Fracture mechanics experiments and occluded crack chemistry modeling validate the mechanism for intergranular stress corrosion cracking (IGSCC) of sensitized Al-Mg alloys: dissolution of discontinuous grain boundary β (Al3Mg2) precipitates activates Al-Mg (α) solid solution dissolution to acidify the crack solution for crack tip hydrogen embrittlement. Slow-rising displacement experiments with precracked specimens establish the strong effect of applied potential on IGSCC kinetics for sensitized AA5083-H131 (S-L orientation, 22 mg/cm2) in neutral NaCl solution. Anodic polarization increases growth rates through enhanced α dissolution for crack acidification and H uptake, whereas cathodic polarization below the pH-sensitive α breakdown potential reduces growth rates by limiting Al dissolution, and thus crack acidification and the overpotential for H production (ηH). Polarization below the β breakdown potential eliminates IGSCC by precluding β dissolution and crack acidification. Cathodic polarization could provide a practical means to mitigate IGSCC in sensitized Al-Mg alloys. As a second validation, AA5083-H131 without β is susceptible to IGSCC in either acidic AlCl3/MgCl2 simulated crack solution or neutral NaCl with anodic polarization, but resists cracking in NaCl at near the open circuit potential and in NaOH with strong cathodic polarization. Growth rates and calculated crack tip H solubility systematically increase with increasing ηH, estimated from crack pH and ohmic drop, to confirm the proposed mechanism. As a corollary, severe IGSCC at near open circuit potential occurs above a critical degree of sensitization, as required to produce a sufficient volume of grain boundary β for dissolution and acidification to trigger intergranular H cracking.
Magnesium solid solution strengthened Al is prone to severe intergranular corrosion (IGC) and intergranular stress corrosion cracking (IGSCC) if the Mg content exceeds about 3 wt% and if the microstructure is sensitized by prolonged thermal exposure.1-8 Anodically active β (Al3Mg2) precipitates on grain boundaries at temperatures between 40°C and 180°C.8-10 The amount, morphology, and distribution of β control the degree of sensitization (DoS), and in turn the kinetics of IGC and IGSCC for Al-Mg alloys exposed to moist environments often with chloride.6-8,11-22 A nitric acid mass loss test (NAMLT, governed by ASTM G6723 ) is used to quantify DoS for Al-Mg alloy development, lot acceptance, and failure analysis. A variety of smooth and precracked specimens, with various loading formats, have been used to characterize the DoS dependence of IGSCC in Al-Mg alloys.1,8,11,13-14,24
The effects of low-temperature sensitization on IGC and IGSCC are important for long-term service of Al-Mg,21 but have not been sufficiently investigated.6,11,13-15 As such, a fracture mechanics study was conducted to characterize IGSCC kinetics for Al-Mg alloy AA5083-H131 (UNS A95083)(1) sensitized at low temperatures between 60°C and 100°C.20,25 As detailed in a companion paper,26 AA5083-H131 is susceptible to severe IGSCC when the NAMLT DoS exceeds 9 mg/cm2 to 12 mg/cm2 for the S-L crack orientation (loading in the thickness direction, S, and crack propagation in the rolling or extrusion direction, L). For immersion in NaCl solution, with the fatigue cracked specimen polarized to −0.800 V versus standard calomel electrode (SCE) (near open circuit potential, OCP), the slow-rising displacement threshold stress intensity for the onset of IGSCC (KTH) decreased from 20 MPa√m to 4 MPa√m, and Stage II subcritical crack growth rate (da/dtII) increased from 1 nm/s to 10,000 nm/s, as DoS rose from 9 mg/cm2 to 54 mg/cm2. These DoS dependencies appeared to be identical for sensitization at 60°C, 80°C, and 100°C. As-received AA5083 (2 mg/cm2) was highly resistant to IGSCC. Sophisticated methods characterized the size, spacing, and morphology of grain boundary β versus DoS for the composition of AA5083-H131 investigated in this research.12,16,24,27-38 For low to moderate temperatures, β forms as discrete, nanoscale particles and/or thin plates discontinuously arrayed on Al-Mg solid solution (α) grain boundaries. These results provide the basis to predict the DoS dependencies of IGC17,39 and IGSCC.20,26
The hypothesized mechanism for IGSCC is coupled grain boundary β-α dissolution and hydrogen embrittlement in the crack tip fracture process zone (FPZ), as proposed by Jones, et al.,14 and schematically illustrated elsewhere.26 A similar coupled dissolution-hydrogen framework and model was recently reported to predict rates of stress corrosion cracking in Al-Zn-Mg-Cu alloys.40 The electrochemical basis for the chemistry of the occluded crack in Al-Mg was established for an IGC-fissure geometry.16-18 Without β, the Al-4%Mg α matrix is passive near OCP, with minimal dissolution, and the crack does not acidify by cation hydrolysis.16-18,41-43 For a sensitized microstructure, grain boundary β rapidly dissolves to increase Al3+ and Mg2+ concentrations in the crack, which begin to destabilize the passive-α surface leading to dissolution, cation enrichment, and hydrolysis. As crack tip pH decreases, α further breaks down and dissolves, leading to near-equilibrium crack solution acidification of about pH 2 from Al3+ hydrolysis42,44-48 and lowered somewhat by Mg2+.47-48 Metallographic cross-section results demonstrate that IGC fissures in AA5083 immersed in NaCl solution are not blunted by this localized corrosion sequence for times relevant to IGSCC propagation and potentials near OCP,17-18,47 unlike fissure widening reported for long time IGC in Al-Cu-Mg (AA2024-T351 [UNS A92024]).49 Crack acidification intensifies cathodic hydrogen (H) production and uptake at the tip, which diffuses into FPZ grain boundary regions on order of 1 μm from the crack tip surface.26 With locally concentrated tensile stress, H promotes inter-β ligament H embrittlement along this α boundary.26,50-52 Cracking of α boundaries provides the connected path through a forest of discontinuously arrayed β on grain surfaces. With this framework, the dependence of IGSCC kinetics on DoS was qualitatively attributed to contributions of grain boundary β to crack electrochemistry, crack tip FPZ stress concentration, and extraordinary H diffusion.20,26,53-54 The crack tip mechanics and H trapping aspects of grain boundary β were considered;26 the remaining challenge is to model the DoS dependence of crack chemistry and H production. The hypothesis is that a critical amount of grain boundary β dissolution triggers Al dissolution that then produces the crack chemistry change (acidification and sufficient H-overpotential) necessary for hydrogen environment assisted cracking (HEAC) in sensitized Al-Mg.
The first objective is to critically test the proposed mechanism of IGSCC by control and modeling of crack electrochemistry. The requirement for β dissolution is probed with IGSCC experiments that stress AA5083-H131 without grain boundary β in solutions that produce various levels of crack tip H uptake into the FPZ. Solutions include neutral NaCl, acidic simulated crack tip solution, and alkaline NaOH with cathodic polarization. The role of crack chemistry in IGSCC is investigated for sensitized Al-Mg in NaCl using applied potential to promote or preclude α and β breakdown. A second objective is to use this crack chemistry formulation, coupled with the DoS dependence of AA5083-H131 β volume fraction12,16,24,27-29,31-38 and H diffusivity,55-56 to model the DoS dependencies of KTH and da/dtII measured in companion work.26
For IGSCC testing, single edge notch tensile (SENT) specimens were machined in the S-L orientation,20,26 and fatigue precracked in moist air to a total depth of about 1.8 mm ahead of a 1.5 mm deep notch, with a final maximum stress intensity (K) of 3.5 MPa√m. Notch mouth opening was 220 μm. A 185 mL Plexiglas† cell housed the SENT specimen under full immersion in three non-deaerated aqueous solutions: (1) 0.6 M NaCl (pH~7.3), (2) 0.01 M NaOH (pH 12), and (3) 2.5 M Al3+, 0.11 M Mg2+, 7.7 M Cl−, (average pH −0.3) mixed from AlCl3 and MgCl2 to simulate the occluded-crack solution expected for sensitized Al-Mg alloys.17 Each solution was circulated at 30 mL/min from a 2 L reservoir at ambient temperature. Bulk solution pH and OCP were monitored before and after each experiment. The NaOH solution was adjusted to maintain pH 12. The grounded SENT specimen was polarized to a constant bold-surface potential (Eapp) during loading using a floating-ground potentiostat. The specimen was loaded under grip displacement control (0.0002 mm/min) to a K of 4 MPa√m and held at the associated load for 10 h. Following this hold, the specimen was loaded at an initial dK/dt of 0.25 MPa√m/h until final fracture. The direct current potential difference (dcPD) method for crack length versus time measurement and the da/dt versus elastic-plastic stress intensity (K) analysis are detailed elsewhere.26 The S-L crack orientation was used to focus on crack chemistry effects on IGSCC, apart from the complication resulting from crack path tortuosity.
For sensitized AA5083-H131 (S-L) in near-neutral NaCl at Eapp of −0.800 VSCE (near the initial OCP of −0.760 VSCE), fracture mechanics experiments established that the fatigue crack tip advanced by IGC during the 10 h hold at a low K of 4 MPa√m.26 During slow-rising K, precrack IGC transitioned to IGSCC with a well-defined KTH and subcritical da/dt versus K that was approximately K independent in Stage II (da/dtII). The rate of IGC, KTH, and da/dtII at K of 15 MPa√m (da/dtK15) each depended on DoS for exposure at 60°C, 80°C, or 100°C.26 With this basis, the effects of Eapp and bulk solution composition on IGC and IGSCC rates are detailed here.
