Grain boundary anodic phases affecting environmental damage

1 Introduction

It has been observed that there are two major types of grain boundary (GB) precipitates in most commercial precipitation-hardened alloys that can affect the intergranular stress corrosion behavior in an aqueous solution environment. These are of the type anodic and cathodic precipitates in relation to their respective matrices. Anodic precipitates usually undergo dissolution in preference to their matrix in an aqueous solution due to galvanic currents, whereas cathodic precipitates help their interface and their surrounding matrix to dissolve. These two types of GB precipitates play an important role in stress corrosion cracking initiation (and growth) kinetics, in an aqueous electrolyte medium.

Here, for the sake of analysis, we choose two alloys for the study, both having anodic grain boundary precipitates (GB-ppt). The 5083 alloy (Al-4.5Mg-0.7Mn) is a commercial alloy, and the other is an experimental Al-2.9Li (with 2.7–2.9 Li) binary alloy. The mechanical behavior of these two alloys markedly differ in the inert (considered as lab air) environment, after aging. For example, 5083 alloy shows a constant yield stress (YS) with a small variation in fracture toughness (K_Ic) with aging time at 175°C. This constancy of YS with aging time is mainly due to the solid solution strengthening of Mg resulting in serrated yielding and Luders band propagation. Fracture toughness is controlled by the second-phase constituent (such as AI₆ (Fe, Mn) and AI₃Fe) and dispersoid (such as Al₂₀Mn₃Cu₂ and AI₁₂Mg₂Cr) particles in the 5083 alloy. It is not affected by GB-β (β is Al₃Mg₂). On the other hand, Al-2.9Li alloy shows YS increasing with aging time to a peak value and then decreases with further aging. Thus, aging behavior AL-2.9Li alloy is similar to that of other precipitation hardening alloys such as 7075 and 2024. In the underage (UA) region of this alloy, slip is planar, and in the overage (OA) region, it is wavy. For this alloy, strengthening comes from ordered δ′ (Al₃Li) precipitates in the matrix. In contrast to the YS, fracture toughness (Charpy K_Ic) decreases with aging time in the UA region, reaches a minimum at peakaged (PA), and levels off in the OA region. In the UA region, the fracture is mixed mode (slip+IG) and transitions to completely intergranular fracture (IGF) in the OA region. The IGF is controlled by GB-δ precipitates via GB void nucleation and growth. Given these mechanical property differences in the two alloys, we analyze the fracture behavior in inert (lab air, K_Ic) and in NaCl (K_Iscc) solution environments.

The experiments are conducted in a constant 3.5% NaCl solution (for AL-2) and 1% NaCl (for 5083) to evaluate stress corrosion properties such as K_Iscc as a function of aging time. Selected fracture surface observations are made to describe the failure behavior. The mechanisms for corrosion crack initiation and growth are examined in the light of GB-ppt dissolution versus hydrogen-assisted cracking (HAC). Experimental details of these alloys are described by Gao and Quesnel (2011), Pao, Holtz, Goswami, and Bayles (2013), and Vasudevan and Doherty (1987). However, not all the data with heat treatment time taken from literature match well with electrochemical results. Hence, we have to interpolate to show the general trend on the behavior.

2 General physical/chemical properties

Al-Li binary alloy is a planar slip alloy with an ordered matrix δ′ precipitate, with very low misfit strain with matrix Al, and forms an anodic GB-δ (ALLi) ductile phase at grain boundaries. The δ phase (Uesugi, Takigawa, & Higashi, 2005) is an intermetallic phase with elastic modulus E(δ)=44.9 GPa and ν(δ)=0.35, compared with E(Al)=71 GPa. δ has 16 atoms/cell, with lattice parameter a=0.6477 nm larger than the matrix AL (a=0.4033 nm). This large lattice parameter difference creates interface dislocations around δ making it incoherent. The precipitates are anodic because the open circuit potential (E_OCP) of the δ phase is -1.96 V, which is lower compared with that of Al=-0.73 V. E_OCP is referred as the potential in an aqueous solution, in which metal surface corrodes freely with no applied driving force. The difference between the E_OCP values of an alloy with respect to the Al matrix gives an indication about the anodic or cathodic character of the alloy phases.

