in non-sensitized stainless steel have occurred [16-19]. In these cases, the local environment and condition of material do not change, although low temperature sensitization could aggravate the susceptibility to SCC. Fig. 10(a) shows the probability of cracking vs. time for SCC of 2" and 4" stainless steel piping that was sensitized at welds and exposed to B WR conditions [20]. Here, the probabilistic nature of the ultimate failures is clear, although the initiation most likely started with the beginning of operation of the plants. Fig. 10(b), from work by Angeliu et al. [21 ], compares sensitized and nonsensitized Type 304 stainless steel in B WR environments vs. the corrosion potential and shows that the crack growth rates for both heat treatments are not significantly different. The crack growth rate for the cold worked materials is increased.
Volume 2: Prediction, Industrial D e v e l o p m e n t s a n d Evaluation
(5. = ~
0). If pit nucleation cannot be regarded as being "instantaneous", thesimplest assumption concerning the pit nucleation rate, n(t) = dN/dt, is that n(t) is proportional to the number of available sites, N o - N(t) [5], which yields: N(t) = N 0 [ 1 - exp(-t/t 0)]
(14)
where to is some characteristic time that depends on the corrosion potential, temperature, and electrolyte composition. A more general and sophisticated calculation of the pit nucleation rate can be made in the following way. Although an extensive database does not exist to support this
62
Prediction of Stress Corrosion Cracking
position, it is postulated that the rate of nucleation of stable pits is proportional to the rate of nucleation of metastable pits, n ~ = (dN/dt)~, as
n(t)=(-~)
sP
= (d-~-)
(15) MP
where parameter ~ is the survival probability for a metastable pit and subscripts SP and MP designate stable pits and metastable pits, respectively. This parameter can be measured experimentally and work is currently underway to calculate the value of ~ in an ab initio manner. Thus, for example, for Type 304L stainless steel, the experimentally measured survival probability has a value of the order of, depending upon the potential and solution conditions 10-2 to 10-4 [21,22]. For calculating the rate of nucleation of metastable pits, the PDM for passivity breakdown can be used [2-5]. In accordance with this model for the constant external conditions we have: N(t)Mp = N 0 erfc(t + b)/erfc(b)
(16)
(ffaVapp) 2RT aXx/2 .
where a = ~/(B~/2ao), b=(Jm/B-D)/(xf2aD) and B=~uX/2exp
A detailed discussion of these equations is available in the original papers [2-5,23] to which the reader is referred. It is important to note that the PDM describes adequately many important properties of passive films. Thus, the PDM predicts that, for a given material, the mean critical breakdown potential, Vc, is a function of halide activity, ax, (e.g. CI-) of the form Vr = P - Q.log(ax)
(17)
where P and Q are constants. These constants can be expressed as explicit functions of the parameters listed in Table 1 [23]. Fig. 1 shows the mean critical breakdown potential for aluminum calculated by Eq. (6) ("PDM") and experimentally determined values by Kaeshe [24] and Bohni and Uhlig [25]. A particularly powerful confirmation of the validity of PDM for describing the pitting of aluminum alloy 2024-T6 was recently published by Fonseca et al. [26], who showed that the sweep rate dependence of the apparent breakdown voltage correctly yields the critical concentration of condensed vacancies at the metal/barrier layer interface, a finding that had been previously demonstrated for nickel [27]. This test of the model is considered to be particularly powerful, because it is an absolute test, in that it does not involve any adjustable parameters.
63
Volume 2: Prediction, Industrial Developments and Evaluation
Table 1 Values of parameters used in the calculating velocity of pit nucleation X f) w Jrn
aD ot 13
Parameters Passive film stoichiometry Mole volume of the oxide per cation Electric field strength Gibbs energy function Critical vacancy flux Critical concentration of vacancies Mean vacancy diffusion coefficient Standard deviation for vacancy diffusivity Constant Constant
Value 3 12.61 106 52 x 103 1.72 x 1012 1015 1019 0.5 x 1019 0.48 -0.01
Units cm 3
V/cm J/mol no./cm2s no. cm 2
cm2/s cm2/s V
-0.1 ILl
--.
-o.2
-0.3
E
m
o
-o.4
-0.5
9 o
Kaesche
9
POM Regression
Uhlig & Bohni
C m
-0.6
-,
o
L o g ( C [ Activity)
Fig. 1. Plot of the mean critical breakdown potential vs. log(Cl- activity) for aluminum in sodium chloride solutions at 25~
It is also important to note that in original version of PDM (Ref. [2] and citations therein), the pit nucleation rate was obtained from the criterion for pit initiation: (Jca - J m ) •
1:)->~
where Jca is the cation vacancy flux, Jm is the rate of annihilation of the metal vacancies at the metal/film interface, t is the observation time, x is the time of dissolution of the pit cap, and ~ is the critical "areal" concentration (#/cm 2) of condensed vacancies that are required for the separation of the barrier layer from the substrate metal. However, criterion Eq. (18) is valid only if all external parameters (temperature, electrolyte composition, pH, corrosion potential, etc.) do not depend on time, and, accordingly, Jca, Jm, and 8, are constants. In the general case, when the external conditions are timedependent, instead of condition Eq. (18), we must use the following criterion:
Prediction of Stress Corrosion Cracking
64
t
[(Jca (t') - Jm (t')]dt' > ~(t)
(19)
T
Details of the generalization of the PDM, for the case of variable external conditions, can be found in Ref. [5].
2.2. Prompt repassivation As noted above, only a small fraction of passivity breakdown events transition into stable pits (about 1 in 10000 for stainless steel in chloride-containing solution), so that stable pitting is a very rare event. The probability that a metastable pit will transition into a. stable pit is known as the survival probability (~), as noted above. The death of a metastable event, the probability of which is 1-~, occurs in a time of milliseconds to seconds by a process known as "prompt repassivation", which is to be distinguished from the "delayed repassivation" ("stifling") phenomenon that is responsible for the eventual death of stable fits at times that may extend to many years. Extensive work by Galvale [28] and Pistorius and Burstein [29], among others, has demonstrated that whether a pit nucleus survives to form a stable pit depends upon geometrical factors and the current density, but the models that were developed tend to be more phenomenological than mechanistic in nature. Nevertheless, the criteria for survival of the embryo are commonly expressed as the product of a geometrical function and the current density exceeding a critical value, which is a useful conceptual statement. There is little doubt that the failure of a passivity breakdown event to transition into a stable pit is due to the failure of the system to establish a viable differential aeration cell, which restricts the local anode predominantly to the nucleus and the local cathode predominantly to the external surfaces. Noting that passivity breakdown begins with the formation of a blister, due to separation of the barrier layer from the substrate metal, via cation vacancy condensation at the metal/barrier layer interface (Ref. [2] and citations therein), the cathode is already restricted to the external surface. Subsequently, as a result of stresses within the barrier layer, the blister fractures, thereby permitting the entry of solution into the embryonic cavity, where both anodic (metal dissolution) and cathodic .(hydrogen evolution) partial reactions are established. If these reactions occur at the same rate, no net acidification occurs within the cavity. However, if current can flow freely from within the cavity to the external surfaces, where it is consumed by oxygen reduction, the establishment of viable differential aeration is assured and the cavity will transition into a stable pit. If the flow of current is impeded, either because the perforation in the blister is not sufficiently large, or because the cathodic reactions occurring on the external surfaces cannot consume the current, viable differential aeration cannot be established and prompt repassivation occurs. Thus, any successful theory for prompt repassivation must address the kinetics of the reactions that occur on the external surface as well as the fracture of the barrier layer cap over the nucleation site.
65
Volume 2: Prediction, Industrial Developments and Evaluation
2.3. Rate of pit propagation The quantitative description of pit (or crack) growth can be regarded as one of the key problems in predicting corrosion damage in many practical systems. This follows from the fact that the calculated corrosion damage that is based only on this (growth) stage can be compared with experiment, in many limiting cases. For example when all pits nucleate "instantaneously", or when the induction time for pit nucleation is much smaller than the observation time, it is possible to ignore the initial stage of pit nucleation when estimating the damage. In addition, if the probability of survival of a corrosion defect is sufficiently high, we must take into account the possibility that a stable corrosion defect (pit or crack) nucleates immediately after the start of operation and propagates without repassivation. In any case, calculations based only on the growth stage yield the most conservative estimate of the service life, kmin, of the system. We can be sure that, if calculation of the service life is based on growth alone, the real service life, t, will at least be not less than kmm. It is well known fi'om both experiment and theory [30] that the dependence of the characteristic dimension of a corrosion cavity (for example, cavity depth, L) on time, t, can be expressed by a simple equation of the following form L=kt
(20)
TM
where k and m are constants, and, often, it is found that m = 1. Thus, it has been shown [20] that, if a single metal dissolution reaction (with the valence 3) takes place within the cavity, the growth rate of a sufficiently deep pit is described with great accuracy as a function of time by Eq. (20), with m-
~t+l 2~+1
(21)
and ~+1 8 1 k = (2~t+l M m ~t + 1 Pm 4F)2~+1 (4FDmCb)2ti+l (i)2&+l
(22)
where & is the anodic transfer coefficient, Dm, Mm, and Pm are the diffusion coefficient, the atomic weight, and the density of the metal, respectively, Cg is the total bulk concentration of anions, and i is the current density calculated in the absence of a potential drop in the cell (the maximum possible current density on the pit internal surface at the given potential of the metal). However, this dependence of L on t cannot be used directly in mathematical calculations for small times, because of the non-physical limit v = d L =kmt mq - ~ , dt
att~O
andm 0 (Eq. (7)), with the ratio between the two being determined by the value of the delayed repassivation constant. This ratio is greater than zero (i.e., dead pits exist), provided that ), > 0. As the living pits grow to greater depths, the depths at which pits die also become larger and at some point a dead pit will exist at a sufficient depth for a crack to nucleate, because K~ exceeds K~scc. Of course, if the value of), is sufficiently high no pits may attain the critical depth before they die and cracks will not nucleate. However, those dead pits that are subcritical at low stress intensity might become supercritical at higher stress intensity (due to an increase in the load or due to a decrease in Kiscc, for example) and hence crack nucleation may still occur. One of the great attributes of DFA is that it allows the calculation of the population of dead (passivated) pits as a function of )' and other system variables through Eq. (9) and hence provides an analytical description of crack initiation. In any event, it is evident, from the above discussion, that a critically important parameter in understanding the nucleation of cracks from pits is the delayed repassivation rate constant, ~. To date, no theory has been developed to calculate ~, in an ab initio manner, but that process is now underway with regard to the prediction of localized corrosion damage on Alloy-22 canisters in the Yucca Mountain high level nuclear waste (HLNW) repository.
3. Modeling crack growth Many models have been developed for describing and predicting the propagation of cracks in metals and alloys in contact with aqueous solutions, but most have emphasized mechanical processes rather than electrochemical phenomenon. Additionally, most models are empirical in nature, in that they do not incorporate clearly def'med mechanisms for the electrochemical propagation of cracks and the predictions are not constrained by the natural laws. In many cases, "calibrating factors" are incorporated that have neither physical meaning nor the possibility of being determined by independent experiment. Although calibration is a legitimate means of determining the value of these factors (provided that they have a well def'med physical meaning and provided that the value is constrained to a range that is physically realistic), some models are evaluated against essentially the same data that were used in the calibration! Indeed, in approaching this whole issue, we should heed Albert Einstein's admonishment that theories should not contain ad hoc additions simply to "make them work". Models that do contain such factors can never be deterministic, no matter what level of massaging is applied. 3.1. Deterministic calculation of lGSCC growth rate
The increasing incidence of IGSCC in sensitized stainless steel components in BWR heat transport circuits (HTCs) has led to the development of radiolysis models for ~alculating the concentrations of electroactive species, such as hydrogen (H2), oxygen (O2), and hydrogen peroxide (H202), as a function of the reactor operating parameters and the concentration of hydrogen added to the feedwater [6-9,36-38]. Furthermore, over the past several years, we have developed powerful codes for calculating ECP and for predicting the damage that accumulates from IGSCC in BWR primary coolant
70
Prediction of Stress Corrosion Cracking
circuits [6-14]. Because 02, H2, and H202 are electroactive, they are instrumental in establishing the ECP of components within the HTC [6-9,39]. Extensive work in many laboratories worldwide has established that sensitized Type 304 SS becomes increasingly susceptible to pitting corrosion, intergranular stress corrosion cracking (SCC), and corrosion fatigue in high temperature aqueous solutions as the ECP is increased above critical values [40]. Constant extension rate tests (CERTs), using round tensile specimens in actual BWR coolant at 288~ [41 ], has led the Nuclear Regulatory Commission (NRC) to adopt a value for the critical ECP (Eicscc) for the occurrence of SCC o f - 2 3 0 mVsHE, although we note that critical potentials as negative as -400 mVsHE have been observed in laboratory studies [40]. A distribution in E~scc in an operating reactor is expected, because of the variability in the degree of sensitization of the steel at welds and because of differences in neutron fluence experienced by invessel components. Because SCC occurs only when ECP > Eicscc [40], a principal goal of water chemistry control protocol for inhibiting cracking in BWR coolant circuits is to displace the ECP to a value that is more negative than the critical value for the component of interest under the prevailing conditions, as noted previously in this paper. In hydrogen water chemistry (HWC), molecular hydrogen is added to the feedwater with the objective of reducing the concentrations of oxidizing species (e.g., O2, H202), and of displacing the ECP in the negative direction. (Note, however, that a reduction in the concentrations of 02 and H202 is not a necessary condition for the displacement of the ECP in the negative direction upon the addition of hydrogen, as shown by mixed potential theory). The original code (DAMAGE-PREDICTOR) [6-9] incorporates deterministic modules for estimating specie (in particular, H2, O2, and H202) concentrations, the ECP [39], and crack growth rate [13-19] of stainless steel components at closely spaced points around the coolant circuit, as a function of coolant pathway geometry, reactor operating parameters (power level, flow velocity, dose rates, etc.), coolant conductivity, and the concentration of hydrogen added to the feed water. The radiolysis module, RADIOCHEM, which is employed to calculate the specie concentrations, is based on a model that was originally developed to describe the corrosion of high-level nuclear waste containers [10]. This model was subject to extensive analysis and predictive evaluation, which involved tracing the reactions contained within the model to their original sources, and in ensuring that the model could reproduce the original observations. The ECP model, the Mixed Potential Model (MPM), which others are now using, makes use of the fact that, for a system undergoing general corrosion (which is the process that establishes the ECP), the sum of the current densities due to all charge transfer reactions at the steel surface must be zero [39]. By expressing the redox reaction currents in terms of the generalized Butler-Volmer equation, which incorporates thermodynamic equilibrium, kinetic, and hydrodynamic effects, and by expressing the corrosion current in terms of either the PDM [2] or as an experimentally derived function [39] (both have been used), it is possible to solve the charge conservation constraint for the ECP. The MPM has been extensively tested against experimental and field data and has been found to provide accurate estimates of the ECP [ 17,42]. For example, Fig. 3 shows a comparison of calculated and measured ECP, [H2], and [02] measured in the recirculation piping of the Leibstadt BWR in Switzerland as a function of the feedwater hydrogen concentration [42]. Given that the potential measurements are probably no more accurate than _+50 mV, the level of agreement obtained in this "blind" HWC mini-test between measured and calculated ..
Volume 2: Prediction, Industrial Developments and Evaluation
71
Redrculation 300
I
.
.
.
.
I
'
9
'
9
I
"
"
"
I 9 9
200
"
I
.
.
.
.
I
I
i
Measured(accuracy = 50 mY) Calculated(accuracy = 40 mV)
I
.,
100 0 > E v o. o100
A
0 IZl
-2OO -300 -400 "600
I
9
,
,
= .... 1
0.0
.
.
.
.
I
0.5
,
,
,
,
I
1.0
,
,
,
,
I
1.5
2.0
Feedwater Hydrogen Concentration (ppm) 1000
I
'
'
'
'
i
'
'
'
'
I
.
.
.
.
I
t
.
.
.
t
t
I
,0o
.
