Development and Validation of an Integrated Intergranular Corrosion/Cracking Model of Al-Mg Alloys for Naval Applications

Transcription

1 Development and Validation of an Integrated Intergranular Corrosion/Cracking Model of Al-Mg Alloys for Naval Applications R. G. Kelly, J. R. Scully, R. P. Gangloff Center for Electrochemical Science and Engineering Department of Materials Science and Engineering University of Virginia Charlottesville, VA Submitted to: Dr. Airan Perez Sea Platforms and Weapons Division, Code 333 Office of Naval Research Arlington, VA November, 2007

2 Statement of Work We aim to develop a cross-discipline, fully integrated, quantitative model of intergranular corrosion (IGC) and its transition to intergranular stress corrosion cracking (IGSCC). Our approach to this generic problem on the impact of corrosion on structural integrity will focus on AA5XXX alloys. We aim to describe the evolution of damage from a distribution of intergranular corrosion sites to multiple growing intergranular cracks (in the presence of applied or residual-secondary stress). The integrated model will consist of a series of individual state-ofthe-art damage models for the controlling processes. These damage process models will be based on fundamental scientific principles at the frontier of current understanding, and each will be substantially advanced by the new scientific understanding developed as part of this proposal. Each process model will be specifically constructed to pass inputs/outputs to one another as needed. Criteria for transition from one corrosion mode to the next will be addressed and incorporated. In addition, the models will provide outputs of damage in a probabilistic framework to allow for quantitative risk assessment. Experimental validations of model components will be performed at the coupon level, and the integrated model will be also validated for a coupon under controlled laboratory conditions. Successful completion of this validation in an initial phase will lay the foundation for further development of the tool for specific engineering applications under 6.2 and higher levels of funding, and will provide clear directions for further scientific breakthroughs. Our ultimate goal is to impact alloy development and component sustainability, in engineering context, with a user-friendly toolkit based on fundamental-mechanism based damage models. Problem Statement The use of AA5XXX alloys is increasing across DOD due to their combination of tunable strength/ductility, weldability, and corrosion resistance. These aluminum alloys offer the ability to decrease the weight of ship/vehicle structure, thus increasing speed, range, and fuel economy. Although the physical and mechanical metallurgy of these alloys is well characterized, there remains a need for a computational tool that predicts the effects of environment (chemical and thermal), alloy composition and heat treatment, and stress on the development of intergranular corrosion and intergranular stress-corrosion cracking (IG/IGSCC) of AA5XXX-type alloys. There is no current model of IG/IGSCC for Al alloys that allows such calculations to be made in a damage tolerant framework that links quantifiable corrosion damage to performance affecting attributes of a naval vehicle. Structure designers need such a tool for material/heat treatment selection and structural detail design to minimize risk and maximize long-term performance. Fleet support engineers need such a tool for improved condition-based maintenance decisions leading to cost-effective sustainability. Although substantial service experience exists, including clear examples of IGC/IGSCC failures, very little of that experience can be used to manage current and future corrosion problems involving these alloys as many of the critical, controlling parameters are either unknown or not quantified for the specific failures. Moreover, a wide range of variables affect this interactive damage mode; as such, behavior can only be described by a modeling approach grounded on newly developed and fundamental science with highly focused experimentation.

3 A comprehensive model of IG/IGSCC must be based on a validated fundamental scientific understanding of the controlling processes rather than an empirical database. While a description of IG/IGSCC based on empirical testing alone may be able to rationalize corrosion damage for the metallurgical and exposure conditions used in its development, it is not robust in that it has no ability to confidently predict damage evolution for other conditions that are unavoidably encountered in long-term naval service. In addition, in the absence of a fundamental understanding, the question of how different the metallurgical and exposure conditions must be to be of importance in a given application cannot be answered. The current state-of-the-art in corrosion science has not provided sufficiently predictive models of IGSCC of Al alloys exposed to marine environments. Critical gaps include scientific understanding of the factors that control the kinetics of controlling parameters. Kinetics that are neither quantitatively characterized nor fundamentally understood include sensitization rates, the rates of IGC along sensitized grain boundaries, development of IG stress-corrosion cracks from IGC damage, the rate of propagation of IGSCC along single grain boundaries, and the development of crack distributions in a grain network. This problem is representative of the state-of-the art challenge facing computational materials science covering multiple length scales. That is, kinetics and damage evolution are governed at atomic through microstructural lengths, but the impact on engineering properties is at meso (crack distribution and macro-growth behavior) and macro-structural response scales. A software tool based on a comprehensive IGSCC model addresses the recommendation of the recent DSB report on corrosion control 1 which advocated that the DOD invest in toolsets which could predict corrosion damage based on a combination of experimental data and model results. Our ultimate goal is to impact alloy development and component sustainability, in engineering context, with a user-friendly toolkit based on fundamental-mechanism based damage models. The path to this goal will result in cutting edge corrosion and materials science that will provide a roadmap for similar problems that limit other alloy systems relevant to high performance naval applications. General Technical Background Aluminum alloys are being considered more strongly for structural applications by all sectors of DOD due to their high strength-to-weight ratio. Non-heat-treatable alloys such as the 5XXX series (Al-Mg, i.e., Al with Mg as the main solid solution alloying element) have a combination of good strength and ductility that are tunable via work-hardening, grain size and crystallographic texture control. In addition, the commercial compositions are weldable as they do not precipitate deleterious phases when the process is conducted appropriately. AA5XXX alloys have been successfully used for hull structure in commercial ferries and US Navy patrol boats. Some recent naval applications of 5XXX alloys include 5456-H116 deckhouses, H116 buoyancy boxes, 5086 small craft, hull structures for the Littoral Combat Ship (LCS) fabricated from welded 5083-H321 and 5383-H Defense Science Board Report on Corrosion Control, Report A767824, Office of the Undersecretary of Defense for Acquisition, Technology, and Logistics, Washington, DC (2005).