In contrast, commercial 5083 is an AL-4.5Mg alloy that undergoes serrated yielding (at low strain rates) giving rise to Luders slip band. This has an anodic GB-β phase (Zielinska-Lipiec, Dubiel, & Czyrska-Filipowicz, 2010) that is brittle because K_Ic<0.5 MPa·m^1/2, E(β)=68 GPa compared with E(Al)=71 GPa. GB-β is called a complex Samson phase with 1168 atoms/cell and a=28.242 nm with respect to (Al)=0.4033 nm. B is an intermetallic Laves phase with E_OCP (β)=-1.33 V, which is also lower than matrix AL. Also, the phase has interface dislocations confined to the GB plane (Goswami & Holtz, 2013; Goswami, Spanos, Pao, & Holtz, 2010). It is clear that the corrosion potential of the anodic intermetallic phases is substantially lower than the matrix and is likely responsible for the dissolution of the precipitates.

The corrosion potential data of the various precipitate phases in aluminum alloys was investigated by preparing individual intermetallic phases and measuring their electrochemical properties. A summary of the electrochemical properties was compiled by Buchheit (1995). Table 1 gives a list of properties for the GB β and δ phases.

Table 1

Physical properties of GB ppts in 5083 and Al-3Li alloys.

3 Materials and methods

The two alloys chosen for the study are 5083 and AL-Li binary alloy with Li content varying from 2.7% to 2.9% Li. These alloys were solution heat treated and aged at 175°C and 177–200°C, respectively, for various aging times to measure YS, K_Ic in lab air, and K_Iscc in 1% NaCl (for 5083) and 3.5% NaCl (for Al-2.9Li) solutions. Fracture toughness in lab air was measured using Charpy samples for AL-Li alloys, whereas compact tension (CT) samples were used for 5083. Precracked double cantilever beam (DCB) samples were used to measure K_Iscc by the load reduction method in NaCl solution. Hence, K_Iscc is related to long incubation time in the load reduction method compared with a rising load method at low K-rate. Selected fracture surfaces were mapped in SEM to observe the mode of fracture in these environments. Some DCB sample fracture surfaces were contaminated with oxides preventing from clear observation of fracture. In such cases, slow strain rate (SSR) tensile fracture samples in NaCl were used. Mechanical property data are taken from several sources of reference (Gao & Quesnel, 2011; Pao et al., 2013; Sanders, 1980; Sanders & Starke, 1989; Tosca, Christodoulou, & Pickens, 1985; Vasudevan & Doherty, 1987, 1989; Vasudevan, Ludwiczak, Baumann, Doherty, & Kersker, 1985). All the SCC tests were conducted without an applied voltage.

4 Estimation of the area fraction (A_f) of GB precipitates

The sensitization heat treatment process in 5083 produces Mg-rich β phase (Al₃Mg₂) on the grain boundaries as well as at any interior heterogeneous nucleation sites such as constituent and dispersoid particles. The free corrosion potential is made up of an anodic reaction and the reverse cathodic reaction leading to equilibrium at a specific potential (termed as open circuit potential, E_OCP), where the number of electrons produced from the anodic reaction equals the number produced from the cathodic reaction, all occurring on the same surface. Mixed potential theory in electrochemistry states that when metals are mixed, both of the metals have their same pairs of processes, but the potential that is observed is due to the anodic branch of the more active metal with the cathodic branch of the more passive metal. The other two branches have their reactions suppressed. Hence, to determine the area fraction of the metals based on the variation in apparent free corrosion potentials, the log of current rather than current density should be plotted. The smooth and well-defined variation in the experimentally determined E_OCP together with the aging of these alloys (and hence area fraction beta phase) suggests that such mixed corrosion potential can be applied here. Gao and Quesnel (2011) have explained the causality of using mixed potential theory to estimate the relative area fractions of the regions with different OCPs, when measured separately. Thus, electrochemistry allows a direct measurement of the nonequilibrium area fraction of the β phase based on a rule of mixtures as the precipitation process proceeds with heat treatment time in the 5083 alloy.

