Investigating the Impact of Accelerated Testing Variables on the Exfoliation Corrosion of AA2060

Heat treatable high-strength aluminum alloys (Al alloys) have long been considered essential components of aircraft design.¹  The need to create more efficient and more cost-effective air transport has driven the development of new high-performance Al alloys with lower density and higher strength, stiffness, and fracture toughness. Corrosion resistance is also a critical property because aerospace structures are exposed to corrosive atmospheric conditions during service.

Al alloys have good resistance to general corrosion as a result of a protective oxide film, but most Al alloys can experience localized corrosion damage such as pitting, intergranular corrosion (IGC), or exfoliation. Accelerated corrosion tests are used to evaluate the susceptibility of alloys to localized corrosion. It is crucial that accelerated tests accurately reproduce the corrosion behavior observed in service because they are used for material lot acceptance and new alloy and temper development.²  Many standardized accelerated tests are currently used for evaluating localized corrosion susceptibility of high-strength Al alloys (ASTM Standards G34, G85, G110, and B117),^3-6  but these tests can produce drastically different results for the same alloy.⁷ 

The focus of the current work is to develop a framework for understanding the discrepancies between accelerated testing methods and for designing improved tests. Accelerated testing for exfoliation corrosion susceptibility of AA2060 (aluminum-lithium alloy)⁸  will be used to illustrate this framework. This alloy has been studied by Moran, et al., who demonstrated that the -T36 temper is susceptible to exfoliation during seacoast exposures, while the -T86 (near peak-aged) temper is resistant.⁷  Considering both a highly susceptible and a highly resistant temper will provide a basis for understanding the mechanisms at work in accelerated tests for exfoliation corrosion of this alloy.

Exfoliation is thought to occur in alloys with an elongated grain structure when voluminous corrosion products form along grain boundaries during IGC. The corrosion products provide a wedging stress that lifts the intact grains above to form exfoliation blisters.⁹ 

The accelerated tests for exfoliation of interest to this work include ASTM G34 (a constant immersion test called EXCO) and ANCIT (aluminum-nitrate-chloride-immersion test, a modified version EXCO).^3,10  Figure 1 shows the behavior of several AA2060 tempers after 1.2 y of seacoast exposure and after EXCO testing.⁷  EXCO results did not correlate well with the seacoast results and even produced an opposite trend in terms of exfoliation resistance of AA2060 tempers. According to EXCO, the under-aged tempers appeared to have superior exfoliation resistance compared to the -T86 temper, but in the seacoast exposure, the -T86 temper was more resistant, as shown in Figure 2.

EXCO fails to reproduce seacoast results for other aluminum-alloys as well. This was the motivation for the development of the ANCIT test. Lee, et al., observed that EXCO failed to produce expected results for 7X50, 2024 (UNS A92024⁽¹⁾), or 2090 (UNS A92090) Al alloys.¹⁰  They thought this may be a result of the extremely low pH of the EXCO solution, and noted that after 24 h of immersion, the solution pH increased to about 3. The group also noticed that after 24 h, there were traces of aluminum ion (Al⁺) resulting from dissolution of the metal. The testing environment for ANCIT was guided by these observations, but there is still not a fundamental understanding of what makes EXCO successful for some Al alloys and ANCIT successful for others.

In order to understand discrepancies in these accelerated exfoliation tests, ANCIT and EXCO were deconstructed in terms of their key testing variables. Four key variables were identified: the addition of oxidizing agents, the testing temperature, and the chloride content and pH of the testing solution. Both EXCO and ANCIT incorporate these variables at some level, but the tests have different values for some variables (e.g., pH of 0.4 vs. pH of 3.2).