"
i9
8
9
o
Calculated
I
c
-r
10
c
I
,
,
,
i
0.0
I
i
i
i
i
0.5
I
|
,
i
i
I
1.0
i
i
,
i
I
1.5
2.0
Feedwater Hydrogen Concentration (ppb)
'i ......... 'i'"' 9 9
JO
100
Measured Calculated
c
c
~
~0
o c
1
9 ,
I
0.0
,
,
9
|
I
0.5
9
l .
.
.
.
I
1.0
|
.
9 ,
,
I
1.5
9 ,
,
i
|
|
2.0
Feedwater Hydrogen Concentration (ppm)
Fig. 3. Comparison of measured and calculated ECP, [H2], and [02] in the recirculation piping system from a "double blind" HWC "mini-test" at the Leibstadt BWR [17].
Prediction of Stress Corrosion Cracking
72
ECP is excellent. Equally impressive agreement was obtained for the hydrogen and oxygen concentrations (Fig. 3, lower plots). Finally, the deterministic Coupled Environment Fracture Model (CEFM) [11-14] estimates the rate of growth of an existing crack in Type 304 SS of a given degree of sensitization under any given set of environmental conditions. The CEFM is deterministic, in that the predicted crack growth rate is constrained by the relevant natural law; the conservation of charge. Furthermore, a basic premise of the CEFM, that current flows from the crack and is consumed on the external surface (Fig. 4), which is embodied in the differential aeration hypothesis for localized corrosion, has been demonstrated experimentally [43]. To our knowledge, the CEFM is the only currently available model that satisfies the conservation of charge constraint explicitly. The high degree of determinism is demonstrated by the fact that the model can be calibrated by a single crack growth rate/ECP/conductivity/temperature/stress intensity/flow velocity datum for a given degree of sensitization of the steel [43,44]. F1 ui d Flo~,
""-
Ox ~ge n T r a n s p o r t
Post t i v e C u r re nt
Post t i v e C u r re nt
02
.,..
20~
,L
Creek Advance
Fig. 4. Coupling of crack internal and external environments. Note that in the steady state, the crack can grow only as fast as the positive current flowing from the crack can be consumed on the external surfaces by oxygen reduction.
As noted above, IGSCC in sensitized austenitic stainless steels in BWR coolant circuits is caused by an excessively high (positive) ECP, due to the presence of oxidizing radiolysis products (02, H202) in the coolant water, and an excessively low Emscc, due to sensitization. In order to reduce the ECP, hydrogen, in the ppm concentration range, is being added to the feedwater of many currently operating B WRs in a strategy referred to as "HWC" [45], which was referred to above. However, HWC has proven to be only partially successful in alleviating IGSCC, because hydrogen is not sufficiently strong as a reducing agent [39]. This issue is illustrated in Fig. 5, where
73
Volume 2: Prediction, Industrial Developments and Evaluation [H2]FW = 0.0 ppm 0.40
'
9
'-
I
'
'
'
J, UP
'
I'
'
'
'
MP uDLD
9
I
.
.
.
.
I
.
.
.
.
I
'
'
9
RS
'
I
'
'
'
'
j p BLP TLP
0.30 0.20 0.10 0.00 ~. -0.10
CC : Core Channel CB : Core Bypass UP : Upper Plenum MP : Mixing Plenum UD: Upper Downcomer LD : Lower Downcomer RS : Recirculation System JP : Jet Pump BLP : Bottom Lower Plenum TLP : Top Lower Plenum
n
oU3 -0.20
-0.30 -0.40 -0.50 -0.60 -0.70
I
'
'
9
I
,
.
.
,
1000
0
I
.
.
.
.
2000
I
,
,
,
3000
,
I
,
,
.
4000
,
I
.
.
.
.
I
5000
.
.
.
.
6000
7000
Flow Path Distance from Core Inlet (cm)
0.40
....
, .... .I. UP MP
0.30 0.20 -
[H2IFW = 1.2 ppm , .... , .... , .... CC" Core Channel CB" Core Bypass UP" Upper Plenum RS MP 9Mixing Plenum UD" Upper Downcomer ~ LD" Lower Downcomer RS 9Recirculation System JP 9Jet Pump
, ....
,_ ~ 'J. ,, UD LD [""~"~ I ~
0.10 -
~
~
0.00
~i
, .... jp
BLP
,I,T TLP .J.
ol~om LowePrlenum
Top Lower Plenum
-0.10 r
oLU -0.20 -0.30 -0.40 -0.50
-0.60 -0.70
i
i
i
*
I
i
1000
!
i
i
I
2000
.
.
.
.
I
.
3000
.
.
.
I
4000
.
.
.
.
l
5000
.
.
.
.
I
6000
.
.
.
.
7000
Flow Path Distance from Core Inlet (cm)
Fig. 5. Predicted ECP vs. flow path distance from the bottom of the core for 0 and 1.2 ppm of hydrogen added to the feedwater of the Leibstadt BWR [17]. Crack growth rate/ECP/ conductivity/temperature/stress intensity/flow velocity datum for a given degree of sensitization of the steel [43,44].
74
Prediction of Stress Corrosion Cracking
predicted ECP is plotted as a function of distance along the primary coolant path for normal water chemistry (NWC) and HWC at a feedwater hydrogen concentration of 1.2 ppm. Noting the NRC's critical potential for IGSCC in sensitized Type 304 SS (-230 mVsnE at 288~ [41 ], we see that no component in the coolant circuit is predicted to be immune under NWC, whereas significant protection if predicted for HWC. However, important regions are predicted to be still unprotected even at this very high feedwater hydrogen level, including much of the downcomer, most of the core bypass, the upper plenum, parts of the lower plenum, and the jet pumps. The situation isnot so critical if the crack growth rate is considered instead (Fig. 6). Thus, with regular inspection, a crack growth rate of 5 x 10-9 cm/s (62 mpy) could probably be tolerated. If so, then part of the recirculation system is not at risk, even under NWC conditions, and most of the coolant path is protected under HWC at a feedwater hydrggen concentration of 1.2 ppm. Indeed, using the ECP = -230 mVsHE as the protection criterion essentially requires that no IGSCC can occur, whereas slowly growing cracks ( 0.001 D" U.l 0.000 0.01
0.1
Distance from crack tip, r (mm)
(a) 0.005 == E
.,n
t
- , , - A a=0.005 mm - ~ A a=0.010 mm - - - A a=0.016 mm
0.004
z_
.o
0.003
Q. ~., 0.002 G
t~ 9>:3 0.001 o" tu 0.000 0.01
,
i
i
,
i
,
" A A A
9
A,A
A
A
--,
,A,
,J
0.1
1
Distance from current crack tip, r (mm)
(b) Fig. 9. Distribution of e.p along a crack tip before and after crack propagation. (a) The end of Kincreasing stage and (b) crack propagation stage at K = 35 MPa.m u2.
4.2. Correction o f strain distribution based on FEM analysis
In previous calculations, the fl value for Gao-Hwang's equation was assumed to be 5.08 or a similar value. However, the theoretical gradient did not agree well with the simulation result in Fig. 1. Therefore, we decided to evaluate the influence of n on fl numerically. The 1T-CT specimen, with E = 183 GPa, YS = 436 MPa, and load = 14 kN (K = 24.2 MPa.m~/2), was employed in the analysis. Five ideal stress-strain properties with n = 1.29, 2.0, 3.0, 5.0 and 7.0 were used for the analysis. Because O'y,E and n have fixed values, fl could be obtained from the values of the gradients. Fig. 11 shows the variation of fl with n. Using the new value of fl obtained in this analysis (0.0921) for the theoretical gradient in the previous RCT model showed a better agreement with the simulation result, as shown in Fig. 12.
118
Stress Corrosion Cracking in LWR Environments 0.020
"~ 0.015 I,.
-n-
ro=10 i~m
-o-
ro=20 l~m
_ A _ ro=30 I~m
W
/
O ,m (n
9
9
co 0.010 D. p, m
>
0.005
o" I,LI
"l"l/l
t o o l . . 0.000 0.000
9
i/i/O/
n
9
I
9
,
9 I
0.005
i j "
~
0.010
I
0.015
0.020
Crack length increment, A a (mm) (a) A
3.0X10 "7
E
/I
E 2.5x10 7 ~
/
2.0x10.7
Z
l~m
-A-
ro=10
-o-
ro=20 l~m
- n n - ro=30 l~m
/
d m 1.5x10 7
~O-"-
9~
;9
oJ O~
~ 1.0x10 7 o
-' o.o/ .~ 5.0x10 "8 0 m !._.
o
0.0 0.0
9 9
I 9
..,
.~I~
9
O 9
A
A
i -
A/
g
,
5.0x10 4
~/ A~
,A"
I
1.0x10 .7
L
I
J
1.5x10 -7
2.0x10 -7
Crack tip strain rate, dEctldt (l/s) (b) Fig. l O. Relationships between: (a) crack length increment, Aa, and crack tip plastic strain, ev, and (b) da/dt and d~r
5. Interpretation of crack growth under or beyond SSY
Results of CGR vs. CTSR for sensitized 304 SS with or without warm rolling, for alloy 182 weld metal in simulated B WR environments, and for alloy 600 in simulated PWR environments showed a consistency between crack growth and the deformation/oxidation interaction represented by CTSR. It has been reported that deformation by cold or warm rolling can significantly increase CTSR and thus accelerate SCC growth in both BWR and PWR environments, for both stainless steels and.Ni-base alloys [2,4]. In the present study, sensitized 304 SS with the higher YS produced by warm rolling exhibited lower CGR than those of their as-sensitized counterparts at similar levels of K (Fig. 3). This might have arisen from the plasticity
Volume 2: Prediction, Industrial Developments and Evaluation
119
8
7
........................................................................................................... i ........................................ ~
A
6 ...................................................................................
~.
~
"
..................
4
3
2-
1
0-
.
,
i
;
1
2
3
i 4
i
i
5
6
7
8
n G
Fig. 11. Variations offl with n based on FE analysis.
=
0.5
/ Theoretical
I .......1o. I /~i
0.4
Gradient
( 13 = 0.0921
i i i i i!l i ii}iil
............................................
= .,..~
-
0.3
O .,..~
0.2
i
i
i ! i ~
K
"-4L
!
",,\ !~.i
i
i
~o no
)
9
i,-
~ ~
i i i l-;-)~--30:; ~..Ai
~ ' ~ ~
i l-'-K=25-~
iil i
","**~
i
= - - K = 15.4 ......
i.i.............
)i i :
i iiiil
ii~-..~ ~"-+._i !
........................i.........i.......i......i..-+.---.=-i............................ i.......-A.-i..........i.......}-.--.T:---i---i.............i > .,..~
0.1
0-~
'............................................ ' I ....... i!.................i.... i..i...Li ......................... i................ ~il
ii~
0.0 0.01
0.1 Distance
1
from Crack Tip, r (mm)
Fig. 12. Modified theoretical gradient for RCT model.
effect, namely, the plane strain and plane stress constraint condition. Because RCT specimens fabricated from as-sensitized 304 SS with the YS o f 174 MPa (low value) did not meet S S Y conditions at experimental levels of K, CGRs were high. Sensitized and warm-rolled 304 SSs have higher Y S that m o v e s the test specimens made o f the steels closer to S S Y conditions. Thus, C G R values are comparable to other K-valid
120
Stress Corrosion Cracking in L WR Environments
CGR data. A similar phenomenon has been observed for 304 SS, where CGR values for RCT specimens were much higher than those for 1T-CT specimens [ 13]. These results implied that YS had a dual effect on SCC growth rates: on the one hand, high YS could increase CGR by increasing CTSR under plane strain conditions, or, on the other hand, it could maintain the low CGR by preventing the transition from plane strain to plane stress condition. Interpretation of the effect of K on CGR for alloy 182 in simulated BWR environments at different dissolved oxygen (DO) levels and on CGR for alloy 600 in simulated PWR environments might also be based on quantification of CTSR, as shown in Figs. 5-8. Decreasing m with increasing DO or with increasing electrochemical corrosion potential (ECP) for alloy 182 clearly showed that the crack tip repassivation abilily during crack growth decreased with increasing DO or ECP. The values of non-K-valid data for Ni-base alloys are not much higher than the values for Kvalid data, in either simulated BWR or PWR environments. Some non-K-valid CGR data for alloy 600 in simulated PWR environments appeared to be lower than the values for K-valid condition data calculated by the modified Scott's model. For alloy 182, some non-K-valid data are lower than K-valid data at similar K levels. These results suggested the possibility that the actual K-validity criteria for EAC tests might dependent on material-environment combinations. Sato and Shoji [ 14] observed that the effects of plasticity in terms of specimen thickness on CGR were quite different for stainless steels in simulated BWR and PWR environments. It was postulated that a variation of crack-tip water chemistry with specimen size might affect CGR as well. Another possibility might arise from the criteria for SSY. It has been suggested that, for low YS materials where strain hardening may occur at the crack tip, some margin exists between actual K-valid SCC conditions and those designated by ASTM E-399-95 [13,15]. Flow stress, which is the mean of YS and ultimate tensile strength, can be used according to ASTM E-647 [16] even though it was originally meant for fatigue crack growth tests. There might not be such a margin for high YS materials. For irradiated materials, where "work softening" can occur, the effective stress must be discounted for the elevation of YS due to irradiation [13]. There was not much enhancement of CGR for apparently non-K-valid specimens, as shown in Figs. 7 and 9. This could be due to the fact that these specimens might not be far from the plane strain condition when the K-validity criteria for SCC tests are used. 6. Conclusions
CGR for austenitic alloys in simulated BWR and PWR environments were interpreted based on crack tip plasticity quantification by both theoretical formulation and FE analysis. The crack tip deformation/oxidation mechanism for crack advance was represented by a unique parameter, CTSR, in conjunction with fundamental oxidation kinetics. FE analysis provided more appropriate results for crack tip stress/strain distribution, especially for LSY conditions where LEFM analytical methods could not be applied directly. The CGR data generated under K-valid and non-K-valid conditions were analyzed. Differences in the effects of crack tip plasticity on CGR were found for different material-environment combinations. Results of FE analysis showed that ev, Aa, and da/dt increased with increasing CTSR. The importance of r was also emphasized by FE analysis due to its effects on both the absolute value and the distribution of crack tip strain. The theoretical crack tip plastic strain equation was modified based upon the FE simulation results. The dependence of fl on the strain hardening coefficient was
Volume 2: Prediction, Industrial Developments and Evaluation
121
proposed, and better agreements were observed at~er correction of fl values based on crack tip plasticity analysis.
Acknowledgements This work was financially supported by the Japan Nuclear Energy Safety (JNES) Organization through the contract of Study of Safe-Related Advanced Technology for Aged NPP (SAT). Part of this work has also been supported by the Grant-in Aid for COE Research (No. 11 CE2003), MEXT.