4 In order to meet the requirements of Sea Power 21, the Navy is designing and building a variety of high-speed ships and craft. The 5XXX aluminum alloys are candidates for hull and structural materials due to their high strength-to-weight ratio, fabricability, cost, and availability. Although other services and industries routinely use aluminum structures in corrosive environments, their methods and experience are not directly transferable to Navy ships, making 5XXX alloys unique technologies for the Navy. The corrosion resistance of the non-heat-treatable 5XXX alloys is in general superior to that of the higher strength precipitation hardened alloys such as AA2024 and AA7075. That said, AA5XXX alloys are susceptible to pitting, intergranular corrosion (IGC) and intergranular stress-corrosion cracking (IGSCC) due to the precipitation of the β phase (Al 3 Mg 2 ) localized to grain boundaries and when greater than 3.5% Mg is present in the alloy. The higher Mg is required for higher strength (via solid solution hardening), but this intergranular precipitation of β, known as sensitization, leads to a preferred path for corrosion and associated stable crack evolution. This corrosion damage can provide nucleation sites for fatigue cracking as well as acting as the precursor for catastrophic fracture, loss of load bearing capability, or stiffness. This general problem of the impact of corrosion on structural integrity is broadly relevant to many DOD systems and was identified as a high priority issue for fundamental research by a recent AFOSR/ONR sponsored workshop held in Arlington, VA (December, 2006). The scientific issues are summarized in the ensuing section on Technical Challenges. In the 1960 s, tempers were developed whose aim was to reduce the susceptibility of the higher Mg-containing 5XXX alloys to sensitization. Temper H116 was developed to reduce the susceptibility to exfoliation in which grain boundaries are preferentially attacked leading to a macroscopic fissuring of substantial consequence. Strain hardening and thermal treatment are used to try to homogenize the β precipitation throughout the grain, leading to lower levels of Mg available for later sensitization by precipitation of β at the grain boundaries. While an improvement, Navy studies [Vassilaros, 1979] have shown that even properly heat-treated materials can be sensitized in service in as little as 5 y when exposed to the normal solar field associated with above deck structures. Recent failures of RAST support blocks made of AA5456-H111 were concluded to be due to in-service sensitization and subsequent IGSCC [Davis, 2002]. Although Alcoa has shown that the addition of Sc can improve the performance of AA5XXX alloys, these new alloys are proprietary and thus not under consideration for use by DOD. In addition, the mechanism(s) of the improvement, as well as the breadth of improvement across important loading and environmental variables, are unclear. Technical Objectives The needed suite of computational tools referred to in the Problem Statement should have the following characteristics: 1. It should be usable to answer questions at several levels of sophistication. 2. It should be founded on a demonstrated, fundamental scientific understanding of the controlling processes rather than an empirical database alone.

5 3. It must be in the form of a probabilistic framework to move from go/no-go decisions to quantitative risk assessment of the impact of corrosion damage of mission capability. This approach allows one to operate with more reasonable performance margins, increasing weapon system readiness and decreasing maintenance costs without an openended increase in risk. The approach recognizes the inevitability of corrosion damage, but allows management of the impact of damage on structural integrity. 4. It should consider the fact that both microstructures and component geometries are intrinsically three-dimensional in shape and stress distribution. 5. It should be modular to allow improvements in understanding and availability of data to improve its performance. In the proposed work, spiral development will be used to mitigate the risks inherent in the multilength scale modeling of processes that are not completely understood from the scientific perspective. Each spiral has a specific technical objective that can be described in terms of the model that results from the work: Spiral 1 Objective: Continuum-Scale Model of IGSCC Times-to-Failure Given an estimate of distribution of corrosion damage size topography, degree of sensitization (DOS), the marine environment, and stress; this Spiral 1 model will provide predictions of the distribution of time-to-failure by IGSCC for an axially-loaded material of known metallurgy. Such a model requires a scientific basis for the following: a. A means of quantitatively assessing prior corrosion damage in terms that are amenable to fracture mechanics modeling to provide input to the NDI community on what to try to measure; b. A means of determining DOS rapidly and quantitatively in a manner that is amenable to the development of a fieldable probe by the NDI community. Spiral 2 Objective: Quantitative Scientific Model of DOS and IGC Damage Evolution, and Continuum-Scale Model of IGSCC Times-to-Failure Given the applied stress, the distribution of grain boundary character, the exposure time, temperature and environment, this Spiral 2 model will provide, for an axially loaded 5XXX alloy of known metallurgy, several predicted probability distributions as outputs: a. the DOS; b. the IGC damage depth and shape characteristics; c. the time-to-igscc formation about IGC and thus improved time-to-failure prediction. This enhancement of the Spiral 1 model will strengthen the predictive capabilities as well as provide the foundation for the fundamental modeling of the components of IGSCC crack propagation. Spiral 3 Objective: Quantitative, Fundamental Model of IGSCC Damage Evolution Given the applied stress, the distribution of grain boundary character, the exposure time, temperature and environment, this Spiral 3 model will provide, for an axially loaded 5XXX alloy

6 of known metallurgy, the predicted probability distributions described for Spiral 2 enhanced as follows: a. Replacing the empirical model of crack velocity as a function of stress intensity and DOS with one that is based on an integrated, mechanistic approach to AD and HE. b. Predicting first a 2-D crack growth rate incorporating HE and AD aspects and then constructing a three-dimensional microcrack distribution using the integrated mechanistic model c. Predicting macro-crack growth kinetics necessary for component level modeling d. Transitioning the integrated mechanistic model into a code to interact appropriately with the DOS model, IGC damage evolution model, and IGSCC formation model to create a Spiral 3 model. Validation Objective: Experimental Validation of Each Element and the Integrated Model. For the purposes of this program, validation will focus on a single alloy and temper (5083 with the temper to be selected in consultation with Alcoa) in a single product form and orientation (plate material with cracking along the ST plane), a single environment (full immersion artificial seawater, constant potential), and sensitization at a single temperature (100ºC). This focus will allow the ability to predict the distributions of the different outputs to be assessed. Validation experiments will focus on measurements of the key predictions for each model. For Spiral 1, the validation will focus on the IGSCC time-to-failure and effective-starting corrosion size distributions for different levels of DOS. For Spiral 2, the validation will focus on the distributions of DOS, IGC damage characteristics (depth, geometry), and crack formation life. For Spiral 3, the validation will focus on microcrack distributions and single crack da/dt vs. K relationship as a function of DOS. Technical Challenges The proposed research aims to improve ship-component design by allowing for optimization of corrosion resistance or by providing quantification of the effects of corrosion that can contribute to decision making, aid condition-based maintenance strategies protocols to guarantee structural integrity, and contribute to corrosion-informed alloy selection/development. These outcomes are enabled by producing mechanism-based descriptions of corrosion and stress-corrosion cracking as well as interactions and transitions, then implementing these scientific results into a software tool useful for Navy engineers. This work targets several cutting edge scientific issues, such as interacting damage mechanisms modeled at the microstructure scale. Critical technical hurdles are as follows: Laboratory experiments must be well designed and highly focused in order to: (a) control and capture the effects of a wide range of electrochemical, metallurgical and mechanical variables known to affect the kinetics of IGC, IGSCC, and transitions; and (b) efficiently establish the distributions of microstructure-scale processes occurring over