Knowing that the open circuit free corrosion potential (Buchheit, 1995) of the matrix phase is approximately -0.73 V (SCE) and the open circuit free corrosion potential of pure β phase is approximately -1.33 V, measurements of the open circuit free corrosion potentials of sensitized alloys represent the growth of the β phase area fraction exposed by preparation of the surface area between these two extremes. Similar to the logic used in mixed potential theory, the free corrosion potential of a system containing area fraction A_m of matrix phase and A_β of the β phase is just the area-weighted average of the potentials of each region,

Naturally, the extreme values from the AL-Mg alloy phase diagram check as expected with a linear dependence of alloy potential on the area fraction of β. Using the fact that A_β+A_m=1, it is possible to estimate the area fraction β, equal to the volume fraction β, because on a two-dimensional cross section created by the crack surface, we can assume that A_f ~vol%:

These are estimates because the OCP values can change with experimental conditions and solution concentrations. Transmission electron microscopy (TEM) work (Goswami et al., 2010; Goswami & Holtz, 2013) show β are also nucleated at the interface of the matrix constituent and dispersoid particles, and hence the estimation of the A_f of β numbers on the GB could be lower, but the trend in A_f versus aging time would be similar. The previously mentioned method of estimating A_β is described in reference (Gao & Quesnel, 2011).

In AL-2.9Li alloy, the GB-ppt is δ phase (ALLi). Here the δ forms only on the GBs and are not found in the matrix at the second phase of the particle sites. The area fraction of this phase was measured in the SEM using liquid Ga-induced fracture surface after various aging treatments (Vasudevan & Doherty, 1987). This is a direct measurement of area fraction of δ. Here we tried to locate flat areas (with respect to the electron beam) of the GB plane for the measurements required by quantitative metallography. δ Precipitates are sensitive to lab air humidity, and the size can decrease because of its reaction with time of exposure, but the A_f would be same.

Figure 1 shows the plot of the estimated A_f of and δ as a function of sensitization/aging time on a log-log scale. A_f varied linearly with heat treatment time as A_f α tⁿ. The trend is similar for both precipitates, with a slope n~0.15. This could be due to similar precipitate growth kinetics. We have included the upper limit vol% calculated by Stumpf (1980) from phase diagrams of Al-Li and Al-Mg (Phillips, 1961) and know the respective densities of pure AL, δ, and β. Phase diagram calculations imply that all the solute Li and Mg in the respective alloys are tied up to form δ and β precipitates, respectively:

Figure 1:

Estimated area fraction of the grain boundary precipitates in AL-3Li (from SEM fracture surface measurements) and 5083 alloys (from electrochemical data) with aging heat treatment; along with selected TEM micrographs.

For AL-2.9Li alloy, at the aging temperature 178°C, the amount of Li in solid solution is approximately 0.05 wt%, which indicates that approximately 2.85 wt% of Li has precipitated. Similarly, for AL-4.5Mg, at 175°C, ~2.5 wt% is in solid solution, indicating that approximately 2.0 wt% of Mg has precipitated. Using Equations 3a and 3b, we can estimate the maximum vol% of δ=21.2% and that of β=8.03% at the heat treatment temperature.

The estimated A_f data merge with the phase diagram estimated values as an upper limit. Dotted lines are extrapolations based on available data. Figure 1 also shows that after the solution treatment of the AL-2.9Li alloys, there is no incubation time for the δ precipitates to form on the GBs. This is due to the low formation energy for δ, approximately -23.4 kJ/mol (Kishio & Brittain, 1979, 1981; Sluiter, Watanabe, deFontaine, & Kawazoe, 1996). Hence, the quenching from solution temperature is not fast enough to suppress the precipitation kinetics of δ (and δ′). X-ray analysis (Fuller, 1990; Whitman, 1990) indicates that the A_f<0.1 for GB-δ formation in the as-quenched condition. In 5083, the as-quenched alloys do not show any β precipitates forming on the GB. Perhaps it takes longer time to form β precipitates. This is perhaps due to the higher formation energy, approximately +15.7 kJ/mol (Starink & Zahra, 1998), for β. Few TEM micrographs show (in the inserts) the GB-ppts in as quenched condition and after longer aging times in both alloys.