Effect of Applied Polarization for Sensitized AA5083-H131
Applied potential significantly affects the IGC and IGSCC susceptibilities of sensitized AA5083-H131 (S-L, 100°C for 175 h, DoS of 22 mg/cm2) during immersion in 0.6 M NaCl. IGC and IGSCC growth rates are exacerbated by anodic polarization, and are diminished by cathodic polarization to below the resolution of the crack monitoring system and perhaps to zero.
Intergranular Corrosion Growth
Figure 1 presents dcPD measurements of IGC growth in sensitized AA5083-H131 (22 mg/cm2) during 10 h of low K (4 MPa√m) loading, followed by slow-rising displacement to 6 MPa√m in 0.6 M NaCl with a range of Eapp. The capability to measure IGC is evident and novel.26 Average rates were determined with regression analysis over the 10 h hold. This rate of IGC was rapid (6.5 nm/s) for sensitized AA5083-H131 anodically polarized to −0.730 VSCE and decreased to 1.4 nm/s when polarized to −0.800 VSCE. Seven replicate experiments on sensitized AA5083-H131 at Eapp of −0.800 VSCE yielded IGC rates between 0.80 nm/s and 2.6 nm/s.26 For Eapp of −0.900 VSCE, growth rate decreased to 0.9 nm/s, and declined to 0.35 nm/s with Eapp set at −0.975 VSCE. Polarizations to −1.020 VSCE, −1.080 VSCE, and −1.100 VSCE yielded much slower growth rates (0.03 nm/s, 0.03 nm/s, and 0.05 nm/s, respectively), which were close to the resolution level of 0.004 nm/s measured in air at 4 MPa√m. The most cathodic polarization of −1.300 VSCE produced an increased rate (0.31 nm/s) during the low-K hold, but without evidence of expected-transgranular cathodic corrosion resulting from crack alkalinity. The cause of this rate increase is not understood and the effect was not further investigated. The growth rate at each Eapp was sensibly constant during the low-K hold up to 10 h; increased da/dt during the rising K experiment was not interpreted because of the potential false da/dt component caused by increasing dcPD from rising crack tip plasticity.26
Intergranular Stress Corrosion Cracking Growth
Rates of IGSCC are plotted versus rising K in Figure 2 for sensitized AA5083-H131 (S-L, 22 mg/cm2) in 0.6 M near-neutral NaCl at various levels of constant Eapp. The elastic-plastic fracture toughness (KJIC of 23.2 MPa√m for S-L26 ) is noted. Two stage da/dt versus K behavior is observed for Eapp of −0.730 VSCE and −0.800 VSCE. At −0.730 VSCE, KTH (defined as the transition from slow IGC growth to rapid IGSCC in Stage I) was 7.9 MPa√m and Stage II growth rate (da/dtK15, or the extrapolated rate if cracking at K of 15 MPa√m is not measured26 ) was 7,200 nm/s. At −0.800 VSCE, KTH increased to 10.7 MPa√m, and da/dtK15 decreased to 3,000 nm/s. Seven replicate experiments at this potential and DoS established that KTH varied from 8 MPa√m to 11 MPa√m, with a single experiment yielding 14 MPa√m, and da/dtK15 varied from 800 nm/s to 3,000 nm/s with a single experiment yielding 10,000 nm/s.26 These KTH are much less than KJIC establishing subcritical IGSCC susceptibility, as confirmed by scanning electron fractography.26 J-integral analysis26 accounted for ligament plasticity at high K during polarization from −0.900 VSCE to −1.300 VSCE. At −0.900 VSCE, Stage I exhibits a shallow slope, obscuring threshold definition, but KTH is likely 10 MPa√m to 15 MPa√m and da/dtK15 is 10 nm/s. Similar behavior is observed for −0.975 VSCE and −1.020 VSCE. For −0.975 VSCE, KTH is between 10 MPa√m and 18 MPa√m, and da/dtK15 is 5 nm/s. For −1.020 VSCE, KTH is 11 MPa√m to 13 MPa√m, and da/dtK15 is 8 nm/s. Growth rates for sensitized AA5083-H131 polarized to −1.080 VSCE, −1.100 VSCE, and −1.300 VSCE equal the low da/dt measured during rising-K loading in moist air. These da/dt levels are characteristic of plasticity-induced dcPD rise superimposed with possible-low-rate HEAC, then stable-ductile tearing at high K.26 There was no evidence of transgranular cathodic corrosion.
Effect of Applied Polarization and Solution Composition for Unsensitized AA5083-H131
Although as-received AA5083-H131 SENT specimens (S-L) without sensitization (2 mg/cm2) resist IGC and IGSCC in near-neutral NaCl solution at −0.800 VSCE,26 each damage mode is promoted by either anodic polarization or bulk-chloride solution acidification.
Intergranular Corrosion Growth
Figure 3 presents dcPD measurements of IGC growth in unsensitized AA5083-H131 (S-L, 2 mg/cm2) versus time during the hold over 10 h at K of 4 MPa√m, then slow-rising displacement to 6 MPa√m, while immersed in: (a) 0.6 M NaCl (pH 7.3) polarized to Eapp of −0.730 VSCE, −0.800 VSCE, and −1.020 VSCE, (b) 0.01 M NaOH (pH 12) at OCP (−1.37 VSCE) and polarized to −1.800 VSCE, and (c) two replicate specimens in AlCl3/MgCl2 (pH −0.2 and pH −0.4) polarized to −1.000 VSCE. Average rates of IGC were determined by linear regression over the first 10 h, and are plotted versus Eapp in Figure 4, including the pH-sensitive breakdown potentials for α and β (EBd-alpha and EBd-beta detailed in the Discussion). Figures 4 and 5 include measured IGC rates for low-DoS AA5083-H131 in the various environments, as well as sensitized AA5083-H131 in 0.6 M NaCl from Figure 1.
Slow IGC propagation was recorded for low-DoS AA5083-H131 in NaCl at −0.800 VSCE and −1.020 VSCE (0.03 nm/s and 0.06 nm/s, respectively, Figures 4 and 5); rates are faster than the resolution limit of 0.004 nm/s recorded for a no-growth case stressed at low K in air. The growth rate increased to 3.6 nm/s as a result of anodic polarization at −0.730 VSCE in NaCl; however, the potentiostat registered suspect values of current and Eapp could have been as high as −0.450 VSCE. Scanning electron microscopy (SEM) suggested that corrosion on the boldly exposed SENT surface caused significantly longer extension of the edges of the crack front compared to the center. Given uncertain Eapp and bold surface corrosion, absolute crack lengths (Figure 3) and growth rates (Figures 4 and 5) are suspect for Eapp of −0.730 VSCE. Nonetheless, IGC is severe in the fatigue crack without β provided that Eapp is anodic to the OCP. For exposure in simulated crack tip solution (AlCl3/MgCl2 at −1.000 VSCE), the average IGC rate during the low-K hold is 1.4 nm/s and 5.2 nm/s for duplicate experiments. These rates are faster than those measured for the neutral NaCl or alkaline-NaOH environments, and are similar to IGC rates for sensitized AA5083-H131 in near-neutral NaCl at Eapp near OCP (−0.730 VSCE to −0.900 VSCE26 ), as shown in Figures 4 and 5. Corrosion was present on boldly exposed surfaces and fatigue crack flanks, which may have contributed to dcPD increase being not strictly representative of IGC. An experiment was also conducted with the AlCl3/MgCl2 solution at Eapp of −0.800 VSCE, anodic to the measured OCP of −0.880 VSCE, but extensive corrosion precluded dcPD interpretation.
For low-DoS AA5083-H131 in NaOH, low-K-hold growth rates are similar for either cathodic (−1.37 VSCE, 0.5 nm/s) or near OCP (−1.80 VSCE, 0.6 nm/s) levels of Eapp. These rates are higher than measured for as-received AA5083-H131 in moist air and cathodically polarized NaCl, yet are much lower than rates for SENT specimens in anodically polarized NaCl and acidic simulated crack solution.