7 three dimensions and the associated macroscopic properties relevant to component-level analysis. Mechanism-based scientific modeling must: (a) capture in quantitative form several highly localized corrosion and cracking processes that can provide outputted damage evolution based on laws that accept key materials and environmental inputs as well as time. Currently, damage evolution is modeled with adjustable parameters necessitated by gaps in understanding and analysis capability - the scientific basis justifying adjustable parameters must become better understood. ; (b) connect these elements of corrosion and cracking in a physically realistic way challenged by the possibility of multiple damage features that interact, and (c) predict macroscopic properties from a distribution of microstructure-level corrosion and cracking. In all aspects of measurement and model building, it is necessary to decide on the proper size scale for observation and modeling, necessitated by fact that each form of damage is governed by processes ranging from the atomic level to the micrometer scale while each material property of importance is probed by a macroscopic measurement. Macroscopic and micrometer scale inputs must enable appropriate descriptions of the governing damage events. Throughout this program, it will be necessary to develop the interdisciplinary collaborations necessary to produce step-out improvement in understanding and modeling. When scientific obstacles become insurmountable, the challenge is to develop physics-informed approximations that can be upgraded with future research to fill in these gaps in knowledge. Technical Approach The damage tolerance approach in solid mechanics assumes the presence of flaws and involves the design of structures using materials that can withstand the design stresses without unstable fracture or plastic instability even in the presence of these flaws. In service, one must also consider a corrosion-induced increase in structure compliance and displacements due, for example, to an array of cracks. In corrosion, the failure of protection systems (i.e., coatings) has been historically difficult to predict, thus a corrosion damage tolerance approach would assume the failure of these systems and focus on the rate at which the damage would evolve from this coating failed condition. Rarely is the earliest initiation of corrosion a structural integrity issue, although it might affect the total time until failure. It is the evolution from that initiation site of corrosion damage that eventually threatens the reliable performance of the structure. The approach described below quantitatively models the corrosion damage evolution processes for AA5XXX alloys exposed to saline environments of interest to DOD, and specifically the Navy. The first phase of the project (FY08-FY12) would provide the scientific foundation for the model components and the integrated model as well as experimental validation under controlled laboratory conditions.

8 The proposed program consists of three spirals of development. Each spiral will produce improved scientific understanding through the discovery of the fundamental, controlling mechanisms underlying the IGC and IGSCC damage processes. This quantitative understanding will be implemented into the form of a series of software tools of increasing sophistication and fidelity. The proposed program is organized around the concept of a model that predicts the effects of environment (chemical and thermal), alloy composition and heat treatment, and stress on the development of IG and IGSCC of AA5XXX-type alloys. The model would provide outputs of damage in a probabilistic framework to allow quantitative risk assessment. The concept is depicted schematically in Figure 1. The input data are applied within the software tool as the drivers for two process models, one that predicts the extent and nature of the IGC damage, and one that uses that damage as input to predict the damage evolution via IGSCC. Software Tool INPUTS DOS σ app t expose Environment IGC Damage Model Corrosion-Induced Damage IGSCC Damage Evolution OUTPUT Probability Distribution of Corrosion Damage Figure 1. Flow chart of approach to IGC and IGSCC modeling. The three spirals of development consist of increasingly sophisticated scientific approaches to the process models. The work of each spiral provides a foundation and guidance for the next in each process model. Spiral 1 will produce a model of time-to-failure by IGSCC using a continuum-scale approach for both process models. Given an estimate of distribution of corrosion damage size topography, degree of sensitization (DOS), the marine environment, and stress, the Spiral 1 model will provide predictions of the distribution of time-to-failure by IGSCC for an axially-loaded material of known metallurgy, as shown in Figure 2. The resulting software tool (developed in collaboration with VEXTEC) will be available for transition to fleet support engineers to assist in the estimation of damage evolution in order to render better decisions regarding repair/replacement of current AA5XXX structures. It will also provide the framework in which the refined models would be used to enhance the predictive abilities of the tool. In Spiral 2 (Figure 3), the IGC Corrosion Damage Model is enhanced by introducing microstructural-scale models that enable prediction of the distribution of β phases on grain

9 boundaries (Prof. Free, Utah) and the distribution of IGC damage given the applied stress, DOS, and the time of exposure in a given environment. The IGSCC Damage Evolution Model will be advanced through the introduction of a microstructure-based model for the time-to-formation of a stress-corrosion crack from previous IGC damage in order to increase the accuracy of the timeto-failure predictions in cases where the time to form the crack is significant. These enhancements of the Spiral 1 model will strengthen the predictive capabilities as well as provide the foundation for the fundamental modeling of the components of IGSCC crack propagation. Spiral 1 Software Tool INPUTS DOS σ app t expose Statistical IGC Damage Model f(σ,dos,t expose ) Corrosion-Induced Damage Continuum-scale Model of IGSCC Time-to-Failure f(σ,dos,igc damage) OUTPUT DOS IGC Damage IGC Damage Database Empirical (da/dt) = f(k,dos) t f a/c DOS Figure 2. Flow chart of Spiral 1 modeling showing the two specific approaches used for the IGC Damage Model (green, on left) and the IGSCC Damage Model (blue, on right). The expected dependencies for each model are indicated as f(variables). In addition, during Spiral 2 the results of grain boundary engineering (GBE) efforts by Prof. Allen at Wisconsin will be analyzed both experimentally and in terms of the microstructural variables that affect β formation, IGC and IGSCC via the Spiral 2 process models. The goal of the GBE is to create grain boundaries that are substantially less susceptible to sensitization and/or less susceptible to corrosion/cracking without sacrificing bulk mechanical properties. Three software tools will result from this work: an enhanced time-to-failure model, a model that predicts the level of grain boundary precipitation and DOS, and a model of IGC damage evolution. In Spiral 3 (Figure 4), further advances are made to both process models. In the IGC Damage Model, the effects of grain boundary misorientation (in terms of the grain boundary character distribution (GBCD)), applied stress, and three-dimensionality of time dependent IGC damage will be incorporated. In the IGSCC Damage Evolution process model, a quantitative-mechanism model would be developed that integrates models of crack advance by anodic dissolution and hydrogen embrittlement. Such integration represents a Grand Challenge in the field of corrosion science. Success would allow the evolution of crack networks in to be predicted as well as the IGSCC velocity vs. K curve.

10 INPUTS Spiral 2 Software Tool OUTPUT GBE Science-based IGC Damage Model f(σ, β distrib, DOS, t exp ) Corrosion Induced Damage Size and a/c Enhanced Continuum IGSCC Model f(σ,dos,igc damage) DOS β distrib t f GBCD T(t) σ app Science-based DOS Model f(gbcd, T(t)) DOS Empirical (da/dt) = f(k,dos) Scientific N Form Model t expose DOS IGC Damage a/c Figure 3. Flow chart of Spiral 2 modeling showing enhancements in IGC and IGSCC modeling as well as additional outputs of IGC model. INPUTS GBE Science-based IGC Damage Model f(σ, GBCD, DOS, t exp ) Spiral 3 Software Tool Corrosion Integrated Induced IGSCC Damage Size Model for Grain and a/c Network OUTPUTS t f β distrib GBCD T(t) σ app t expose Science-based DOS Model f(gbcd, T(t)) Cracking on Single Grain Boundary Hydrogen Anodic Embrittlement Dissolution IGSCC Model IGSCC Model f(σ,β distrib, GBCD, t exp ) f(σ,β distrib, GBCD, t exp ) DOS DOS IGC Damage Figure 4. Flow chart for Spiral 3 modeling. Inputs are the grain boundary engineering methods, resulting in a GBCD, the time-temperature profile, the applied stress, and the time of exposure. Outputs are predictions of: time-to-failure distributions, 3-D crack networks, v(k) curves, and the geometry of the IGC damage.