5 General mechanical behavior in inert environment

Figure 2 shows the YS stress variation with aging time for AL-2.9Li and 5083 alloys. AL-2.9Li is an age-hardening alloy (Sanders, 1980; Vasudevan & Doherty, 1987, 1989; Vasudevan et al., 1985) in which ordered δ′ (AL₃Li, L1₂, misfit strain [~-0.08%]) is the strengthening precipitates. The YS increases with aging time at 177°C to a peak strength around aging time of 24 h, and with further aging, the YS decreases. Strengthening occurs because of dislocation cutting δ′ in the UA region giving planar slip and by dislocation bypassing in the OA region yielding wavy slip.

Figure 2:

Yield strength variation with aging time for AL-2.9Li and 5083 alloys.

In contrast, 5083 alloy is a solid solution alloy. Mg in solid solution contributes to the YS (Gao & Quesnel, 2011; Pao et al., 2013), and hence it is independent of heat treatment time at 175°C. This is because the %Mg in solid solution at 175°C heat treatment temperature is constant (~2.5 wt%) with time and yields in “strain aging” phenomena.

In both alloys, GB-ppts form during heat treatment, and their size and volume fraction increase with time. However, 5083 alloy does not show any GB precipitate free zones (PFZs) whereas AL-2.9Li alloy shows wide GB-PFZ with aging time.

The PFZs in Al-3Li binary alloy form by a solute Li depletion mechanism, and its growth was described as a solute diffusion-controlled process in α-Al (Jha, Sanders, & Dayananda, 1987), coupled with the δ formation/growth at the GB. Such a PFZ formation is absent in 5083 alloy because no Mg depletion was observed at the GB (Goswami et al., 2010; Uesugi et al., 2005; Zielinska-Lipiec et al., 2010). In addition, the growth of β is due to the collector plate mechanism with Mg diffusing along the lattice and dislocation. The activation energies associated with GB diffusivity in both alloys is nearly the same: 118 kJ/mol for Al-Li alloy (Fugita, Horita, & Langdon, 2004) and 115 kJ/mol for Al-Mg (Fugita et al., 2004). Thus, one can infer that the GB diffusivity for these alloys is not very different. However, GB diffusivity can vary depending on the misorientation angle.

In 5083 alloy (Gao & Quesnel, 2011; Pao et al., 2013), both YS (Figure 2) and fracture toughness K_Ic (Figure 3A) in the inert (lab air) environment are not affected by the sensitization treatment. As the heat treatment time at 175°C is increased, A_f of GB- increase and saturates after approximately 300 h. GB-β is in the form of a platelet (Goswami & Holtz, 2013; Goswami et al., 2010) that grows with time within the GB-plane. GB-β does not seem to take part in the fracture process of 5083 in an inert environment. This could be due to the interfacial energy of β/Al~0.8 J/m² being similar to the surface energy of Al~0.9 J/m² (Bernstein, Goswami, & Holtz, 2012). One can only infer that strong β/Al interface can suppress its contribution to fracture (Bernstein et al., 2012), giving rise to dimpled transgranular fracture. The dimpled transgranular fracture in lab air (Figure 3A, insert) indicates that the fracture is of ductile mode and is due to second-phase intermetallic constituents and not from GB-β. Unfortunately, fracture surface that does not show β is being fractured for reasons unknown, although pure β is brittle with K_Ic<0.5 MPa·m^1/2 (Zielinska-Lipiec et al., 2010).

Figure 3:

(A) 5083: K_Ic and K_Iscc dependence on A_f of GB-β along with selected fracture micrographs. In comparison with data from LME for the alloy 7075-T651: K_Iscc (Hg) ~1.5 MPa·Vm, K_Iscc (Ga) ~1.7 MPa·Vm. (B) Environmental contribution to stress intensity K_env variation with A_f of β in 5083 alloy.

Fracture toughness K_Ic in AL-2.9Li alloy (Figure 4A) decreases in the UA region, reaching a minimum at PA, and is unchanged in the OA region. K_Ic is affected by the GB-δ because of its weak interfacial energy. The interfacial energy of GB-δ/AL~0.5 J/m² (Fujikawa, Izeki, & Hirano, 1986) was measured and estimated using electrical resistivity method. The electrical resistivity method gives a resistivity combination of both matrix-δ′ and GB-δ phases. The formation of δ will have had a major effect on the resistivity behavior, causing it to fall to much lower values than if only δ′ had been produced during the 200°C aging. Incidentally, the interfacial energy of matrix-δ′ is approximately 0.045 J/m², significantly lower than the GB-δ phase. Analogously, in low-carbon steels containing sulfide particles, void nucleation and growth fracture are due to the low interfacial energy of sulfide particles (Martin, 1980).

