Underfilm corrosion on polyurethane-coated aluminum alloy 2024-T3 containing dissimilar metal fasteners.

Panels of AA2024-T3 were fastened using Cu and Cu-Ni fasteners, pretreated, and coated with around 25 [micro]m of polyurethane film. The pretreatments were based on typical conversion coating processes. Samples included alkaline-cleaned, deoxidized, and conversion-coated. The conversion coatings included chromate and cerium-based conversion coatings. Samples were tested by scribing around the fastener and on the AA2024-T3 (matrix), exposing them to HCl vapor for 15 min, and subjecting them to a filiform corrosion test by exposure to 82% relative humidity at 40 [+ or -] 1 [degrees]C for up to 1000 hr. Inspection showed that the filiform corrosion that developed from a scribe around the fastener was an order of magnitude larger than from a scribe on the matrix (scribe on AA2024-T3). With respect to the surface treatments, the amount of filiform corrosion for scribes on the matrix was greatest on the alkaline-cleaned samples, followed by the deoxidized samples, with the conversion-coated samples performing the best. The amount of filiform corrosion for scribes around the fasteners was highest for the deoxidized sample.

Keywords: Corrosion testing, corrosion, corrosion protection, polyurethanes, accelerated testing, chromate, chromate replacements, conversion coating, cerium-based conversion coatings

**********

Underfilm corrosion such as blistering, filiform corrosion, and the development of a general underfilm corrosion front from a defect in a paint system is a widespread phenomenon for a range of metals in a range of service conditions and applications. It is of major interest in metal finishing since it leads to a reduction in cosmetic appearance in nonloadbearing applications, and may lead to loss of structural integrity in loadbearing applications.

Underfilm corrosion is often associated with galvanic coupling between dissimilar metals (1) and occurs commonly where either dissimilar metals are joined or the same metal is joined using a fastener manufactured from a different metal. For example, corrosion on aircrafts is often initiated around fasteners. (1-9) This occurs for several reasons. First, the fasteners are stress concentrators (stress can be raised by a factor of 3 (2)) and initiate more damage to the paint system at these points. Second, cracking of the paint system means that there is a defect pathway for corrodents and moisture to come into contact with the underlying metal, resulting in corrosion processes that can lead to pitting and intergranular attack and ultimately fatigue failure. (1) Finally, the fasteners are often made of a different metal than typical aluminum alloys used for fuselage and wing skins. Steel is a common choice, but it is plated with cadmium to reduce the galvanic effects of corrosion on the surrounding aluminum; but it can nevertheless provide little protection if breached or if in a crevice environment. (1)

In the aircraft industry, the link between corrosion around fasteners and the loss of fatigue lifetime is well documented. (1-5) Furthermore, the effect of underfilm corrosion on aircraft skins made from aluminum alloy 2024 is also well documented. (1) In these instances, substantial filiform corrosion was detected around fasteners on the fuselage and lower wing skin, causing a reduction in the lifetime of the components affected.

The move towards extending the operational lifetime of aircraft beyond their original design limits means that close monitoring of the aircraft health is becoming increasingly important. (10,11) The Aloha Airlines incident highlights the catastrophic impact of corrosive operating environments on aging aircraft. In this incident, the age of the aircraft, which had experienced nearly 90,000 cabin pressurization cycles, led to a combination of adhesive bond failure, corrosion, and fatigue damage of lap joints in the upper skin. This was identified as the likely cause of failure and removal of a section of fuselage skin during flight. (2,10)

[FIGURE 1 OMITTED]

Understanding the progress of corrodents through a damaged film and the consequent development of corrosion can lead to better design of fastened joints and/or better remedial treatments for corrosion when it does occur around the fasteners. Corrosion in joints is likely to be driven by galvanic coupling between dissimilar metals and acidic conditions that develop in occluded volumes at the site of corrosion, such as pits, crevices, and areas where exfoliation has occurred.