11 The software tool development would transition the integrated mechanistic model into a code to interact appropriately with the DOS model, IGC damage evolution model, and IGSCC formation and growth. The remainder of this Technical Approach section describes a more detailed plan of work for each of the process models as well as the integration of the process models into a software tool. Within each process model, the characteristics of each spiral are detailed. Intergranular Corrosion-Induced Damage Modeling The Spiral 1 has two main foci: (a) development of a statistical model of IGC damage evolution as well as the database of damage needed for Spiral 1 in the IGSCC Modeling, and (b) development of the kinetics of grain boundary dissolution, including the development of the basis for an electrochemical DOS measurement. In Spirals 2 and 3, IGC modeling will focus on prediction of the distribution of IGC damage given the distribution of β phases on grain boundaries, applied stress, DOS, and the time of exposure. In Spiral 2, the basic model will be established. In Spiral 3, additional sophisticated aspects such as effects of GBCD, applied stress, and three-dimensionality of time dependent IGC damage would be incorporated into the basic model. The model would be validated by comparison to experimental results on IGC damage acquired using the ASTM G 67 method over a range of sensitization levels. Spiral 1: Statistical IGC Damage Model Task 1.1: Create range of IGC damage as f(dos, t), quantify in terms of effective a, a/c, b in 3D, and model the damage evolution. Goal: Our goal in this task is to be able to control and model the three-dimensional geometry of an IGC site on the surface of a tensile sample that will then be subjected to constant load SCC testing. The geometry is defined in Figure 5: a is the damage depth in the direction of crack propagation, b is the height of the damage in the direction of the applied load, and c is the width of the damage perpendicular to the applied load. The stress intensity factor that transitions this damage into a crack is controlled by these geometric parameters and the applied load. The Continuum-Scale Model of IGSCC requires this geometry as input. In this task, the evolution of this damage and its dependence on DOS will be characterized and modeled, allowing the Spiral 1 inputs (DOS and t expos ) to determine the damage that is used by the Continuum-Scale Model along with the applied stress to predict time-to-failure. Method: In order to predict distributions, a sufficiently large amount of data is required. Thus, it is important to have a methodology that efficiently creates a large number of IGC damage sites and subsequent IGSCC events. Two approaches will be used. Both involve limiting the area of attack to a site or sites within the gage length of a dog-bone style tensile sample. The difference between the two methods is the size scale of the area probed. The first method we will use a coating to limit the regions available for IGC attack to regions ca. 6 x 6 mm. The microstructure distribution will determine the geometry of the damage site, but the control of the area of attack will allow the mechanics to remain well-defined. Multiple crack initiation sites will be generated, thus increasing the number of data points obtained per sample.

12 σ c b a Figure 5. σ Definitions of geometry of corrosion damage with respect to loading axis. In the second method, we will use custom-shaped mini-capillaries to limit the area of attack. The shape of the damage site in the plane of the tensile sample (i.e., dimensions b, c) will be controlled through manipulation of the geometry of a small (200 µm diameter) glass tube. Fundamentally, this method is the same as the first, but each of these areas will be much smaller than for the first method (0.03 mm 2 vs. 36 mm 2 ). This method will allow specific areas of the microstructure to be probed. A light etching would reveal the β phase distribution for grain boundaries intersecting the surface. The microcapillary could then be used to focus the attack on areas of differing β phase distribution. This method will be especially useful in Spiral 2 and 3 when localized electrochemical and damage kinetics are probed. In Spiral 1, three shapes will be studied, a circular area and two ovals of different eccentricity. The oval shapes will be created by the heating the end of the glass tube and flattening it with controlled pressure. The depth of the corrosion attack (i.e., dimension a) will be controlled by the length of exposure time to the ASTM G67 solution. It has been reported [Craig, 1972] that the mass loss is linear with time for both sensitized and unsensitized material, although the slopes (i.e., the rates) are dramatically different between the two. Four levels of DOS (as determined by ASTM G67: 0, 25, 50, and 80 mg/cm 2 ), and three exposure times (1, 2, and 5 days) will be used to prepare samples for the IGSCC testing. Five replicates will be used for each condition, resulting in 60 samples being used to populate the damage distributions. From each sample using method 1, we expect between 5 and 10 areas of IGC to initiate cracks upon IGC testing. Thus, the database would contain between 300 and 600 data points. After exposure, each sample will be cleaned according to G-67 and the region of damaged surface analyzed by confocal laser scanning microscopy (CLSM). The CLSM will give a quantitative image of the damage available via line-of-sight. The analysis of the topographic image so created will be analyzed in terms of a, b, and c for input into the fracture mechanics model. A subset of samples will be sacrificed for analysis via cross-sectional metallography to determine the extent of undercutting of the surface. The samples will then be exposed to the test environment at constant potential in a constant load apparatus and monitored for failure. Posttest analysis will confirm the geometry of the IGC damage at which each crack initiated.

13 The Statistical IGC Damage Model will also utilize the results from the IGC damage experiments as the source of the data for its underlying database. The Spiral 1 IGC model will produce the outputs of interest to the Continuum-Scale IGSCC model in terms of the dimensions of the IGC damage: a, b, and c as well as their ratios (a/c, b/c, and a/b). The model will use a brick model along the lines of that used Ruan et al. for AA2024-T3 [Ruan (2004)] and AA7178- T6 [Zhao (2007)]. Whereas the goal of Ruan s model was the calculation of the minimum intergranular corrosion path, the focus of the proposed work will be the distributions of the dimensions of IGC damage. For intergranular corrosion in aluminum alloys, it has been found [Zhao (2007), Zhang (2002)] that the time-dependence of the depth of penetration, d, along a given plate processing direction (i.e., L = rolling or longitudinal direction, T= transverse, S= short) follows a power law: n d = d o t Eq. 1 In addition, in high strength aluminum alloys, the value of d o, which can be interpreted as the corrosion rate along the grain boundary, varies with the direction of growth in the rolled material, with the order being d o (L) d o (T) >> d o (S) for both AA7178-T6 [Huang (2007)] and AA2024- T3 [Zhang (2002)]. Thus, for the proposed work, the geometry evolution of the damage will be analyzed in terms of power law functions of the time of exposure, for example: a = a o t n a a = c c o t b Eq. 2 Both a o and n will be likely functions of DOS and exhibit a distribution of values. The fundamental parameters underlying the mean values and the distributions will be part of the focus of Spiral 2. Non-linear regression of the CLSM and metallography data will allow determination of these parameters in this Spiral. Distributions of each parameter will be taken into account in the implementation of the model into the software tool. Validation of the Spiral 1 IGC Damage Model will be performed in collaboration with Prof. Moran (USNA) by comparing the predicted distribution of damage geometry with that measured on new samples that are corroded for different times and at different DOS that are bracketed by the range of each of these variables used in the development of the database. Milestone: Report on Statistical Model of IGC Damage, ver. 1: December 31, 2010 Task 1.2: Develop an electrochemical approach for the quantification of the amount and reactivity of grain boundary beta (i.e., an electrochemical DOS measurement). Goal: Based on fundamental materials science and corrosion science, develop a rapid electrochemical approach that quantifies the sensitization of AA5XXX.