Figure 4:

(A) K_Ic and K_Iscc variations with A_f of GB-δ precipitates in AL-2.9Li alloys along with selected fracture micrographs. (B) Environmental contribution to stress intensity K_env variation with A_f of δ in AL-2.9Li alloy.

In Figures 3A and 4A, K_Ic is plotted as a function of A_f (taken from Figure 2) of the GB precipitates for the two alloys. Figure 3A shows that the K_Ic in 5083 is independent of A_f, whereas in AL-2.9Li (Figure 4A), K_Ic α 1/A_f of the GB precipitates. In the case of AL-2.9Li alloy, GB-δ is an incoherent oblate spheroid that grows in size and number (A_f) with aging time. In the early stage of aging where δ is very small in size, the fracture is transgranular. As δ grows in size and A_f with aging time, fracture increasingly becomes mixed mode (TG+IG) from severely UA (planar slip) to PA (wavy slip). This mix mode fracture comes partly from the matrix planer slip decohesion in UA and GB-δ. With longer aging time past the PA condition, the fracture is completely along the GB with void nucleation and growth process at the interface of δ. Fracture surfaces at low and higher aging times are shown in the inserts of Figure 4A. One can replot Figures 3A and 4A to see the trend due to the contribution of environment to K. Then K takes the form K_env=K_Ic-K_Iscc, as in Figures 3B and 4B as a function of A_f. Here K_env increases with A_f for 5083 (K_Ic is constant with A_f), whereas it decreases for Al-2.9Li alloy (K_Ic decreases with A_f). There is a lower and an upper plateau in Figure 3B. This can be due to the delay in forming enough A_f (and size) of β that can dissolve to contribute to a measurable K_Iscc. The upper plateau is where β is nearly continuous, giving rise to dissolution as A_f does not change. Such a trend in Figure 3B is in agreement with incubation time for crack initiation decreasing with A_f of β (Gao & Quesnel, 2011). In contrast, K_env decreases with A_f of δ for the Al-2.9Li case (Figure 4B) because the δ formation begins from the stage of as quenched condition. K_env eventually tends to plateau at higher A_f>0.17 of δ. The background decrease in K_Ic affects the overall fracture in NaCl. This mechanical fracture in the background seems to reduce to environmental effect to a minimum past PA, suggesting that mechanical fracture due to δ is the primary cause and dissolution of δ is the secondary cause. This can be observed on the fracture surface in PA with dimples after exposure to NaCl (insert in Figure 4A) in an SSR sample, where GB-δ has dissolved. At very low A_f, the dissolution rate can be severe, and the fracture surface was covered with oxide.

Figure 5 illustrates schematically the GB fracture process of AL-2.9Li alloy for the severely UA and OA cases in lab air. In the UA condition, fracture is mixed with some slip and IG fracture, whereas in the OA case, it is completely IG fracture with GB microvoid and coalescence. The schematic in Figure 5 gives the UA and PA fracture with the corresponding fracture surfaces showing mix mode in the UA and IG voids in PA and OA conditions. One can show a similar trend observed in Figures 3A and 4A if we had plotted K_Ic versus heat treatment time.

Figure 5:

Fracture process in AL-2.9 Li alloys with GB-δ ppts under inert lab air environment. In the UA, fracture surface is mixed TG+IG and in PA (and OA) it is IG.

6 Stress corrosion

When these two alloys are tested under NaCl solution environment, the threshold K_Iscc behavior with A_f of GB precipitates vary significantly from the trend in K_Ic. This is also shown in Figures 3A and 4A. For comparison, the lab air data are shown.