Intergranular Stress Corrosion Cracking Growth
Figure 6 presents measured IGSCC rate versus rising K for as-received low-DoS AA5083-H131 (S-L, 2 mg/cm2) stressed in several environments, and Figure 7 shows da/dtK15 as a function of Eapp compared to rates for sensitized (22 mg/cm2) AA5083-H131. KJIC denotes the initiation of stable crack growth by microvoid damage and the J-integral accounted for plasticity at high K.26
Crack growth kinetics for low-DoS AA5083-H131 in neutral NaCl demonstrate the effect of applied potential on IGSCC. Values of da/dt during rising K at −0.800 VSCE in Figure 6 somewhat exceed the moist air resolution limit (♦), but IGSCC did not occur until well above KJIC where crack growth progressed by both ductile rupture and IGSCC under rising (♦) then fixed (★) displacement.26 The experiment at −1.020 VSCE was prematurely terminated (Figure 6) and the point in Figure 7 is plotted with an upward arrow, assuming no substantial acceleration of growth to K of 15 MPa√m. Growth rates for as-received AA5083-H131 polarized to −0.730 VSCE were essentially constant near 9 nm/s while K increased from 7 MPa√m to 10 MPa√m (Figure 6). At this point applied polarization ceased, the specimen corroded at OCP (about −0.840 VSCE), and da/dt dropped to near 1 nm/s while applied K increased to 12 MPa√m. At this K, an Eapp of −0.730 VSCE was reapplied for the remainder of the experiment and crack growth accelerated to 90 nm/s. For this potential, KTH appeared to be below 6 MPa√m prior to polarization interruption, and da/dtK15 was 43 nm/s after Eapp of −0.730 VSCE was reestablished. Fractography revealed significant corrosion product coverage, yet regions were identified as likely IGSCC.20,26
Slow-rising K loading of as-received AA5083-H131 in acidic AlCl3/MgCl2 solution promoted multi-stage IGSCC for replicate experiments (Figure 6). Threshold stress intensities are 4.0 MPa√m and 7.5 MPa√m, and da/dtK15 are 50 nm/s and 200 nm/s. The KTH for each experiment is much less than KJIC, establishing subcritical stress corrosion cracking without β present on α boundaries.26 The faster cracking experiment was exposed to an aged solution (pH −0.2) and exhibited uniform corrosion on the boldly exposed SENT surfaces, which may have affected the accuracy of dcPD measurements. The slower growth rate experiment was performed in fresh AlCl3/MgCl2 solution (pH −0.4) and exhibited corrosion damage restricted to the flanks of the crack, just ahead of the notch. SEM analysis was not possible because of crack wake corrosion.20
Specimens of as-received AA5083-H131 loaded in NaOH solution exhibit similar low rates (~1.3 nm/s at 15 MPa√m) of crack extension, at the upper bound of the resolution limit defined by moist air plasticity (Figures 6 and 7).26 Part of these rates may be attributed to cathodic corrosion and perhaps low-rate HEAC.26 Fracture surfaces were compromised by extensive corrosion product and SEM analysis could neither confirm nor refute intergranular cracking.20
Experimental results yield two key observations relevant to: (a) testing the proposed hypothesis regarding the role of β amount (DoS) in crack chemistry change (acidification), and (b) mitigating IGSCC in Al-Mg alloys. First, for sensitized AA5083-H131 in NaCl solution, both fatigue crack tip IGC at low K (Figure 1) and IGSCC at higher K (Figures 2 and 7) are eliminated by cathodic polarization at Eapp of −1.050 VSCE and −1.300 VSCE. This finding is robust because slow-rising K loading, the S-L crack orientation, high DoS (22 mg/cm2), and a fatigue crack promote severe IGC and IGSCC in NaCl near OCP. Because IGSCC is mitigated by modest cathodic polarization, either galvanic coupling between Al-Mg and a coating or paint system, or imposed current, could provide a practical means for IGSCC control. The possibility of HEAC in sensitized AA5083-H131 stressed at Eapp below −1.300 VSCE was not investigated; however, the low DoS microstructure did not exhibit resolvable cracking in NaOH solution for Eapp as low as −1.800 VSCE (see Figures 6 and 7). Environmental H cracking under severe cathodic polarization is considered in an ensuing section, but this issue requires additional H-uptake55-56,58 and cracking experiments. Prior work suggested the importance of polarization in IGSCC of AA5083-H321, either unsensitized or mildly sensitized (6 mg/cm2) and stressed in 0.6 M NaCl.14,59 Jones, et al., found that growth rates were faster in anodically polarized AA5083-H321 (4 nm/s and 3 nm/s at 25 MPa√m in unsensitized and DoS of 6 mg/cm2 microstructures, respectively) compared to rates at OCP (0.1 nm/s and 0.25 nm/s).14 These data were used in a coupled environmental fracture model of IGSCC in Al-Mg;59 predictions suggest that cathodic polarization is beneficial, but interaction of DoS and Eapp plus model uncertainties were not investigated. A similar benefit of cathodic polarization on IGSCC in Al-Mg-Cu (AA2024-T351) was reported without detailed consideration of crack tip dissolution and H uptake interaction.49
The second set of experimental observations establishes that IGSCC in Al-Mg is dominated by an acidic crack solution, consistent with the coupled crack dissolution and FPZ H embrittlement mechanism. For AA5083-H131 without grain boundary β, crack tip H uptake can be enabled by either anodic polarization and cation hydrolysis during exposure in near-neutral NaCl (Figure 2), or by use of highly acidic simulated crack solution as the bulk environment (Figure 6). For sensitized Al-Mg in neutral NaCl near OCP, grain boundary β dissolution triggers pH reduction for crack tip H production and uptake leading to α-boundary embrittlement. This scenario is eliminated by sufficient cathodic polarization (Figures 6 and 7), which stifles α and β breakdown to minimize dissolution and crack acidification. Collectively, these results validate the coupled crack dissolution and HEAC mechanism. Quantitative analyses are discussed to support this conclusion.
Occluded Crack Electrochemistry for 0.6 M NaCl Solution
Background on Intergranular Corrosion Rate Modeling
The environment in the occluded fatigue crack is governed by crack geometry coupled to distributed anodic and cathodic reactions that establish IR-affected crack tip potential (Etip), pH, and Cl− concentration, which in turn control the overpotential for crack tip H production (ηH = Etip − EH/H+) and H solubility.16-18,40,42-46,58,60-66 Modeling of unstressed IGC fissure penetration rate in Al-Mg in NaCl, as a function of Eapp and DoS,17-18,47 provides a basis for interpreting IGSCC. This model used Faraday’s law coupled with the measured current densities in Figure 8 from potentiodynamic scans of α and β phases in quiescent, acidic, simulated fissure AlCl3/MgCl2 solution or near-neutral 0.6 M NaCl solution.14-15,17-18,20 The breakdown potential (Ebd) of each phase-solution combination is identified in Figure 8. The highly reactive anodic electrochemical behavior of β, present with α in alloys such as AA5083, was affirmed for dilute (0.01 M) NaCl solution over a wide range of pH (from 2 to 10).67 With the inputs from Figure 8, Bumiller and Kelly estimated that an IGC-fissure tip in sensitized AA5083 (DoS of 24 mg/cm2 to 49 mg/cm2) must contain highly acidic solution (2.5 M Al3+, 0.11 M Mg2+, 7.7 M Cl−, pH −0.12) in order for the tip to corrode at measured rates, while flanks are passive and the tip remains sharp.17-18 This composition is the simulated crack solution used as the bulk environment in the fracture experiments in Figure 6.
Model-predicted rates of bold-surface IGC47 agree with measured rates of fatigue crack IGC for sensitized AA5083-H131 (22 mg/cm2) at low K in NaCl with various Eapp, as shown in Table 2. This agreement validates the IGC-electrochemical basis for crack chemistry modeling relevant to IGSCC. Both measured and predicted IGC rates decline with decreasing Eapp, and two regimes of behavior are suggested (Figure 4). IGC growth is substantial at Eapp of −0.730 VSCE and declines as Eapp falls to −0.975 VSCE. The IGC rate decrease is substantial between Eapp of −0.730 VSCE and −0.800 VSCE, consistent with Ebd-alpha of −0.790 VSCE for neutral NaCl (Figure 8). Less steeply falling IGC rate between −0.800 VSCE and −0.975 VSCE is consistent with reduced, but still significant, α dissolution because Ebd-alpha is −0.870 VSCE for acidic solution, and β remains active for some fissure/crack acidification. IGC growth rate tends toward 0 for Eapp less than −0.975 VSCE to between −1.020 VSCE and −1.050 VSCE, consistent with Ebd-beta of −0.960 VSCE (neutral 0.6 M NaCl) and −1.015 VSCE (acidic AlCl3/MgCl2 solution, Figure 8). Three points are notable. In acidic fissures at Eapp above −0.800 VSCE, the α passive film is unstable, and it follows that the role of stress in IGC is small. This film instability may explain why corrosion growth rates for the crack tip at low-static K are comparable to predictions of the stress-free IGC model (Table 2).47 Second, finite IGC rates measured for cathodically polarized precracked specimens (Eapp of −0.975 VSCE to −1.300 VSCE) are likely caused by α corrosion in an alkaline crack resulting from cathodic polarization not considered in the IGC model. Third, although β is active for a wide range of Eapp and pH,67 α dissolution is critical in IGC. The favorable comparison of IGC rates in Table 2 suggests that interaction of crack tip potential with Ebd-beta and Ebd-alpha controls crack chemistry and IGSCC, consistent with the dissolution-HEAC mechanism.