References [1] P.M. Scott, P. Combrade, On the mechanism of stress corrosion crack initiation and growth in alloy 600 exposed to PWR primary water, in: Proc. 1l th International Conference on Environmental Degradation of Materials in Nuclear Power Systems-Water Reactors, American Nuclear Society, Chicago, 2003, pp. 29-36. [2] T. Shoji, Progress in the mechanistic understanding of BWR SCC and its implication to the prediction of SCC growth behavior in plants, in: Proc. 1l th International Conference on Environmental Degradation of Materials in Nuclear Power Systems-Water Reactors, American Nuclear Society, Chicago, 2003, pp. 588-598. [3] T. Shoji, S. Suzuki, R.G. Ballinger, Theoretical prediction of SCC growth behavior, in: Proc. 7th International Symposium on Environmental Degradation of Materials in Nuclear Power Systems-Water Reactors, NACE International, Houston, 1995, pp. 881-889. [4] T. Shoji, Z.P. Lu, Q.J. Peng, Experimental and numerical approaches for characterizing the crack growth behavior of alloy 600 in PWR primary water, and life time predictions for welded structures, in: T.S. Mintz, W.H. Cullen, Jr. (Eds.), Proc. Conference on Vessel Penetration Inspection, Crack Growth and Repair, vol. 1, NUREG/CP-0191, Nuclear Regulatory Commission, Washington, DC, 2003, pp. 309-335. [5] Y.C. Gao, K.C. Hwang, Elastic-plastic fields in steady crack growth in a strain-hardening material, in: D. Francois (Ed.), Advances in Fracture Research: Fracture 1981, vol. 2, Pergamon Press, Oxford, 1981, pp. 669-682. [6] W.W. Gerberich, D.L. Davidson, M. Kaczorowski, J. Mech. Phys. Solids 38 (1990) 87-113. [7] S. Namatame et al., Proc. Zairyo-to-Kankyo A201 (2000) 35-38. [8] M. Itow, Y. Abe, H. Sakamoto, S. Hida, K. Takamori, S. Suzuki, The effect of corrosion potential on alloy 182 crack growth rate in high temperature water, in: Proc. 8th International Symposium on Environmental Degradation of Materials in Nuclear Power Systems-Water Reactors, American Nuclear Society, La Grange Park, 1997, pp.712-719. [9] 2001 Annual report: Evaluation Technology for Stress Corrosion Crack Growth of Ni Base Alloy, JAPEIC, Tokyo, March 2002. [ 10] 2002 Annual report: Evaluation Technology for Stress Corrosion Crack Growth of Ni Base Alloy, JAPEIC, Tokyo, March 2003. [11] 2003 Annual report: Evaluation Technology for Stress Corrosion Crack Growth of Ni Base Alloy, JAPEIC, Tokyo, March 2004. [12] P. Scott, C. Benhamou, An Overview of recent observations and interpretations of IGSCC in nickel base alloys in PWR primary water, in: P. Ford, G. Was, L. Nelson (Eds.), Proc. 10th International Conference on Environmental Degradation of Materials in Nuclear Power Systems-Water Reactors, NACE International, Houston, 2001. [13] P.L. Andresen, K/size effects on SCC in irradiated, cold worked and unirradiated stainless steel, in: Proc. 1lth International Conference on Environmental Degradation of Materials in Nuclear Power Systems-Water Reactors, American Nuclear Society, Chicago, 2003, pp. 870-886.
122
Stress Corrosion Cracking in LWR Environments
[14] T. Sato, T. Shoji, Effects of specimen size and thickness on CGR in high temperature waters, in: Proc. 1 l th International Conference on Environmental Degradation of Materials in Nuclear Power Systems-Water Reactors, American Nuclear Society, Chicago, 2003, pp. 862-869. [15] Standard test method for plain-strain fracture toughness of metallic materials, ASTM E39990, 1995 Annual Book of ASTM Standards, vol. 03.01, ASTM, Philadelphia, 1995. [16] Standard test method for measurement of fatigue crack growth rates, ASTM E-647-95, 1995 Annual Book of ASTM Standards, vol. 03.01, ASTM, Philadelphia, 1995.
123
The role of hydrogen and creep in intergranular stress corrosion cracking of Alloy 600 and Alloy 690 in PWR primary water environments- a review Fred H. Hua a, Ra61 B. Rebak b a Bechtel SAIC Company, LLC, 1180 Town Center Drive, Las Vegas, NV 89144, USA b Lawrence Livermore National Laboratory, 7000 East Ave., MS 631, Livermore, CA 94550, USA
Abstract Intergranular attack and intergranular stress corrosion cracking (IGSCC) of Alloy 600 in pressurized water reactor (PWR) steam generator environment has been extensively studied for over 30 years without rendering a clear understanding of the essential mechanisms. The lack of understanding of the IGSCC mechanism is due to a complex interaction of numerous variables such as microstructure, thermomechanical processing, strain rate, water chemistry and electrochemical potential. Hydrogen plays an important role in all these variables. The complexity, however, significantly hinders a clearer and more fundamental understanding of the mechanism of hydrogen in enhancing intergranular cracking via whatever mechanism. In this work, an attempt is made to review the role of hydrogen based on the current understanding of grain boundary structure and chemistry and intergranular fracture of nickel alloys, effect of hydrogen on electrochemical behavior of Alloy 600 and Alloy 690 (e.g. the passive film stability, polarization behavior and open-circuit potential) and effect of hydrogen on primary water SCC (PWSCC) behavior of Alloy 600 and Alloy 690. Mechanistic studies on the PWSCC are briefly reviewed. It is concluded that further studies on the role of hydrogen on intergranular cracking in both inert and primary side environments are needed. These studies should focus on the correlation of the results obtained at different laboratories by different methods on materials with different metallurgical and chemical parameters.
I. Introduction Intergranular (IG) attack and IG stress corrosion cracking (IGSCC) of Alloy 600 in pressurized water reactor (PWR) steam generator environment has been extensively studied for over 30 years without rendering a clear understanding of the essential mechanisms [1]. The lack of understanding of the IGSCC mechanism is due to a complex interaction of numerous variables such as microstructure, thermomechanical processing, strain rate, water chemistry and electrochemical potential [1 ]. As far as the materials are concerned, most studies have been conducted with commercial alloys,
124
Stress Corrosion Cracking in LWR Environments
which not only had heat-to-heat differences in composition but also experienced different thermomechanical histories. Among many metallurgical and chemical variables that may influence the behavior of grain boundaries, the grain boundary precipitation, the segregation of impurities to grain boundary, the depletion of chromium near the grain boundary and the oxide film stability have received extensive studies. The complexity significantly hinders a clearer and more fundamental understanding of the mechanism of hydrogen in enhancing intergranular cracking via whichever mechanism. Further studies on the role of hydrogen on IG cracking in both inert and primary side environments are needed and should focus on the correlation of the results obtained at different laboratories by different methods on materials with different metallurgical and chemical parameters. This review is focused on the role of hydrogen and creep in assisting IGSCC of Alloy 600 and Alloy 690 in nuclear steam generator primary side environments. A comprehensive review was recently published on the different submodes of SCC and the influencing variables from the secondary side [2].
2. Grain boundary structure and chemistry of relevant nickel alloys Although the chemical composition of Alloy 690 does not differ significantly from that of Alloy 600, the resulting microstructure and IGA behavior can be very different [3,4]. For instance, Angeliu and Was [3] showed that Alloy 600 type materials exhibited M7C3 on the grain boundary and M7C3 and M23C6 intragranularly, while Alloy 690 type of materials only exhibited M23C6precipitation. It is also reported that the Alloy 690 type of materials exhibited more Cr depletion than Alloy 600 type on an absolute scale but less Cr depletion when the comparison was relatively made (e.g. shown by percentage of Cr depletion). This is consistent with the thermodynamic calculation, which showed that less severe Cr depletion should be expected with precipitation of Cr23C6than with Cr7C3 [5]. 2.1. The role of intergranular carbide precipitations
Bruemmer and Henager [6] assumed that grain boundary carbide precipitation acts as low energy dislocation sources, producing more homogeneous plastic deformation and modifying the local stress state. The presence of a semi-continuous array of IG carbides may continuously blunt the crack tip and reduce the SCC crack propagation. Hertzberg and Was [7] separated the effect of carbon in solution from the effect of carbon present as grain boundary carbides and demonstrated better resistance to IGSCC for materials with grain boundary carbides than with carbon in solution. 2.2. Effect of grain size
It has been found [8-10] that coarse-grained Alloy 600 (ASTM 8). Sung and Was [ 11] found IGSCC of a simulated Alloy 600 in 360~ high purity water was dominant for small grain sized samples (30 ~tm) over large grain-sized samples (130 lam). The effect of small grain size was to enhance diffusional creep processes and decrease grain boundary segregation of impurities due to a larger grain boundary area per unit volume [12]. The effect of hydrogen on cracking of simulated
Volume 2: Prediction, Industrial Developments and Evaluation
125
Alloy 600 materials, however, was found controversial [11,13,14]. For small grainsized materials (30 ~tm), increasing hydrogen overpressure from 100 to 200 kPa increased the % of IG fracture by 61%, while for large grain-sized materials (130 ~tm), the % of IG fracture decreased from 13.9 to 6.1%. On the other hand, Symons' data [ 15] for hydrogen-charged and uncharged Alloy 690 specimens in air showed that aging (TT) after solution annealing (MA) increased the grain size and grain boundary carbide precipitation and, therefore, increased the propensity of IG cracking. Hydrogen is more effective in embrittling the material when the carbide size and carbide coverage increases. Aging Alloy 690 increased grain boundary carbide size and the extent of grain boundary M23C6carbide coverage. Therefore, aging increased the degree of hydrogen embrittlement of the alloy [ 15]. The results from fracture toughness testing showed that J~c decreased by hydrogen charging from about 480 kJ/m2 for uncharged specimens to about 210 kJ/m for solution annealed materials. For solution annealed followed by aged materials, J~c was further reduced to 90 kJ/m2 [15]. The morphological study showed that as the hydrogen content increased, the fracture morphology changed from mainly transgranular ductile fracture for uncharged MA materials to mainly IG brittle cracking for TT charged materials [ 15]. The two contributing factors explaining the fact that the increased grain size is responsible for increased hydrogen embrittlement are: (i) increased grain size increases the local stress at the particles, which would result in a lower strain necessary to decohere the grain boundary/carbide interface, as proposed by Goods and Brown [ 16]; and (ii) hydrogen is preferentially trapped at the carbide/matrix interface, reducing the strength of the interface more than at the matrix/matrix interface, as proposed by Young and Scully [ 17]. 2. 3. Effect of carbon content
Sung and Was [11] observed that addition of 300 ppm carbon in solution to a controlled purity Ni-16Cr-9Fe alloy significantly increased the strength and resistance to IGSCC in 360~ high purity water. The main effect of carbon doping to Ni-16Cr-9Fe type and Ni-30Cr-9Fe type nickel alloys was found to enhance the formation of the corrosion products, in particular the Ni(OH)2. The carbon-free Alloy 600 was found to have a Cr203-rich film. Doping carbon increased the film thickness from 38 to 281 nm with a large amount of increase in Ni(OH)2. In the case of Alloy 690, doping carbon increased the film thickness slightly from 46 to 58 nm with somewhat more Cr and Fe found in the film [18]. 2.4. Effect of chromium content
Increased Cr content up to ~19% in Alloy 600 and Alloy 750 did not appreciably change their microstructure, such as the grain size and distribution of inter-and intragranular carbide precipitation, in hydrogenated primary water at 350-360~ and in steam at 400~ [ 19]. The mechanical properties were not significantly changed by the increased chromium content either. However, the increased Cr content significantly improved the resistance to IGSCC in hydrogenated primary water. On the other hand, the increased Cr content did not significantly improve the SCC resistance to caustic SCC [19]. It was concluded that the increased resistance to SCC of Alloy 600 by
126
Stress Corrosion Cracking in LWR Environments
increasing Cr content could not be attributed to the changes in microstructure but to the improvement in passive film [19]. Noel et al. [20] by using ehronoamperometric method, showed that the increased Cr content only slightly improved the repassivation ability of the passive film of nickel alloys in primary water and the improvement seemed to be too small to explain the significant improvement in primary water SCC (PWSCC) resistance of Alloy 690 over Alloy 600. Using constant extension rate test (CERT) and constant load methods, Angeliu et al. [ 18] showed that in both argon and high purity hydrogenated water (18 MD and 16 cc/kg hydrogen, 0.1 MPa, O2 < 10 ppb) increasing Cr from 5 to 16 to 30 wt.% decreased the % of IG fracture. As Cr content increased, the steady-state creep rate decreased. Extrapolation of the incomplete curves revealed several orders of magnitude difference in creep rate at same stress level due to chromium alone. As demonstrated by Angeliu et al. [21 ] on controlled purity N-Cr-Fe low carbon alloys, the effect of Cr is to decrease the creep rate at 360~ as a result of solid solution strengthening. The soluble carbon has a similar but much stronger effect. Increased Cr content also increased the thickness of the surface film as observed by Angeliu et al. 18]. Increasing Cr content from 16 to 30 wt.% increased the thickness of the film from 46 to 58 nm with an enrichment in Cr203. Note that the role of carbon in thickening surface film is to enrich Ni(OH)2, rather than Cr203 [ 18].
2.5. Effect of chromium depletion The phenomenon of Cr depletion along the grain boundary after grain boundary carbide precipitation is well known [3,22-24]. Norring et al. [24] attempted to correlate the grain boundary microstructure and chemistry with IGSCC of Alloys 600 and 690. Six different Alloy 600 and Alloy 690 tube materials were tested in hydrogenated (5.9 kPa, 50 ml H2/kg H20) 365~ deaerated (02 < 5 ppb) water by using RUB method. All Alloy 690 specimens were crack-free even after extremely long exposure time (>23000 h), while most Alloy 600 specimens showed IGSCC after 12000 h or less. Transmission electron microscopy (TEM), energy-dispersive X-ray (EDX) and atom probe results showed that all Alloy 690 exhibited IG and intragranular carbide precipitation of the type M23C6 that were more or less globular, while The M7C3 type of carbide precipitation was predominant in Alloy 600 in all conditions. Cr depletion was observed at the grain boundaries both in Alloy 600 and Alloy 690. In Alloy 690, the Cr depletion was more pronounced after TT at 715~ (~6%) than in MA condition (~3%). Cr depletion was found in Alloy 600 with M23C7type of grain boundary precipitation. In Alloy 600 only with M7C3 type grain boundary precipitation, no Cr depletion was observed.
2. 6. Effect of segregation of impurities More grain boundary carbides seemed to be detrimental to IGSCC as suggested in the rising load tests on X-750 in 93~ water with cathodic charging performed by Miglin and Domain [25] where it was found that the cracking propensity of HTH treated samples was reversed, suggesting that the role of grain boundary segregation should be considered. While the IGSCC of Alloy X-750 has been attributed to grain boundary P segregation [26], Airey [27] found that P present at grain boundary of Alloy 600 following thermal treatment provided the most resistant structure. Comet et al. [28]
Volume 2: Prediction, Industrial Developments and Evaluation
127
showed that for solution-annealed samples of high purity Ni-Cr-Fe, the addition of P and C tend to decrease the degree of IG cracking during cathodic charging, further supported by tension tests of precharged samples of high purity Ni-Cr-Fe doped with P and heat-treated so as to segregate P to the grain boundary [4]. Was [4] was able to show that the ultimate tensile strength increased monotonically with increasing P at the grain boundary. On the contrary, Caceres et al. [29,30] found that among P, Si and B added to Alloy 600, only P segregated to grain boundary aRer thermal treatment at 704~ for 16 h and promoted IG fracture when M7C3 grain boundary carbides were present. Sung and Was [ 11] comprehensively studied the role of grain boundary chemistry in pure water IGSCC by using controlled purity Ni-16Cr-9Fe alloys and performing SCC tests on a series of controlled purity nickel alloys in 360~ high purity water with oxygen 1050~ was attributed to a high carbon content in solution and larger grain size [31 ]. In Symons' work [ 15] on hydrogen effect on mechanical properties of Alloy 690, the author found that the segregation of S was independent of heat treatment. After MA, the grain boundary P segregation was slightly lower for lower P containing material (0.006 wt.%) than higher P containing material (0.013 wt.%). However, after TT, the grain boundary P segregation of the low P containing alloy increased to the same level as the high P containing alloy. In high P containing (~0.013 wt.%) material, the heat treatment did not affect the grain boundary segregation of P. On the other hand, Stiller et al. [32] observed only small amount of grain boundary P segregation. 2. 7. Effect of special grain boundary (SGB) The coincident site lattice boundary (CSLB) has been shown to affect high temperature intergranular oxidation [33] and grain boundary carbide density [34]. Cheung et al. [35] showed that the CSLB morphology influenced the IGSCC behavior of Alloy 600. A decrease in total fraction of IGSCC in alloys with enhanced CSLB levels has been observed [36]. The creep rate could be decreased due to the presence of a high fraction of CSLBs [37]. It has been demonstrated [35] that alloys with CSLBs have a greater (2 to 14 times higher) resistance to IGSCC than the high angle boundaries in all samples tested with the greatest benefit at low strain (72% Ni, 1417% Cr, 6-10% Fe [3-5]. The yield strength of the alloy varies from 213 to 517 MPa. Normally this alloy is mill annealed at 885~ and final annealed for 4-6 h followed by air cooling. Nevertheless such a treatment can be varied depending on its purpose. The Alloy 600 works with some variation at 315~ and 15.5 MPa in pure water [3]. The primary water SCC (PWSCC) appears in the lower part of each nozzle that is fabricated in Alloy 600 and welded to the internal vessel head surface with dissimilar material such as Alloy 182. There are typically 40-90 penetrations per vessel that may include some spare penetrations which are not fired with CRDM or through core insmanentation of PWR [6]. 2. Models and modelling SCC nucleation and propagation are very complex phenomena. SCC is one modality of environment-assisted cracking (EAC) besides corrosion fatigue and hydrogen embrittlement, depending on several variables that can be classified in microstructural, mechanical, and environmental terms [7,8]. Microstructural variables are: (i) grain boundary microchemistry and segregation, M; (ii) thermal treatment, TT, that can cause intragranular and intergranular metallic carbide distribution; and (iii) grain size, gs, and cold work, CW, or plastic deformation. The second two variables fix another variable such as the yield stress, O'vs. Mechanical variables are: (i) residual stress, ~; (ii) applied stress, era (a tension stress and geometry can be summarized as a stress intensity factor, KI); and (iii) strain o~ and strain rate ob . Environmental variables include: (i) temperature, T; (ii) [HI § or pH; (iii) solution or water chemistry, SC; (iv) inhibitors or pollutants in solution; (v) electrochemical potential, V; and (vi) partial pressure of hydrogen, Pm [9]. Environmental cracking susceptibility can be expressed as [ 10]: SCC = f (M, TT, gs, CW, KI, 8 , ~', T, pH, SC, V, PH2)
(1)
Fig. 1 summarizes the main processes by which the above conditions at grain boundaries lead to SCC [11 ]. There are several models to express these phenomena mathematically: (i) the slip dissolution/film rupture by Ford and Andresen [13]; (ii)the enhanced surface mobility theory by Galvele [14]; (iii) coupled environment fracture model by Macdonald and Urquidi-Macdonald [15]; (iv) the internal oxidation mechanism by Scott and Le Calvar [16]; (v) numerical model by Rebak and Smialowska [17] and by Seung-gi and Il Soon Hwang [ 18]; and (vi) hydrogen-induced cracking models by Shen and Shewmon, and Magnin et al. (see Ref. [ 10]). For a comprehensive review of several of these models, see Ref. [10], and for hydrogen action models see Refs. [19,20]. Two kinetic models, including an empirical-probabilistic model and a deterministic strain rate damage model [21 ], were chosen to develop the model presented in this work.