6.1 5083 alloy

Figure 3A shows the K_Iscc decreasing with A_f of GB-β in 5083 alloy (Pao et al., 2013). There is an upper shelf plateau (here A_f of β is very low), and a lower shelf corresponds to when A_f has saturated. In between these limits, K_Iscc α (1/A_f). Even in 1% NaCl solution at A_f<0.03, the fracture is transgranular similar to that in lab air. At A_f>0.06, fracture is completely intergranular. The SEM fracture surfaces are shown in the inserts. In between A_f ~0.03–0.06, fracture starts off in mixed mode transitioning to intergranular mode with increasing A_f. This trend can be related to the β dissolution in NaCl. At low A_f<0.03, where the spacing between β is very large, the alloy is not very sensitive to NaCl environment and K_Iscc is same as K_Ic. With increasing A_f with heat treatment time, β spacing decreases. Here the fracture mode changes from the mixed mode of IG+TG (at low A_f) to the IG mode (at higher A_f) due to the β dissolution at GB. Spacing in between β has been interpreted as due to HAC (Crane & Gangloff, 2012; Jones, 2003). This may not be occurring because K_Ic~K_Iscc at very low A_f of β. This can also be just a transgranular mechanical fracture (Osaki, 1974) due to the applied stress. See insert on Figure 3A at K_Iscc~20 MPa·m^1/2, showing IG with edges tearing. When A_f>0.06, the fracture is completely IG and is due to just active path β dissolution, and K_Iscc reaches a minimum value around 4–5 MPa·m^1/2. Inserts show the mode of fracture from TG to completely IG. Interestingly, as a comparison, liquid metal embrittlement (LME) results in 7075-T651 alloy with Hg, and Ga gives similar K_Iscc, around 1.5–2.0 and 1.7 MPa·m^1/2, respectively (Benson & Hoagland, 1989; Chu, Liu, Luo, & Qiao, 1999; Speidel, 1971; Wheeler, Hoagland, & Hirth, 1989). LME thresholds are also comparable with 5083-NaCl K_Iscc~3 MPa·m^1/2 data for heat treatment times longer than 300 h where A_f>0.06. K_Iscc values at longer aging times are still higher than Griffith’s fracture values K (Griffith) ~0.22 MPa·m^1/2. Factors that can affect this difference could be the oxide delaying the reaction at the crack tip and the plasticity suppressing brittle fracture. Because β can dissolve without the need of an applied stress, then applied stress can just help in keeping the crack tip open to allow for the access of NaCl electrolyte to allow the galvanic dissolution of β until its solubility limit is reached in NaCl solution, a process similar to LME. Figures 6 and 7 give the schematic description of the mechanism for the two cases, where β is discontinuous (Figure 6 with some finite spacing between ) and continuous (Figure 7 with negligible spacing between β).

Figure 6:

Fracture process in 5083 alloy with discontinuous GB-β In NaCl environment.

Figure 7:

Fracture process in 5083 alloy with continuous GB-β precipitates in NaCl solution.

7 Electrochemistry

If the chemistry within the crack tip region reaches a critical condition because of the reactions between the β phase and the NaCl solution, then the crack can form intergranularly. This depends on the size and A_f of the β phase. A hypothetical sketch of the development of critical local chemistry during crack initiation period when β reacts with NaCl is illustrated in Figure 8, which results in a decreasing pH (~3) at the crack tip.

Figure 8:

Chemical reactions in NaCl solution of 5083 alloy.

Al³⁺ ions from dissolving β phase can participate in the cathodic reaction:

AL 3 + + 3e- → Al, and

the Mg²⁺ ion can dissociate the H₂O to release H⁺:

Mg 2 + + H 2 O- → Mg(OH) + + H +

The β phase on the grain boundaries is polarized to higher values than the breakdown potential, which would then lead to the rapid dissolution of the precipitates. The dissolution of the precipitates leads to the generation of the Al³⁺ and Mg²⁺ ions, which hydrolyze subsequently leading to a low pH environment at the GB, setting up conditions for crack nucleation/propagation along the GB. Thus, after the dissolution of β, H⁺ is released from the hydrolysis reaction of Mg³⁺ with water.

E _OCP corrosion potential of 5083 is influenced by the degree of sensitization time. As this time increases, E_OCP decreases monotonically from -722 to -751 mV. Because the E_OCP potential of pure β phase in 3.5% NaCl is lower than Al, -1330 versus -730 mV (Buchheit, 1995; Jain, Lim, Hudson, & Scully, 2012; Searle, Gouma, & Buchheit, 2001) (Figure 10), this decrease in E_OCP for longer sensitization time can be attributed to an increase in the amount of the β phase exposed on GBs. This implies that the electrochemical anodicity of sensitized 5083 alloy increases with the amount of the β phase and thus with sensitization time. Breakdown potential is insensitive to sensitization time (Jain et al., 2012), suggesting that the passive film stability is not affected.

