Operando Assessment of Galvanic Corrosion Between Al-Zn-Mg-Cu Alloy and a Stainless Steel Fastener Using X-ray Tomography

Al alloy 7050 (Al-Zn [6.7 wt%]-Cu[2.6 wt%]-Mg[2.6 wt%]; UNS A97050⁽¹⁾) was developed to obtain a combination of strength and fracture toughness resistance greater than that provided by other common high-strength alloys such as AA7075 (UNS A97075).¹  AA7050-T7451 is precipitate strengthened by the formation of MgZn₂-η phase in the reaction sequence.^2-6  

Precipitates form by nucleation and growth from a supersaturated solid solution during natural or low-temperature aging. The η phase is highly soluble in the matrix and will readily dissolve during homogenization.⁷  In the partially recrystallized and artificially aged microstructure, η and η′ are usually found on high angle and subgrain boundaries, respectively.^8-10 

Tempering has a significant effect on the microstructure of Al alloys.^9,11-12  In 7xxx series alloys, the T6 temper reaches peak strength but possesses poor stress corrosion cracking (SCC) resistance because of a high density of GP zones and the formation of η′ precipitates along grain boundaries. Overaging treatments, such as the T73 and T76 tempers improve the short transverse SCC resistance by increasing grain boundary precipitate size and spacing, and by modifying their composition. However, these tempers result in a loss of strength.¹³  Aging can also effect the Cu content in these Al alloys. Overaging can result in higher Cu content on grain boundary precipitates such as η, producing a Cu-depleted zone which can facilitate intergranular corrosion (IGC) and SCC.¹⁴  The T7 (overaged) condition results in an increase in the amount, coarseness, and distribution of η precipitates on the grain boundaries.¹¹  The T7451 temper is achieved by solution heat-treating, stress-relieving through controlled stretching, and artificially overaging the alloy between T73 and T76.⁹  The T7451 temper provides an advantageous combination of strength, SCC, general corrosion resistance, and fracture toughness.^12,15 

AA7050-T7451 contains coarse intermetallic particles (IMPs), strengthening precipitates, and dispersoids. Coarse IMPs in the range of 5 μm to 30 μm in diameter (S-Al₂CuMg, β-Al₇Cu₂Fe, and Mg₂Si) form during solidification and are often found in clusters or stringers aligned parallel to the rolling axis. Dispersoids (Al₃Zr) form during homogenization, control recrystallization, and occur at temperatures above the Al-solidus.¹⁶  Dispersoids pin grain boundaries, preventing them from growing during solution heat treatment (SHT) and function as a noble particle that will not sustain large cathodic currents and may be too small to adversely impact corrosion kinetics.¹⁷  In contrast, coarse IMPs can greatly influence the corrosion properties of these alloys.¹⁸  Alloying with Fe and Si can have strong detrimental influences on fracture toughness and corrosion properties as a result of the formation of brittle intermetallics such as S-Al₂CuMg and β-Al₇Cu₂Fe.^19-21  These IMPs are often located in recrystallized grains or at their grain boundaries and can lead to dealloying of the particle leading to localized Cu enrichment.^5,22  MgZn₂ and Mg₂Si are both active particles with high self-dissolution rates which leave behind surface cavities.^23-24  MgZn₂ greatly affects the corrosion properties of AA7050-T7451. As the IMP is very active: the open-circuit potential (OCP) is −1.4 V_SCE and the breakdown potential is −1.14 V_SCE in neutral 0.5 M NaCl.⁵ 

The effect of Cu content on the microstructure corrosion has been well documented.^{5-6,14,22,25-26}  Cu is distributed throughout the microstructure in different ways: (1) in the Al matrix, (2) hardening precipitates (η phase), and (3) in coarse IMPs such as S-Al₂CuMg. A Cu-containing 7xxx series alloy will exhibit coarser, more widely spaced η at grain boundaries as compared to an Al-Zn-Mg alloy.²⁷  The growth rate of the η phase was shown to be higher in Cu-containing alloys.²⁸  Cu increases the GP zone solvus from 120°C to 160°C, resulting in homogeneous precipitation at aging temperatures above 149°C for a Cu content of 1.4 wt%.²⁹  It has been observed that MgZn₂ can become enriched with highly soluble Cu and Al, locally depleting the matrix of Zn, Mg, and Cu. This especially occurs in the T6 temper,³⁰  which was found to have increased the corrosion potential by about 250 mV with 17 at% Cu enrichment in η phase.^6,14  IGC in acid can be attributed to MgZn₂ or Cu depletion.¹⁷  Moreover, in the T74 temper IGC is suppressed because of the overaging treatment as stated above. The presence of S-Al₂CuMg is common in AA7050-T7451 because of its increased Mg and Cu content. S is initially anodic to the AA7050-T7451 matrix but after dealloying occurs, the Cu remnants become high surface area cathodes for oxygen reduction.³¹ 