Over the years the metal finishing and paint industries have developed a number of tests to evaluate the ability of surface treatments and paint systems to resist underfilm corrosion. (12) These tests usually involve scribing through a painted specimen and exposing that specimen to an environment that will facilitate the development of corrosion under the paint film adjacent to the scribe. There are few published tests, however, that evaluate underfilm corrosion, which is driven by galvanic coupling from a fastener/metal system.

In this study, a filiform corrosion test was combined with a galvanic couple by using dissimilar metal fasteners to join two sheets of aluminum alloy, thus producing a model lap joint. The model lap joints were finished in a range of metal finishing treatments including either alkaline cleaning, deoxidation, and chromate or cerium-based conversion coatings. Hence, the effect of both the galvanic coupling and metal finishing treatment on the development of filiform corrosion around a fastened aluminum joint was examined. The aluminum alloy was 2024-T3 (AA2024-T3) and the fasteners included copper and copper-nickel alloy. After the metal finishing treatment, the fastened samples were coated with a clear polyurethane topcoat. A clear polyurethane was chosen so that the corrosion could be observed through the topcoat.

Chromate conversion coatings were chosen as an example of the current conversion coating technology, as applied to AA2024-T3. (13) Characterization of this type of conversion coating (14,15) with a chromate-based deoxidizer treatment (16,17) has been reported previously, as well as the filiform corrosion resistance. (18) Cerium-based coatings were chosen as a Cr-free alternative to chromate conversion coatings. Cerium-based conversion coatings were applied to a range of metals, including tinplate, (19) magnesium (AZ91), (20) zinc (Zn), (21,22) and aluminum alloys (23-29) including AA2024-T3. (30,31) The accelerated cerium-based coatings described herein have been previously reported to show good corrosion resistance on a range of aluminum alloys. (32-34) Characterization of these conversion coatings using a chromate deoxidizer (35) or a cerium-based deoxidizer, (36) as used here, has also been described elsewhere along with filiform corrosion studies. (37)

EXPERIMENTAL

Alloys

AA2024-T3 sheet material was used for the manufacture of all fastened samples. Analysis by inductively coupled plasma (ICP) atomic emission spectroscopy showed that the following alloying elements were present at or above 400 ppm: Cu 5.3%, Mg 1.6%, Mn 0.6%, Fe 0.2%, Si 0.06%, and Ti 0.04% wt%.

Surface Preparation

Samples of AA2024-T3 were cut into 7.6 cm x 2.5 cm lengths. Four holes were drilled in the end of each sample and pairs of samples were fastened together using metal rivets, as depicted in Figure 1. The samples were then fastened with either four Cu fasteners or four Cu/Ni alloy fasteners, which were 4 mm in diameter. After fastening, the samples underwent surface treatment. For each fastener type, one set of samples was given a silicated alkaline clean treatment (Gibson Process 204B), the second set was alkaline cleaned and then deoxidized, and the final set was alkaline cleaned, deoxidized, and conversion coated with either a chromate- or cerium-based (cerate) conversion coating.

The procedure used for the conversion coatings is outlined in Table 1. Both the cerium-based and chromate procedures utilized the Gibson 204B alkaline clean treatment. The chromate procedure used Henkel, Parker + Amchem No. 7 deoxidizer followed by Henkel, Alodine 1200s conversion coating. The cerium-based procedure utilized a rare earth deoxidizer: (N[H.sub.4])[.sub.4] Ce(IV)(S[O.sub.4])[.sub.4].2[H.sub.2]O (63.46 g), KFHF (1.56 g), [Na.sub.2][S.sub.2][O.sub.8] (8.63 g), [H.sub.2]S[O.sub.4] (56 ml), and HN[O.sub.3] (163 ml) for every two liters; (38) followed by a cerium-based solution: Ce[Cl.sub.3].7[H.sub.2]O (0.035 M), [H.sub.2][O.sub.2] (0.1 M), and a bismuth accelerator. (39) For brevity, the chromate and cerium-based conversion coatings are referred to as CrCC and CeCC, respectively. After surface treatment, all samples were spray-coated with a polyurethane topcoat to an average thickness of 25.6 [+ or -] 3.6 [micro]m.