145
Volume 2: Prediction, Industrial Developments and Evaluation Weak g r i n boumlarY p'oduci~ by predpitate tree zone /" ' ~ r
Acceleration of corrosive attack ~ ejection of AcxzeleralJonof corrosive passivaung spedes attack bY attraction of -~ ........ depassiva~lngspecies
Acceleration of corrosive attack by catalytically active precipitates
Grain boundary w e a ~ by bubUes (b)
~
" / "
\.
.-..=,,,,,
(d)
9 n, 0r nuclear reaction 9gas formation e.g. CH 4
eo.naary~
brittle precipitam
by 9
~"-'~
/
MXtX M 9
t~r Boundary. weakener by segregalon of soluzes with Weakened boondary ( m - m ) ..... , , O 4,-MMm/Mm~m'MmM M " ~ (~ MMmlmMM
Corrosive electro,re _ produced by dissolution of solute at grain boundary to give anions
(hi
-
Boundary weakened by preferential grain boundary diffusion of environmental species (x) -"x x x i / / x xx Xl~/X
(3 ~
X X~rXX ~
\
N
(g)
"
(e)
(f)
Wedging stress produced by corrosion product forces (produce exfoliatJon) =
O
L-Corrosion
products
(:It .4-
Hydrogen embd~ement
accentuated _by poisons (such
--e.O
as sulfide) which are enriched at grain boundary
M SIS
U
sis
Fig. 1. The processes starting from (a) to (k) range from the mostly chemical to the mostly mechanical [ 11]. The empirical-probabilistic model is derived from the general dependencies of timeto-failure shown in Eq. (2) treated statistically: (2)
tf = d[H+ ]'m o " e x p ( R ~ )
where tf = time to failure, cr = stress, n = exponent of stress, Q = thermal activation energy, T = absolute temperature (K), R = gas constant, [H +] = hydrogen ion activity, m = exponent of hydrogen ion activity, and d = constant [ 11 ]. The model proposed in Ref. [22] is a simplification of Eq. (2) that can be converted into a form more convenient to use as"
tf = A
tre f
o
exp
Q
1
1
(3)
where A = non-dimensional material constant reflecting the effect of material properties on time to 1% PWSCC, tref = time to selected fraction of PWSCC for a reference case, error= reference value of stress, and Trr f = reference value of temperature. The 2-parameter Weibull statistical distribution describes the variation of PWSCC as time function as:
,.
]
(4)
Stress Corrosion Cracking in L WR Environments
146
where F = fraction of population of components under consideration all susceptible to the same failure mode that experience PWSCC, t = time normally given in effective full power years (EFPY), b = Weibull slope, a fitted parameter determined by analysis of failure data, and 0 = Weibull characteristic time that corresponds to the time when 63.2% of the components have experienced PWSCC. This parameter can be written as tf = fi%:
0=
t,% (0.0101) l/b
(5)
Eqs. (4) and (5) combined yield Eq. (6):
F = -1 exp[- 0.010 l(tl-~/,)~ ]
(6)
The value of tl*/, together with an appropriate value for the Weibull slope, b, determine the complete prediction for PWSSC as a time function using Eq. (6). More detail on this model, plus several examples that were solved, are given in Refs. [ 11,22]. The strain rate damage model is essentially a semi-empirical model theory of SCC, where strain rate rather than stress is considered to be the main mechanical variable. The main parameter of this model is the damage parameter, D, that includes the initiation and propagation stages of the cracks. It begins essentially from a semiempirical theory of SCC, based on the analogy with Tresca criterion to plastic flow. It formalized the strain rate as a moving factor in a damage model that allows quantitative predictions on serviceable life which in turn depends on SCC. A damage function is defined as a mode linked to a component submitted to a strain rate history. When this damage function reaches a critical value, it can predict the SCC. The critical value of this damage function depends on the material in question and environment D = ~otA
[ ~"(t)]p dt,
[D] = [length]
(7)
where t = time, ~ (t) = total strain rate, A and p = parameters that depend on materialenvironment combination. In Eq. (7), the strain is divided into elastic and a non-elastic: i(t) = i =
+ i.
(8)
It is then necessary to adjust the experimental true stress-true swain data in accordance with Eq. (8). This can be accomplished using the Bodner-Partom constitutive equation that assumes Eq. (8) where the applied uniaxial stress o is related to the non-elastic strain rate g, by 2Do
d, = - ~ e x p
[
-~l / Z ) 2 "
]
(9)
Volume 2: Prediction, Industrial Developments and Evaluation
147
where Do is a constant, n is a temperature-dependent material parameter, and Z is a function related to strain hardness. When thermal recovery is neglected, the hardness function Z is such that Z = Z 1 -- ( Z 1 - Z 0 ) e x p ( - m
Wp)
(lO)
where the inelastic strain energy density is Wp=l ado~,
(11)
The temperature-dependent constants in the above equations might be written as: a
n=--+b T m = m o T + Co
Zo = Zl(mt T + cO
(12) (13) (14)
Hence the list of material constants in Bodner-Partom's model include Do, a, b, ZI, m~, Cl, mo, and Co. Thus, the model needs at least three values of stress and strain at two different strain rates at each of two temperatures as the minimum data set to determine these constants. In brief, in this model, we have formalized the concept of strain rate as a driving force in a damage model that permits quantitative predictions of stress corrosion lifetimes through a damage function defined as dependent on the strain rate history of a component. SCC is predicted when this damage function reaches a critical value. The critical damage value depends on both the material in question and the environmental condition of interest. The principal advantage of this model is that it's not necessary to distinguish between cracking initiation and propagation [21,22]. More detail on this model, plus modelling examples, are given in Ref. [22]. 3. Proposed model Staehle [11] has proposed a 3-dimensional diagram in accordance with Fig. 68 of Ref. [ 11]. It shows the thermodynamic conditions to occur at the modes of PWSCC in Alloy 600. The base is the 2-dimensional potential-pH (known as Pourbaix) diagram for this material in primary water at high temperature (300 to 350~ (Fig. 2). It superimposes the corrosion submodes based on experimental data from the literature. Submodes are determined by regions of potential where the different modes of surface material-environment interactions can occur, such as SCC, pitting, generalized corrosion, and passivation. The third dimension is the "useful strength" of the material as affected by the environment at that point, the strength fraction. Staehle [11] explained that the third variable could be a crack velocity for the vertical coordinate, instead of the strength fraction, because the data are sparse and the component experiments with reference to this diagram used different methods of loading states and data analysis.
148
Stress Corrosion Cracking in L WR Environments
Fig. 2. Pourbaix diagram for Alloy 600 at ~300~ used as a base for submode regions of the 3dimensional (V-pH-strength fraction) diagram [ l 1]. It is proposed that the model be flamed over the same Pourbaix (V-pH) diagram for Alloy 600 in the typical environment, namely water at high temperature. Over this diagram is plotted the region where the SCC submodes can occur. Firstly, over one of these regions will be coupled a strain rate damage model that can describe the damage parameter evolution with time and an empirical-probabilistic one that can describe the time to failure, normally expressed in terms of EFPY as a function of a total stress at the material surface, its temperature and parameters depending on environment-material combination and thermomechanical treatment of the alloy. Then, we will test the model using data from the literature plus data obtained using the new slow strain rate tensile (SSRT) test equipment installed at CDTN in Brazil [12]. Thus, this model could be used for a Brazilian nuclear power plant taking into consideration the plant materials and the characteristics of its design and operation, including the heat material fabrication processes, material composition, plant thermomechanic history, primary water chemical composition, and operational temperature conditions at this plant.
149
Volume 2: Prediction, Industrial Developments and Evaluation
4. Preliminary results A computer worksheet was created to plot an empirical-probabilistic model to be fed with data. This is represented by Eqs. (3)--(6), as was done in Refs. [ 11,22]. Fig. 3 was created using data of Table 1 from Ref. [23] with b-Weibull slope parameter equal to 1.5 to check the reproduction of the model.
Fraction of Population of Components (F) ] - , - F(Oconee #1 ) ~
F(Oconee #3) ---,- F(TMI#1 ) ]
0.1
u.
0.01
0.001
0.0001
i
1
10
100
t(EFPY) Fig. 3. Diagram showing plotted curves for three nuclear plants referred in Table 1 of Ref. [23].
If it is known how long a plant has operated in the submode IIIscc (see Fig. 2), this length of time can be used to couple the curves of Fig. 3 with Pourbaix diagram and thus to estimate a parameter F by Eq. (4) that represents the fraction of population of all components susceptible to the same degradation submode that experiences PWSCC.
5. Analysis and discussion The above empirical model serves as a highly practical method for the prediction of PWSCC. Using Eqs. (2) and (6), a higher F is expected for the lower pH. It is necessary to relate a damage initiation with the variations of pH and V in the PWSCC domain. These values are usually below an equilibrium borderline for Ni/NiO [16,19]. It is therefore necessary to verify the suppositions through empirical tests. Referring to the plants considered in Fig. 3, it is desirable to know which pH and V values should be employed for each of them to operate efficiently. The crack growth rate presents a dependence with pH shown in Eq. (15) from Ref.
[25]: dD = CGR = c ( P H ) ~ dt ~,-~J
for 7.5 0.95, the crack growth rate appears to be predominantly controlled by AK. Nonetheless, a threshold K must be met at these conditions, and this threshold increases with R [2,9]. Threshold data are not readily available, and may be affected by crack size and test conditions. Crack growth has been measured at K ~ 30 MPa~/m with R = 0.82 and 20 MPa~/m with R = 0.5 [10,11]. Threshold AK values from 3-5.5 MPa~/m have been reported. However, Beavers and Jaske [7] suggested that AK z 10 MPa~/m is required to sustain crack growth. They also observed that at lower AK crack growth rates decrease and cracks may even become dormant. Small surface cracks may experience fewer
223
Volume 2: Prediction, Industrial Developments and Evaluation
surface constraints and closure effects. Parkins [6] suggested a threshold AK ~ 2 MPa~/m for the earliest stages of crack growth. Much higher AK values required for sustained crack growth suggest that extrinsic factors such as time-dependent environmental factors and closure effects influence the mechanical driving forces required for crack growth. It thus becomes apparent that not one mechanical or environmental driving force dominates the crack growth process, but rather that their interactions determine the crack growth rate observed in the field. Fig. 1 illustrates schematically the interactions that may be present based on laboratory and field observations. The effects of K and AK on crack growth consist primarily of the generation, concentration, and emission of dislocations, which are promoted by the presence of hydrogen. Corrosion is essential for the generation of hydrogen, and results in the formation of a protective iron carbonate and/or iron oxide film. Cyclic loading may influence the integrity of this film and, together with the effect of strain, promote dissolution. The removal of material under cyclic loading conditions sustains a flow of metal, which supports the formation of dislocations. Crack growth occurs through the emission of dislocations, the rupture of metal bonds, and the removal of material. Crack growth on pipelines has been observed to diminish in the field as well as in the laboratory, leading to many small cracks only some of which (510
24.4 Yes 18.4 Yes 43.5 No Test interrupted. No corrosion or cracking
48.9 24.7 27.7 44 19
29 35.6
Secondary cracking No Yes Yes No No No No
Fig. 4. Surface of SSRT specimen and cross-section of cracks from Test 11 showing extensive crack initiation from a near-neutral pH CO2-HCO3- solution.
4.3. Hydrogen sources evaluation This mechanism relies on the additional hydrogen release from H C O 3 - compared to the hydrogen produced from cathodic reduction. A series of preliminary experiments and calculations were made to compare those two sources of hydrogen for one case at 15~ The experiments were designed to compare the metal weight loss along with the
Volume 2: Prediction, Industrial Developments and Evaluation
241
Fig. 5. Cross-section of line pipe taken from field failure showing near-neutral pH cracking. (Note: morphology is consistent with laboratory cracks from Fig. 4.)
corresponding hydrogen expected based on analysis of FeCO3 precipitation measurements. The results showed that an order of magnitude increase in total hydrogen produced via the mechanism proposed here as compared to that expected from the hydrogen reduction reaction alone. The amount of hydrogen will depend upon the temperature and other species that can contribute to the process, but these calculations did support the supposition that this additional hydrogen source can significantly increase adsorbed hydrogen available for diffusion into the steel.