Cu replating on the surface of the Al alloy can greatly impact corrosion properties of the alloy substrate.^6,32-34  Cu may become oxidized and made available for plating by a number of pathways. For instance, if Cu particles become mechanically detached, they can dissolve at their own corrosion potential in solution and electrochemically replate on the alloy surface.^22,24,35-36  The alloy Cu content is important as the composition of the bulk matrix, depleted zones, and η precipitates change with aging. This decreases the electrochemical potential window between Mg(Cu)Zn₂ dissolution and the matrix pitting potential. This decrease in the potential window decreases IGC and exfoliation susceptibility.⁶  Overall, Cu plays a complex role, lowering intergranular stress corrosion cracking (IGSCC) susceptibility while raising the tendency for general corrosion. Moreover, Cu’s effect on fissure growth is unclear. Dissolution of Cu-rich phases may leave a Cu-rich surface that is cathodic to the matrix,²⁴  increasing general and local corrosion by providing a site which supports fissure growth.¹⁸  The 7xxx series Al alloys are usually protected with a multifaceted coating system.^37-38 

Defects in the corrosion protection system are common because of the harsh environment operating conditions of aircraft, and can become enhanced at complex joining/fastener locations which trap electrolyte into tight crevices. The geometry leads to occluded local environment formations where a galvanic cell is established between an Al alloy plate and high-strength steel fastener.¹⁸  While the danger of galvanic corrosion has been recognized, little work has been done to assess the extent of corrosion damage and characterize the damage morphology of AA7050-T7451 and stainless steel at rivet sites.^39-40  Significant previous work has investigated the localized corrosion of high-strength Al alloys and developed an electrochemical framework for IGC attack.^{1,4,7,18,41-42}  However, there is a dearth of literature concerning the unique factors and environments present in a rivet hole and how they can influence the overall damage morphology. Previous work limits the understanding of damage morphologies to a few specific test environments: Point Judith marine seacoast atmospheric exposures,⁴³  full immersion in 0.6 M NaCl,⁴²  EXCO solution,⁴⁴  and 0.6 M NaCl + H₂O₂.⁴⁵  There has been very few detailed investigations into crevice and rivet environments formed in marine atmosphere solutions and how the microstructural details above interact with the local chemistries.

Literature has shown that galvanically-induced localized corrosion in Al alloys can be strongly influenced by environmental parameters such as pH, chemical species, and concentration.^46-48  The depletion of dissolved oxygen inside a crevice initially results in the formation of a differential aeration cell. Metal ions accumulate in the crevice solution as a result of anodic dissolution, and acidification follows with their subsequent hydrolysis. Then, a corresponding migration of chloride ions into the crevice occurs in order to preserve electroneutrality. The primary cathodic reaction occurs outside the crevice while hydrogen evolution or proton reduction occurs inside the crevice solution.^49-50  Cottis reported variable behavior where the pH was acidic near the crevice mouth and alkaline in the deeper region of a crevice on AA7475 (UNS A97475).⁵¹  The mechanism of alkalization was deemed unknown. However, it can be speculated that the anode and cathode lack sufficient anodic and cathodic separation. There is consequently greater OH⁻ production by the cathodic reaction than H⁺ by hydrolysis. In the case of crevice corrosion in dissimilar metal rivets, the local cathode supports cathodic reactions throughout crevice lengths. This can affect the pH gradient and differs from conventional crevice corrosion where the cathode is located only or primarily outside the crevice.

A final important aspect of crevice corrosion in a rivet hole is the damage morphology. Corrosion damage in an Al crevice typically has been assessed by metallographic cross section and imaging. However, this process is destructive and time consuming. Moreover it may be difficult to assess whether damage sites initiated together or separately and their pathway relative to microstructure. The corrosion volume and oxidation charge can be assessed from the area fraction, enabling qualitative investigation of cathodes. Synchrotron x-ray computed tomography (XCT) is a non-destructive technique which enables three-dimensional (3D) analysis of microstructural defects and corrosion damage volume at the micrometer length scale.^52-59  The technique is based on the differences in absorption coefficients of the materials along the path of transmission x-ray beams through the sample. The absorption coefficient is correlated to atomic number, enabling easy differentiation of different materials and phases.⁶⁰  Three-dimensional analysis is generated by acquiring many two-dimensional (2D) micrographs while rotating 180° to create a 3D volume.⁵⁴ 

While laboratory XCT has been used previously to study IGC rates in Al alloys,^53-54,61  the resolution was not high enough to characterize corrosion propagation path.⁶²  In recent years, there have been studies using synchrotron radiation XCT to investigate localized corrosion.^54,63  Knight, et al., previously used synchrotron XCT to examine IGC of bare Al alloys 7050 and 2024 (UNS A92024), with no dissimilar metal couple, in service-like conditions and showed that the growth of one fissure was constrained by that of another fissure at OCP.⁶¹  Ghahari, et al., also utilized XCT to for investigation of pitting corrosion in stainless steel under different applied currents and potentials,⁶⁴  while Davenport, et al., correlated the distribution of Y in a Mg alloy with corrosion propagation.⁶⁵  These studies have been contributory in demonstrating the effectiveness of XCT to analyze corrosion morphology. However, XCT has rarely been applied to the galvanic corrosion damage assessment between an Al alloy and stainless steel in a tight crevice.