Filiform Corrosion Experiments

PREPARATION OF SAMPLES: Prior to filiform testing, all samples were scribed using a scalpel blade around the rivet and on the sample away from the fasteners, which is referred to as the matrix (Figure 1). The samples were then exposed to HC1 vapor for 15 min. A corrosive environment of 82% humidity was produced by placing a saturated solution of ammonium sulfate, (N[H.sub.4])[.sub.2]S[O.sub.4] (> 82 g/L), inside a plastic container. The container was then sealed and allowed to come to equilibrium inside an oven at 40[degrees]C for two days. The samples were then inserted into the container.

FILIFORM MONITORING: The alkaline-cleaned, deoxidized, and cerium-based conversion-coated samples (both Cu and Cu/Ni fastened) were all exposed for a total of 1000 hr. From the eight remaining chromate conversion-coated samples, two were exposed for 250 hr, two were exposed for 500 hr, two were exposed for 750 hr, and two were exposed for 1000 hr. The treatment given each sample is summarized in Table 2.

Every two days, photographs of the samples were taken to track the development of filiform corrosion. With these photographs, it was possible to determine the site density, the average filament length, the corrosion number, and the area of corrosion. The site density was defined as the number of filament initiation sites per scribe length and was determined using optical microscopy for both the scribes on the matrix and those around the fastener. The average filament length was simply the average length of the 10 longest filaments. The corrosion number was defined as:

corrosion number = site density x average filament length (1)

In addition to the corrosion number, the area of corrosion along the matrix scribe and around the fastener was also determined using the image analysis software, Image Pro Plus. Both the corrosion number and the area of corrosion were used as a measure of the overall extent of corrosion.

In the Results Section, the corrosion areas are expressed as area per unit length of scribe. For a circular scribe around a fastener, there is more available area/unit length of scribe for filament growth than for a linear scribe, and the ratio of the area is described (see Appendix) by 1 + 1/2[r.sub.1], where [r.sub.1] is the radius from the center of the fastener to the scribe and 1 is the length of the corrosion away from the scribe.

RESULTS

Description of Four Samples

Figures 2-5 show the evolution of filiform corrosion from the scribe on the matrix at 250, 500, 750, and 1000 hr for alkaline-cleaned, chromate-based deoxidation, CeCC, and CrCC samples, respectively. Figures 3-5 include magnifications of the scribe to show the corrosion more clearly. The low magnification images are numbered as Figures 3a, 4a, and 5a, and the high magnification images are numbered as Figures 3b, 4b, and 5b. Comparison of Figures 2-5 indicates that the alkaline-cleaned samples developed the greatest amount of corrosion. At 72 hr (not shown) the alkaline-cleaned sample had individual sites developing from the scribe; however, by 250 hr many more sites were generated, and many of these sites had merged to form a broader corrosion front. As a result, the initiation site density decreased during the course of the test but the overall corrosion was greater. The filiforms then developed from the corrosion front which itself grew from the scribe. The corrosion front is synonymous with "undercutting" or "creepage," terms used to describe lifting of paint near a scribe. This degree of corrosion has been observed previously with alkaline-cleaned samples. (40,41) The initial site densities for all the samples were similar, as shown in Table 3.

The deoxidized surface as well as the CeCC and CrCC surfaces all showed isolated filaments developing from the scribe, but were otherwise indistinguishable in the optical images. After a preliminary period, where a number of sites were initiated, no further initiation was observed during the course of the test, so the site density remained constant throughout. The filaments themselves grew slowly and remained isolated.

Figure 6 displays the development of corrosion around the CrCC Cu (Figure 6a) and Cu-Ni (Figure 6b) fasteners as a function of time up to 500 hr. Like the scribes on the matrix, corrosion started at individual sites along the scribe. As the exposure time increased, the amount of corrosion increased and new sites developed. In some places along the scribe, these new sites merged into a common corrosion front, which was not observed for the CrCC on the matrix. Thus, the amount of corrosion was greater per unit length of scribe around the fastener than on the matrix. The development of underfilm corrosion from the scribe around the Cu-Ni fastener was similar for the CeCC and CrCC coatings, but greater for the CeCC coating compared to the CrCC for the Cu fasteners. In the case of the alkaline-cleaned and deoxidized samples, after initiation, many of the filaments merged and developed into a common corrosion front emanating from the scribe around the fastener. These undercut the polyurethane topcoat.