It is important to consider the various localized corrosion mechanisms for Al-Cu-Li alloys because exfoliation initiates from other forms of corrosion such as IGC. Intergranular attack in these alloys is generally attributed to the effects of hardening phases that form on grain boundaries. For alloys like AA2060, the primary hardening precipitate is the hexagonal T₁ phase (Al₂CuLi).^11-13  Both Connolly and Buchheit have reported that the T₁ phase tends to form on subgrain boundaries and matrix dislocations.^11,14  The precipitation of T₁ results in the creation of a Cu depleted zone; however, Buchheit, et al., showed that the T₁ phase is anodic with respect to both the matrix and the Cu depleted zone. They suggested that the corrosion mechanism was selective dissolution of the T₁ phase rather than of the Cu depleted zone.¹⁴ 

Many Al-Cu-Li alloys have a transition in IGC susceptibility during aging where under-aged tempers are more vulnerable to IGC than peak-aged and near peak-aged tempers. However, IGC susceptibility increases again upon over-aging. The improvement in IGC resistance at near peak-aged tempers correlates with an improvement in stress corrosion cracking and exfoliation resistance.^7,13  Recent work by Ott, et al., revealed some of the changes that occur in Al-Cu-Li alloy grain boundary chemistries upon aging. They found that in naturally aged AA2050, grain boundaries were depleted in Cu and highly enriched in Li compared to the matrix, which would lead to preferential dissolution of grain boundaries. The grain boundaries of the over-aged sample were enriched in Cu, which would create a Cu depleted area around the grain boundaries that is susceptible to attack.¹³ 

The AA2060 used in this work was provided by Alcoa Inc. in the form of a 3.5 cm thick plate. This material was received in the -T36 (under-aged) and -T86 (near peak-aged) tempers, and the composition is shown in Table 1.⁸  All samples were polished to a 1200 grit finish, cleaned ultrasonically in deionized water, and rinsed in ethanol before testing. Keller’s etch was used to reveal grain boundaries before optical microscopy and scanning electron microscopy (SEM).

EXCO, a constant immersion test, was performed according to ASTM G34³  for the standard 4-d period, as well as modified testing times. AA2060-T36 samples were exposed for 4 d, 7 d, and 28 d, while AA2060-T86 was tested for 6 h, 4 d, and 7 d.

The EXCO testing environment included a solution of 4 M sodium chloride (NaCl), 0.5 M potassium nitrate (KNO₃), and 0.1 M nitric acid (HNO₃). This provided 0.6 M nitrate ion (⁠⁠), which is an oxidizing agent. The pH of this solution was 0.4, and the testing temperature was 25°C. Figure 3 shows the sample dimensions used for this test. Post-testing analysis included visual examination of exposed surfaces, as well as cross sectioning for optical microscopy. SEM was also used for select samples.

ANCIT is a modified EXCO test that does not currently have an ASTM specification. This accelerated exfoliation test was developed to replace EXCO for some aluminum alloys.¹⁰  The ANCIT test solution was 4 M NaCl, 0.6 M KNO₃, and 0.0224 M AlCl₃, resulting in a solution pH of 3.2. This test operated at a temperature of 52°C. Both AA2060-T36 and AA2060-T86 were tested in ANCIT for the standard 2-d period and also for an extended time of 7 d. After testing, samples were examined visually and cross sectioned for optical microscopy.

A half-fractional factorial design of experiments (DOE) was used to investigate the impact of key testing variables on corrosion potential (E_corr) and polarization resistance (R_p). Five factors were considered in this DOE, requiring a total of 16 experiments. Replicates were not used in this study as the intention was to screen several factors in an efficient manner. A more in-depth study of the parameters found to be significant will be a topic of future work.

A summary of the five factors in this DOE are shown with their low and high values in Table 2. These were chosen after considering the testing environments of EXCO and ANCIT and include chloride concentration [Cl⁻], solution pH, an added oxidizing agent [], testing temperature, and alloy temper. Minitab^†, a commercially available statistical software package by Minitab Inc., was used to generate a testing matrix from the high and low values of the five factors. A half-fractional factorial design was used to reduce the number of experiments necessary while maintaining good resolution.

Electrochemical measurements were performed in a standard three-electrode cell using a saturated calomel reference electrode (SCE) and a platinum counter electrode. Corrosion potential (E_corr) and polarization resistance (R_p) were measured using a linear polarization scan starting at 0.05 V vs. open-circuit potential (OCP) and scanning down to −0.4 V_OCP at a rate of 0.5 mV/s. The scan was preceded by a 24-h OCP delay to allow the system to approach steady state. The solution for each of the 16 tests was outlined by the DOE, which specified a value for [Cl⁻], [], solution pH, testing temperature, and alloy temper.