The objective of this study was to examine the macro location and number of damage sites formed along with mesoscale galvanic corrosion morphologies that develop under conditions representative of galvanic coupling between a rivet and plate in an aerospace structure. Synchrotron XCT was used to track in 3D in situ or operando corrosion damage evolution in a simulated rivet geometry. These measurements enabled analysis of corrosion as a function of macroscale (e.g., rivet geometry) and meso- and microscale factors (e.g., spacing and size of constituent particles). A combination of electrochemical techniques was used to simulate the damage assessed. As the volume lost to corrosion was tracked and equated to anodic charge, the analysis presented a unique opportunity to quantify which cathode reaction rates were potent to contribute to the damage morphology. A further understanding of the sources of cathodic reactions which account for electrochemical galvanic corrosion damage is thereby obtained. Subsequent studies on inhibiting cathodic sites might be guided by these findings.

The material used in this study was machined from Al-Zn-Mg-Cu alloy plate (AA7050-T74511) 50 mm in thickness from the LT surface with the rivet hole in the LT plane and long axis parallel to the S orientation. A Type 304 stainless steel (UNS S30400) wire was embedded into the plate, simulating a rivet. A 3D rivet schematic with grain structure and orientation is shown in Figure 1. The microstructure of AA7050-T7451 has elongated grains in both the longitudinal (L)- and long transverse (T)-directions and are thinned in the short transverse (S) forming a pancake-like microstructure. The grains range from 22 μm to 1,250 μm, 15 μm to −265 μm, and 12 μm to 112 μm in the L-, T-, and S-directions, respectively. Aspect ratio, width over length, was determined to be 0.07, 0.51, and 1.41 in the L-, T-, and S-directions, respectively.

While Type 304 stainless steel was used in the operando XCT exposures, Type 316 stainless steel (UNS S31600) was used in the pH measurement studies and cathodic reaction rate studies conducted in the laboratory. The cathodic reaction rates of Type 304 and 316 stainless steel are found to be nearly identical. The composition of these materials are shown in Table 1.

The grain structure, particle size, and spacing of AA7050-T7451 was analyzed and is discussed below. Particle analysis was conducted using micrographs obtained with a JEOL JSM-6700F^† scanning electron microscope (SEM). These images had sufficient contrast to enable constituent particles to be distinguished. The threshold was adjusted so that the particles appeared black and the background appeared white. ImageJ^†,66  was used to calculate the size of the particle. The grain structure of the alloy was revealed by exposing samples to Keller’s etchant which was made up of 190 mL deionized water, 5 mL nitric acid, 3 mL hydrochloric acid, and 2 mL of hydrofluoric acid. The samples were etched for 25 s to reveal the grain structure of the alloy in the SL, TS, and LT orientations.

The simulated rivet had physical length scales accessible by tomography but scalable to a full rivet. The x-ray fastener geometry was a 1 mm diameter cylindrical AA7050-T7451 pin with a 250 μm rivet hole embedded with a Type 304 stainless steel wire. The rivet hole was 500 μm deep. The gaps between the stainless steel wire and the AA7050-T7451 varied from 1 μm to 50 μm, depending on the 360° location and the tilt of the pin. The area ratio inside the hole of AA7050-T7451 to stainless steel was 0.89:0.87. The entire pin was encapsulated with a plastic tube that contained filter paper with saturated potassium sulfate (K₂SO₄) controlling the relative humidity (RH) to approximately 98%. The plastic was sealed with an Al stopper at the top. A schematic of the simulated fastener is shown in Figure 2. Two pin samples with two different solutions were tested in operando, a 0.4 μL droplet of 4.0 M NaCl or 2.0 M MgCl₂ was placed on the surface on the pin. The electrolyte concentrations of the droplet on the surface of the pin equilibrated to the droplet’s equilibrium concentration of 0.5 M NaCl and 0.5 M MgCl₂ at 98% RH.

Potentiodynamic and potentiostatic polarization were conducted in simulated rivet environments to observe the anodic and cathodic electrochemical behavior of AA7050-T7451, the constituent phases of the alloy (Al₂CuMg, Al₇Cu₂Fe, MgZn₂), Type 316 stainless steel, pure Cu, and Cu replated on AA7050. The solutions used were simulated rivet solutions of 0.5 M NaCl, pH 5.5; 0.5 M MgCl₂, pH 5; and 0.5 M NaCl + NaAlO₂, pH 8. These environments were chosen to represent different regions in a macrocrevice where acidified pH may not exist because of the lack of anode and cathode separation. The environment may dictate the type of damage to occur. In an alkaline environment, intragranular fissure damage may prevail, while in an acidic environment, IGC may prevail. Electrochemical testing was performed using a three-electrode cell with a platinum mesh counter electrode and a saturated calomel reference electrode (SCE). Hydrogen gas was collected using an electrolyte filled burette with an inverted funnel attached centered over the sample. The samples were exposed through a 1 cm² knife edge washer. Selected experiments were deaerated using nitrogen for 1 h prior to testing and through the duration of the test. OCP was measured for 30 min prior to each test to ensure corroding conditions at the surface had been established. Scan rates of 0.5 mV/s were utilized for potentiodynamic polarization. A rotating disk electrode (RDE) was used to match the diffusion boundary layer thickness to the water layer thickness for some cathodic polarization experiments. The experimental current density was area corrected using the entire rivet geometry. All electrochemical tests were conducted three times to ensure statically reliable results. The average standard deviation of corrosion potential (E_corr) and corrosion current density (i_corr) were all found to be within ±5%.

pH measurements inside a simulated crevice were conducted. A Type 316 stainless steel was bolted into an AA7050-T7451 plate (6.35 mm). The cathode to anode ratio was approximately 1:1. The fastener was cross sectioned and attached with a plastic plate with set screws into the AA7050-T7451 cross section. This enabled the plastic to be removed quickly after exposure and the pH to be measured inside the rivet hole. The sample was immersed in 4 M NaCl for 62 h following the same test protocol of the simulated rivets. After the conclusion of the exposure, the plastic was removed and universal pH indicator and pH paper were used to determine the pH inside the rivet hole, in cross section, with the Type 316 stainless steel bolt present.