[FIGURE 2 OMITTED]

It was interesting to note that when the scribe itself was cut over the actual Cu or Cu-Ni fastener, the corrosion initiation under the polyurethane was at the matrix/fastener boundary nearest the scribe mark and it progressed away from the scribe. Once initiated, the corrosion formed filaments and quickly spread around the boundary between the fastener and the aluminum matrix. Filaments that did develop from the scribe had very tortuous paths. In these circumstances it was easier to measure the total area of corrosion around the fastener, rather than try to measure the filament lengths.

[FIGURE 3A OMITTED]

[FIGURE 3B OMITTED]

Measurement of Filiform Corrosion

Figure 7 shows the growth in the filament length from the scribe on the matrix for all four surface treatments as a function of time. Measurements were made on the 10 longest filaments and the error bars on the alkaline-cleaned samples reflect the maximum deviation from average. The error bars for the other processes were a similar percentage of the total length and, thus, correspondingly smaller. All samples exhibited an increase in filament length as a function of time. At 1000 hr, the deoxidized, CeCC, and CrCC samples had filament lengths around 2.0 [+ or -] 0.5 mm. At 1000 hr of exposure, the average filament length of the alkaline-cleaned sample reached 6.0 [+ or -] 1.5 mm. The filaments on this sample were also wider than those of the deoxidized or conversion-coated samples.

The results presented in Figure 7 are similar to those reported in the literature. For example, Leth-Olsen and Nisancioglu (41) found that the rate of filament growth decreased with increasing time for a range of alkaline-cleaned aluminum alloys including AA1080, AA1050, AA1200, AA8006, and AA3005, which they attributed to a loss of chloride from the filament tip. Their filament lengths were in the range 1-10 mm. On the other hand, Williams and McMurray found that the filament length increased linearly with time on polished surfaces. (42) In a later study, however, they found that the area of corrosion reached a limiting value with a hydrotalcite inhibited paint system, which they attributed to the removal of chloride from the filiform head. (43)

The corrosion number at 1000 hr for the different treatments is displayed in Table 3, and was obtained using equation (1) and the initial site density. This was necessary since the site density of the alkaline-cleaned sample decreased during the filiform test due to the amalgamation of filaments along the scribe as corrosion progressed. Since the initial site density was similar for all samples, the data in Table 3 simply reflect the average filament length; hence, the corrosion number is much larger for the alkaline treatment. The corrosion numbers were similar for the deoxidation and conversion-coating processes.

[FIGURE 4A OMITTED]

[FIGURE 4B OMITTED]

The closest study to this is that of Mol, (40) who used the same chromate treatments and a similar cerium-based treatment. Mol found filament lengths at 1000 hr similar to those reported here for alkaline-cleaned, CeCC, and CrCC specimens (Table 2). In the previous study, however, it was unclear what effect deoxidation had on improved filiform corrosion resistance as part of the conversion-coating process. This is addressed in the current article. The corrosion numbers at 1000 hr were also close to values reported by Mol where a similar pretreatment, deoxidizer, and chromate conversion coating were used. (40) In the current study, however, the formulation of the cerium-based is different since a Bi-accelerated solution (39) was used, whereas a Cu-accelerated solution was used in the previous study. (34,44) The Bi-accelerated cerium-based process is much milder than the Cu-accelerated process, which is reflected in the shorter filament lengths and smaller corrosion numbers (which in this study were both less than the chromate process).