14 Method: All Spirals of the proposed program require a quantitative measure of DOS. The accepted method for characterizing the DOS of AA5XXX alloys is ASTM G67 [ASTM, 1999], an immersion test in which a sample is exposed to concentrated nitric acid for 24 hours. The mass loss is determined and considered to be a quantitative measure of sensitization. These values range from ~5 mg/cm 2 to >150 mg/cm 2 for heavily sensitized material [Craig, 1972]. Much of the mass loss is due to grain fall-out as the IGC separates grains from one another. Although accepted by industry, the method is destructive and it is not known to what extent it is controlled by the amount of β at the grain boundaries as opposed to the activity of that β and the grain boundary area between the β particles. The contribution of grain fallout makes it difficult to apply mass loss results as inputs to an IGSCC model as grain fallout occurs well behind the advancing IGC front which is the portion of attack that is most relevant to the cracking phenomenon. An electrochemical measurement of DOS would provide a rapid (<< 1 hr), quantitative characterization for the electrochemically active regions of a material. Due to the low charge passed, the method could be applied to a sample before testing to ensure a direct connection between the DOS and the subsequent IGC and IGSCC. A successful electrochemical DOS method requires discovery of a combination of environment and potential for which the Al(Mg) is passive while the β electrochemically dissolves rapidly. Under these conditions, the dissolution of the β can be measured via the electrochemical current, with larger degrees of sensitization leading to either more β on the boundaries and/or the development of a more active form of β there. We will deconstruct the AA5XXX alloy into the key phases, making electrodes of Al with Mg in solid solution (Al(Mg)) as well as Mg 2 Al 3 (β). Both can be made by standard metallurgical processing, with the method for casting of β recently reported by Searles et al. [Searles, 2002]. It has been shown [Searles, 2001] that in aerated 0.5 M NaCl, β shows a passive region limited by a pitting potential over a wide range of ph. In such a high chloride concentration, the pitting potential is below the open circuit potential of the Al(Mg) matrix. Thus, when β is present at the grain boundaries of AA5XXX alloys, it is polarized to the open circuit of the Al(Mg) matrix and dissolves rapidly leading to the observation of IGC. Under stress, the IGC becomes IGSCC. We will first establish the electrochemical kinetics of the Al(Mg) and β as a function of solution composition and concentration of Mg in the Al(Mg) in order to establish a fundamental, quantitative understanding of the electrochemical behavior of each phase separately. Solution composition variables to be studied include ph (3 to 14), [Cl - ] (0 to 4 M), oxygen level (0 to fully saturated), and temperature (25 to 50 ºC). The electrochemical kinetics so generated will be used in several ways in the research program: 1. Mixed potential theory will use the kinetics to quantitatively rationalize the rate of IGC as a function of the ratio of matrix area to β area in collaboration with the Spiral 2 IGC Damage Model development. 2. The kinetics will be used as input data for the anodic dissolution mechanism model in Spiral The effects of dissolved Mg level on the dissolution kinetics will be assessed. This portion of the study will be important in understanding how IGC propagates between

15 particles of β on grain boundaries. Based on TEM studies of grain boundaries, it is known that a continuous film of β is not a prerequisite for IGC [Davenport, 2006]. Such input information is critical for both the IGC Damage Model of Spiral 2 and the Anodic Dissolution IGSCC Model in Spiral Hydrogen evolution kinetics will be used as input data for the Hydrogen Embrittlement IGSCC Model in Spiral 3. From these results an optimized solution composition for the quantitative measurement of DOS will be selected. This solution will have the following characteristics: 1. The Al(Mg) will have an open circuit potential (OCP) that is at least as negative as the β. 2. Both the Al(Mg) and the β will have low corrosion rates at their respective OCP. 3. Upon polarization, the β will undergo rapid dissolution at potentials lower than the pitting potential of the Al(Mg) One can hypothesize that a mildly alkaline, reducing solution without [Cl - ] will be a good candidate. The alkalinity and reducing nature of the solution will lower the OCP of the Al(Mg) while having little effect on the OCP of the β whose OCP does not depend strongly on ph [Searles, 2002]. The lack of chloride should remove the issue of localized attack of the matrix or the β. The additional of a complexant for Mg 2+ may enhance the sensitivity. After optimization of the test solution composition, optimization of the test protocol will be performed. Several protocol options will be considered in collaboration with Prof. Moran (USNA) including cyclic potentiodynamic testing, potentiostatic testing for a set time, and electrochemical impedance. All of these methods are amenable to later development of the electrochemical method into a sensor or NDI tool. Such development is well outside the scope of the proposed work, but the requirements of a fieldable probe will be taken into account during the downselection process for the protocol. This work will also be coordinated with the grain boundary engineering (GBE) project supervised by Prof. Todd Allen at Wisconsin and the Spiral 2 DOS modeling project supervised by Prof. Mike Free at Utah (described under separate proposals). Materials that Prof. Allen has processed to have a lower DOS will be scrutinized with the electrochemical method developed here to provide a rapid screening of different GBE approaches. The measurement and modeling of β precipitation along grain boundaries performed by Prof. Free will both provide insight into the DOS measurements made as well as help link the measurements and predictions of β distribution in the Spiral 2 DOS modeling to the Spiral 2 IGC Damage Model by providing the electrochemical behavior of the different levels of β distribution. After optimization of the test protocol, a series of validation measurements will be performed in which AA5083-H116 samples, sensitized for times between 1 and 30 days at 100 ºC will be used. Each sample will be tested both in ASTM G67 as well as with the electrochemical test. The mass loss from the G67 test will be compared with the electrochemical metric of DOS. In addition, analysis of the corrosion topography will be performed and the results from the two types of tests compared.