XCT measurements were performed at the Diamond Manchester Imaging Branchline (I13-2) at Diamond Light Source, United Kingdom, on an actively corroding sample. The x-ray transparent AA7050-T7451 cylinders were exposed under NaCl or MgCl₂ for various exposure times. The droplet covered the entire top of the pin and wicked into the crevice between the AA7050-T7451 hole and stainless steel rivet. Measurements were performed using a pink beam with a spread of energies between 6.6 keV and 20 keV, with a primary peak at 8.3 keV and a PCO edge detector with a 100 μm CdWO4 scintillator, a 2× relay optic, and a 4× objective, giving an effective pixel size of ∼812 nm. All of the XCT scans were performed at high resolution, taking 3,600 projections (radiographs) over 180° with continuous motion of the stage, giving a data collection time of approximately 7 min. Twenty dark scans (no beam) and 30 flat scans (full beam without sample) were taken for background subtraction and normalization of the image.⁽²⁾ The camera was checked with the dark field scans (images without the beam). The shutter was closed for balance of exposure times. The specs used for this operando XCT experiment enabled resolution of 0.8 μm, allowing local pitting attack to be monitored with time and depth.

Volume loss from corrosion can be used to calculate the anodic charge associated with the fissure damage. The anodic charge associated with corrosion damage, Q_A (C), can be determined using Equation (1):

where V_XCT is the volume loss from corrosion, found by the XCT (cm³), F is Faraday’s constant³³  (96,500 C/cm³), ρ is the density of AA7050-T7451 (2.83 g/cm²), and EW is the equivalent weight of AA7050-T7451. The EW was 9.77 g/equivalent assuming congruent dissolution and 9.54 g/equivalent assuming incongruent dissolution. Incongruent dissolution refers to the circumstance when Cu does not dissolve with the Al, Zn, and Mg equally.

Hydrogen gas was collected using an inverted burette filled with solution. The H₂ collected for select exposures was converted cathodic charge using the ideal gas law and Faraday’s law using Equation (2).

where z is 2 for H₂ evolution, P is the pressure inside the burette (1 atm [101.325 kPa]), V is the volume of gas in the burette (cm²), A is the area of the electrode (cm²), R is the ideal gas constant, and T is temperature.

The cathodic charge,   Q_{i  net  cathodic}, was determined using the integrated signal of the current as a function of time transient at each potential using Equation (3).

where i_c (E_couple) is the cathodic current at E_couple, the galvanic couple potential. E_couple was derived from finite element analysis (FEA) reported elsewhere.⁶⁷ 

The corroded area from the XCT was quantified using contrast analysis. At each slice the image was individually adjusted. Ring artifacts were subtracted at each slice as they appeared in different locations throughout the slices. This was necessary to ensure the contrasted area was physical damage and not artifacts. Each fissure was then individualized per 10. There were 470 and 380 slices in the 4 M NaCl and 2 M MgCl₂, respectively. The simulated rivet was cross-sectioned using a FEI Quanta 3D FEG^† focused ion beam at the Analytical Instrumentation Facility at North Carolina State University to verify area and depth of corrosion fissures.

Shape descriptors were calculated to geometrically describe the corrosion fissures in the 4 M NaCl case. The applicable shape descriptors were form factor and aspect ratio. Form factor reflects the smoothness of the corrosion fissure. A value close to one indicates a smooth and round shape and a value close to zero indicates an elongated shape. The form factor and aspect ratio were calculated using Equations (4) and (5).⁶⁸  The aspect ratio is defined using the largest diameter and the smallest diameter orthogonal to it. Form factor and aspect ratio were determined and analyzed using ImageJ.^†,66,68  