[FIGURE 5A OMITTED]

[FIGURE 5B OMITTED]

As an alternative to the corrosion number, the growth in corrosion area was also examined. This parameter includes any corrosion front that develops from the scribe and takes account of the filament width. Neither of these factors is included in the corrosion number or average filament length. Thus, the development of the corrosion area along the scribe as a function of time is plotted in Figure 8 for the matrix. The area of corrosion per unit length of scribe increased as a function of time; however, the rate of increase appeared to slow at longer times, displaying a similar trend to that of Figure 7. Corrosion area was recently used by Le Bozec et al. (45) for a range of reasons primarily related to the difficulty of distinguishing different forms of underfilm corrosion as well as noting that it is not only the filiform corrosion which represents underfilm degradation. Unlike Figure 7, where the filament lengths appear to stabilize at around 6 mm for the alkaline-cleaned samples, the corrosion area appeared to increase. This increase is attributed to the amalgamation of filaments into a corrosion front which undercuts the polyurethane near the scribe.

[FIGURE 6 OMITTED]

A further distinction between the average filament lengths and the area is seen by comparing the deoxidized sample with the conversion-coated samples in Figure 7 to that in Figure 8. The filament lengths for the deoxidized and conversion-coated samples were similar in Figure 7, whereas the area of corrosion for the deoxidized samples was clearly greater than that of the conversion-coated samples in Figure 8. This indicates that the conversion coat provides an increased resistance to filiform development beyond deoxidation alone.

Corrosion Around Fasteners

The development of corrosion around the fasteners is displayed in Figure 9. The most notable aspect is that the area of corrosion from a scribe around the fasteners was between one and two orders of magnitude larger than that from the scribe on the matrix. For example, the alkaline-cleaned specimen displayed an area of corrosion per unit length of scribe of approximately 60 [mm.sup.2]/mm around the fastener, compared with approximately 8 mm2/mm around the scribe on the matrix. The bigger area of corrosion is due to two factors: (a) the galvanic couple between the fastener and the matrix and (b) the greater available area for corrosion and filament growth due to the curvature associated with the scribe around the fastener. Based on the ratio formula described in the Experimental Section and Appendix, the filament would need to grow to five millimeters for the available area to be twice the area from a linear scribe. As can be seen from Figure 6, some parts of the corrosion front have grown some 5 to 7 mm away from the scribe and therefore would have roughly twice the available area over a liner scribe. In these circumstances, not all the area around the scribe, however, was consumed with the corrosion front; but the area that was consumed is still 10 times larger than the scribe on the matrix, indicating that the area affected plays a secondary role. This suggests that the galvanic effect is the dominating contribution to the dramatically higher corrosion area observed around the fastener.

[FIGURE 7 OMITTED]

For the deoxidized surface, the area of corrosion around the Cu-Ni fastener was larger than that of the alkaline-cleaned sample, which is the opposite trend of the scribe on the matrix. At the end of the test, the area of corrosion around the fastener was nearly 60 times the area for the filiforms from the scribe on the matrix.

The area of corrosion around the conversion-coated Cu-Ni fasteners was less than half that of the deoxidized surface, demonstrating a distinct advantage with the conversion coating on the surface. Like the deoxidized surface, the area of corrosion was still around 60 times that of the scribe on the matrix. The CrCC specimens appeared to have less corrosion than the CeCC samples.

[FIGURE 8 OMITTED]

In Figure 10, the corrosion for the deoxidized, CrCC, and CeCC coated specimens is compared for Cu and Cu-Ni fasteners. The deoxidized surface had the greatest area of corrosion, being larger than either conversion-coated surface (cf Figure 9). The CeCC appeared to have slightly greater corrosion than the CrCC on the Cu-Ni fastener.

The area of corrosion for the CeCC and CrCC coatings was larger for the Cu fastener than for the Cu-Ni fastener. This suggests that the potential difference generated by the galvanic couple is greater for the Cu-AA2024-T3 pair than the Cu-Ni-AA2024-T3 pair (Figure 10). In the case of the CrCC, there was only slightly greater corrosion on the Cu fastener compared to the Cu-Ni fastener. For the CeCC, the area of corrosion was significantly greater on the Cu fastener.