At longer heat treatment times, the β phase becomes continuous along the GB and is expected to be more susceptible than that in the as-received condition. With the continuous precipitation of the β phase along the GB, β dissolves rapidly and can lead to rapid crack nucleation/propagation. Thus, at heat treatment times greater than approximately 300 h, the selective anodic dissolution of β is the major process of failure. At lower heat treatment times, the β phase is discontinuous, and the dissolution of the β phase will occur discontinuously along the GB. The local chemistry established by this discontinuous β could cause lower dissolution rate in between the β-free region along the boundary because it is not likely to be polarized anodically to the rest of the matrix. Thus, the β dissolution mechanism largely accounts for the SCC observation in Figure 3A and B.

Figure 9 schematically describes the reaction of GB-δ in AL-2.9Li alloy with NaCl solution. Electrochemical tests in NaCl solution on AL-2.9Li alloys show that OCP is nearly constant in the UA region (OCP changed from -990 to -1023 mV) up to 80 h of aging and then dramatically decreases (from -1023 to -1374 mV) in the OA region. The alloy may be showing general corrosion in the UA region along with δ dissolving. This is observed as surface reaction is severe in the UA region (see insert in Figure 3A). Also, in the UA region, the alloy undergoes planar slip that can react with NaCl solution to give slip line dissolution (Chen & Duquette, 1992). In the OA region, the alloy is becoming more active as the aging time increases (or A_f of GB-δ increases) and OCP is reduced severely. At longer aging times, beyond PA, the OA region enhanced the anodic dissolution of δ as noted by increased OCP (Chen & Duquette, 1992; Niskanen, Sanders, Rinker, & Marek, 1982). Thus, the corrosion resistance of the alloy was reduced with the increasing aging time because of the presence of increased A_f of active δ (Figure 10). The pitting potential of the alloy was not sensitive to aging (Chen & Duquette, 1992; Niskanen et al., 1982), suggesting that passive film stability is not affected. Figure 9 describes the basic galvanic reactions of δ-ppt with NaCl solution. Both β and δ phases show similar electrochemical behavior in NaCl solution.

Figure 9:

Chemical reactions in NaCl solution of AL-2.9Li alloy.

Figure 10:

OCP variation with heat treatment time for 5083 and AL-2.5Li alloys.

Figure 10 indicates that in the early stages of heat treatment, the 5083 alloy seems to be more anodic in general relative to AL-2.9Li. Figure 11 shows that K_Iscc decreases with E_OCP for both 5083 and AL-2.9Li alloys, with a large decrease in 5083 mainly due to β dissolution and a modest decrease in AL-2.9Li alloy due to the dissolution of δ and possibly slip dissolution and the background K_Ic. One can note that the changes in OCP for 5083 I are much smaller (~-30 mV) compared with AL-Li (~-320 mV). Other factors such as size and shape of δ versus β could also play a role in the rate of dissolution, interfacial energy, and H⁺ generation. The E_OCP values for pure β and δ are also shown on the plot. The overall mechanism could be GB-ppt dissolution coupled with HAC.

Figure 11:

K _Iscc variation with OCP for 5083 and AL-2.9Li alloys.

These anodic reactive phases (β and δ) with respect to the matrix (cathode) allow the net rate of chemical reaction to occur (Figures 8 and 9), and this is observed to be related to the A_f of these phases. The rate of chemical reaction determines the rate of pH change at the crack tip. As pH becomes more acidic, it provides the crack tip pH and potential to stimulate H⁺ production that can drive the cathode reaction to consume the electrons and create more H⁺, which diffuse into the crack tip deformation region of the alloy. This consumption of the electrons drives the dissolution of the anodic phases. These anodic phases have both AL³⁺ ion and Mg²⁺(from β) and Li⁺ (from δ) ions, which hydrolyze to dissociate H₂O to create more H⁺. Thus, HAC can occur in the regions of the GB with no phases to give rise to brittle fracture. This means the overall SCC behavior is probably a combined effect of anodic dissolution and HAC, although it is difficult to separate them. Fracture surfaces are not very discernable as it is covered by oxides.