As confirmed by electron dispersive spectroscopy analysis, AA7050-T7451 has three types of coarse constituent particles: Al₇Cu₂Fe, MgSi₂, and Al₂CuMg. One micrograph from each plane from the middle of the plate (∼25 mm) can be observed in Figure 3. The bright white spots are the constituent particles. The constituent particles are observed to be aligned in stringers parallel to the L-direction. It was observed that the size and distribution of the constituent particles varied with plate depth and with plate orientation (Figure 4). On the LT surface the average particle size was 2.6 μm, 3.3 μm at eighth-thickness, 3.8 μm at quarter-thickness, and finally 3.4 μm at mid-plate thickness. At mid-plate the average diameter of the particles was 2 μm. The measured number of particles per 1 mm² on the surface was 290, 503, and 424 on the LT, LS, and TS planes, respectively. At mid-plate, the observed number of particles per 1 mm² was 246, 350, and 292 on the LT, LS, and TS planes, respectively. For all three orientations, the number of constituent particles per 1 mm² had a decreasing trend until quarter-thickness, then changed to an increasing trend at the mid-plate depth. The mean particle diameter increased from the surface to the quarter-plate depth, then a decrease in particle size was observed at the mid-plate depth. At the surface the mean particle diameter was 2.6 μm, 2.3 μm, and 2.6 μm at the LT, LS, and TS planes, respectively. At the mid-plate depth, the mean particle diameter was 3.3 μm, 2.8 μm, and 3.1 μm at the LT, LS, and TS planes, respectively. The cumulative size distribution show that the microstructure is heterogeneous, which agrees with previous findings.⁶⁹ 

A typical horizontal slice through an x-ray tomogram for the sample exposed to 4 M NaCl and a secondary electron image of the LT surface are shown in Figure 5. This shows a horizontal section viewed from the top of the cylindrical AA7050-T7451 test piece. The large gray disk is the AA7050-T7451. The bright blocky and round particles are Mg, Fe, and Cu-rich constituent particles and the large central white circle is the stainless steel wire. The circular bands to the lower right of the stainless steel are reconstruction artifacts. The dark corrosion fissures were observed in AA7050-T7451 around the rivet hole and extended from 1 μm to 500 μm in depth. The corrosion damage in both the 4 M NaCl and 2 M MgCl₂ were identified by phase contrast and verified using SEM. The fissures are seen to be intragranular and the relationship to grain structure is shown below.

After 62 h, four separate fissures were observed in the 4 M NaCl exposure: one at the mouth of the rivet, one at the middle of the rivet, and two at the bottom of the rivet. The first fissure was located 1.1 μm deep, with respect to crevice depth, the second fissure was located 22.7 μm deep, the third fissure was located 286.5 μm deep, and the last fissure was located 159.0 μm deep.

In the 2 M MgCl₂ exposure, three fissures were observed after 60 h; one was seen at the rivet mouth and two at the bottom of the rivet. In the 2 M MgCl₂ exposure, the first fissure was located 13.1 μm deep, the second fissure was located 105.3 μm deep, and the last fissure was located 251.0 μm deep into the 500 μm rivet hole. The two exposures were reconstructed in 3D using Avizo Fire^†, enabling 3D differentiation of the corrosion damage over time.

The four fissures in the 4 M NaCl exposure and the three fissures in MgCl₂ exposure can be observed in a 3D reconstruction in Figures 6 and 7. The 3D reconstruction shows the simulated rivet with a transparent AA7050-T7451 matrix where the blue cylinder is the Type 304 stainless steel wire and the red 3D regions around the wire are the corrosion fissures. The samples were exposed for 62 h under the 4 M NaCl exposure and 60 h under the 2 M MgCl₂ exposure. The 62 h and 60 h damage morphologies can be observed in relation to the constituent particles, which are shown in green (Figure 8). It can be observed that the first fissure (top of the pin) was the largest fissure in the 4 M NaCl exposure, while the second fissure (middle of the pin) was the largest in 2 M MgCl₂ (Figures 6 and 7). Both exposures had corrosion sites located at the top and bottom of the simulated rivet with no relationship with respect to the constituent particles in the 4 M NaCl exposure.

In the 4 M NaCl exposure, the corrosion fissures did not obviously follow bands of IMP clusters. The fissures corroded intragranually. Multiple fissures developed over the same time period in both 4 M NaCl and 2 M MgCl₂ exposures. The first corrosion fissure initiated after 12 h in the NaCl exposure and after 4 h in the MgCl₂ exposure. The fissure area was tracked over time and depth for each sample. The 2D area of corrosion as a function of depth and time can be observed in Figure 9 for the 4 M NaCl exposure and in Figure 10 for the 2 M MgCl₂ exposure. The 4 M NaCl exposure contained more damaged volume than 2 M MgCl₂ exposure. The 2D area of coarse constituent particles at each slice for each exposure was determined using the threshold method as discussed above. The area fraction of constituent particles varied as a result of the changing constituent particle area fraction with plate thickness. The constituent particle area on each slice was determined as a function of rivet depth for the 62 h 4 M NaCl exposure and the 60 h 2 M MgCl₂ exposure. The area fraction of the constituent particles observed was about 0.40% area fraction of AA7050-T7451 pin for the 4 M NaCl exposure and about 0.15% area fraction of AA7050-T7451 pin for the MgCl₂ exposure.

As a control, laboratory XCT was utilized to study a simulated fastener without the stainless steel wire present and also with the use of a cylindrical inert 250 μm pin⁽³⁾ crevice former. These simulated fasteners had the same geometry and were exposed to a 0.4 μL droplet of 4 M NaCl and 2 M MgCl₂ in a desiccator at 98% RH for 62 h and 60 h. These samples were used as a control to compare to the operando simulated fasteners with stainless steel. In the operando exposures, it was unclear whether corrosion damage occurred because of constituent particles, crevice corrosion, or the stainless steel. The control samples showed no damage in any of the exposures on the surface and through the entire 500 μm rivet hole. This indicated the corrosion was a result of the stainless steel fastener which also becomes a driving force for the Cu replating to initiate.