DISCUSSION

Corrosion on the Matrix

The development of filiform corrosion from a scribe has been studied extensively in recent years. On alkaline-cleaned surfaces the growth of filiforms is rapid and characterized by only a superficial attack of the underlying metal. This is attributed, in some instances, to electrochemically active surface layers which readily undermine the polymeric topcoat, (41,46,47) and in other cases to poor adhesion of the topcoat to the basic residual oxides left on the surface after alkaline cleaning. (17,18,48,49) The rapid growth of filaments in this instance is attributed to the poor adhesion of the polyurethane to the alkaline-cleaned surface of AA2024-T3, which comprises basic oxides and silicates, (49) although reactive surface layers cannot be ruled out.

Deoxidation etches the surface, providing a mechanical key for adhesion of the polyurethane to the substrate metal, including "scalloping" and also "texturing" down to the tens of nanometers scale, (49) which was recently attributed to localized etching around impurities. (50) The progress of a filament therefore requires the generation of a greater amount of corrosion product to lift the polyurethane, resulting in deeper attack of the metal. (18,37,40,41) This type of attack is often called "successive pitting attack." Hence, the propagation rate of the filaments on the deoxidized surface was much less and the average filament was considerably shorter.

On the basis of filament length alone, the conversion coatings did not appear to further reduce the susceptibility to filiform corrosion (Figure 7). The area measurements, however, suggested that the presence of the conversion coatings provided additional resistance to filiform corrosion. Furthermore, there was no distinguishable difference between the cerium-based and the chromate process, which is in agreement with previous studies. (40) The greater sensitivity of the area is probably due to the inclusion of the corrosion front from the scribe (only the length of the 10 longest filaments is measured for the average length or corrosion number) and/or the width of the filaments. The additional protection could be provided through two avenues: either from additional surface texturing from the conversion coatings, which increases the surface area and provides an additional source of keying for the topcoat; or a chemical effect due to the inhibition of corrosion under the filiform head or blocking of oxygen reduction in the filament tail. Atomic force microscopy (AFM) of both the cerate (36) and the chromate coatings (13,51) indicates that the surface area of the coatings may be quite high due to topography, suggesting that mechanical keying to the paint coating might promote better paint adhesion. In terms of corrosion protection, chromate is well documented as an anodic and cathodic inhibitor (51-54) and, on aluminum, cerium ions act as cathodic inhibitors, (26,55) suggesting that electrochemical inhibition may also play a part in the greater degree of protection offered by the conversion coatings.

[FIGURE 9 OMITTED]

Corrosion Around Fasteners

In the presence of the galvanic couple between the fastener and the matrix, the amount of corrosion is at least an order of magnitude worse than that for corrosion emanating from the scribe on the matrix. This can be understood by considering the fundamental mechanism of filiform corrosion. The corrosion process begins with a defect in a paint layer over a metallic substrate. The defect allows ingress of moisture and other components critical for corrosion such as chloride ions and oxygen. Blistering and delamination result as an attack of the substrate begins. A differential aeration cell is established since the area immediately under the paint defect is richer in oxygen than the tip of the blister. This results in a cathodic site immediately under the paint defect, an anodic site at the tip of the blister, and an electrical potential between them. As the potential is established, corrosion proceeds by preferential dissolution of the substrate at the anode, forming individual filaments. The mechanism is shown schematically in Figure 12. When the galvanic effects of the dissimilar fastener were added to the mechanism, the potential difference between anode and cathode was increased. This greatly increased the rate of dissolution of the substrate at the tip, dramatically accelerating the overall rate of corrosion. Hence, the area of corrosion per unit length of scribe was more than an order of magnitude greater around the fastener than on the matrix.