After E_corr and R_p were measured, these data were entered into the commercial software for analysis. The average effect of single factors and two-factor interactions on E_corr and R_p were calculated. The threshold for statistical significance was determined using Lenth’s pseudo-standard error (PSE), which is a standard method for studies without replicates. To calculate the threshold, Lenth’s PSE was multiplied by the t-value at α = 0.05 and degrees of freedom, df = 5. This method is described in detail elsewhere.¹⁵ 

EXCO results after 4 d and 7 d of testing (Figure 4) did not correlate well with observations at seacoast (Figure 2). The -T36 temper, which has been shown to exfoliate during seacoast exposures,⁷  formed no exfoliation blisters visible to the naked eye. In contrast, the -T86 temper, which is resistant to exfoliation at seacoast, had a flaky appearance. Cross sections after 7 d of testing revealed grain lifting on the T/2 plane of both tempers (Figures 5[a] and [b]), which is an indication of exfoliation. In addition, cross sections, shown in Figures 5(c) and (d), show that attack from the ST plane was IGC for the -T36 temper and selective grain attack (SGA) for the -T86 temper. SGA was identified by locating partially dissolved grains where dissolution did not cross over grain boundaries. An example of SGA is circled in Figure 5(d).

Although there were no exfoliation blisters visible to the naked eye on the AA2060-T36 sample after 7 d of testing, the cross section shown in Figure 5(a) indicated that exfoliation had initiated. Testing time was extended to 4 weeks for this sample in order to determine if more significant exfoliation attack would form with longer exposure time. Figure 6(a) shows that large exfoliation blisters formed on the T/2 surface of AA2060-T36 after 4 weeks of EXCO testing. The cross section in Figure 6(b) confirms the presence of significant grain lifting.

Additional testing was performed on the -T86 temper to better understand the early stages of attack in this material. Testing time was only 6 h, and Figure 7(a) shows the attack on the T/2 surface after this period. Blisters on this sample were small and difficult to see without magnification, but SEM imaging confirmed the presence of blisters on the T/2 surface (Figure 7[b]). The optical microscopy showed grain lifting associated with a blister (Figure 7[c]).

Neither AA2060-T36 nor -T86 showed signs of exfoliation blisters after 2 d of testing, but they both were covered in powdery corrosion product (Figures 8[a] and [c]). However, the tempers were easily distinguishable after 7 d of testing. The -T36 temper formed severe exfoliation attack on the T/2 surface, and cross sections confirmed the presence of significant grain lifting (Figures 8[b] and 9[a]). The -T86 temper also experienced severe corrosion attack, but no exfoliation blisters were visible. The cross section for the -T86 temper showed deep pitting (Figure 9[c]), and selective grain attack was revealed at higher magnification (Figure 9[d]).

A half-fractional factorial design of experiments was used to find the average effect of the four key testing variables on electrochemical parameters. Figures 10(a) and (b) show the Pareto chart of effects for the corrosion potential (E_corr) and polarization resistance (R_p), respectively. It was found that factor C, the addition of an oxidizing agent (0.6 M ⁠) to the testing solution, was the only variable to have a statistically significant impact on E_corr. Factor A (solution pH), factor C (the addition of an oxidizing agent), and the AC interaction all had a statistically significant effect on R_p. Table 3 shows a summary of the average effects for all single factors and two-factor interactions. A positive effect indicated that an increase in that factor led to an increase in the corresponding measured outcome (E_corr or R_p). A negative effect meant that an increase in that factor resulted in a decrease in the measured outcome.

Exfoliation behavior for AA2060-T36 and -T86 in EXCO did not agree with the seacoast results for this alloy reported by Moran, et al.⁷  The -T36 temper, which has been shown to form exfoliation within 1.2 y of exposure at seacoast, did not form blisters that were visible to the naked eye after 7 d of EXCO testing. In contrast, the -T86 temper experienced more severe attack than the -T36 temper and had a flaky appearance.

A better understanding of the attack morphology was gained using optical microscopy. Both samples exhibited the grain lifting associated with exfoliation,¹⁰  indicating that exfoliation had initiated in both tempers. This was confirmed for the -T36 temper as large exfoliation blisters formed on this sample with extended testing time (4 weeks). However, exfoliation blisters were not expected in the -T86 temper as there was no evidence of intergranular attack in this material after EXCO testing, and exfoliation is usually thought to begin with IGC.¹⁰  Figure 5(d) showed that corrosion did not proceed along grain boundaries in the -T86 temper, and the observed attack was better explained by selective dissolution of susceptible grains or possibly selective dissolution of subgrains. Because this material was rolled as a part of its thermomechanical processing, subgrains could be quite narrow in the S direction. It is possible that attack along narrow subgrains could produce wedging forces similar to that generated by IGC. EXCO testing of the -T86 temper for 6 h demonstrated that small blisters formed quickly in this material, providing additional evidence that the attack morphology could be considered exfoliation.