5. Summary and conclusions In this paper, a simple mechanism for near-neutral pH cracking of line pipe steels is proposed that does not require the complicating factors of additional environmental components that have been collected in the field. The mechanism development was focused initially on the field reports that FeCO3 is present at nearly all field failure sites underneath disbonded coating. One of the key features of the proposed mechanism shows how hydrogen can be generated on the surface of line pipe steel from HCOawithout changing the bulk pH of the solution. It also shows how dissolved oxygen contributes to an increase in hydrogen generation. While not discussed here, the presence of other cations probably allows the solution pH to be lowered toward pH 6.4 similarly to the same process as adding CO2 to HCO3-. So, in some cases where the dissolution of CO2 is the rate limiting step to providing the requisite buffered solution, other cations may assist in achieving the right environmental conditions for cracking. The mechanism was validated by using tapered tension tests and SSRT tests. By selecting various environments that would be predicted by this mechanism to cause cracking and those that would not, we were able to show that cracking can be mined on and off in the laboratory according to the proposed mechanism. This provides us some confidence at this time that this mechanism may be the underlying driving force for near-neutral pH cracking. The crack morphologies generated in the laboratory were consistent with those found in the field, which has not been consistently the case with previous investigators. In addition, near-neutral pH cracking apparently occurs under some very specific environmental conditions because it has not been widely reported. For example,
242
s c c and Hydrogen Embrittlement of Pipeline Steels
anecdotally it has been described as a low-temperature and northern North American problem and not a concern in warmer climates. However, the initial work reported here indicates that the possibility for near-neutral pH cracking via this mechanism at temperatures approaching 45~ The fact that the tests reported here indicate that the cracking can be turned on and off as desired also supports the field observations because cyclic environmental conditions are well known in the field and would therefore be consistent with the variability observed. It is also instructive here to speculate on additional implications for pipeline operators based upon our insights about the mechanism that has been described. The near-neutral pH cracking process described here is consistent with a lack of, or at least inconsistent, cathodic protection on the line. The presence of adequate cathodic protection would limit the near-neutral mechanism for two reasons. First, the lack of iron cations caused by cathodic protection would therefore limit the hydrogen released from HCO3-, and second, effective cathodic protection would provide local alkalization and would, therefore, promote a shift in pH toward the other buffered region of pH of Fig. 1, possibly setting the stage for high pH cracking instead. Moreover, it is not hard to visualize in some cases that the pH might cycle from one buffered region to the other, i.e., from 6.4 to 10.3, depending upon intermittent cathodic protection, the ability to drive the pH more acid with CO2 or other cations, and how dilute the solution is during these processes. With higher solution concentrations it would be expected that any shifts in pH would take longer time. There have been field reports of both nearneutral and high pH cracking having occurred on the same line. With the mechanism described here, and from Fig. 1, both modes of cracking on the same line is not inconsistent as pH shitts can occur over time with the end point tending toward each buffered region. References [I] [2]
[31 [4] [51 [61 [7] [81
R.R. Fessler, Stress Corrosion Cracking Gap Analysis, PRCI Report GRI-8293, Pipeline Research Council International, September 2002. R.N. Parkins, A review of stress corrosion cracking of high pressure gas pipelines, CORROSION/2000, NACE International, Houston, 2000, paper no. 00363. B. Delanty, J. O'Beimc, Oil Gas J. 90(24) (1992) 39-44. B. Gu, W.Z. Yu, J.L. Luo, X. Mao, Corrosion 55 (1999) 312-318. E.A. Charles, R.N. Parkins, Corrosion 51 (1995) 518-527. F. King, T. Jack, W. Chen, M. Wilmott, R.R. Fessler, K. Krist, Mechanistic studies of initiation and early stage crack growth for near-neutral pH SCC on pipelines, CORROSION/2000, NACE International, Houston, 2000, paper no. 00370. R.N. Parkins, A review of stress corrosion cracking of pipelines in contact with near-neutral (low) pH solutions, PRCI Report PR-232-9701, Pipeline Research Council International, 1999. M. Pourbaix, Lectures on Electrochemical Corrosion, NACE International, Houston, 1995, p. 42.
243
A mechanistic study on near-neutral pH stress corrosion cracking of pipeline steel B.T. Lu, J.L. Luo Department of Chemical and Materials Engineering, University of Alberta, Edmonton, Alberta T6G 2G6, Canada Abstract
The roles of dissolved hydrogen in material and anodic dissolution in the near-neutral pH stress corrosion cracking (SCC) process of a pipeline steel was investigated. A modified Devanathan-type dual cell was used in SCC tests to independently control both the dissolved hydrogen concentration in specimens and the anodic dissolution current density on the test surface. The experimental results showed that both dissolved hydrogen in the pipeline steel and anodic dissolution promoted crack initiation over a wide potential range. At high anodic dissolution rates, the crack initiation is retarded owing to the dissolution of the potential crack nuclei. The experimentally-measured crack growth rate decreased almost monotonically with increasing applied potential, but the anodic dissolution had no significant impact on crack growth rate. These results clearly indicate that the crack propagation process is controlled mainly by a hydrogen-induced crack mechanism. A preliminary chemo-mechanical model was proposed for the near-neutral pH SCC based on theoretical analysis and electrochemical measurements. I. Introduction
Transgranular stress corrosion cracking (SCC) on pipelines in dilute groundwater with near-neutral pH differs from "classical" intergranular SCC that has beenfound in high pH concentrated carbonate (CO3 2-)-bicarbonate (HCO3-) solutions [1,2]. The most distinguishing features of near-neutral pH SCC are: (i) it occurs in metal-environment combinations that experience active dissolution rather than the active-to-passive transition, (ii) lateral dissolution on crack sides is substantial, and (iii) cyclic or dynamic loading is necessary for cracking to occur [2]. The "classical" intergranular SCC has been well accepted as being controlled by an anodic dissolution mechanism, but the mechanism of near-neutral pH SCC is poorly understood. The test data obtained with a slow strain rate tensile (SSRT) technique or cyclic loading indicated that cathodic polarization or hydrogen precharging of test specimens degraded the SCC resistance of pipeline steels [3-6]. Because of this, hydrogen-induced cracking (HIC) was usually accepted as the mechanism of near-neutral pH SCC [2,3,7]. On the other
244
SCC and Hydrogen Embrittlement of Pipeline Steels
hand, 73% SCC events in the field were reported in pipelines having polyethylene tape coating [1]. The high electrical resistance of such coating makes it difficult for the current of cathodic protection to reach the surface of a pipe under the disbonded coating. This implies that SCC events are more likely to occur at locations where cathodic protection is insufficient. That is, under certain conditions, cathodic protection inhibits near-neutral pH SCC. Also, some laboratory studies have shown that the SCC resistance does not decrease monotonically with gradual shiffing of the applied cathodic potential to more negative values [5,8]. This effect may derive from the dominant role of anodic dissolution in the near-neutral pH SCC process. As pointed by Parkins [7], it is difficult to believe that anodic dissolution on crack walls does not occur at the crack tip and, that consequently, it is difficult to ignore the involvement of anodic dissolution to the near-neutral pH SCC process at potentials that are close the open circuit potential, Ecorr. Although many researchers agree that both hydrogen dissolved in steel and anodic dissolution affect SCC in pipelines [5,7-9], the precise role of these two processes in near-neutral pH SCC is insufficiently understood. In traditional SCC tests using SSRT or cyclic loading, anodic dissolution on the test specimen surface is suppressed when cathodic polarization that provides hydrogen charging to the specimen is applied. This means that the rate of the anodic dissolution of steel and the process of hydrogen dissolution in steel are interdependent of each other [2,5,6]. This phenomenon makes it difficult to identify the exact contribution of dissolved hydrogen and anodic dissolution to the SCC process. Therefore, an improved experimental technique is required for SCC studying. In the past, most research efforts have been aimed at understanding the roles of hydrogen and anodic dissolution in the process of crack propagation [2,4,5-8]. Less attention was focused on their roles in the process of crack initiation [6]. However, the crack initiation period otten comprises the major part of service lifetime of pipelines [2,7]. A series of studies [10-12] has concluded that the crack advance mode during transgranular SCC is discontinuous. The need to study the crack initiation stage becomes more apparent from the fact that the same stress-environment interactions take place at a crack tip during crack propagation. In other words, each single crack extension event can be considered to be a crack re-initiation process [ 13]. Therefore, a comprehensive understanding of the mechanism of crack initiation might be crucial to understanding the mechanism of crack propagation, especially aspects concerning the roles of dissolved hydrogen and anodic dissolution in SCC development. In addition, for practical purposes, such an understanding is required for (i) the improvement of SCC resistance of steels, (ii) the prediction of the service lifetime of pipelines, and (iii) the optimization of pipeline protection techniques. o
2. Experimental procedures 2.1. SCC tests
X60 pipeline steel was used as our test material. The width and thickness of specimens at their gauge length were 25.4 mm and 1.2 mm respectively. Prior to SCC testing, the specimens were ground using SiC paper of grits up to 1200, and sequentially cleaned and rinsed with de-ionized water and acetone. Cyclic loading was provided with an MTS machine at a loading frequency of 0.1 Hz and a stress ratio (R =
Volume 2: Prediction, Industrial Developments and Evaluation
245
O'min/O'max) of 0.5. In all tests, the maximum applied stress per cycle (O'max) was 434 MPa (1.05 specific minimum yield strength, SMYS). For the independent control of hydrogen concentration in specimens and anodic dissolution rate on the surface of specimens, the modified Devanathan-type dual cell (Fig. 1) was used in SCC experiments. A cyclically-loaded specimen was inserted between two parts of the cell. Hydrogen concentration in specimens was controlled by applying a constant potential, Eapp, in a range from -0.6 to-1.2 VSCE at the charging side (denoted here as "Eapp side"). On the opposite side of the specimen, an anodic potential could be applied to control anodic current density. However, in this study, only Eco~ was adopted, and the side was denoted as "Ecor~side." When hydrogen atoms generated at the Eapp side reached (through diffusion) the specimen surface at the Er side, they did not oxidize because Er (---0.74 VscE) was lower than the oxidizing potential of hydrogen. In such experiments, the dissolved hydrogen concentration was controlled by the level of applied potential at the Eapp side. Due to the extreme thickness of the specimen, the concentration of hydrogen across the test section of the specimen was considered being uniformly distributed. That is, the concentration of dissolved hydrogen near the surface layers at both Eapp and E~o~rsides was considered essentially the same. At the Er side, the anodic current density was independent of Eapp, and was equal to the self-corrosion current density of steel. However, at the Eapp side, the environment condition was similar to that observed during regular SSRT tests [2-5,8], and the hydrogen concentration and the anodic dissolution rate were interdependent.
Fig. 1. Devanathan-type dual cell. The test solution on both sides of the cell was the NS4 solution prepared with analytical reagents and de-ionized water. The chemical composition of the NS4 solution in g/l was 0.122 potassium chloride (KCI), 0.483 sodium bicarbonate (NaHCO3), 0.137 calcium chloride (CaCl2), and 0.131 magnesium sulphate (MgSO4"7H20). During tests, the solution was deaerated by the continuously bubbling N2 + 5% CO2 through it to make the pH of the solution equal to ~6.7. Before cyclic loading was applied, the specimen was charged with hydrogen for more than 3 h to reach stabilization. All tests were conducted at room temperature using a three-electrode cell with a saturated calomel electrode (SCE) as the reference electrode (RE) and a platinum wire as the counter electrode (CE). A Pine potentiostat was used for the control of potential during
246
SCC and Hydrogen Embrittlement of Pipeline Steels
hydrogen charging and a Gamry potentiostat (EG&G Inc.) was used for electrochemical measurements. Tests were interrupted at various intervals for crack measurements. The number of cracks initiated and the size of cracks were measured using a replica technique. 2.2. Electrochemical measurements
Polarization curves for X60 steel were obtained with the same experimental set-up that was used for SCC tests. All measurements were made at the oxidizing side of the deaerated NS4 solution after the steady-state hydrogen permeation rate was reached. The solution on the hydrogen-charging side was 0.1 M NaOH + 150 ppm As203. A charging current density of 10 mA/cm 2 was used to create a high hydrogen concentration. Hydrogen permeation rates through the specimen were measured with the same dual cell (Fig. 1). To create a hydrogen concentration close to the concentration observed in pipelines exposed to a near-neutral pH environment, the de-aerated NS4 solution in charging cell and applied potentials in the range from -1.2 to -0.6 VSCEwere used. The solution on the oxidizing cell was 0.3 M H3BO3 + 0.075 M Na2B407 with pH 8.4. Hydrogen atoms generated on the surface of one side of the specimen facing the charging cell diffused through the specimen and were oxidized electrochemically on the opposite side, which was maintained at the potential of 0.2 VSCEto oxidize hydrogen atoms passing through the specimen. The current density of hydrogen permeation, iH-p~meatng, was measured at the oxidizing side and used as the indicator of the concentration of hydrogen dissolved in steel, C H (C H oc/H-permeating) [7]. 3. Effect of dissolved hydrogen and anodic dissolution on SCC process
When loading cycles reached a certain number, cracks began to appear on the specimen surface. Both the crack density, Dc (defined as the number of cracks in a unit area), and the crack size increased during the test. To evaluate crack initiation kinetics, the crack initiation rate, b c, was defined as dDc/dt. Results indicated that crack initiation rates were virtually independent of the test duration and on the ratio between the crack initiation rate and applied potential. Crack initiation rates increased with decreasing applied potential. Because the concentration of hydrogen in steel increased with shifting an applied potential to more negative values, this implies that dissolved hydrogen could promote crack initiation. At potentials below -0.8 VSCE, the lower crack initiation rate at the Eapp side indicated that the inhibition of anodic dissolution decreased the number of cracks initiated. The extrapolation of the curves of crack densities vs. test duration to the zero crack density, enable us to determine the crack initiation lifetime t i - that is, the incubation time for crack initiation. Fig. 2 shows that ti at the Er side, where the average rate of anodic dissolution was almost constant, decreased with increasing concentration of dissolved hydrogen when the applied potential was decreased. Also, Fig. 2 shows that the trend of crack initiation lifetime at the Eapo side was slightly different. At E~pp< -0.8 VSCE, the initiation lifetime ti determined from the Eapp side was higher than ti at the Er side. This indicates that the inhibition of anodic dissolution retarded crack initiation. At potentials close to Er the steel displayed a lower crack initiation lifetime that was close to ti on the opposite side. This was due to the fact that the anodic dissolution rates
Volume 2: Prediction, Industrial Developments and Evaluation
247
on both sides were very close at these potentials. At the potential o f - 0 . 6 VSCE, the crack initiation rate increased again. This suggests that the crack initiation was retarded by the dissolution of crack nuclei due to the relatively high rate of anodic dissolution.
1,6
-
--O--Eo, side
/
O
/
--o-- E side
/ /
o
/
1.2
o v
~- 0.8
01t 0.4
-1'.2
.~~ 9 '
f
Eoor,
-1'.o
'
-o.8
'
-o.6
Applied potential (VscE)
Fig. 2. The effect of applied potential on crack initiation lifetime.
1.8 1.6
--o- ~
side i
........o ........E
side
E 1.4 E ~'o 1.2 ,TO t~ t-
I
1.0 0.8
%
o~ 0.6 O
o 0.4 0.2
Eoorr I
I
-1.2
-1.0
a
I
~=:~,
-0.8
I
-0.6
Applied potential (VscE)
Fig. 3. The effect of applied potential on crack propagation rate.
In this study, the crack propagation rate, da/dt, was defined as the slope of the curve of the maximum surface crack size, am~x, vs. the duration of the test. Fig. 3 shows that crack propagation rates in steel on both sides of a specimen increased with decreasing applied potential. At the same applied potential, the difference in crack growth rates on different sides was not remarkable, especially at potentials close to Eco~. This suggests that the process of crack propagation was controlled mostly by the contents of dissolved hydrogen.