16 Milestone: Report of DOS sensitization method: December 31, 2010 Spirals 2 and 3: Science-Based IGC Damage Modeling Method: A scientific IGC model must accept T(t) for sensitization, alloy and heat treatment and later GBCD and applied stress and output a distribution of IGC damage dimensions as a function of exposure time. The fundamental premise of the IGC model is that the dissolution rate of β in acidic solutions is greater than that of the Al(Mg) solid solution in grain interiors and the pitting potential is also less in neutral or alkaline solutions containing Cl - [Searles, 2001; Davenport, 2006]. A second key assumption is that Mg in solid solution in Al grain boundary regions (i.e., Al(Mg) gb ) between β precipitates also possess enhanced dissolution rates in acids and a lower pitting potential in alkaline solutions containing Cl - [Brillas, 1998]. Recall that TEM studies of grain boundaries, it is known that a continuous film of β is not a prerequisite for IGC [Jones, 2004; Davenport, 2006]. Thus the dissolution rate of Mg rich grain boundaries as well β particles along boundaries will preferentially corrode leading to IGC when corroded at the open circuit of the Al(Mg) matrix [Scully, 2002]. However, the paths and network of corroded grain boundaries will depend on a particular grain boundary s β coverage, the grain shape, alloy, heat treatment and exposure time. The heat treatment, alloy composition, and alloying microstructure define the DOS which in turn relates to the over distributions of boundaries with high and low β and Al(Mg) gb coverages while time of exposure defines the extent of IGC damage. Sophisticated additions to this description include the crystallographic texture (GBCD) and coincident site lattice (CSL) description of the specific grain orientations [Scully, 2002]. It is well known that high angle boundaries are often the most easily sensitized [Scully, 2002]. Hence, the clusters of connected high β coverage will depend on the alloy, microstructure, DOS, as well as the CSL description of high angle boundaries. In the model high β coverage boundaries and Al(Mg) gb boundary regions may by assigned at random or assigned according to some rule based on characterization of grain orientations provided from Professor Todd Allen (U of W- Madison). A previously developed quantitative model applied to the case of sensitized stainless steels [Scully, unpublished] would be used as the basis for the IGC model of 5XXX alloys with appropriate modifications to accurately describe IGC in a 5xxx alloy. The model would also benefit from previous studies in sensitized materials using bond percolation theories applied to IGC [Wells, 1989; Wells, 1992, Gaudett, 1993]. The model could contain two aspects and simulations could be performed in two modes. The first aspect would involve surface spreading of IGC at the interface between the 5XXX alloy and the corroding solution as previously performed in the case of pitting [Organ, 2005; Organ, 2007]. The second aspect would involve IGC penetration with depth into the material. The two modes could either involve a general dissolution or a pitting mode, each of which can lead to IGC damage. The model would be constructed via the following task steps: 1. Grain shape in equiaxed or brick wall configuration would be established in a 0.26 by 0.26 cm square 2-D grid representing the planar surface. A similar grid would be developed on a 2-D slice into the material perpendicular to the metal/electrolyte surface

17 using consistent grain shapes and orientations. In earlier models, a grain was assumed to be a square of 32x32 µm with a grain width of 2 µm. Any texture rules (CSL) would be obtained as inputs in spiral 3 and used to assign boundaries a given misorientation in step 2 as a precursor to β coverage assignment. 2. β coverage and Al(Mg) gb compositions would be assigned at random to the grain boundaries in the grid (Spiral 2) and then again according to some rules (Spiral 3) based on misorientation (spiral 3). 3. The grain boundary dissolution kinetics established in Spiral 1 as a function of composition (β and Al(Mg) gb ) will be used as input data for the anodic dissolution mechanism based IGC model in Spirals 2 and 3. Active dissolution would be simulated in each 2 µm grain segment using a library of data defining the active dissolution rate as a function of acidity. Pitting would occur as a statistically distributed parameter at each Cl - level while active dissolution rate of the β phase would be purely deterministic. These E-i relations would be assigned to all boundaries and reassigned based on local potential and ph. The matrix consisting of a low Mg content in the Al(Mg) solid solution would be assumed not to corrode. 4. Pitting or IGC by active dissolution would be initiated at a random site on the grain boundary. 5. Once the pit is formed or active dissolution occurs a current is emitted. Subsequently, the IR drop and Cl - (ph) accumulation would be calculated using numerical recipes previously developed [Organ, 2007]. 6. The potential and ph over each grain boundary segment would be determined and a dissolution rate would be assigned. The sum of the dissolution rates determines the subsequent IR drop and Cl-(pH) level. 7. Spreading of corrosion across surfaces would occur as a function of potential drop (inhibiting) and acidification/halide accumulation mechanism (enhancing pitting of boundaries). The size distribution of connected clusters of corroded grain boundaries would be calculated. 8. Penetration would occur without interactions along high coverage boundaries or by interactions (Spiral 3) induced by interactions between grain boundaries via hydrogen diffusion in the solid state. An Ohmic resistance function would inhibit dissolution-based penetration with increasing penetration depth using various functions to account for increased dissolution with depth. 9. The surface spread and depth of penetration information would be reconstructed to give a 3-D representation of damage as a function of the inputs discussed above. Outputted information on IGC a/c ratios as a function of alloy, DOS, and microstructure (GBCD) would be provided to other aspects of the model, (a/c)/ t rates could also be determined. After optimization of the modeling protocols, a series of validation measurements will be performed in which AA5083-H116 samples, sensitized for times between 1 and 30 days at 100 ºC will be used. Each sample will be tested both in ASTM G-67 and the electrochemical test method to be developed in Task 2. The mass loss from the G-67 test will be compared with the electrochemical metric of DOS. In addition, analysis of the corrosion topography will be performed and the results from the two types of tests compared.

18 Figures 6 and 7 show the input data for a similar approach to IGC applied to a Fe-Cr alloy with sensitized grain boundaries. The distribution of sensitization levels is shown in Figure 3 assuming 80 hrs of sensitization. 10 % Cr Current density(a/cm 2 ) % Cr 12 % Cr 14 % Cr 16.8 % Cr 18 % Cr V (V) SCE Figure 6. Estimated current densities as a function of potential as a function of Cr content. Figure 7. The grain boundary configuration for sensitization time of 80 hrs (based on Gaudett,1993). An example of the outputted IGC damage distribution along the surface of the alloy is shown in Figure 8 as a function of corrosion time. Similar computations would need to be performed on a plane perpendicular to the surface using different rules as discussed above. In Figure 9, the distribution of planar damage across a metallic surface is shown and the size of the largest cluster is calculated. In this project a/c and d(a/c)/dt would be calculated. (d) (a) Figure 8. Snapshots of the simulated intergranular damage on an surface with no pit at V app = V SCE and sensitization time = 80h at (a) t= 0 s, (b) t= 1,225 s.