Al-Mg Alloy Crack Tip Potential and pH
Because a quantitative crack chemistry model is not available for Al-Mg in NaCl, the Eapp dependence of crack chemistry is estimated by comparing polarization kinetics for α and β (Figure 8) with the extent of cation dilution, ohmic difference, and crack tip pH. The aim is to estimate crack tip ηH, which governs IGSCC by HEAC.40,66,68-69 The pit stability product estimates the crack tip dissolution current density (itip) necessary to maintain cation enrichment without dilution for a given diffusion length (do).62-64,70 The critical stability product (idcrit) has not been reported for Al-Mg, but a value on the order of 5 mA/cm is reasonable.71-72 Three one-dimensional diffusion distances (do) are possible for the SENT specimen: (Orientation A) transport down the notch plus fatigue and IGC/IGSCC cracks (do≥3.3 mm), (Orientation B) transport across the crack front, from bold surface to mid-thickness (do = 3.3 mm), and (Orientation C) transport to quarter-thickness along the crack front (do = 1.65 mm). The latter two distances provide the shortest and presumably controlling path for cation dilution. Based on idcrit of 5 mA/cm and do of 3.3 mm (Orientation B), a current density of 15 mA/cm2 or larger maintains a solution saturated in Al3+ and Mg2+. This current density Faradaically corresponds to an IGC rate of 5.3 nm/s, which is close to that value of the IGC rate (6.5 nm/s) measured in the fatigue crack at Eapp of −0.730 VSCE (Table 2). As such, localized dissolution is sufficient to sustain saturated Al3+ and Mg2+ in the solution at the fatigue crack at mid-thickness. For do of 1.65 mm (Orientation C), the IGC growth rate must exceed 10.7 nm/s. This rate is somewhat larger than the fastest value measured for sensitized AA5083-H131 (Table 2), suggesting that ion diffusion parallel to the crack front dilutes Al3+ and Mg2+ at the quarter-thickness location. Because idcrit for Al may be as low as 2 mA/cm,71 it is estimated that mass transport will not dilute crack tip cation enrichment from α and β dissolution.
Second, ohmic drop along the crack depresses Etip below Eapp. Etip is estimated for Orientations A, B, and C, assuming a rectangular slot with passive walls and active tip.65 Two cases are modeled: IGC at the fatigue crack during a hold at K of 5 MPa√m and Stage II IGSCC propagation at K of 15 MPa√m. For fatigue crack tip IGC, slot opening equals the crack tip opening displacement (CTOD = 0.5 μm for K = 5 MPa√m73 ) for Orientations B and C, and the average of CTOD and crack mouth opening displacement (CMOD = 8 μm for K = 5 MPa√m74 ) for Orientation A. For IGSCC in Stage II, the length of typical crack growth ahead of the fatigue crack and IGC is 2.3 mm. For an elastic K of 15 MPa√m, blunted CTOD is 4 μm73 and CMOD is 21.8 μm.75 The average slot height is 12.9 μm for Orientation A, and equals CTOD (4 μm) for Orientations B and C. Solution conductivity is the average of 93 mS/cm for 80% saturated solution (2.4 M Al3+ and 0.11 M Mg2+) and 45 mS/cm for saturated (3.1 M Al3+ and 0.14 M Mg2+) solution in equilibrium with α.17-18 The itip is the current density associated with the IGC rate measured at the Eapp of interest, and wall current density is assumed to be 85% of the value for passive α (17 μA/cm2) in order to account for local cathodic reactions.17 The maximum ohmic drop is limited by the OCP at the crack tip, which is assumed to be OCP of α in the acidic solution (−1.050 VSCE17-18 ). Calculated Etip is plotted versus Eapp in Figure 9 for fatigue/IGC at low K (top) and fatigue/IGC/IGSCC at high K (bottom). The ohmic reduction is small for each orientation of the fatigue-IGSCC tip at K of 15 MPa√m (30 mV to 40 mV at Eapp = −0.800 VSCE), and for Orientation A of the fatigue-IGC crack tip at low K (10 mV at all Etip above the α OCP), as governed by relatively large slot openings. High ohmic drop (200 mV to 250 mV at higher Etip) is predicted for Orientations B and C because of small CTOD (0.4 μm, top plot) at low K. The Etip pertinent to Stage II IGSCC at K of 15 MPa√m is summarized in Table 3, and Orientation B is emphasized as a limiting case.
Several factors could increase the crack-ohmic difference to reduce Etip below the values shown in Figure 9. First, it is assumed that neither H2 bubbles nor corrosion product in the crack hinder ion transport and solution resistance. This complication can be important, but is stochastically complex.40 Second, CTOD is calculated based on large strain blunt crack continuum mechanics.73 However, recent results suggest that CTOD is reduced as a result of crack tip strain hardening from geometrically necessary dislocations that are present to accommodate the crack tip strain gradient.76-79 Third, IGSCC in a 3D microstructure may include a Mode II sliding component that produces rough-surface asperity contact and increased impedance. Use of the S-L orientation in the present work minimized this effect of crack path tortuosity. Finally, x-ray computed tomography observations demonstrate that the atmospheric-environment IGSCC tip-front in a highly anisotropic microstructure of AA5083 plate is composed of an array of multiple parallel (but not co-planar) cracks sized on the order of 1,000 μm long in the S direction.80 Macrocrack advance in the L direction requires rupture of the ligament, of order 25 μm high, between each microcrack. Local K and CTOD are reduced by this morphology of microcracks, relative to the far-field macroscopic K used in Figures 2 and 6. However, the magnitude of this reduction was not quantified for the nearly-planar array of cracks in the highly anisotropic microstructure of AA5083.80 It is premature to conclude that the Stage II character of da/dt is controlled by the multiple-crack geometry for anisotropic AA5083 and S-L orientation,80 or that Etip estimated for a simple slot (Figure 9) is significantly reduced by microscopic crack geometry. This is particularly true given the potential larger differences in crack tip stress field and CTOD resulting from strain gradient plasticity76-79 or crystal plasticity.81 Modeling of each of these issues is under development and application to predict HEAC in Al-Mg is not presently possible. Future research is required to extend the modeling presented here, as recognized in a similar study of Al-Zn-Mg-Cu alloys.40
Crack pH was not measured for sensitized AA5083 as a function of Eapp and DoS, but four results provide a basis for estimating this relationship. First, studies with Al-Zn-Mg-Cu in neutral NaCl demonstrate that measured crack tip pH increases with decreasing Eapp from −0.700 VSCE to −1.000 VSCE.42,46 Regression analysis shows that crack tip pH depends on Eapp (in VSCE) according to: pH = −22.421Eapp − 14.241.20 Second, unlike Al-Zn-Mg-Cu without β, crack solution in sensitized Al-Mg alloys is enriched in Mg2+ and Al3+ during β and α dissolution. Bulk solution measurements showed that the pH of 2 M AlCl3 solution (pH 1.04) slightly decreased to 1.02 with up to 0.01 M Mg2+ addition, but decreased sharply with increasing MgCl2 addition above 0.01 M. The pH is 0.94, 0.80, 0.65, and 0.50 at Mg2+ concentrations of 0.15 M, 0.3 M, 0.6 M, and 1.4 M, respectively.74 Third, solution experiments established cation solubilities in equilibrium with α (3.1 M Al3+ and 0.14 M Mg2+) and β (2.0 M Al3+ and 1.3 M Mg2+).17-18 Fourth, analysis of IGC in highly sensitized AA5083-H131 (22 < DoS < 49 mg/cm2) suggests that the fissure solution is likely 3.1 M Al3+ and 0.14 M Mg2+.47 The Mg2+ concentration used for crack modeling is 0.34 M for a DoS of 22 mg/cm2, which exceeds solubility and is justified in an ensuing section. This solution is an upward estimate from the composition of 2.4 M Al3+ and 0.11 M Mg2+ previously reported.17-18,20 Coupling these results, pH equals 0.8 in a sensitized AA5083 (22 mg/cm2) crack with Eapp of −0.730 VSCE as a result of Mg2+ enrichment, reduced from the Mg2+-free pH of 1.5 given by the Al-Zn-Mg-Cu correlation. With this calibration point, the crack tip pH at any Eapp is estimated by the slope of −22.4 pH/V. This assumed value of Eapp at −0.730 VSCE is a reasonable assumption, but not critical for Al-Mg crack pH estimation provided that all analyses that follow are based on this same correlation. Finally, ηH is calculated from Etip (Figure 9 bottom) and crack tip pH; calculated values are summarized in Table 3 and provide the basis for validating the HEAC mechanism for IGSCC in Al-Mg. This approach is superior to da/dtII correlation with crack pH, which ignores varying Etip that contributes to ηH.40,46,82
Intergranular Stress Corrosion Cracking Mechanism Validation
where DH-EFF is the trap-sensitive H diffusivity, measured to equal between 1.3 × 10−10 cm2/s and 6 × 10−9 cm2/s (23°C) for the plate of AA5083-H131 used in the present study and independent of DoS.55 CHσ is the FPZ hydrostatic stress enhanced concentration of diffusible H (CH-Diff) in equilibrium with crack tip ηH from Table 3 and located at xcrit.66,84 Equation (1) is based on discontinuous cracking over a distance, xcrit, from H damage sites in the FPZ back to the crack tip. An increment of growth occurs when diffusion enables CHσ to exceed the local-stress dependent critical level for H embrittlement, CH-crit. Time per crack advance corresponds to that necessary for H to diffuse from CH-Diff at the crack tip surface to CHσ at xcrit. Literature analysis establishes that xcrit is 0.9 μm for HEAC in a wide variety of alloys.83 Prediction of absolute da/dtII versus Eapp through Equation (1) is not possible because neither the ηH dependence of CH-Diff nor CH-crit is known for a sensitized grain boundary in Al-Mg.26,58 To test the dissolution-HEAC model for IGSCC, crack tip ηH (Table 3) is correlated with CH-crit/CHσ calculated from measured da/dtK15 using Equation (1).