[FIGURE 10 OMITTED]

[FIGURE 11 OMITTED]

From the optical microscopy, it is clear that a corrosion front developed for a considerable length of the scribe for the deoxidized and conversion-coated specimens, unlike the scribe on the matrix where individual filaments were observed to grow from the scribe. For the alkaline-cleaned samples, the amount of corrosion was clearly more around the fastener (compare Figures 9 and 10 with Figure 8) due to a nearly continuous corrosion front emanating from the scribe. For the scribe on the matrix, however, the corrosion front was not continuous. For all samples, the development of a corrosion front was probably assisted by the presence of a much more efficient cathode in the form of a Cu or Cu-Ni fastener compared to the AA2024-T3 along the exposed scribe on the matrix.

[FIGURE 12 OMITTED]

In addition to a greater amount of corrosion product, the growth rate of corrosion was different around the fastener compared to the matrix. In the latter case, Figure 8 suggests that the growth slows as a function of time, i.e., as the filament length increases. "Filament stagnation," as it is termed, is believed to be a result of a decreased oxygen supply with an increasing diffusion path. (41,48) Furthermore, a reduction in the amount of chloride at the tip due to chloride incorporation into the corrosion product within the filament tail also contributes. (18,37,41,48) In the fastener case, the growth of the area under corrosion appeared to be linear (Figures 9 and 10), suggesting that stagnation was not occurring.

The delay in the onset of stagnation was probably due to a number of factors. First, coupling of the Cu or Cu-Ni fastener to the AA2024-T3 means that at the mixed potential of the couple, the AA2024-T3 was actively corroding at its pitting potential; therefore, the rate of dissolution of material was much higher than under the scribe on the matrix. Second, the backscatter electron image of Figure 11 showed significant corrosion under the paint adjacent to the fastener, including exfoliation corrosion, as is often observed around fasteners. (1) While pitting corrosion was observed in filaments, (18,37,40,48) exfoliation corrosion had not been seen in previous studies of filiform corrosion on CrCC AA2024-T3. (18,37) The generation of considerable corrosion product results since exfoliation corrosion causes general lifting of the polyurethane topcoat, allowing much greater access to air and moisture than could be obtained by diffusion through a filament tail.

The increased susceptibility of the deoxidized surface to corrosion around the fastener compared to the alkaline-cleaned surface was a further indication that the drivers for underfilm corrosion at the fastened joint were different than those along the scribe on the matrix. For the deoxidized surface, the deoxidation treatment removed an Al-Mg-Si oxide layer present on the surface after alkaline cleaning. At the mixed potential of the surface adjacent to the fastener, the Al-Mg-Si-oxide layer may provide greater protection of the underlying aluminum than the very thin oxide left after deoxidation. Potential-time curves of native oxides on Al-Mg binary alloys show a lowering of the corrosion potential with increased Mg content, whereas increased Cu content made the corrosion potential more positive. (56) The effect for AA2024-T3 is more complex due to its alloy composition, but the presence of an Mg-rich oxide on the surface appears to favor some passivation of the matrix alloy.

For the conversion coatings, a protective barrier oxide in the form of a cerium oxide for cerium-based or mixed Cr(III)/Cr(VI) for the chromate adds to the protective oxide system. These coatings provide protection by inhibiting anodic dissolution of the aluminum surface. In the case of the chromate layer, the release of soluble Cr(VI) from the coating, which, for pitting, is known as an anodic inhibitor, (52,57) results in protection of the matrix. Cerium ions that may be released from the cerium-based coating, however, are well known as cathodic inhibitors. (26,58,59) While cathodic inhibitors may work to some degree at the cathodic site behind the head of a filament on the matrix, they probably offer limited ability to migrate in enough concentration to the fastener where the majority of the cathodic reaction is taking place and passivate the fastener. Hence, the performance of the cerium-based coating is poorer than the chromate coating for the Cu fastener and is likely to reflect the relative contribution of the cathodic region immediately behind the filament head to the contribution from the fastener.

CONCLUSIONS

Underfilm corrosion from a scribe was studied on AA2024-T3 and around Cu and Cu-Ni fasteners used to join AA2024-T3 matrix. For the AA2024-T3, it was found that the underfilm corrosion was best measured using the area of corrosion. For AA2024-T3, underfilm corrosion in the form of filiform corrosion was worst for the alkaline-cleaned sample, followed by the deoxidized sample, and then the chromate and cerium-based conversion coatings.