19 Normalized cluster length % of grain boundaries with 12.6 % Cr or less Figure 9. Normalized cluster length (largest corroded cluster length initial corroded cluster length) as a function of percentage of grain boundaries containing below 12. 6% Cr or lower than that at V app = V. Milestone: Report on Science-based Damage Model ver 1 (Spiral 2): December 31, 2010 Milestone: Report on Science-based Damage Model ver 2 (Spiral 3): December 31, 2012 Intergranular Stress-Corrosion Cracking Modeling The Spiral 1 focus is on developing: (a) the model framework for predictions using continuum fracture mechanics, and (b) the highly focused set of data on IGSCC necessary to guide and validate the model for laboratory conditions, as well as for inputs to other tasks in this proposal. The Spiral 2 focus is to enhance this model by developing an analytical description of the stress corrosion crack formation stage of life, governed by interaction of the local topography of IGC damage (modeled by others on the team) with either high fidelity continuum or microstructurescale crystal plasticity descriptions of local stress. A damage mechanism-based failure criterion is developed based on both grain boundary dissolution and hydrogen embrittlement, to define crack formation. This failure criterion will be approximated in Spiral 2. In Spiral 3, a quantitative-mechanism model will be developed for crack network evolution by combined HEE well as by AD components so that a microcrack distribution will be predicted through a sensitized 5XXX alloy based on a grain by grain assignment of da/dt =f(k I, DOS). The model will be validated by comparison to experimental results on IGSCC measured over of sensitization levels. Spirals 1 and 2: Continuum-Scale Model of IGSCC Goal The overarching objective is to develop a continuum-scale model that predicts the effect of IGC on IGSCC life, as a critical factor that governs structural integrity for 5000-series aluminum alloys under stress in marine applications.

20 Method: The impact of corrosion damage on structural integrity, and the necessary capability to predict the mechanical performance of a corroded component, is a critical problem facing engineers responsible for ensuing sustainable, economic, and safe operations of high performance military systems. In the present study, this issue is embodied in the interaction of intergranular corrosion (IGC) and intergranular stress corrosion cracking (IGSCC) in a series aluminum alloy relevant to ship applications. In this system, localized corrosion along a network of grain boundaries will result in a highly roughened surface due to grain pullout as well as an underlying intergranular fissure network through the thee-dimensional microstructure. This corrosion state will, at a point in component life, transition to a distribution of microcracks, driven by either fatigue or IGSCC. This cracking will evolve with load cycles or time into a single or few dominant cracks that continue to grow to the point where component integrity is compromised, for example by either catastrophic fracture, stiffness loss or plastic collapse. Figure 10 shows an example in which a fatigue crack has initiated and grown (top to bottom) from a precorroded surface (corrosion to crack interface is outlined) in a series aerospace alloy with an exfoliated surface (top). The engineering challenge is to characterize a so-called as-is state at a point in service, then perform a quantitative analysis of the future fitness-forpurpose of the structure as IGC and IGSCC evolve. The scientific challenge is to understand the fundamental damage mechanisms, to model each so as to produce quantitative predictions of damage progression, to connect transitions between corrosion and cracking, and to develop an overarching model of macroscopic mechanical properties Figure 10. Fatigue crack propagation emanating from surface corrosion damage at the top of the SEM fractograph. necessary for the engineering analysis. The corrosion to fatigue interface is marked. [Kim, Burns The proposed research follows this and Gangloff, 2007] strategy. Tasks 1.5 and 1.6 Predict distributions of IGSCC life and effective IGC damage size from the continuum perspective. Integrate in a free-standing code. The goal of these two Spiral 1 tasks is to develop a first-generation model to predict IGSCC time to failure for various precorrosion conditions. The primary intent is to rapidly develop the proper, but necessarily approximate, framework for engineering applications and to guide the scientific research in Spirals 2 and 3. The proposed Spiral 1 approach to this problem has been developed in the context of fatigue failure, as illustrated in Figure 11 for 7075-T6 aluminum alloy [Kim et al., 2007; Fawaz, 2003], but not for time-dependent stress corrosion cracking.

21 For the stress corrosion problem, time-to-failure replaces cycles-to-fatigue-failure and applied stress is constant not cyclic; the hypothesis is that the modeling approach is otherwise equivalent. Modeling can be carried out in two formats: (a) prediction of the distribution of time-to-failure for a known (measured or corrosion science model-predicted) distribution of starting corrosion damage geometry produced by a given environmental pre-exposure, or (b) prediction of a distribution of initial corrosion damage depth and shape, reverse calculated from measured time to failure. This latter approach is shown in Figure 11 where measured fatigue life for two replicate specimens at one stress, was used to predict two precorrosion depths and aspect ratios characteristic of the corrosion exposure. Modeling was validated by comparing: (a) predicted initial damage size to measured corrosion size (shown in the cross-sectional micrograph) at a single stress, and (b) specimen life measured experimentally at a lower stress to the model prediction using the initial damage size predicted at the higher stress. This continuum approach has been validated as effective for fatigue T6511 EXCO (6h) LS H2O/N2 (RH > 95%) R = Hz a/c Crack, Variable (IV) 200 µm Maximum Stress (MPa) AFGROW prediction a = 238 µm; c = 380 µm a = 258 µm; c = 412 µm a = 247 µm; c = 392 µm (mean) AFGROW prediction a = 267 µm; c = 428 µm a = 205 µm; c = 328 µm a = 230 µm; c = 367 µm (mean) Experiment 0 Calculated endurance limit Propagation Cycles to Failure Figure 11. (Left) Localized corrosion damage in 7075-T651, associated with grain boundaries, and the expected regime of crack formation proximate to this defect site. (Right) Fracture mechanics model predictions (dashed lines) of the fatigue life of 7075-T651, corroded to produce the feature shown on the left, then stressed cyclically in moist air. Measured fatigue life is shown by data points at two stress levels. [Kim, Burns and Gangloff, 2007] This foundation provides the starting point for the proposed Spiral 1 task. Total IGSCC life is composed of a crack formation component, required to convert a blunt and complex shape of IGC topography into an organized intergranular stress corrosion crack (see the shaded region in the micrograph of Figure 11), plus a crack propagation component required to grow this crack to an endpoint. In order to predict a distribution of IGSCC failure times, the following inputs must be defined: Distribution of IGC surface depth and shape. This input will be provided, initially based on experimental measurement from the IGC task in Spiral 1 and later based on predictions from the IGC model introduced in a Spiral 2 IGC task and enhanced to account for additional complexities in Spiral 3. The particular challenge is to develop a distribution of corroded surface topography that includes grain pullout and an