The coupled environmental fracture model (CEFM)59,85-86 was not used to assess the Eapp dependence of IGSCC because of several limitations.20 First, this CEFM was neither calibrated nor validated for sensitized Al-Mg. Second, interpretation and determination of the crack tip net anodic current that couples with bold surface cathodic reactions is central to the CEFM, but controversial.87-89 Moreover, this coupling is not relevant to experiments under potentiostatic control in which the vast majority of the polarizing current is from the counter electrode, not the external surface of the working electrode. Third, this model lacks a FPZ failure criterion, but rather equates da/dtII to the product of the squared microfracture event distance, reciprocal crack front length, and fracture-event frequency (from measured noise transient or crack tip strain rate analysis). CEFM-calculated microfracture distances for low DoS Al-Mg are unrealistically large (50 μm to 2 mm) and not supported by modern crack tip stress field analysis or a pertinent microstructural distance.59 Hydrogen diffusion is not sustainable over such large distances for measured rates of IGSCC in sensitized Al-Mg.53,55
Effect of Applied Potential for Sensitized AA5083-H131
The Eapp dependence of ηH and its relationship to crack growth rate (Table 3) quantitatively validates the coupled dissolution and HEAC mechanism for IGSCC of Al-Mg in neutral NaCl. Figure 10 shows the Eapp dependence of measured da/dtK15 from Figure 6 for sensitized AA5083-H131 (22 mg/cm2), with crack tip ohmic drop corrections indicated by arrows for the three diffusion orientations (Table 3). This comparison establishes that both α and β dissolution are critical to IGSCC. Considering the rapid growth rates at Eapp of −0.730 VSCE and −0.800 VSCE, Etip for each orientation is well above Ebd-beta for the neutral and acidic chloride solutions, and somewhat above Ebd-alpha for acidic solution. With the exception of Orientation C, at Eapp of −0.730 VSCE, Etip is between Ebd-alpha for acidic solution and Ebd-alpha for neutral solution. This comparison suggests that grain boundary β dissolves for each Eapp in this range, which destabilizes (or activates90 ) passive α for intensified dissolution and acidification. In this role, β dissolution and the onset of acidification shift Ebd-alpha to below Etip to stimulate Al3+ and Mg2+ enrichment leading to: (a) an acidic crack with pH of −1.5 at Eapp = −0.730 VSCE and pH of 0.1 at Eapp = −0.800 VSCE, and (b) the corresponding substantial ηH levels of −0.68 V and −0.62 V, respectively, for Orientation B in Table 3. Figure 11 shows the correlation between these rapid levels of da/dtK15 and the two most negative values of ηH for ion-diffusion Orientation B; essentially identical results are obtained for Orientations A and C.
IGSCC rate severely drops at Eapp between −0.800 VSCE and −0.900 VSCE in Figure 10, corresponding to Etip between −0.843 VSCE and −0.938 VSCE, respectively (Orientation B in Table 3). The da/dtK15 decreases because Etip falls below Ebd-alpha of −0.870 VSCE for the acidic electrolyte (Figure 10); α dissolution rate is reduced to passive-Al dissolution, but β dissolution continues. Tip pH increases to 2.3 or higher, resulting in a reduced driving force for H uptake (ηH = −0.58 V in Table 3). Figure 11 shows that this small reduction in ηH correlates with a 3 order of magnitude fall in da/dtK15. Growth rate is low but constant (6 nm/s) for Eapp of −0.900 VSCE to −1.020 VSCE (Etip of −0.938 VSCE to −1.050 VSCE). In this potential range, the crack tip is acidified by β dissolution only, as the passive α dissolution rate continues to fall (Figure 8). Eapp of −1.000 VSCE to −1.020 VSCE aligns with Ebd-beta of −1.015 VSCE for acidic solution and is somewhat below Ebd-beta for neutral chloride. Accordingly, when Eapp is lower than about −1.020 VSCE, crack tip potential approaches the lower bound OCP typical of α in acidic chloride (−1.050 VSCE). Etip for each such Eapp is below both Ebd-alpha and Ebd-beta, which limits dissolution of α and β. Critically, polarization below Ebd-beta (−1.020 VSCE) precludes the acidic crack chemistry necessary for ηH that is sufficient for HEAC and finite da/dtII (Equation ). This condition is the justification for the trend line in Figure 11.
Applied cathodic polarization below −1.020 VSCE should not lower Etip below −1.050 VSCE, and crack pH continues to rise so ηH continues to be less negative, yielding a lower driving force for HEAC. It is not possible to specify the precise Eapp for immunity to IGSCC, as low rate HEAC could persist in alkaline crack solution. Moreover, bold surface H production should increase with decreasing Eapp in this regime, suggesting that cathodic hydrogen embrittlement could occur. This consideration is assessed in an ensuing section by experiments conducted with as-received AA5083 stressed in a variety of neutral and alkaline electrolytes with highly cathodic polarization.
Equation (1) provides a means to project the ηH dependence of occluded crack tip H solubility (CH-Diff) in the form of CHσ/CH-crit. Figure 12 shows that calculated CHσ/CH-crit depends strongly on ηH. For this result, xcrit is 0.9 μm and DH-EFF is 1.0 × 10−8 cm2/s, justified in the companion paper26 and representative of the fast H diffusivities measured for sensitized AA5083-H131.46,53,58 The correlation in Figure 12 is fundamental because da/dtII dependence on ηH (Figure 11) is: (a) amplified by the inverse error function in Equation (1), where a small change in the argument results in a large change in the function when CHσ/CH-crit is just above 1.0, or (b) damped where a large change in the argument results in a small change in the function when CHσ/CH-crit is 5 or more. Figure 12 is unaffected by these factors. Consider two points. First, CHσ/CH-crit is 1.02 to 1.01 as Eapp falls from −1.080 VSCE to −1.300 VSCE with measured da/dtK15 of 1.3 nm/s to 1.0 nm/s. By model definition, Stage II da/dt = 0 when CHσ/CH-crit < 1.0; Figure 12 affirms that HEAC is eliminated as a result of crack tip CHσ falling below a critical level. Validated definition of the Eapp where this occurs is clouded because these low growth rates fall to the resolution limit of dcPD measurement of cracking. For Eapp above −0.900 VSCE, CHσ increases sharply with increasingly negative ηH, traced to increasingly severe crack acidification (Table 3). While CHσ/CH-crit in Figure 12 could be analyzed to yield the relationship between CH-Diff and ηH at the crack tip, this analysis was not pursued as a result of uncertainties surrounding CH-crit and the magnitude of crack tip hydrostatic stress which yields CHσ.26,69 The relationship between H solubility and ηH is complex, but several features are likely, as inferred from results for Fe and Ni alloys:66,68,84 (1) H solubility rises from essentially 0 as ηH becomes more negative, (2) dCH-Diff/dηH progressively rises as ηH becomes more negative, (3) a plateau in H solubility exists above an absolute level of ηH, (4) CH-Diff is higher for acidic compared to basic solutions at constant ηH, and (5) Mg2+ and Al3+ may affect CH-Diff. Figure 12 is consistent with several of these features.