248
SCC and Hydrogen Embrittlement of Pipeline Steels
4. The roles of dissolved hydrogen and anodic dissolution in SCC development
Parkins [2,14] pointed out that SCC is controlled by the local plastic strain rate or creep, and can occur when the local plastic strain rate is higher than a certain critical level. The SEM analysis showed that the trace of plastic tear always exists on the fracture surface of cleavage-like transgranular SCC [2,4,15]. Therefore, we believed that creep or plastic deformation at the crack tip activated by the crack advance plays a key role in crack propagation and also that cracks would become dormant when the creep rate at the crack tip cannot maintain crack propagation [2,7,16]. Experiments [ 17,18] showed that the creep of pipeline steels at room temperature was accelerated by cyclic loading. These results are thought to explain why near-neutral pH SCC takes place only under the action of cyclic loads [2,7]. Crack propagation rates increased monotonically with the hydrogen concentration, but Were almost independent of anodic dissolution on the specimen surface (Fig. 3). Crack initiation lifetime at the Er side decreased with decreasing applied potential (Fig. 2). These observations suggest that HIC was one of the major mechanisms in the near-neutral pH SCC development. Hydrogen-enhanced room temperature creep has been observed in iron and carbon steels [ 19-23]. After hydrogen was removed from test specimens, the creep behaviour returned to its original state. When a stress gradient exists, dissolved hydrogen can reduce the flow stress and facilitate the localized surface plastic deformation [ 16,23-25]. Recent experimental observations [ 19] indicate that hydrogen charging can promote the development of local stress and triaxial stress gradients. Moreover, plastic deformation can also accelerate a hydrogen release [26]. This suggests that hydrogen-induced plasticity might be the crack growth mechanism of near-neutral pH SCC, i.e., hydrogen dissolved in steel can promote localized plastic deformation resulting in cracking [2,7]. Fig. 2 also shows that anodic dissolution at potentials less than -0.7 VSCE can promote the crack initiation. Because a pit(s) can serve as a stress concentrator and also supply local environments promoting crack initiation [27,28], it seems reasonable to relate the effect of anodic dissolution on the crack initiation and pitting corrosion. Two types of cracks were detected in near-neutral pH SCC [ 15,29]: "pit-cracks" and "non-pit-cracks." The former originated at pits while, the latter was related to localized plastic deformation. The number of non-pit-cracks on specimens that were cut from original external pipe surface was much higher than the number of pit-cracks [15]. Therefore, a mechanism other than pit-induced cracking must be proposed for anodic dissolution-promoted SCC. In field situations, the SCC of pipelines occurs at potentials between Er and -0.85 VSCE. The latter is the potential of full cathodic protection. According to Parkins [2,7], crack growth rates at potentials close to Er are at least one order of magnitude lower than crack growth rates predicted by Faraday's law. To explain this difference, Gu et al. [30] and Mao et al. [8] proposed a mechanism termed "hydrogen-facilitated anodic dissolution." They believe that the anodic dissolution rate at the crack tip is much higher than that at the crack side owing to the synergistic effect of the local stress concentration and hydrogen at the crack tip [31 ]. To demonstrate the effects of dissolved hydrogen and applied stress on anodic dissolution in this study, polarization curves were obtained and polarization resistance under various conditions was measured. In these tests, continuous hydrogen charging with/H-charging 10 m A / c m 2 w a s provided to create the high concentration of hydrogen =
Volume 2: Prediction, Industrial Developments and Evaluation
249
similar to the hydrogen concentration found at crack tips. According to Fig. 4, neither hydrogen charging nor cyclic tensile loading affect the polarization curve of steel. An exception is the small decrease in Er that results from the limited effect of dissolved hydrogen and applied stress on anodic dissolution in steel exposed to deaerated groundwater. If an uncharged and unloaded specimen represents the crack side, and another hydrogen charged and cyclic loaded specimen represents the crack tip, the difference between E~orrat the crack side and E~o~rat the crack tip would be ~ 10 mV (or slightly higher). Therefore, it can be expected that both the crack tip and the crack side suffer substantial anodic dissolution, in which case the difference in anodic dissolution rates at crack sides and the crack tip would be insignificant. This relationship makes it difficult to use the hydrogen-facilitated anodic dissolution mechanism as an explanation for why the cracking process is maintained without becoming a pit. Because CH in the surface layer of an uncracked specimen is much less than CH at the crack tip, the influence of dissolved hydrogen on anodic dissolution is less significant for crack initiation than for crack propagation. It has been indicated that near-neural pH stress corrosion cracks are more likely to initiate at the edge of specimens [ 15], where the restriction of plastic deformation is weaker. This suggests that the localized microplastic deformation plays an important role in the crack initiation process. Lian and Meletis [13], Jones et al. [32,33], and Gutman [34] showed that anodic dissolution could enhance the mobility of dislocations in surface layers and promote localized surface plasticity, which is now known as a chemo-mechanical effect. In line with the chemo-mechanical model that was originally developed using non-equilibrium thermodynamics and dislocation kinetics [ 13,32-34], the anodic dissolution-induced creep rate, Akc , is correlated to the anodic current density, ia, as follows:
-Lvio
(l)
where L is related to the phenomenological coefficient in the Onsager relation and o = e./g and 6 and ?' are normal strain and shear strain respectively [35]. Anodic dissolution-induced creep was observed in various metallic materials [34,36,37]. It was found that the creep rate increases with the application of anodic current and recovers its initial value after the anodic current removed [32-34]. Revie and Uhlig [36], Jones et al. [32,33], and Gu et al. [30] believe that anodic dissolution promotes the formation and diffusion of vacancies in surface layers, which in turn attenuate the strainhardening subsurface and enhance localized surface plasticity. Weakening interatomic bonds due to subsurface vacancies result in the local reduction of the energy barrier for dislocation nucleation, i.e., in the reduction of the chemical potential of dislocations in the surface layers [34,35]. A chemical potential gradient between the surface layer and the bulk material is likely to cause an additional dislocation flux in the surface layer [34], resulting in the degradation of flow stress or microhardness of materials [35]. Like hydrogen-promoted creep, anodic dissolution-enhanced plasticity is expected to play a role in crack initiation. Duquette [38] and Lian and Meletis [13] found that both the intensity and the localization of surface plastic deformation were enhanced by anodic dissolution, and consequently the slip bands were more likely to be attacked.
250
SCC and Hydrogen Embrittlement of Pipeline Steels
Fig. 4. The effect of applied potential on the dissolved hydrogen concentration, the anodic dissolution rate and the SCC damage evolution.
Using the concepts of hydrogen- and anodic dissolution-enhanced plasticity, a preliminary chemo-mechanical model for near-neutral pH SCC was proposed in this study. According to the model, near-neutral pH SCC includes the following steps: 1. Dislocation slip in surface layer is induced by cyclic loading and produces fresh bare-metal surface that is vulnerable to attack by corrosive species, 2. The formation of slip bands is promoted by anodic dissolution on the surface owing to the chemo-mechanical effect, and the localization of microplastic deformation in surface layers is enhanced because the fresh bare-metal surface is more likely to be attacked, 3. Because the slip bands are surrounded by a large elastic matrix, the local tensile stress is produced in the direction perpendicular to the slip band [13,32], which causes dissolved hydrogen to diffuse into and enrich localized plastic zones. Localized plastic deformation is also likely to accelerate local hydrogen evolution [26], and the movement of dislocations can promote the ingress of hydrogen atoms into steel [25]. The enrichment of hydrogen in the surface layer
Volume 2: Prediction, Industrial Developments and Evaluation
251
will, in turn, promote the movement of dislocations in slip bands as well as the development of local stress and stress gradients [19]. In this way, the localized surface plasticity will be further enhanced, 4. Cracks initiate when accumulated plastic deformation reaches a critical level, and 5. After a crack has been initiated, the crack tip enters an unaffected zone, and a new plastic zone is formed at the crack tip because of stress concentration. The above procedure will repeat, causing crack propagation. Because both the level and the gradient of triaxial stresses at the crack tip are much higher than those around slip bands on an unnotched surface, a much higher local hydrogen concentration at the crack tip is observed [8,31]. Therefore, the HIC mechanism dominates in the crack propagation process while anodic dissolution does not play an important role in crack propagation. As mentioned above, the crack growth rate cannot be estimated using Faraday's law [2,7]. Plastic deformation during cyclic loading results in mechanical fatigue damage that does not depend on applied potentials. Under in-service conditions, when a pipeline transports natural gas or oil, internal pressure fluctuations result in cyclic loading with a very high stress ratio (0.8-0.95) and a rather low loading frequency (several cycles per day). Therefore, it was concluded that the damage due to fatigue is negligible [39], and that the total damage resulting in crack initiation is determined as the sum of damage components due to dissolved hydrogen-promoted plasticity, DH, and to anodic dissolution-promoted plasticity, DA. Fig. 4 shows the effects of applied potential on anodic current density and hydrogen permeation. It is well known that the concentration of dissolved hydrogen in steel is proportional to the steady-state hydrogen permeation current density. Because of that relationship, the rate of damage evolution under a given loading condition related to dissolved hydrogen,/)n, increases with increasing hydrogen permeation current density. On the other hand, the damage evolution rate related to anodic dissolution, /)A, increases initially with the applied potential because the effect of anodic dissolution-promoted plasticity increases. However, when the anodic current density is sufficiently high, the crack initiation decelerates and even stops due to possible crack nuclei dissolution. Accordingly, the relationship between damage evolution rate due to anodic dissolution and applied potential has a bell shaped form. The total damage, D, under a given stress condition is estimated as l
D = D z + D A =
~(D A +
D z )dt
(2)
0
According to the definition of damage, the level of total damage would be equal to 1 when t - t~. To a first approximation, damage evolution rates are assumed to be timeindependent, in which case SCC initiation lifetime is given as
t, = 1/(/5. + hA). The general trend of the relationship between ti and E~pp that was determined theoretically (Fig. 4) was in agreement with the observed experimental relationship between t~ and Eapp (Fig. 2). The correlation between the near-neutral pH SCC
252
SCC and Hydrogen Embrittlement of Pipeline Steels
resistance and an applied potential predicted by the model of SCC proposed in this study agreed well with the results of SSRT tests reported in the literature [2,3,5,7,8,30,40,41].
5. Conclusions 9 The Devanathan-type dual cell technique can be used for SCC tests to control the dissolved hydrogen concentration in specimens and the anodic current density on the test surface independently of each other. 9 Both anodic dissolution and dissolved hydrogen play important roles in the initiation of near-neutral pH SCC, but the crack growth process is likely to be controlled by a HIC mechanism rather than an anodic dissolution. 9 A preliminary chemo-mechanical model was proposed for near-neutral pH SCC development. According to this model, SCC is controlled by a synergistic mechanism of hydrogen- and anodic dissolution-induced plasticity.
Acknowledgements This project was supported by a Strategic Project Grant of the Natural Sciences and Engineering Research Council of Canada, Syncrude Canada Ltd., Dow Chemical Canada Inc., IPSCO and NOVA.
References [1] Stress Corrosion Cracking on Canadian Oil and Gas Pipelines, Report of the Inquiry, NEB Report MH-2-95, National Energy Board, Calgary, 1996. [2] R.N. Parkins, W.K. Blanchard Jr., B.S. Delandy, Corrosion 50 (1994) 394-412. [3] R.B. Rebak, Z. Xia, Safruddin, Z. Szkiarska-Smialowska, Corrosion 52 (1996) 396--406. [4] S.D. Liberto, G. Gabetta, Mater. Sci. 33 (1997) 411-420. [5] B. Gu, W.Z. Wu, J.L. Luo, X. Mao, Corrosion 55 (1999) 312-319. [6] M. Puiggali, S. Rousserie, M. Touzet, Corrosion 58 (2002) 961-970. [7] R.N. Parkins, A review of stress corrosion cracking of high pressure gas pipelines, CORROSION/2000, NACE International, Houston, 2000, paper no. 00363. [8] S.X. Mao, B. Gu, N.Q. Wu, L. Qiao, Phil. Mag. 81A (2001) 1813-1831. [9] M. Elboujdaini, Y.-Z. Wang, R.R. Revie, Initiation of stress corrosion cracking on X-65 linepipe steel in near-neutral pH environment, in: J.R. Ellwood (Ed.), Proc. 2000 International Pipeline Conference, vol. 2, ASME, New York, 2000, pp. 967-978. [10] R.C. Newman, K. Sieradzki, Scripta Metall. Mater. 17 (1981) 621-625. [ 11] W.F. Flanagan, P. Bastias, B.D. Lichet, Acta Metall. Mater. 39 (1991) 695-705. [12] F.P. Ford, Corrosion 52 (1996) 375-395. [13] K. Lian, E.I. Meletis, Corrosion 52 (1996) 347-355. [ 14] R.N. Parkins, Corrosion 46 (1990) 178-194. [15] B. Lu, J. Luo, B. McCrady, Near-neutral pH SCC initiation and early propagation of X70 pipeline steel, in: Proc. 4th International Pipeline Conference (IPC2002), part B, ASME, New York, 2002, pp. 1821-1828. [ 16] R.M. Rieck, A. Atrens, I.O. Smith, Metall. Trans. A, 20A (1989) 889-895. [ 17] S.H. Wang, Y. Zhang, W.X. Chen, J. Mater. Sci. 36 (2001) 1931-1938. [18] W.X. Chen, S.H. Wang, Room temperature creep behaviour of pipeline steels and its influence on stress corrosion cracking, in: Proc. 4th International Pipeline Conference (IPC2002), part B, ASME, New York, 2002, pp. 1895-1902.
Volume 2: Prediction, Industrial Developments and Evaluation
253
[19] T. Zhang, W.Y. Chu, K.W. Gao, L.J. Qiao, Mater. Sci. Eng. A347 (2003) 291-299. [20] L.V. Spivak, Y.Ye. Skryabina, L.D. Kurmayeva, L. V. Smimov, Phys. Met. Metallogr. 66 (1988) 122-128. [21] H. Hideki, K. Hisashi, H. Yasunori, J. Jap. Soc. Mater. Sci. 38 (1989) 668-674. [22] R.A. Oriani, P.H. Josephic, Effect of hydrogen on the room temperature creep of spheroidized 1040 steel, in: Prospect in Hydrogen in Metals, Collected Papers on the Effect of Hydrogen on the Properties of Metals and Alloys, Pergamon Press, Oxford, 1986, pp. 609-614. [23] N.N. Kinaev, D.R. Cousens, A. Atrens, J. Mater. Sci. 34 (1999) 4931-4936. [24] W.Y. Chu, S.Q. Li, C.M. Hsiao, S. Y. Zhu, Corrosion 37 (1981) 514-520. [25] J.D. Hirth, Metall. Trans. A, 11A (1980) 861-890. [26] I.I. Dykyi, I.M. Protsov, Mater. Sci. 29 (1993) 569-574. [27] T.K. Christaman, Corrosion 46 (1990) 450-453. [28] Z.F. Wang, A. Atrens, Metall. Mater. Trans. A, 27A (1996) 2686-2691. [29] M. Elboujdaini, Y.Z. Wang, R.W. Revie, R.N. Parkins, M. T. Schhata, Stress corrosion crack initiation processes: pitting and microcrack coalescence, CORROSION/2000, NACE International, Houston, 2000, paper no. 00379. [30] B. Gu, J.L. Luo, X. Mao, Corrosion 55 (1999) 96-106. [31] L.J. Qiao, J.L. Luo, X. Mao, J. Mater. Sci. Lett. 16 (1997) 516-520. [32] D.A. Jones, Corrosion 52 (1996) 356-362. [33] D.A. Jones, A.F. Jankowski, Scripta Metall. Mater. 22 (1993) 701-705. [34] E.M. Gutman, Mechanochemistry of Materials, Cambridge International Science Pub., Cambridge, 1998. [35] B. Lu, J.L. Luo, Chemo-mechancal effect in erosion-corrosion of carbon steels, Presented at CORROSION/2004, held March 29-April 1, 2004, New Orleans, LA, USA. [36] R.W. Revie, H.H. Uhlig, Acta Metall. 22 (1974) 69-75. [37] B. Gu, W.Y. Chu, L.J. Qiao, C.M. Hsiao, Corros. Sci. 36 (1994) 1437-1445. [38] D.J. Duquette, Corrosion 46 (1990) 434--443. [39] B.T. Lu, X.L. Zheng, D.B. Li, Weld. J. 75 (1993) 72s-80s. [40] W. Zheng, R.W. Revie, O. Dinardo, F.A. MacLeod, W.R. Tyson, D. Kiff, Pipeline SCC in near-neutral pH environment: effects of environmental and metallurgical variables, in: Proc. 9th Symposium on Pipeline Research, Catalog No. L51746, AGA, Arlington, 1998, paper no. 22. [41] J.A. Beavers, C.L. Durr, S.S. Shademan, Mechanistic studies of near-neutral pH SCC on underground pipelines, in: L. Collins (Ed.), Materials for Resource Recovery and Transport, Metallurgical Society of CIM, Montreal, 1998, pp. 51--69.
255
The role of hydrogen in EAC of pipeline steels in near-neutral pH environments J. Been
a, H. Lu a, F. King a, T. Jack a, R.