There were four corrosion sites in the 4 M NaCl exposure: one on the surface and the other three in the lower half of the rivet producing fissure type damage. The 2 M MgCl₂ case had three corrosion sites: one on the surface and two at the bottom of the rivet. Damage was in the form of fissures somewhat aligned with the L-direction bands of IMPs.

Recall that intragranular fissures were observed in Figures 5 and 8. Electron backscatter diffraction (EBSD) was conducted on the operando XCT samples to reveal the grain structure. Figure 11 shows the grain structure of the 4 M NaCl surface and Figure 12 shows the grain structure of the 2 M MgCl₂ surface with their corresponding SEM image and XCT surface images. The grain structure is dissimilar for each exposure. In the 4 M NaCl exposure, the grains are pancake-shaped grains, while in the 2 M MgCl₂ exposure the grains are thinner and longer. These grain structures produced slightly different damage morphologies. In the 4 M NaCl exposure, the damage was in the form of round-like corrosion fissures, while in the 2 M MgCl₂ exposure the damage was in the form of elongated corrosion fissures that followed the grain shape.

The samples were also characterized in cross section using a focused ion beam. The first fissure, top fissure, was cross-sectioned on the 4 M NaCl exposure. The cross section revealed the fissure was 105 μm deep. The 4 M NaCl pin was cross-sectioned approximately 300 μm from the edge of the pin. Fissure 3, at the bottom of the 2 M MgCl₂ exposure, was also cross-sectioned. A depth of 260 μm from the top LT plane was examined. The 2 M MgCl₂ pin was cross-sectioned approximately 370 μm from the edge of the pin. The corrosion fissure was found to be 94 μm in depth. The 4 M NaCl exposure cross section is shown in Figure 13. The cross-sectional analysis verified the XCT.

The volume loss from the corrosion sites revealed by operando exposures was calculated for various exposure times. The volume loss was an important parameter to determine. This enabled analysis of the required anodic charge to produce the damage that was observed. The volume and anodic charge as a function of time for the 4 M NaCl and 2 M MgCl₂ can be observed in Figure 14. This was calculated using Faraday’s law from Equation (1). In the 4 M NaCl exposure, the first fissure had the largest volume loss and the second fissure had the smallest volume loss. In the 2 M MgCl₂ exposure the first fissure had the smallest volume loss and the second fissure had the largest volume loss. The total anodic charge is the sum of the anodic charge for all of the individual fissures. The total anodic charge was 14 mC and 9 mC for the 2 M MgCl₂ exposure assuming congruent dissolution with all elements in AA7050 dissolving as the same time.

The form factor gave indication of smoothness and circularity of the corrosion fissures. Circularity is defined when all of the points on a surface of revolution are the same distance away from a common axis or center point.⁶⁸  Form factor and aspect ratio of the fissures in the 4 M NaCl exposure varying with depth can be observed in Figure 15. The first fissure was the least circular fissure of the four fissures, while the second fissure had the highest form factor, indicating a smoother shape. The average form factor was 0.15, which indicated that the cross section of the corrosion fissures varied from an ideal circle. The form factor varies with surface irregularities but not with elongation, and the aspect ratio has the opposite shape attributes. The aspect ratio describes length/breadth of the corrosion fissure. This indicates the size of the fissure will not change the numerical value of aspect ratio. The average aspect ratio of the four corrosion fissures in the 4 M NaCl exposure was 2.45. The second and third fissures had the lowest aspect ratio. This means that the major axis of the corrosion fissure was larger than the minor axis, which indicated these corrosion fissures were more elongated than the first and fourth corrosion fissures.

The pH gradient inside a dissimilar metal couple of AA7050-T7451 and a Type 316 stainless steel bolt was measured in a larger scale rivet/plate assembly. The pH was found to be between 7 and 8 at the top of the rivet hole and around 5 to 6 near the bottom of the rivet hole. No acidic sites were found.⁽⁴⁾ These results guided the environments chosen for the cathodic reaction rate studies.

Cathodic polarization studies were desired to provide indication of the different possible dominant cathodes and their reaction rates. Cathodic and anodic polarization of AA7050-T7451, pure aluminum, pure Cu, Type 316 stainless steel, Al₇Cu₂Fe, Al₂CuMg, and Cu replating on AA7050-T7451 were characterized with respect to cathodic kinetics in 0.5 M NaCl + NaAlO₂, pH 8. Pure Cu and Type 316 stainless steel were measured to the electrolyte film using a RDE with the intent to match to diffusion boundary layer thickness of 100 μm in a droplet or thin film covering a rivet. The cathodic and anodic polarization curves are shown in Figure 16. Type 316 stainless steel, pure Cu, Cu replated on AA7050-T7451, and, to a lesser extent, Cu containing IMPs all contribute to the pertinent cathodic current.