This analysis for Al-Mg suggests that IGSCC in Al-Zn-Mg-Cu should be impacted by crack acidification stimulated by dissolution of active grain boundary precipitates. A similar conclusion was recently reported.40 IGSCC in peak aged AA7075 (UNS A97075) and AA7050 (UNS A97050) is reduced by modest cathodic polarization below OCP; da/dtII fell from 120 nm/s to 3 nm/s as Eapp decreased from −0.620 VSCE to −1.280 VSCE for AA7075-T651 in 1 M NaCl,44-45 and from 90 nm/s to 1 nm/s as Eapp decreased from −0.495 VSCE to −0.760 VSCE for AA7050-T651 in 0.05 M NaCl solution inhibited with 0.5 M Na2CrO4.91-92 The grain boundary precipitate in these alloys is η (Mg(Zn,Cu)2), with a low (1 at% to 3 at%) Cu content for peak aged AA7075.93-94 The breakdown potential of η in 0.5 M NaCl (pH 7, deaerated) is −1.14 VSCE to −1.10 VSCE for Cu contents of 0 at% to 8 at%,95 and is likely lower for acidic NaCl following the behavior of β (Figure 8). As such, η is active for the potential range that favored IGSCC and the crack should acidify. As Eapp falls over the range noted for AA7075-T651, Etip falls from −0.800 VSCE to −1.120 VSCE45 and crack tip pH rises from 2 to 12,20,46 providing for reduced ηH and decreasing da/dtII by the HEAC mechanism. This scenario was validated by thermal desorption spectroscopy measurements of crack-wake H concentration, which decreased as Eapp and da/dtII fell for peak aged AA7050 (S-L).91-92
The da/dtII for IGSCC in Al-Zn-Mg-Cu stressed in neutral NaCl near OCP (about −0.75 VSCE95 ) falls by 4 to 6 orders of magnitude because of mild overaging45,91,96-97 and IGC is similarly reduced.98 This benefit of overaging is substantially reduced for Al-Zn-Mg without Cu, suggesting an important role of precipitate composition.43,91,97 Specifically, the Cu content of boundary η rises from 1 at% to 3 at% for peak aged AA707593-94 to 10 at% to 30 at% for the overaged temper, depending on alloy Cu level and heat treatment.93,98-99 The Ebd-eta for Mg(Zn,Cu)2 with 25 at% Cu is −0.940 VSCE in neutral deaerated 0.5 M NaCl.95 Speculatively, higher Cu content of overaged η increases Ebd-eta, above Etip (about −0.8 VSCE) associated with bold surface OCP for near-neutral NaCl. Because this Etip is somewhat above Ebd-eta for Cu-enriched η, the role of boundary precipitate stifled crack acidification is not quantitatively supported. Moreover, experiments demonstrated that IGSCC in resistant overaged AA7050 was not exacerbated by either: (a) anodic polarization from −0.55 VSCE to −0.30 VSCE for neutral NaCl,92 or (b) injection of simulated acidic crack solution into the crack tip.46 These experiments involved Na2CrO4 inhibited NaCl, so the relevance to crack acidification in near-neutral NaCl is in question. The lack of correspondence between Etip versus Ebd-eta could be explained by Cu plating on crack surfaces after η dealloying,95 leading to reduced acidification not encountered in Al-Mg. Additionally, 7xxx alloys exhibit intragranular η‘ precipitation and a grain boundary precipitate-free zone (PFZ). Peak to overaging changes these features, reduces deleterious-heterogeneous slip localization,100-102 alters trap-sensitive H diffusion proximate to grain boundaries,103 and affects PFZ characteristics.43 Recent work concluded that decreased rates of crack tip anodic and cathodic H-production reactions, as a result of the increased Cu content of boundary η from overaging, dominate these other contributions to IGSCC.99 However, this work did not detail crack pH and ηH changes with aging and Eapp. The role of crack chemistry in IGSCC is important for Al-Zn-Mg-Cu, but not as dominant as observed for Al-Mg; additional work is needed.40
Intergranular Stress Corrosion Cracking in AA5083-H131 Without Sensitization
The environmental conditions that promote IGSCC in low DoS AA5083-H131 without grain boundary β also validate the coupled crack tip dissolution and HEAC mechanism for IGSCC in sensitized Al-Mg. Considering IGSCC in near-neutral 0.6 M NaCl solution at Eapp of −0.800 VSCE, subcritical cracking only occurred in low DoS AA5083-H131 (2 mg/cm2) when stressed at very high K (>25 MPa√m); the da/dtK15 of 1.5 nm/s is low (Figure 6) and equals the upper bound of dcPD resolution. In contrast, IGSCC was produced at KTH of 5 MPa√m to 7 MPa√m and da/dtK15 of 50 nm/s to 250 nm/s for this microstructure stressed in either NaCl with anodic polarization (Eapp = −0.730 VSCE), or in acidified AlCl3/MgCl2 solution at mild cathodic polarization (Eapp = −1.000 VSCE). For neutral NaCl, the geometry of the fatigue precrack favors local acidification upon anodic polarization. Literature confirmed that Al-Mg without β is susceptible to HEAC when stressed in several acidic environments.14,52,104 The current experiments confirm the requirement for an acidic crack environment to enable H production for HEAC of Al-Mg and affirm the coupled dissolution-HEAC mechanism. Grain boundary β is not required for stress concentration, local constraint, or H trapping during IGSCC in AA5083-H131.26
Crack chemistry and ηH were estimated following the previously outlined approach in order to quantitatively interpret the experiments with low DoS AA5083 in the environments of Figure 6; calculated crack tip potential, pH, and ηH are given in Table 4. Input parameters for Etip calculation were estimated for unsensitized AA5083-H131 in NaCl at K of 15 MPa√m with 0.3 mm of IGSCC extension. For mass transport Orientation B: do is 3.3 mm, CTOD is 4 μm, and other parameters were previously cited. Because of similar IGSCC growth rates, the parameters used for unsensitized AA5083 in AlCl3/MgCl2 are the same as those used for sensitized AA5083 stressed in NaCl at K of 15 MPa√m (crack extension of 2.3 mm and CTOD of 4 μm).
Crack pH is considered for each environment. For NaCl solution, the correlation between Eapp and crack tip pH developed for 7xxx alloys is used,20,42,46 with two important changes. First, acidification is not enhanced as a result of Mg2+ because reactive β is not present in low DoS AA5083-H131;26 as such, crack tip pH = −22.421Eapp − 14.241.20 For Eapp of −0.730 VSCE, crack solution is assumed to be enriched to the solubility of Al3+ (3.1 M) in equilibrium with dissolving α because even a low current density of 0.5 μA/cm2 produces this Al3+ concentration in the crack after 10 h of exposure, assuming no loss of Al3+ via mass transport. This current density is exceeded at Eapp of −0.730 VSCE because Etip (−0.762 VSCE, Table 4) is above the neutral pH Ebd-alpha (−0.790 VSCE, Figure 8), which falls to −0.870 VSCE as the crack tip acidifies. This Al3+ concentration is associated with a crack pH of −1.2.46 At Eapp of −0.800 VSCE, the situation is different for β-free Al-Mg because the associated Etip (−0.821 VSCE) is below the neutral pH Ebd-alpha (−0.790 VSCE). Without β to stimulate/activate α dissolution, crack pH is neutral to slightly alkaline provided that passivity limits dissolution so that Al3+ is below the level of about 0.1 M necessary for crack acidification from pH 8 toward 3.42,46 Passivity must be strong because a current density of 0.015 μA/cm2 produces 0.1 M Al3+ in an occluded crack of the geometry modeled here, and this Al2+ concentration is the boundary above which crack acidification is significant.42,46 None-the-less, the argument that crack pH is well above 5 for β-free Al-Mg is justified by: (a) a higher breakdown potential for Al (−0.790 VSCE or higher) in neutral chloride,20,90 (b) the absence of bold-surface IGC for as-received AA5083 in 0.6 M NaCl at Eapp as high as −0.760 VSCE,105 and (c) the near-zero rate of IGC (0.03 nm/s) measured for a fatigue precrack in as-received AA5083 (2 mg/cm2) in 0.6 M NaCl at Eapp of −0.800 VSCE (Figure 5). There is no justification to support crack acidification in Al-Mg without either a reactive grain boundary precipitate, such as β, or anodic polarization; however, a lower limit on crack pH of 5 is included in Table 4 for sensitivity analysis in the context of IGSCC. The Al-Zn-Mg-Cu pH correlation is assumed accurate at Eapp = −1.020 VSCE and crack pH is 8.6. For the crack in acidic simulated crack solution, the pH is bounded by the level governed by Al dissolution to reach the solubility of 3.1 M Al3+ (pH −1.2) and that dominated by crack tip cathodic reaction and isolation from the bulk solution (pH 4). Because there was no evidence of cathodic corrosion within the crack for this acidic crack solution, the average of these bounds is used. For NaOH solution, Etip and pH were estimated assuming crack tip isolation from bulk solution resulting from corrosion product.20
Predicted crack tip ηH in Table 4 for AA5083-H131 without boundary β correlate with crack growth rate in excellent agreement with the relationship between ηH and either da/dtK15 or CHσ/CH-crit established for sensitized AA5083-H131 (22 mg/cm2). Figure 13 shows that the H overpotentials associated with anodic polarization (ηH = −0.61 V) and simulated crack solution (ηH = −0.74 V) are sufficiently large to support substantial da/dtK15 for low DoS AA5083-H131. In contrast, ηH of −0.13 to −0.30 V for Eapp of −0.800 VSCE is insufficient to enable HEAC; measured da/dtK15 is low. The growth rate versus ηH relationship for the low DoS case aligns with that developed for sensitized AA5083-H131. Values of CHσ/CH-crit, calculated from Equation (1) with measured da/dtK15, increase with increasing ηH as shown in Figure 14, paralleling the behavior for sensitized AA5083-H131. In this calculation, DH-EFF of 1 × 10−10 cm2/s was used for AA5083 without grain boundary β, compared to DH-EFF of 1 × 10−8 cm2/s for sensitized AA5083-H131. A mechanism for the lower diffusivity of H when β is not present on grain boundaries is presented elsewhere.26 This approach is justified because measured da/dtK15 for as-received AA5083-H131 in NaCl (Eapp of −0.730 VSCE) and AlCl3/MgCl2 (−1.000 VSCE) solutions agree with absolute growth rates predicted using Equation (1). Collectively, these findings validate the crack tip dissolution-HEAC model for IGSCC, where the centrally important acidified crack solution is controlled by Eapp-dependent β dissolution for sensitized Al-Mg, and by either acidic simulated crack solution or anodic polarization for Al-Mg without grain boundary β.