Sutherby b
a NOVA Chemicals Corporation, 2928- 16th Street NE, Calgary, Alberta T2E 7K7, Canada b TransCanada Pipelines, 4 5 0 - 1st Street SW, Calgary, Alberta T2P 5H1, Canada
Abstract
Metal dissolution and hydrogen absorption are thought to interactively affect crack growth on gas pipelines in near-neutral pH soil environments. Although it is well established that hydrogen enters the steel, its role in the crack growth mechanism is not clear. The paper reviews the literature on the role of hydrogen, and then discusses the results of electrochemical polarization and corrosion experiments, slow strain rate tensile (SSRT) tests, hydrogen permeation tests, and cyclic loading experiments in simulated groundwater solutions consisting of normal and heavy water. The kinetic isotope effects resulted in reduced corrosion rates and slower crack growth rates in the heavy water. The results are discussed with reference to mechanistic implications for pipeline crack growth in the field. I. Introduction
Environmental cracking of gas pipelines in near-neutral pH soil environments has been recognized since the mid-1980s, and has commonly been referred to as "low-pH stress corrosion cracking" [ 1]. This form of transgranular cracking occurs at the opencircuit potential (OCP) in dilute bicarbonate (HCO3-) groundwater solutions of pH in the range of 5.5 to 7.5. Such an environment can be encountered underneath coatings that shield the steel from cathodic protection. The mechanism of near-neutral pH cracking is poorly understood. Both dissolution and hydrogen are believed to be involved [2]. Cracks in the field often display extensive corrosion of the crack walls and are filled with corrosion products. However, researchers have not favoured dissolution as the main crack growth mechanism, since general corrosion rates are of the order of 10-m mm/s, whereas crack growth rates of the order of 10-8 mm/s have been measured in the field. Only rapid straining or scratching experiments could provide sufficiently high dissolution rates [3]. Similarly high strain rates are not likely to be encountered by pipeline cracks. A pertinent role of hydrogen has been well accepted, with the main indicators being the quasi-cleavage crack appearance and the observation that hydrogen enters the steel in hydrogen permeation
256
SCC and Hydrogen Embrittlement of Pipeline Steels
tests where the steel is exposed to groundwater solutions. Assuming that crack growth is driven interactively by dissolution and hydrogen, neither the role of hydrogen nor rate-determining step is clear. 2. Overview of the literature
In near-neutral pH groundwater solutions, hydrogen is generated by corrosion reactions or by a small quantity of cathodic protection current that enters the shielded coating disbondment. Only a small fraction of the cathodic hydrogen produced on the charging surface diffuses into the steel [4]. The hydrogen that is absorbed by the steel is transported to the highly-stressed plastic zone at the crack tip by diffusion or dislocation movement. Once there, the hydrogen enhances the crack growth rate by embrittling the area. A number of embrittling mechanisms have been suggested. The lattice decohesion mechanism and the hydrogen-enhanced plasticity mechanism are most probable for pipeline steel in a groundwater environment [3]. In the lattice decohesion mechanism, hydrogen as a solute decreases the cohesive metal bond forces. Dissolved hydrogen occupies interstitial sites of the metal lattice and has a significant lattice expansion effect [5]. When the tensile stresses ahead of the crack tip are greater than the hydrogen-weakened cohesive strength, the crack propagates. In the hydrogenenhanced plasticity mechanism, hydrogen assists the plastic flow at the crack tip by facilitating the flow of dislocations or by easing the generation of dislocations. Qiao et al. [6] showed that hydrogen concentrated at the crack tip in response to plastic strain and a high dislocation density. The concentration increased with the static load [6,7]. Both hydrogen embrittling mechanisms may be operative, and both have the ultimate result of lowering the applied stress at the crack tip required for crack growth. 2.1. Effect o f strain and strain rate
The need for a cyclic stress emphasises the importance of strain in crack propagation [2]. Low-strength steels are not prone to hydrogen embrittlement under a constant load but severe cracking can occur when the stress intensity factor (K) is increased at a slow to modest rate [8]. The effect of strain has been attributed to enhanced hydrogen uptake and to hydrogen damage within the plastic zone that results from active plastic deformation. An effect of strain on the absorbed hydrogen concentration was demonstrated by Chen et al. [9], who found that the hydrogen permeability of X-70 steel was enhanced when loaded beyond 2/3 of its yield strength. Since elastic stresses do not appear to affect the hydrogen diffusion coefficient, the above observation implied an increase in the absorbed hydrogen concentration. Loading past the yield strength yielded significant plastic deformation, which halted further increases in hydrogen permeability by increased hydrogen trapping [9]. An essential role of strain and strain rate has been observed by other researchers [10-12]. The effect of strain rate was found to be most pronounced at intermediate values, with high strain rates resulting in ductile failure [13]. The latter was attributed to insufficient time for corrosion processes to generate hydrogen or for diffusion of hydrogen into the metal.
Volume 2: Prediction, Industrial Developments and Evaluation
257
2.2. Synergism between stress, hydrogen, and corrosion Experiments that have demonstrated an effect of hydrogen have predominantly involved slow strain rate tensile (SSRT) tests. However, these tests also allude to synergistic effects of hydrogen and the corrosive environment on crack growth. Precharging samples with hydrogen did not affect the SSRT test results in air, irrespective of strain rate, thereby indicating the need for a continuous source of hydrogen from cathodic protection or corrosion reactions [14]. Tests conducted over a range of potentials showed a reduction in ductility at OCP, which was attributed to facilitated entry of hydrogen in the presence of low corrosion rates [14,15]. Indications that absorbed hydrogen may in turn promote corrosion rates have been presented by Gu et al. [14], who showed that precharged steel samples displayed higher anodic currents. Calculations of the effect of hydrogen interacting with stress and dissolution indicated a possible increase in the crack growth rate by a factor of 7. Facilitated dissolution was attributed to a decrease in the atomic cohesion due to the presence of hydrogen in the plastic zone [5,14]. In the absence of hydrogen, strain was shown to increase the dissolution rate by an order of magnitude, presumably the result of continuously emitted dislocations at the crack tip surface [16]. Also, applied stresses may lower the crack tip electrochemical potential as compared to stress-free surfaces, thereby increasing the crack tip dissolution rate [ 17].
2.3. Expectation for the kinetic isotope effect Deuterium (D) can be used in laboratory experiments to identify the role of hydrogen in a complex process such as environmentally assisted cracking. D, a hydrogen (H) isotope, contains a neutron in addition to a proton and, therefore has twice the atomic mass of H. Whereas H and D share the same chemical behaviour, the difference in mass can influence chemical equilibria and the relative rate at which chemical and physical processes occur. For example, the dissociation constant of H20 at 25~ is an order of magnitude greater than that of D20, leading to slightly higher pH in D20 [ 18]. The kinetic isotope effect is particularly pronounced when the rate-controlling step of a chemical reaction is characterised by bond breaking. For the corrosion of steel in near-neutral pH groundwater solutions, the anodic dissolution mechanism probably involves a rate-determining oxidation step of the intermediate FeOH [ 19]. Although this is not a bond-breaking step, the reaction rate may be affected by the presence of the D isotope, thereby affecting free corrosion potentials. The active corrosion of steel is controlled cathodically by the reduction of water to form hydrogen, which does include a bond-breaking step. The latter is expected to result in a readily-measured isotope effect. Both adsorption and desorption of hydrogen atoms may also show a kinetic isotope effect based on the adhesive bonding energy with the surface. Physical processes such as diffusion also are influenced by a difference in mass. For two atoms with the same kinetic energy, the relative rates of travel will be inversely proportional to the square root of the ratio of their masses. This means that the diffusion coefficient for D is expected to be 0.7 (or x/89 ) times that of the hydrogen atom. The remainder of this
258
SCC and Hydrogen Embrittlement of Pipeline Steels
paper discusses the results of a number of experiments in simulated H20 and D20 groundwater solutions.
3. Experimentalprocedure 3.1. Materials
Specimens were prepared from X-65 line pipe steel taken from the excavation of a crock rupture site on an operating pipeline system. Compact tension (CT) specimens used in the cyclic loading tests were machined from X-70 line pipe steel that originated from the site of a hydrotest failure resulting from environmental cracking. The standard test solution consisted of simulated groundwater containing NaHCO3 (0.840 g/l), NaCI (0.016 g/l), and Na2SO4 (0.044 g/l) in either H20 or D20. The cyclic loading tests used a similar solution, with NaHCO3 (0.375 g/l), KHCO3 (0.128 g/l), CaC12"2H20 (0.063 g/l), and MgSO4"7H20 (0.034 g/l) dissolved in H20 or D20. All solutions were purged with 5% CO2 in nitrogen. 3.2. Electrochemical experiments
Corrosion rates were measured on polished rotating electrodes using electrochemical impedance spectroscopy (EIS) and linear polarization resistance (LPR) at 120 to 2340 rpm. The electrode surfaces were polished to 1000 grit. The corrosion current was calculated using the Stern-Geary equation assuming Tafel slopes of 120 mV/dec. Steady-state water reduction currents were obtained between-800 and-1500 mVscE on rotating electrodes at 400 rpm. Tests were performed under several electrode surface conditions: (i) "as received" millscale covered, (ii) polished to 1000 grit, parallel to the pipe surface, and (iii) polished to 1000 grit, perpendicular to the pipe surface. The polished surfaces were cathodically cleaned at-1500 mVscE for 3 min, which, in some tests, was followed by pre-oxidization at-700 mVscE for 60 min before the test. The potential was scanned either from-800 mVscE or-1500 mVscE in 20 mV increments, with a hold time of 30 min at each potential. 3.3. Hydrogen permeation tests
Hydrogen permeation tests were carried out for "as received" millscale-covered steel surfaces that were polished to 1000 grit using a test cell based on the design and technique described by Devanathan and Stachurski [20]. The charging side was left at the freely corroding potential. The data were analyzed to determine the diffusion coefficient of hydrogen in the steel (Dn) and the hydrogen concentration in the charging surface of the steel (Ca~ The latter was calculated from the hydrogen permeation current density (ie): it7/3 f , o
ip = n--"'H"t4
(1)
L
where F is the Faraday constant, n = 1 for the oxidation of hydrogen, and L is the specimen thickness.
259
Volume 2: Prediction, Industrial Developments and Evaluation 3.4. Mechanicaltests
SSRT tests were performed at OCP under displacement control equivalent to the strain rate of 8.4 x 10 -7 s -1. Time to failure (tf) and strain to failure (6f) were estimated from stress-strain curves. Percent reduction in area (%RA) was determined by inspection of the failed specimen through scanning electron microscopy (SEM). Cyclic loading tests were performed on compact tension (CT) specimens according to ASTM E 1820. Some tests were performed with a millscale-covered surface placed in close proximity to the crack and galvanically coupled to the CT specimen. The galvanic current flow between the crack area and the millscale was measured. (Note: Mechanical conditions are summarized in Figs. 8 and 9.) The test frequency was 0.001 Hz. Crack growth was continuously monitored by the potential drop technique.
4. Kinetic isotope effect in the corrosion of polished steel Fig.1 presents the corrosion rates based on the EIS and LPR measurements as a function of the rotational speed. The ratio of corrosion rates obtained in D20/H20 was 0.56 + 0.06, regardless of the rotation rate. This insensitivity to the rotation speed supports the hypothesis that a bond-breaking step in the cathodic reduction process controls the rate of the overall corrosion process.
5. Surface effect on hydrogen evolution in HzO and D20 solutions A number of studies were conducted to determine the effect of surface condition and orientation on the rate of reduction of H20 and D20. As shown in Fig. 2, the rate of H20 reduction on millscale-covered electrodes was approximately an order of magnitude greater than that on a polished and cathodically-cleaned electrode. Millscale
0.5 E E
o
o E,S ]
9
"
LPR
/
[] /
H20
[]
0.4
[]
D
tll
: 0.3 o w
/
,,m,,,
II
o 0,2 o 0.1
,
0
I
1000
=
i
,
2000
~
I
3000
Rotational Speed, rpm Fig. 1. Corrosion rates on a rotating polished line pipe steel electrode.
SCC and Hydrogen Embrittlement of Pipeline Steels
260
duplicate millscale tests
10-3 (I
,r e, 10-4
cleaned / pre-oxidized
,i,,e
e" L.
0=
1 O.S
cathodically ~c cleaned
10"6 ,
-1600
I
,
-1400
I
,
~~176
I
-1200
,
I
-1000
-800
Potential, m V s c E
Fig. 2. Steady-state reduction currents for 3 different surfaces at 400 rpm in H 2 0 .
10 "2
-
9 9 km
10-3
3
[]\ 9
i9
\ratio
O O4 m
200 mV/dec.) and ill-def'med Tafel slopes were observed, which are characteristic of a potential-dependent number of active surface sites [22].
261
Volume 2: Prediction, Industrial Developments and Evaluation
Fig. 3 compares of the rates of reduction of H20 and D20 on polished and cathodically-cleaned surfaces orientated perpendicular to the rolling direction. The rate of reduction of H20 was higher than that of D20 by a factor of 1.5-3. It is unknown whether the complex dependence of the current ratio on potential is a real phenomenon or an experimental artefact. Millscale-covered electrodes showed a similar higher rate of reduction of H20 over D20, although the isotope effect was less apparent for polished surfaces orientated parallel to the rolling direction. For both 1-120 and D20, the rate of hydrogen evolution was 1-5 times faster on the surface parallel to the plate rolling direction compared to the perpendicular orientation. The parallel surface may have exhibited greater activity because of a larger surface area of catalytic Fe3C-containing pearlite grains. The higher rates of hydrogen evolution both on the millscale-covered electrodes and for the orientation parallel to the rolling direction for polished electrodes suggest that water reduction would be favoured on the pipe Surface rather than in the crack itself.
6. Kinetic isotope effect in the absorption and diffusion of hydrogen Table 1 summarizes the kinetic isotope effect on the flux of diffusible hydrogen through a steel coupon. For polished specimens, the rate of hydrogen breakthrough is faster and the hydrogen permeation current, ia, is greater in H20 than in D20 (Fig. 4). The ratio of the diffusion coefficients obtained in the experiments was 0.6, which is in good agreement with the expected result of 0.7. Available literature values for the diffusion coefficients of H and D are 4.0 x 10-7 and 2.8 x 10-7 cm/s respectively, which compare favourably with the values in Table 1. The current that f'mally permeated the sample is affected by the rate of corrosion and the rate of hydrogen absorption on the charging site, followed by the rate of diffusion through the sample and f'mally by the rate of oxidation on the detection site. The ratio of the hydrogen permeation current in D20 and H20 is 0.5 + 0.1. Both a lower DH and a lower ia in D20 resulted in similar CH~ in D20 and H20 solutions (Eq. (1)). The ratio of average CH~ concentrations in D20 vs. H20 is 0.81 + 0.21. Although the rate of hydrogen evolution was higher on the millscale-covered electrodes (Fig. 2), the rate of hydrogen permeation was lower than for polished specimens (Fig. 4). The most likely explanation for the lower hydrogen permeation currents on millscale-covered surfaces is a surface-blocking effect. This conclusion is consistent with the observed increase in permeation current with time for the millscalecovered surfaces (Fig. 4), since the porosity of the millscale layer has been shown to increase with time upon exposure to solution [21 ]. Table 1 Summary of hydrogen permeation results (average of 2-3 results) Surface
Solution
Dn (10-7 cm~/s)
ia (ix.A)
Polished
5.04
1.99
Cri~ (mol/m3) 0.73
Polished
H20 D20
3.00
0.93
0.59
Millscale
H20
0.44
1.16
0.38
MiUscale
D20
0.20
0.48
0.27
262
SCC and Hydrogen Embrittlement of Pipeline Steels
H20 < 2,
~ o l i s h c ~ ! I
=.
H20 " cale
D O, polished
0
D20, millscale 0
25,000
50,000
75,000
Time, s Fig. 4. Hydrogen permeation currents through polished and millscale-covered coupons in H20 and D20.
7. Kineticisotopeeffectsin the presenceof dynamicloading SSRT tests are frequently used to evaluate a material's resistance to environmentally assisted cracking by using the degree of reduction in cross-sectional area upon failure (%RA) as an indicator. Our results are summarized in Table 2. Stress-strain curves (Fig. 5) demonstrate that the presence of hydrogen decreases the ductility significantly, but does not affect the strength of the steel. The %RA, el, and tf are smaller in H20 relative to D20. The ratio of the difference between the %RA in D20 and that in air to the %RA in H20 and that in air was 0.83 + 0.17. Although it may be premature to draw mechanistic conclusions based on this ratio, the ratio does compare fairly well with that obtained for CH~ in the hydrogen permeation tests, suggesting that the absorbed diffusible hydrogen concentration may be responsible for the loss of ductility. This argument is supported by the literature on the critical role of hydrogen in SSRT test failures [3,12,13]. The rates of corrosion, absorption, and diffusion are likely also affected by the effect of strain, but it is not clear what the isotope effect on these rates might be.