Potentiostatic polarization was conducted at E = −0.77 V_SCE for the Type 316 stainless steel and E = −0.80 V_SCE for the Cu replated on AA7050-T7451. These potentials were chosen using FEA conducted on the simulated pin geometry to represent the galvanic couple potential distribution in a galvanic couple, E_couple, extracted from current and potential distribution.⁷⁰  The FEA model showed limited IR drop through the 500 μm rivet hole. The FEA model determined the galvanic couple potential distribution was consistent with typical galvanic couple potentials observed both utilizing zero resistance ammeter measurements and overlying anodic and cathodic polarization curves. The environments that simulated the inside crevice were deaerated stagnant solutions to simulate the restriction of oxygen diffusion into a crevice and a confided space area without flow. The pin was simulated utilizing cathodic reaction studies open to air using RDE with δ = 100 μm. The current was converted to cathodic charge by integrating the current and normalizing it to the assumed wetted area of the simulated pin. The cathodic current density and cathodic charge density with respect to time can be observed in Figure 17.

The cathodic charge has three components: (1) the charge associated with the macro stainless steel rivet, (2) the charge associated with Cu replated on the surface and dealloyed S-phase, and (3) the charge associated with hydrogen evolution reaction. The cathodic charge was obtained from potentiostatic and RDE exposures over 62 h. The charge associated with the hydrogen evolution reaction was found to be 0.27 mC from Equation (2) and gas collection. The charge associated with the macro stainless steel rivet was found to be 11 mC, while the charge associated with replated Cu and dealloyed S-phase was found to be 17 mC. The cathodic charge capacity for each of the contributing cathodic reactions is shown in Figure 18. The total cathodic charge capacity for the environments considered was found to be 28 mC over 62 h, meaning the estimated cathodic charge was greater than the anodic charge or Q_C > Q_A. Charge calculations enabled a comparison of anodic charge from XCT to the cathodic charge, which will be discussed in the next section.

The key results of these rivet galvanic couples were: (1) multiple fissures were extended in size and length over the period of exposure albeit instantaneous corrosion rate could not be verified to be non-zero at the same instant time (Figures 9 and 10), and (2) fissures did not follow obvious clusters of IMPs in the 4 M NaCl exposure (Figure 8). This was interpreted to be the result of the presence of one or more strong cathodes associated with sites other than IMPs.

While both exposures had the same chloride concentration, the damage morphologies in the simulated rivet varied between exposures. The cross section (Figure 13), 3D reconstructions (Figure 8), and shape descriptors (Figure 15) of the 4 M NaCl exposure showed the fissure to have a more circular shape that did not seem to follow bands of constituent particles. In contrast, the 2 M MgCl₂ exposure was observed to have more elongated corrosion fissures that followed bands of consistent particles in the rolling direction (Figure 7). The EBSD on the simulated pin surface showed a different grain structure for the two exposures (Figures 11 and 12). The 4 M NaCl simulated rivet was sectioned from the surface of the AA7050-T7451 plate, while the 2 M MgCl₂ simulated rivet was sectioned from the middle of the AA7050-T7451 50 cm thick block. The grain structure of the AA7050-T7451 may be similar to the fissure morphology, in both cases, as might be expected for intragranular corrosion. However, the corrosion fissures were not cathodically limited to one corrosion site over the duration of the experiment. Fissures were able to grow through the entire depth of the 500 μm deep rivet hole as a result of the presence of several strong cathodic sites.

When stainless steel is coupled with AA7050-T7451 there are multiple cathodic reactions occurring. There are four key cathodic contributors: (1) Cu replating on the surface of AA7050-T7451, (2) dealloyed S-phase, (3) the stainless steel rivet, inside the crevice as well as externally above the crevice, and (4) a local hydrogen evolution reaction, which is a cathodic reaction local to a fissure, which may be an acidic site. In the simulated fastener plate system, multiple corrosion fissures grew simultaneously over the period of exposure (Figures 9 and 10). The question remains as to the governing cathodes that support a high number of fissures. Previous work has shown that with no stainless steel present, corrosion fissures are stunted by the growth of another fissure at the same cathodic site.⁶¹  This indicated that the corrosion rate was cathodically limited. In these exposures the corrosion fissures did not follow obvious clusters of constituent particles, suggesting the presence of the stainless steel rivet and a newly developed strong cathode, such as provided by Cu replating, were important. However, corrosion may be dominated by the cathodic reaction occurring on the stainless steel dissimilar metal, as the removal of this pin resulted in suppression of corrosion.

Pure aluminum and Al-Cu-Zn-Mg alloys are viable substrates for cathodic reactions supporting local corrosion.^71-74  However, the low exchange current density on Al and the insulating oxide film limits electron transfer reaction on aluminum alloys.⁷⁵  Cu replating significantly enhances the cathodic reactions supporting local corrosion and was found to cover approximately 50% of the surface.^23,73  There are two sources of Cu in AA7050-T7451: (1) Cu in solid solution, and (2) Cu present in precipitated and constituent particles such as S-phase. Cu replating occurs on the surface of many Cu-containing alloys during corrosion.²¹  This is a result of dealloying of Cu containing second phase particles in the matrix, as well as the release of Cu present in solid solution.^30,76 

The cathodic polarization data of Figure 16 in 0.5 M NaCl + NaAlO₂, pH 8, showed the cathodic reaction rates of various possible cathodes within the galvanic couple range, which was determined by various zero resistance ammeter exposures as well as FEA modeling. The cathodic reaction rate at potential E = −0.9 V_SCE of the Type 316 stainless steel and Cu replating on AA7050-T7451 were 3 × 10⁻⁵ A/cm² and 5 × 10⁻⁵ A/cm², respectively. When the RDE was used to simulate a 100 μm electrolyte layer, the current density increased by two orders of magnitude. The RDE reaction rate was comparable to previous work conducted by Policastro, et al.,⁷⁷  on polarization under droplets of 0.6 M NaCl at 80% RH on UNS S13800. This analysis indicated that the Type 316 stainless steel and replated Cu had a significant effect on the cathodic kinetics when AA7050-T7451 was coupled with Type 316 stainless steel under full immersion and thin film conditions.