8 Summary

The following list gives a few salient points from the analysis of the SCC behavior of the two alloys. Interestingly, most of the available data in the literature were concerned with crack growth than initiation. The lack of connection between electrochemistry and K_Iscc and good matching fracture surface analysis makes it difficult to analyze the results.

• Binary planar slip AL-Li alloy

  • In an inert (lab air) environment, matrix δ′ affects YS over the entire aging curve. Dislocations cut the coherent δ′ below PA (planar slip region), and the YS increases from as quenched (AQ) to PA. Past the PA, in the OA region, YS decreases due to dislocations bypassing the incoherent δ′.

    The A_f of GB-δ ppts strongly affects the static fracture toughness in lab air and K_Ic α (1/A_f) below PA (UA region), whereas in the OA region, K_Ic is independent of A_f. This is due to the weak δ/matrix interface resulting in GB void nucleation/growth.

    The static fracture in lab air shows dimple fracture surface in the AQ condition, slip line fracture with IG voids at intermediate aging times, and complete IG with voids in the OA region.

  • Stress corrosion fracture initiation K_Iscc is mostly due to the dissolution of δ (assisted by HAC) with a weak dependence on A_f of δ below PA and K_Iscc α (1/A_f). In the AQ condition, fracture surface is masked by Li-oxide showing strong reaction with NaCl. At very low A_f, the fracture mostly consists of some slip fracture and IG, and at high A_f, the SCC fracture is completely IG and the fracture surface seems to maintain the dimples without δ, suggesting that deformation preceded before δ dissolution. NaCl has a large effect on K_Iscc when the A_f is very low, and this environmental effect decreases as A_f is increasing with aging time. These alloys being chemically very active are prone to pitting. Overall, the K_Iscc mechanism is affected by the background K_Ic of the alloy at A_f<0.17. The environmental contribution to SCC is mostly due to the dissolution of δ at high A_f>0.17.

• Commercial 5083 solid solution (serrated yielding) alloy

  • GB-β ppts do not affect YS as well as K_Ic in lab air. Mg in solid solution affects the YS and is constant with sensitizing time because the amount of Mg at the 175°C temperature is constant.

    In NaCl environment, there are three regions of SCC behavior. Environmental effect on K_Iscc increases, with increasing A_f, to reach a minimum plateau value of approximately 4–5 MPa·m^1/2. This value is close to the LME threshold observed in 7075-T651 alloy with Hg or Ga, and the dissolution of β mechanism is similar to that of LME.

    At very low A_f of β, near the unsensitized condition, SCC is negligible because K_Ic=K_Iscc. This also suggests HAC is negligible and the failure is mechanical with minimal effect from NaCl.

  • At low sensitization times from approximately 5 h–300 h, where β is discontinuously precipitated, failure is mixed with mechanical tearing (in the β free regions) and dissolution of β, as observed on the fracture surface. In this region, K_Iscc α (1/A_f).

  • At longer heat treatment times greater than 300 h, failure is mostly due to the anodic dissolution of GB- (similar to LME), and K_Iscc is independent of A_f. The A_f of GB ppts δ and β shows an inverse relationship with K_Iscc=C/A_f.

  • C=300 (lab air) and C=50 (NaCl) for AL-2.9Li.

  • C=0 (lab air) and C=60 (NaCl) for 5083 alloy (the unit of C is MPa·m^1/2 per unit A_f of GB precipitates).

  • The large differences in slopes of K_Ic, K_Iscc versus A_f, can in part be due to the differences in the background fracture behavior in lab air and dissolution (in NaCl) kinetics affected by the interface strength of δ and β and their electrochemical behavior.

  • Environmental contribution to SCC, K_env=K_Ic-K_Iscc, increases with A_f of β for 5083 and decreases with A_f of δ for AL-2.9Li alloy.

  • E_OCP decreased with heat treatment time for both the alloys with negligible changes in pitting corrosion. K_Iscc also varied inversely with E_OCP for both alloys as K_Iscc α (1/E_OCP) because of the differences in the electrochemical properties of δ and β, their size and A_f, and their interfacial energies. How these parameters affect the electrochemistry and crack initiation is not clear.