Table 2 Summary of slow strain rate tests results (average of 3 results) Environment
%RA
ef (%)
tf (h)
Air
69.5
17.9
62.9
H20 02~
44.2
13.8
48.3
48.6
15.0
51.9
Volume 2: Prediction, Industrial Developments and Evaluation
263
700 600 500 400
[
m
W W 0 I._
D20
300
Air H20
200 100 0
5
10
15
20
Strain, % Fig. 5. Stress-strain curves for polished steel samples in air, H20 and D20.
8.0x10-5 -
I
II
d o c 7.5xl 0-5 H20
in 7.0x10 -se"
.
o 6.5xl 0.5 -
6.0x10-5
0
D20 200
400
600
800
Time, h Fig 6. Resistance drop at OCP in H20 and D20.
Cyclic loading experiments were performed in D20 and H20 to obtain additional information on the relative roles of corrosion and hydrogen effects. The results are summarized in Figs. 6-9, with the corresponding crack growth rates presented in Table 3. The observed crack growth rates were smaller in D20 solutions to those in H20 solutions, with an average isotope ratio of 0.40 _ 0.16. The observed isotope effect suggests that the environment is important in crack growth over the entire range of test conditions. More rapid crack growth in H20 led to higher values of K and stress intensity factor range (AK - Km~, - Kmm) near the end of the run, which further promoted crack growth and decreased isotope ratios. In region IV of Figs. 7 and 9, the
264
SCC and Hydrogen Embrittlement of Pipeline Steels I
9x10 -5
II
III
VI
IV
a;
o c
.~_
H20
J
8x10 -5
w rr
.c 7x10"s
D20
2 ~
6x10.5 I
0
200
40O
,
I
8OO
60O
Time, h Fig. 7. Resistance drop of a CT specimen coupled galvanically to millscale surface in H20 and D20.
II
50-
Kmax, H20 '1-"
40-
. . . . . . . . . .
Kmax, D20 30-
AK, H20
20
AK, D20
10 0
!
0
i
20O
I
i
41111
I
61111
~
I
800
Time, h Fig. 8. Stress intensity factors at OCP in H20 and D20 (data from experiments in Fig. 6). crack growth rate became much lower in 020 upon decreasing AK from 18 to 13 MPa~/m. The reason for the greater drop in crack growth rate in D20 is not clear. If the isotope ratio were indicative of the crack growth controlling mechanism, a drop in this ratio would suggest a change in this mechanism or the involvement of other factors such as closure effects. With time, the crack growth rate increased and an isotope ratio of 0.54 was observed. Galvanic coupling to millscale-covered surfaces increased the crack growth rate by a factor of ~5 in both H20 and 020. Whereas the higher rate of hydrogen evolution on millscale-covered surfaces suggested that water reduction would be favoured outside a crack, a small cathodic current to the CT specimen suggested the opposite in H20. At the beginning of the experiment, this current was anodic, but it slowly decreased over a period of days to become cathodic. This may have resulted from the reduction of
265
Volume 2: Prediction, Industrial Developments and Evaluation
II
60
III
IV
V
I
~
so
9
40-
~
30"--
~
2o
/
K~max,D20 K, H20 _....•
) 1o 0
zxK, D20 I
0
Kmax, H20
2OO
,
41111
I
61111
i
I
81111
Time, h Fig. 9. Stress intensity factors (Km~x and zkK) of a CT specimen coupled galvanically to a millscale surface in H20 and D20.
Table 3 Kinetic isotope effects on crack growth rates, based on Figs. 6-9 Stage I II II III IV IV V
Figures 6 and 8 6 and 8 6 and 8 7 and 9 7 and 9 7 and 9 7 and 9
da/dt in H20 (ram/s)
5.89 x 10-6 9.2 x 10-7 1.21 x 10-6 4.75 x 10-6 3.10 x 10-6 3.10 x 10-6 4.95 x 10-6
da/dt in D20 (ram/s) 4.62 x 10-7 4.62 x 10-7 2.38 x 10-6 3.39 x 10-7 1.68 x 10-6 1.83 x l0 -6
O20/ n20
0.50 0.38* 0.50 0.11 0.54 0.37*
* Ratio of crack growth rates corresponding to higher K and Mr.
millscale catalytic surface sites and pitting corrosion affecting the surface activity. In D20, the galvanic current remained anodic, but continued to decrease. The mechanism by which the millscale increased the crack growth rate is not clear. It may involve increased dissolution and/or an increase in the supply of hydrogen, while maintaining a sharp crack tip.
8. Conclusions The observed isotope effect in the cyclic loading tests suggests that the environment plays a key role in crack growth in near-neutral pH environments. Galvanic coupling to millscale surfaces increased the crack growth rate, implying that the corrosion process is critical. Millscale surfaces were shown to catalyze the evolution of hydrogen, suggesting that water reduction would be favoured outside of the crack. The mechanism by which millscale surfaces affected the crack growth rate is not known. An important role of corrosion does not rule out an equally, if not more, important role of hydrogen. Synergistic effects of corrosion processes and hydrogen on crack growth coupled with
266
SCC and Hydrogen Embrittlement of Pipeline Steels
absorption and diffusion make it difficult to clearly differentiate between the relative roles of each factor and to determine the controlling or rate-determining step. Cyclic loading experiments avoided the large amounts of plastic deformation experienced in SSRT tests, which had an apparent effect on the crack growth mechanism. A smaller isotope effect in SSRT tests suggests an important role for hydrogen, and implies that these tests are not representative of cyclic loading tests.
References [1] Stress Corrosion Cracking on Canadian Oil and Gas Pipelines, Report of the Inquiry, MH2-95, National Energy Board, Calgary, November 1996. [2] J. Been, H. Lu, R. Eadie, G. Shen, R. Sutherby, The role of stress intensifiers in near-neutral pH corrosion fatigue of line pipe, CORROSION/2004, NACE International, Houston, 2004, paper no. 04552. [3] R.N. Parkins, J.A. Beavers, Corrosion 59 (2003) 258-273. [4] W. Chen, S.-H. Wang, F. King, T. Jack, M.J. Wilmott, Hydrogen permeation behavior of X-70 pipeline steel in a near-neutral pH soil environment, in: J.R. Ellwood (Ed.), Proc. 2000 International Pipeline Conference, vol. 2, ASME, New York, 2000, pp. 953-960. [5] H.J. Flitt, J.O'M. Bockris, Int. J. Hydrogen Energy 6 (1981) 119-138. [6] L.J. Qiao, J.L. Luo, X. Mao, Corrosion 54 (1998) 115-120. [7] M. Puiggali, S. Rousserie, M. Touzet, Corrosion 58 (2002) 961-970. [8] R.P. Gangloff, B.P. Somerday, D.L. Cooke, Understanding crack tip strain rate effects in hydrogen embrittlement for improved fitness-for-service modeling, in: Proc. CORROSION/96 Research Topical Symposia, NACE International, Houston, 1996, pp. 161-175. [9] W. Chen, M. Wilmott, T. Jack, Hydrogen permeation and surface deterioration of pipeline steel exposed to neutral pH SCC environment, NOVA Chemicals Report No. 01311, Calgary, 1999. [ 10] P. Andrews, M. McQueen, N. Millwood, Corrosion 57 (2001) 721-729. [11] S.D. Liberto, G. Gabetta, Mater. Sci. 33 (1997) 411-420. [12] D.L. Friant, B. Bayle, C. Adam, T. Magnin, Stress corrosion cracking of pipeline steels in simulated ground water: from mechanisms to a ranking test, EUROCORR'2000, The Institute of Materials, London, 2000 (CD-ROM). [13] S.X. Mao, J.L. Luo, B. Gu, W. Yu, Hydrogen facilitated anodic dissolution type stress corrosion cracking of pipeline steels in coating disbondment chemistry, in: Proc. 1998 International Pipeline Conference, vol. 1, ASME, New York, 1998, pp. 485-492. [14] B. Gu, J. Luo, X. Mao, Corrosion 55 (1999) 96-106. [15] R.N. Parkins, The involvement of hydrogen in low pH stress corrosion cracking of pipeline steels, in: Proc. EPRG/PRCI 12th Biennial Joint Technical Meeting on Line Pipe Research, held May 17-21, 1999, Groningen, Netherlands. [16] B.D. Lichter, H. Lu, W.F. Flanagan, Strain-enhanced dissolution: a model for transgranular stress-corrosion cracking, in: Proc. 2nd International Conference on Environment Sensitive Cracking and Corrosion Damage, Nishiki Print Inc., Hiroshima, 2001, pp. 271-278. [17] O.V. Kurov, Corrosion 49 (1993) 315-319. [18] J.J. Katz, Deuterium and tritium, in: Kirk-Othmer Encyclopedia of Chemical Technology, 3rd ed., vol. 7, John Wiley & Sons, New York, 1979, pp. 539-553. [19] J.O'M. Bockris, A.K.N. Reddy, Modem Electrochemistry: An Introduction to an Interdisciplinary Area, vol. 2, Plenum Press, New York, 1973, pp. 1080-1093. [20] M.A.V. Devanathan, Z. Stachurski, J. Electrochem. Soc. 11 (1964) 619-623. [21] Z. Qin, B. Demko, J. No~l, D. Shoesmith, F. King, R. Worthingham, K. Keith, Corrosion, in press. [22] E.R. Vago, E.J. Calvo, J. Electroanal. Chem. 339 (1992) 41-67.
267
The roles of crack-tip plasticity, anodic dissolution and hydrogen in SCC of mild and C-Mn steels D. Delafosse, B. Bayle, C. Bosch SMS Division, UMR CNRS 5146, Ecole Nationale Supdrieure des Mines, 158 cours Fauriel, 42023 Saint-Etienne, France
Abstract
A two-part study of the stress corrosion cracking (SCC) and hydrogen-induced SCC of structural steels was conducted. The use of various microstructure-environment-mechanical loading combinations allowed the highlighting of specific roles of anodic dissolution, hydrogen and crack-tip plasticity in cracking processes. The first part of the study deals with mild steels exposed to a dried methanolic solution. It was shown that micro-galvanic coupling between carbides and ferritic matrix was dominant during crack initiation, and remained important during crack propagation. In the second part, the near-neutral pH SCC of two pipeline steels exposed to simulated groundwater was investigated. In this case, hydrogen strongly affected the resistance to SCC in the presence of MnS precipitates. The specific role of crack-tip plasticity in SCC was highlighted using specially-designed micro-notched tensile specimens. I. Introduction
Environment-sensitive fracture of carbon or micro-alloyed steels occurs in a variety of environments. In many instances, both anodic dissolution and hydrogen are involved in the fracture process, and they also interact with crack-tip plasticity. From an experimental point of view, separating their specific roles in the process of cracking is not straightforward, especially when the roles vary during crack propagation. In this paper, we present results obtained from two projects that were recently completed. The use of various microstructure-environment-mechanical loading combinations allowed the highlighting of specific roles of anodic dissolution, hydrogen and crack,tip plasticity in the fracture process. In particular, the study of stress corrosion cracking (SCC) of mild steels exposed to a methanolic solution (Section 2 of the paper) showed that micro-galvanic coupling between carbides and the ferritic matrix prevailed during crack initiation and remained important during crack propagation. In addition, some hydrogen effect on SCC in the methanolic solution was observed. Section 3 describes the "near-neutral pH SCC" of two pipeline steels exposed to simulated groundwater. In this case, hydrogen strongly affected the resistance to
SCC and Hydrogen Embrittlement of Pipeline Steels
268
SCC when MnS precipitates were presented. In Section 4, a specific role of crack-tip plasticity is highlighted using the results obtained with micro-notched specimens. 2. Mild steels in dried methanolic solution
2.1. Experimentalprocedure A number of SCC cases have been observed in the last few years on mild steel structures which were in the contact with methanol (CH3OH) used as a moisture scavenger in propane gas tanks [ 1,2]. Stress corrosion cracks were located in the lower parts of cisterns, a zone bathed with methanol. Methanol degradation produces formic acid (HCOOH) and water in variable amounts, in both commercial and denatured methanol. Water also comes from methanol pollution and hydraulic proof test of cisterns. In this study, SCC tests were conducted in a methanolic solution containing 150 ppm HCOOH and 0.3% H20, typical for in-service conditions. Before testing, the solution was dried with one of standard procedures. Also, 0.1 M lithium perchlorate (LiCIO4) was added to the solution in order to increase its electrical conductivity. The measured pH was --4. Three steels were used for studying microstructural effects on SCC (Table 1). The A48AP steel had a banded ferritic-pearlitic microstructure, whereas P420M and TUE360 were ferritic steels with carbide precipitates. A48AP and P420M steels were hot-rolled as recommended for cistern manufacturing. Heat treating of P420M caused carbide coalescence at grain boundaries. Some A48AP specimens were austenitised and then cooled at various rates to obtain mixed microstructures of ferrite, bainite and pearlite. (Note: TUE360, a newly-designed pipeline steel of low-sulphur (S) grade, was also used in Section 3 for comparison with the pipeline steel X52, both of which are widely-used as components of oil and natural gas pipelines.) SCC behaviour was studied with a slow strain rate tension (SSRT) technique at the strain rates of 1.0 x 10 -7 and 6.6 x 10 -7 s -l u s i n g smooth and pre-notched (120 ~tm in depth) specimens. After testing, the fracture surfaces were examined by optical and scanning electron microscopy (SEM). Table 1 Chemical compositions and typical mechanical properties of the steels Steel A48AP
C Mn P (wt.%) (wt.%)(ppm) 1.83 1.48 140
S (ppm) 14 MPa are the only solutions able to meet all requirements for natural gas transmission pipelines longer than 3000 km [1]. However, it is known that susceptibility of steels to environmentally-assisted cracking (EAC), caused by the synergic action of an environment and tensile stresses, increases with their strength. This means that the use of high-strength steels for pipelines
292
SCC and Hydrogen Embrittlement of Pipeline Steels
operating under a high internal pressure will increase the risk of EAC. As a result, EAC will be a critical factor for improving the integrity of high-pressure pipelines. The cracking behaviour of traditional pipeline steels with yield strength of ~450 MPa (65 ksi) is well known from both field experience and laboratory studies. Mechanisms that were proposed to describe the EAC behaviour of such steels include: (i) carbonate (CO32-)-bicarbonate (HCO3-) stress corrosion cracking (CB-SCC); (ii) near-neutral pH stress corrosion cracking (NN-SCC); and (iii) hydrogen embrittlement caused by the application of cathodic protection (CP). However, in the literature there is not enough data about the EAC behaviour of newly-designed high-strength pipeline steels, including micro-alloyed X 100 steel. In the mid-1990s, four Italian companies, namely Eni Gas & Power S.p.A., Snam Rete Gas, Saipem, and Snamprogetti, formed an Eni group that has sponsored two projects on long-distance natural gas pipelines to verify design criteria, materials behaviour and construction techniques. During the period 1995-1997, research activities were aimed at evaluations API 5L X70 and X80 pipeline steels. Between 1998 and 2001, research activities focused on X100 pipeline steel. Most recently, in 2003, a new Eni Development Project was initiated on high-pressure pipelines. The project is being executed mainly within the same Eni group collaborating with research teams at the University of Bergamo and Centro Sviluppo Materiali on a number of specific technological issues. This project aims to evaluate worldwide technological solutions and practices by means of full-scale and laboratory tests. Research activities at the University of Bergamo deal with the EAC behaviour of X 100 steel. In this paper, some preliminary results that were obtained using two experimental heats of X 100 steel are discussed. 2. Experimental procedures 2.1. Materials
The chemical compositions and longitudinal mechanical properties of X100 steels tested in the study are given in Tables 1 and 2. Steels were produced through a thermomechanical treatment with accelerated cooling in order to obtain a microstructure that includes ferrite, martensite and bainite.
Table 1 Chemical composition of X 100 steels Steel A B
C Mn Si P S Cr 0.07 1.96 0.34 0.035 0.007 0.03 0.07 1.97 0.33