Like EXCO, ANCIT failed to predict the exfoliation behavior of AA2060 within the standard testing time (2 d). However, after 7 d of testing, severe exfoliation was observed on the -T36 temper. The attack on the -T86 temper was very severe after ANCIT testing, but the morphology was pitting rather than exfoliation. It is not yet well understood why exfoliation on AA2060-T36 forms must faster in ANCIT than in EXCO, but there are only a few differences between the two tests that could be responsible. The chloride and nitrate concentrations are the same in both tests, but the solution pH is higher in ANCIT (3.2 vs. 0.4), the temperature is higher (52°C vs. 25°C), and there is a small addition of aluminum chloride to the ANCIT testing solution (0.0224 M). Investigating the impact of temperature, pH, and AlCl₃ on the attack rate and morphology of AA2060 will be an area of future work.

The design of experiments showed that the addition of an oxidizing agent to the testing solution (0.6 M ⁠) had a statistically significant impact on both corrosion potential and polarization resistance. In general, the presence of resulted in an increase in E_corr and a decrease in R_p. The decrease R_p corresponds to an increase in corrosion current, i_corr, as these two parameters are inversely proportional. Using mixed potential theory, the simultaneous increases in E_corr and i_corr indicate an increase in cathodic kinetics. The cathodic kinetics of this system are usually limited by the diffusion of oxygen to the corroding surface, but the addition of 0.6 M provided a cathodic reaction faster than oxygen reduction, leading to an increase in both E_corr and i_corr.

It was also shown that the solution pH had a statistically significant effect on polarization resistance. Making the pH more acidic resulted in lower R_p, which corresponds to a higher i_corr. This may have been a result of thinning of the oxide layer on aluminum at low pH, which would increase the diffusion limiting current of oxygen reduction on aluminum and increase i_corr.

The interaction of solution pH and the added oxidizer also had a statistically significant impact on polarization resistance, where a simultaneous increase in solution pH and addition of the oxidizer resulted in a decrease in R_p. When pH was considered alone, an increase in pH (more alkaline solution) actually led to an increase in R_p. However, the interaction of factors sometimes leads to a different result than those factors individually.

These results demonstrate the potential for tuning an accelerated test to provide particular electrochemical kinetics by increasing or decreasing the levels of added oxidizer and solution pH. Other authors have reported that some aluminum alloys have IGC susceptibility only within a particular range of potentials.^16-18  Exfoliation of AA2060-T36 initiates with IGC, and it is possible that exfoliation will form within a specific potential range as well. In that case, the ability to adjust electrochemical kinetics by changing the levels of oxidizing agent and solution pH would be helpful in accelerated test design. Areas of future work will be identifying the potential range for exfoliation susceptibility of AA2060 and designing modified accelerated tests with better correlation to seacoast exposure results.

• Neither EXCO nor ANCIT predicted the exfoliation behavior of AA2060 tempers at seacoast during the standard testing time.

• Exfoliation formed in AA2060-T36 after 4 weeks of EXCO exposure.

• AA2060-T86 experienced some grain lifting after 4 d of EXCO exposure, most likely as a result of corrosion product buildup along susceptible subgrains. Small blisters were observed on this sample even after just 6 h of EXCO testing.

• Exfoliation of AA2060-T36 occurred more quickly during ANCIT testing (7 d) than in EXCO testing (4 weeks).

• Out of the five factors used in the design of experiments, adding nitrate, an oxidizing agent, to the testing solution had the most significant impact on corrosion potential. The polarization resistance was most affected by the solution pH, the addition of an oxidizing agent, and the interaction of these two factors.

This work is supported by Rolls-Royce and the Office of the Undersecretary of Defense Corrosion University Pilot Program under the direction of Mr. D. Dunmire. The authors would also like to recognize Srishti Shrivastava for her assistance with microscopy.