Figures 13 and 14 show that, consistent with HEAC, Stage II da/dt and CHσ/CH-crit each rise as the H overpotential becomes more negative for both β-free and sensitized AA5083-H131. Somewhat lower da/dtK15 levels (and the associated lower ratio of CHσ to CH-crit) are produced for the low DoS microstructure at fixed crack tip ηH. This difference is speculatively explained based on crack tip mechanics and H trapping/diffusivity tied to grain boundary β, not related to crack chemistry, as developed in the companion paper.26 In essence, elimination of grain boundary β for the low DoS case results in reduced local stress concentration; as such, CHσ is reduced and CH-Crit is increased, consistent with the trends in Figures 13 and 14.
Figure 6 shows that AA5083-H131 without boundary β resists IGSCC in NaCl and alkaline NaOH solutions at highly cathodic Eapp. This behavior is explained by the small ηH (−0.32 V to −0.39 V in Table 4, Figure 13, and Figure 14), even with Eapp well below −1.0 VSCE. Similar crack tip potential and pH (−1.200 VSCE at pH 10 in Table 4) are projected for an isolated crack tip at either OCP or polarized to −1.800 VSCE; the calculated ηH is relatively small (−0.39 V), which correlates with small CHσ/CH-crit of 1.2 (Figure 14) and is consistent with low measured da/dtK15 in Figures 6 and 13. The lack of HEAC in the alkaline environment affirms the importance of an acidic crack tip for the amount of H uptake required for IGSCC. Without an acidic crack tip, limited H production and uptake may at best support low rate HEAC (~0.1 nm/s to 1.3 nm/s) in AA5083-H131 at high K. Low crack growth rates in these resistant material-environment scenarios are of the same order as the maximum reported level of water vapor HEAC.26 This analysis suggests that cathodic prevention of IGSCC in sensitized Al-Mg alloys will not produce hydrogen embrittlement above that typical of HEAC in moist air. Long-term cracking experiments are required to examine the impact of bold-surface H production on cracking.
The relationship between crack tip H solubility and ηH in Figure 14 must be validated for Al-Mg alloys in acidic and alkaline solutions.66,84 Pending such results, which are challenging to obtain for H in Al alloys,55,58 the dependencies of da/dtK15 (Figure 13) and CHσ/CH-crit (Figure 14) on crack tip ηH provide a basis to predict the DoS dependence of IGSCC rate.
Degree of Sensitization Dependence of Intergranular Stress Corrosion Cracking in AA5083-H131
This study demonstrates that: (a) crack acidification is necessary for IGSCC by HEAC in Al-Mg, and (b) intergranular H embrittlement is possible without β at grain boundaries. Given this validation, the crack tip dissolution and HEAC mechanism is used to model the DoS dependence of IGSCC in AA5083-H131. IGC47 and IGSCC26 rates are reported versus DoS in two companion papers, and a critical NAMLT of 9 mg/cm2 to 12 mg/cm2 is identified for the onset of IGSCC in near-neutral NaCl solution at Eapp (−0.800 VSCE) near OCP.26 The S-L orientation minimizes the effect of crack path tortuosity on IGSCC rate, which is not captured by a chemistry model. For the lot of AA5083-H131 in Table 1, the volume of grain boundary β is quantitatively known as a function of DoS for the time-temperature conditions used in the crack growth study.12,16,27-38 This section tests the hypothesis that the critical DoS for IGSCC is governed by a threshold volume of boundary β, which enables threshold-crack enrichment in Al3+ and Mg2+ from β-α dissolution, as necessary to lower crack pH and raise ηH above the threshold of −0.58 V (arrow in Figure 14) or −0.53 V (Figure 13).
All grain boundary β rapidly dissolves at Eapp of −0.800 VSCE17-18,67 to stimulate α dissolution, which collectively establish crack solution Al3+ and Mg2+ concentrations. Occluded crack solution volume is approximated by a rectangular slot defined by the blunted CTOD and flank area of separated boundary faces. The focus is on the near-tip solution composition over a distance of order 25 times CTOD behind the tip. Several assumptions are necessary: (1) HEAC during rising K is governed by the fatigue crack chemistry established during the 10 h hold at K of 4 MPa√m, where CTOD is 0.4 μm for AA5083-H131, (2) the high ohmic drop predicted for Orientation B (Figure 9 top) is relieved by ionic transport along Orientation A, (3) Al3+ concentration is not diluted during rising K following pit-stability product analysis for dissolving α, and (4) cation concentration from dissolved β is an upper bound because the amount of this phase is relatively small and dilution could occur as the slot opens. Refined analysis is not possible, because the blunt crack CTOD model may overestimate the opening displacement of an intergranular crack, and H2 bubbles or corrosion product may impact Etip. With these assumptions, Table 5 shows the DoS dependence of the concentrations of Al3+ and Mg2+ provided by full dissolution of the volume of β precipitates per grain boundary area. β volume is estimated using the equivalent-mean radius of β spheres (rmean from Equation  in Crane and Gangloff26 ) and number of β precipitates per grain boundary surface area (nd), each measured for the lot of AA5083-H131 used in the present study and sensitized without prior additional solution heat treatment and quench (SHT/Q):37
ρβ and MWβ are the density (2.3 × 106 g/m3) and molecular weight (129.0 g/mol) of Al3Mg2.20 The Al3+ and Mg2+ molarities are 3 times and 2 times the molarity of dissolved β, respectively. The DoS dependence of cation contributions from dissolution of the α area of the crack wake at Eapp of −0.800 VSCE is presented in Table 6 and summed with the cation concentrations from β dissolution (Table 5). For each composition, near-tip pH was estimated from the Mg2+-adjusted pH versus Eapp correlation,20,46 and Etip was calculated using the ohmic difference model.65
For β-free AA5083 (DoS of 0 mg/cm2 to 3 mg/cm2 in Table 6), the range of crack pH is 5 to 8, given the lack of boundary β present to destabilize α passivity at Eapp of −0.800 VSCE, as discussed with regard to Table 4. At the other extreme for high DoS between 22 mg/cm2 and 49 mg/cm2, the estimated concentrations of Al3+ and Mg2+ demonstrate that grain boundary β plus crack wake α dissolution are sufficient to produce saturation-level cation concentrations. In this calculation it is assumed that: (a) cations from α dissolution cannot exceed solubilities in equilibrium with this phase (3.1 M Al3+ and 0.14 M Mg2+, as indicated by the underlined and parenthetically enclosed concentrations in Table 6),17-18 (b) cations from β dissolution cannot exceed solubilities in equilibrium with this phase (2.0 M Al3+ and 1.3 M Mg2+),17-18 (c) Al3+ from α can exceed solubility associated with β because Al3Mg2 is removed by dissolution, and (d) Mg2+ from β can locally exceed the solubility dictated by α. Ebd-alpha is −0.870 VSCE for concentrated acidic chloride solution,17-18 and Al3+ plus Mg2+ from β dissolution activate otherwise passive Al.90 That is, in the presence of the amounts of grain boundary β shown in Table 5, α passivity breaks down at Eapp of −0.800 VSCE and stoichiometric dissolution of Al-4%Mg at any rate above about 0.6 μA/cm2 yields cation concentrations that exceed the α-based solubility of Al3+ (2.0 M) and Mg2+ (0.14 M) for the crack geometry used in Equation (2), as noted by underlined concentrations in Table 6. Given the amount of grain boundary β present for the DoS of 16 mg/cm2 microstructure, it is reasonable to expect that ionic enrichment in the crack follows the behavior of the higher DoS cases.
For DoS of 10 mg/cm2, the amount of boundary β is finite, but reduced compared to higher DoS; as such, the Al3+ and Mg2+ concentrations from β dissolution are 0.077 M and 0.052 M, respectively (Table 5). This Al3+ concentration should reduce crack pH from 8 to 9, to 4 to 3,20,42 but this low level of Mg2+ does not cause additional pH fall.64 β-stimulated acidification should decrease the Ebd-alpha from −0.790 VSCE in neutral NaCl toward −0.870 VSCE.6,14 Alternately, a passive surface of Al is activated for dissolution when Al3+ exceeds 2 mM to 8 mM, with this threshold falling somewhat as Cl− rises.90 From either perspective, significant α dissolution is triggered by Al3+ from the amount of β dissolution estimated for DoS of 10 mg/cm2.
The exact Al3+ concentration from α dissolution is not known. Assumed levels of 0.5 M and 1 M correlate with crack-solution pH of 3.0 and 2.2, respectively, from the pH versus Al3+ correlation.42,46 Alternately, the correlation with Eapp of −0.800 VSCE for Al-Zn-Mg-Cu yields a crack solution pH of 3.7.42,46 These values bound crack pH for DoS of 10 mg/cm2 in Table 6.