The anodic charge calculated based on the volume loss determined by XCT enabled a comparison study to the cathodic charge, Q_c, associated with the cumulated net cathodic charge determined from of the stainless steel and other sources at the galvanic couple potential determined by the FEA modeling.⁶⁷  The cathodic charge for each of the contributing cathodic reactions can be seen in Figure 18. The total cathodic charge for the environment considered was found to be greater than the anodic charge, Q_C > Q_A. The simulated pin geometry was reconsidered using various crevice environments to study the various cathodic contributors. The cathodic current has four important contributors: the stainless steel above the crevice, the stainless steel inside the crevice, hydrogen evolution in an acidified fissure, and any replated Cu or dealloyed S-phase (Equation [6]). In addition, each of these locations has a specific solution chemistry. This analysis aimed to determine the cathodic charge associated with each of these cathodic contributors or reactions.

where ∑Q_anode(XCT) is the anodic charge associated with volume loss as a result of corrosion found using XCT, ∑Q_cathode is the cumulated net cathodic charge, Q_{copper replated} is the cathodic charge associated with Cu plating on the surface on AA7050-T7451, ∑Q_{dealloyed S phase} is the cathodic charge associated with dealloyed S-phase leaving behind a cathodic particle to the bulk matrix, ∑Q_{remote macro cathode (SS)} refers to the cathodic charge of the macro stainless steel wire, and ∑Q_H2 is the associated charge from the local hydrogen evolution reaction.

The cathodic charge that had the biggest contribution in these environments was the charge associated with Cu replating on 50% of the surface and dealloyed S-phase accounting for 0.3% area fraction, followed by the charge associated with the Type 316 stainless steel macro rivet. The accumulated H₂ from the hydrogen evolution reaction contributed less than 10% of the overall cathodic charge. However, the XCT control samples with no Type 316 stainless steel wire showed no corrosion sites, which may suggest that the stainless steel rivet is needed for the initiation of Cu replating, which then takes over the primary cathodic contributor.

As stated above, a need exists to examine environments typical of crevices. Various conditions were explored. All surface (above crevice) environments were examined using RDE and ambient aeration to simulate a thin water layer and all crevice environments (inside) were deaerated to simulate oxygen depletion inside a crevice. Once the potentiostatic library in various environments was complete, the rivet simulated by a pin geometry was considered in various pH gradients and environments. The most realistic environment that was considered was the condition where the mouth of the crevice in an alkaline pH and the inside of the crevice was considered to be neutral. Therefore, pH 9 RDE data were used on the surface and deaerated pH 7 and pH 5.5 data used for inside the rivet hole. The total cathodic charge capacity for the environments considered was found to be 28 mC over 62 h (Figure 18).

This type of analysis enabled a study on which precise macro and micro cathodes control the galvanic corrosion between AA7050-T7451 and Type 316 stainless steel central to damage; this enables targeted suppression of certain cathodes to mitigate corrosion damage. The Cu replating had a greater contribution to the total cathodic charge than the stainless steel rivet. This was an interesting observation as no corrosion sites grew when the stainless steel was removed. This may indicate that the stainless steel provides the key driving force for Cu to replate on the surface, which in turn is subsequently the key cathodic reaction site.

• •

  The corrosion of AA7050-T7451 was not observed to be cathodically limited to one fissure in a simulated fastener/plate galvanic couple arrangement as multiple fissures grew over the same time period of 62 h for the NaCl exposure and 60 h for the MgCl₂ exposure. The corrosion path was found to be intragranular and did follow obvious clusters of IMPs. Cathodic polarization studies on selected phases present in AA7050-T7451, as well as Type 316 stainless steel, were investigated and compared to the galvanic couple range from E_couple experiments in crevice solutions of various pH and concentrations. The main contributors to the cathodic currents over the galvanic potential range were the stainless steel and AA7050-T7451 with replated Cu.

We gratefully acknowledge support from the Office of Naval Research under the contract ONR: N00014-14-1-0012 with William Nickerson as contract manager. X-ray data were collected and analyzed with assistance from Andrew du Plessis and Weichen Xu of the University of Birmingham and Mark Basham and Trevor Rayment of Diamond Light Source. We also appreciate the collaboration on the FEA modeling done by Chao Liu and Dr. Robert Kelly, as well as the fatigue studies by Noelle Co and Dr. Jimmy Burns. Focused ion beam work was performed in part at the Analytical Instrumentation Facility (AIF) at North Carolina State University, which is supported by the State of North Carolina and the National Science Foundation (Award No. ECCS-1542015).