22 intergranular fissure network with the 3-dimensional shape approximated by depth (a), and aspect ratios a/c and a/b where c is surface corrosion length and b is surface corrosion height. Material properties in an algorithm that defines crack formation time (t Form ) vs. corroded surface topography and applied stress. This algorithm cannot be well defined based on current scientific understanding, but can be developed experimentally, as will be done in Spiral 1. A fundamental description of this part of the problem will be modeled in Spiral 2. Crack growth rate (da/dt) vs. applied stress intensity factor (K), assuming that microstructural and microchemical influences on crack propagation are negligible and continuum mechanics similitude applies. Stress intensity solution (K), assuming continuum elasticity but irregularly-shaped surface cracks that may interact with others as well as the 3-dimensional corrosion front. Each of these inputs depends on the degree of sensitization (controlled by grain boundary engineering) and prior environmental exposure. The time-based evolution of IGSCC, from a corrosion modified-equivalent initial flaw size (CM- EIFS) is predicted with the following well-known relationship [Williams, 1973] applied to this challenging corrosion to crack geometry. t Total = t Form + * a CM EIFS da α f ( σ, a ) β Eq. 3 where α and β are material properties in the IGSCC da/dt vs. K relationship and σ is the applied stress. In Spiral 1, the distribution of CM-EIFS will be established by a combination of reverse prediction from measured t Total and physical measurement. It is proposed that this parameter, plus experimental definition of t Form, will adequately account for the complicating effects of a complex corrosion front topography, multiple microcracks that may interact, microstructural and microchemical effects on crack propagation not described by K, α, and β. These cutting edge issues will be addressed in Spiral 2. The necessary details of this model, developed in Task 1.5 (see above), will be coupled with the inputs from the IGC tasks to produce user-friendly software in Task 1.6. This code will employ state of the art continuum fracture mechanics methods, paralleling user friendly codes such as AFGROW [Harter, 2004] and FASTRAN [Newman and Abbott, 2005] used currently for fatigue modeling. Milestone: Report on predictions of distributions of IGSCC life and effective IGC damage size from the continuum perspective December 31, 2010 Task 1.4 Determine IGSCC formation and propagation properties

23 The goal of this task is to experimentally measure the IGSCC properties of the selected alloy, AA5083 in a relevant temper such as T116 or T321; as a function of degree of sensitization, prior corrosion exposure time and geometry, and applied stress. The tensile yield strength of this alloy is in the range of 190 to 230 MPa. These experiments are necessary for close integration with the proposed IGC studies, to exploit state-of-the art environmental cracking methods leading to significantly improved results, and to produce sufficient data for statistical analysis. The proposed approach to these experiments is detailed as follows. Anisotropic grain structure in plate aluminum alloys, including 5083, critically affects IGSCC behavior. Specimens will be designed to focus IGC and subsequent IGSCC on planes parallel to the L and T direction of rolled plate. A ~5 cm thick plate of 5083 will be studied to facilitate uniaxial loading, parallel to the thickness (S) direction. This macroscopic crack orientation should simplify the experimental and modeling challenges, by restricting macroscopic mixed mode and crack branching [Bovard, 2007], but is none-the-less critical to marine applications. For example, IGSCC in sheet 5083 used in hull applications propagates in a mixed-mode crack geometry that involves stressing parallel to the S-direction. [Bovard, 2007] Extension of the proposed research to this more complex situation can follow as basic scientific understanding is developed. The S-T surface of the ground and polished gauge section (~1 cm x 2 cm by 4 cm) of an IGSCC tensile specimen will be precorroded in the IGC task employing two approaches with multiple corrosion patches per each specimen as described in IGC Damage Model Task 1.1. The geometry of the larger patch area, particularly orientation with regard to microstructure, will be varied in initial experiments to determine the optimal specimen precorrosion to produce Mode I and S-L oriented IGSCC cracks amenable to modeling. In each patch, we expect to produce a topography developed from grain pull-out and an underlying intergranular fissure network. We anticipate characterizing the IGSCC behavior as a function of these variables: 3-4 DoS x 2-3 pre-exposure times x 2 applied stress levels x 4 replicates, requiring 50 to 100 experiments and yielding 250 to 1,000 stress corrosion cracks available to guide and validate modeling with statistically definable confidence. Each specimen will be stressed in chloride solution at fixed electrode potential, for a fixed time averaging ~20 days. The experiment will be interrupted prior to specimen fracture. Each specimen will be sectioned and the size/shape of the largest 5 to 10 IGSCC cracks determined. The results of these experiments; crack depth and aspect ratio produced by stressing for a given time and as a function of DoS, precorrosion exposure time, and applied stress; will be used to guide and validate the Spiral 1 modeling. These data will be directly relevant to other tasks in this proposed program.

24 A set of experiments will be designed to establish the importance of the crack formation period, and to quantify t Form, for two applied stress levels, but a single DoS and single precorrosion exposure time. Two approaches will be used to monitor the formation of IGSCC about corrosion damage on the microstructural scale. The IGSCC specimen with controlled-small patches of corrosion damage will be instrumented to employ the direct current electrical potential difference method to continuously monitor crack formation and early growth from a precorroded area [Gangloff et al., 1992]. This approach is capable of resolving ~10 µm crack extension about a defined corrosion patch, but the method has never been implemented for this purpose. To augment this technique, the loading profile will be altered with addition of limited low amplitude load cycles (or pulses) that should produce bands on the crack surface visible in SEM or optical microscopy [Hahn and Pugh, 1981; Kim, Burns and Gangloff, 2007]. Examples of such bands are shown in Figure 3 for a fatigue crack emanating from a microstructural defect of about 10 µm diameter (left) and from localized corrosion damage of about 200 µm depth, each in T um Figure 12. Crack surface marker-bands, produced from load pulses programmed to delineate the microscopic position of a fatigue crack emanating from (left) a small constituent particle cluster and (right) an irregularly shaped corrosion front; each in 7075-T651. [Kim, Burns and Gangloff, 2007] Milestone: Report on IGSCC formation and propagation properties December 31, 2010 Task 1.3 Determine macroscopic (da/dt)=f(k,dos) The goal of this task is to establish the macroscopic crack growth rate (da/dt) versus the elastic crack tip stress intensity factor for IGSCC in the selected alloy. These data are necessary input to the continuum modeling in Task 1.5. This material-environment property for 5000-series aluminum alloys depends on degree of sensitization, as shown in Figure 13 (Bovard,2007), and applied electrode potential (Jones, et al., 2001), but not on Mg segregation (Jones, et al., 2001) for Al-Mg alloys. It is necessary to measure crack growth rates in the proposed study in order to: (a) augment limited literature findings, (b) characterize the exact alloy used in all tasks of the proposed study, including a data base for assessment of the effectiveness of grain boundary engineering, and (c) emphasize the lower stress intensity regime, for K between 2 and 8 MPa m, pertinent to modeling of IGSCC formed about precorrosion damage at the ~200 µm and above size scale. Moreover, crack growth rate vs. DoS and applied electrode potential data provide a