The predicted DoS dependence of crack tip ηH is combined with the dependence of da/dtK15 on ηH (Figure 14) to predict the Stage II rate of IGSCC in AA5083-H131 for each DoS in Table 6. Predicted da/dtK15 values are in excellent agreement with the experimental DoS dependence,26 each for stressing in neutral NaCl at Eapp of −0.800 VSCE, as shown in Figure 15. For a DoS of 10 mg/cm2, ηH of −0.46 V to −0.37 V is below the critical level of −0.53 V (Figure 13) to −0.58 V (Figure 14) necessary for HEAC and measurable da/dtK15, while the ηH for 16 mg/cm2 (−0.61 V) is above this critical level and sufficient to produce fast IGSCC. This prediction agrees with the measured-DoS threshold of 9 mg/cm2 to 12 mg/cm2 for sensitized AA5083-H131 (S-L).26 Steep critical-DoS behavior is caused by sufficient grain boundary β to stimulate α dissolution and crack acidification, coupled with the inverse error function term in the da/dtII model (Equation ) where a small increase in CHσ above CH-crit results in a sharp increase in growth rate. Thereafter, increasing DoS from 16 mg/cm2 to 50 mg/cm2 produced only a mild increase in da/dtK15 because the change in crack pH and ηH is modest (Table 6) resulting from modest increase in boundary β, and the inverse error function term dampens the increase in da/dtII with rising CHσ/CH-crit when this ratio exceeds about 5. Model predictions in Table 6 and Figure 15 strongly support the existence of a critical DoS for the onset of IGSCC, as well as the unexpectedly low value of about 10 mg/cm2 for the lower bound S-L orientation. Quantitative SEM measurements37 demonstrate that a low DoS of 10 mg/cm2 produces a sufficient amount of grain boundary β to enable α dissolution and crack acidification for a strong-negative ηH.
Crack ηH was examined for alternate microstructures of AA5083-H131 where the amount of grain boundary β was reported versus DoS.31,36-37 Table 7 shows that re-solution heat treatment and quench prior to sensitization modestly decreases the volume of grain boundary β for a DoS of 10 mg/cm2, but increased β content by 1 to 2 orders of magnitude for DoS of 22, 39, and 49 mg/cm2 compared to the values from Table 5. The β size, number density, and spacing were obtained by SEM of Ga-revealed grain boundary surfaces.31,37 The concentrations in Table 7 suggest that the critical DoS near to 10 mg/cm2 is not affected by the additional SHT because α dissolution provides the dominant Al3+ concentration that controls crack pH. Moreover, the onset of IGSCC at about 10 mg/cm2 should be even more dramatic for the SHT/Q case. The very large amounts of grain boundary β for the higher DoS-SHT/Q cases may or may not be statistically significant, and should at most modestly enhance already rapid da/dtII because of the damping effects of Al3+ solubility in equilibrium with α and the inverse error function term in Equation (1). Experiments are required to characterize the IGSCC resistance of AA5083-H131 with SHT/Q prior to sensitization.
As a second example, it is important to define the effect of sensitization temperature on the critical DoS for IGSCC. Experimental da/dtK15 in Figure 15 show that a critical DoS is suggested for sensitization at 100°C (between 10 mg/cm2 and 22 mg/cm2) and 60°C (between 9 mg/cm2 and 15 mg/cm2); however, measurements for 80°C sensitization are not sufficient to establish a transition. Grain boundary β size and coverage were reported for the plate of AA5083-H131 in Table 1 sensitized without SHT/Q at 80°C37 or 70°C.36 Table 8 contains the model results for β-only dissolution and crack cation enrichment following the method used for 100°C sensitization (Table 5). For 80°C and 100°C, grain boundary β was spherical based on transmission electron microscopy (TEM) and with number density measured by the Ga separation method.31,37 A DoS of 10 mg/cm2 produces similar grain boundary β volumes that yield similar Al3+ and Mg2+ concentrations for sensitization at 100°C and 80°C. From the analysis leading to Figure 15, IGSCC is not expected, consistent with the single measurement of a low da/dtK15 (□) at a DoS of 15 mg/cm2 (80°C). β volume increases as DoS rises at 80°C, albeit less so compared to the 100°C case, and thus it is not possible to definitively determine if the critical DoS for IGSCC rises with decreasing sensitization temperature. For example, the β volume for the 80°C DoS of 22 mg/cm2 (7.8 × 10−22 m3/μm2) is smaller than that (1.4 × 1021 m3/μm2) for 16 mg/cm2 (100°C), which was argued to produce sufficient crack tip ηH and rapid IGSCC. The impact of this lower β content on disruption of α passivity and acidification for the 80°C/22 mg/cm2 case is not known with sufficient precision to establish if the critical DoS for 80°C sensitization is below 22 mg/cm2.
Crack chemistry estimates for sensitization of AA5083-H131 at 70°C (Table 8) show that the Al3+ concentration rises by a factor of 27 for increasing DoS from 10 mg/cm2 to 22 mg/cm2, consistent with a critical DoS in this range as verified in Figure 15 for sensitization at 60°C. However, absolute values of the cation concentration estimates are 575 times (10 mg/cm2), 39 times (22 mg/cm2), and 9 times (39 mg/cm2) lower compared to the values estimated for 80°C sensitization in Table 8. For 70°C, β size and coverage were measured by TEM,36 rather than SEM of a Ga-separated surface.31,37 The β morphology was imaged as very thin (4 nm to 15 nm) ellipsoidal-capped plates, which increased in radius with fixed center-to-center spacing of 200 nm,36 at odds with the β spheres (50 nm to 270 nm diameter) suggested by TEM and the Ga attack studies.31,37 This smaller size of the β plates compared to β spheres for the same plate of sensitized AA5083-H131 is the cause of the cation concentration difference shown in Table 8. Because 60°C and 100°C sensitizations produced a similar-measured IGSCC response (Figure 15), it is unlikely that sensitization at 70°C produced the uniquely low amount of grain boundary β reflected in Table 8. While the Ga study was limited to sensitization at 80°C and higher, no fundamental reason was offered to justify a transition from β spheres to thin plates at the lower (70°C) sensitization temperature.31,36-37 This difference in β morphology and volume available for dissolution, as well as absolute identification of β phase in the Ga experiments, must be understood before the effect of sensitization temperature on crack chemistry and IGSCC kinetics can be broadly established by theory and experiment.
Critical cracking experiments and occluded crack electrochemistry analysis confirm the hypothesized mechanism for intergranular stress corrosion cracking (IGSCC) of sensitized Al-Mg alloys: dissolution of discontinuously arrayed grain boundary β (Al3Mg2) precipitates destabilizes/activates passive α (Al-Mg solid solution) dissolution leading to occluded crack solution acidification, which is necessary for crack tip H production/uptake and H embrittlement of α boundary segments between β. When β dissolution is eliminated by cathodic polarization, or minimized by low degree of sensitization (DoS) below a critical level, IGSCC is mitigated for slow-rising stress intensity (K) loading in neutral NaCl solution. Alternately, β-free AA5083-H131 is susceptible to IGSCC when the crack is acidified by anodic polarization or bulk environment composition. In essence crack tip acidification from either increasing grain boundary β volume or anodic polarization is critically necessary for IGSCC by hydrogen environment embrittlement (HEAC). Specific conclusions are as follows:
For sensitized AA5083-H131 (22 mg/cm2, S-L orientation), the rate of intergranular corrosion (IGC) growth of the fatigue crack at low K decreases, threshold stress intensity for the onset of IGSCC (KTH) increases, and Stage II rate of IGSCC growth (da/dtK15) decreases with applied polarization (Eapp) that is increasingly cathodic to the open circuit potential. Each damage mode appears to be eliminated at Eapp of −1.020 VSCE and lower for slow-rising K loading, and correlates with crack tip potential falling below the breakdown potential for β (−1.015 VSCE).
AA5083-H131 without grain boundary β (2 mg/cm2, S-L) resists IGC and IGSCC at fixed potential (Eapp = −0.800 VSCE) near OCP in 0.6 M NaCl because α dissolution is minimal and the crack tip is at best only mildly acidic. IGSCC is promoted by anodic polarization or simulated occluded crack solution (AlCl3 + MgCl2 at pH −0.3) as the bulk environment. This microstructure is not susceptible to hydrogen cracking under severe cathodic polarization in alkaline NaOH.
Stage II crack growth rate quantitatively correlates with the overpotential for H production (ηH) based on estimated occluded crack tip pH and ohmic drop, and crack tip H solubility calculated from measured Stage II crack growth rate sensibly correlates with crack tip ηH. This result validates the coupled dissolution and HEAC mechanism for IGSCC in Al-Mg; however, absolute values of IGSCC rate cannot be predicted because of the uncertain dependence of H solubility on ηH, pH, and Cl− for Al-Mg.
Severe IGSCC above a relatively low-critical DoS of about 10 mg/cm2 for the susceptible S-L orientation is governed by achievement of a necessary critical crack electrochemistry when the amount of dissolved active grain boundary β is sufficient to destabilize the passive Al-Mg solid solution, leading to crack tip solution enrichment in Al3+ and Mg2+ for sufficient acidification and H uptake for grain boundary embrittlement.
Applied cathodic polarization could provide a means to mitigate IGSCC in sensitized Al-Mg alloys, provided that long-term experiments validate accelerated test measurements and crack electrochemistry-based modeling of very low-rate HEAC.
This research was sponsored by the Office of Naval Research with Dr. Airan Perez as the Scientific Officer. AA5083-H131 was provided by the Alcoa Technical Center, Prof. Nick Birbilis shared unpublished data on grain boundary β characteristics,37 and Dr. R. Goswami provided unpublished TEM results.33 Data on the effect of magnesium on pH were supplied by Mary Lyn Lim.75 These important contributions are gratefully acknowledged.
UNS numbers are listed in Metals and Alloys in the Unified Numbering System, published by the Society of Automotive Engineers (SAE International) and cosponsored by ASTM International.