In the case of the fasteners, the amount of corrosion was 10 to 60 times greater than from the scribe on the AA2024-T3. The greatest amount of corrosion was for the deoxidized sample, even greater than the alkaline-cleaned sample. For the scribes in proximity to the fasteners, the chromate and cerium-based performed similarly for the Cu-Ni fastener, but the chromate had less corrosion than cerium-based for the Cu fastener.

ACKNOWLEDGMENT 

The authors would like to thank Dr. Damian Fullston for useful suggestions relating to the area available for filament growth form a curved scribe.

References

(1) Corrosion of Aluminum and Aluminum Alloys, Davis, J.R. (Ed.), ASM International, Chapt. 3 and 4, 1999.

(2) Pitt, S. and Jones, R., Eng. Failure. Anal., 4 (4), 237 (1997).

(3) Liao, M., Bellinger, N.C., and Komorowski, J.P., Int. J. Fatigue, 25, 1059 (2003).

(4) Lepine, B.A. and Holt, R.T., Can. Aero. Space J., 43 (1), 28 (1997).

(5) Terada, H., Int. J. Fatigue, 23, S21 (2001).

(6) Hagemaier, D. and Kark, G., Mat. Eval., January, 25 (1997).

(7) Hagemaier, D. and Nguyen, K., Mat. Eval., January, 91 (1994).

(8) Thompson, J.C., Mat. Eval., December, 1398 (1993).

(9) Hagemaier, D., Mat. Eval., May, 682 (1982).

(10) Brown, A.S., Mat. Perf., September (1992).

(11) Schmitt, G.F., SAMPE J., 43 (1), January/February (1998).

(12) Hahin, C. and Buchheit, R.G., "Filiform Corrosion," in Corrosion: Fundamentals, Testing and Protection, ASM International, Vol. 13A, p. 248, 2003.

(13) Buchheit, R.G. and Hughes, A.E., "Chromate and Chromate-Free Coatings," Section 4b American Society for Materials, Materials Park, OH, Vol. 13A, Corrosion: Fundamentals, Testing and Protection, pp. 720-736, 2003.

(14) Lytle, F.W., Greegor, R.B., Bibbins, G.L., Blohowiak, K.Y., Smith, R.E., and Tuss, G.D., Corros. Sci., 37 (3), 349 (1995).

(15) Hughes, A.E., Taylor, R.J., and Hinton, B.R.W., Surf. Interface Anal., 25, 223 (1997).

(16) Hughes, A.E., Taylor, R.J., Nelson, K.J.H., Hinton, B.R.W., and Wilson, L., Mater. Sci. Tech., 12, 928 (1996).

(17) Moffitt, C.E., Wieliczka, D.M., and Yasuda, H.K., Surf. Coat. Technol., 137, 188 (2001).

(18) Mol, A.J., Hughes, A.E., Hinton, B.R.W., and van der Zwaag, S., Corros. Sci., 46, 1201 (2004).

(19) Mora, N., Cano, E., Polo, J.L., Puente, J.M., and Bastidas, J.M., Corros. Sci., 46, 563 (2004).

(20) Ardelean, H., Fiaud, C., and Marcus, P., Mater. Corros., 52, 889 (2001).

(21) Aramaki, K., Corros. Sci., 46, 1565 (2004).

(22) Hinton, B.R.W. and Wilson, L., Corros. Sci., 29, 967 (1989).

(23) Aballe, A., Bethencourt, M., Botana, F.J., Cano, M.J., and Marcos, M., Mater. Corros., 53, 176 (2002).

(24) Dabala, M., Ramous, E., and Magrini, M., Mater. Corros., 55, 381 (2004).

(25) Arnott, D.R., Hinton, B.R.W., and Ryan, N.E., Mat. Perform., 26, 42 (1987).

(26) Hinton, B.R.W., Arnott, D.R., and Ryan, N.E., Mat. Forum, 9, 162 (1986).

(27) Bethencourt, M., Botana, F.J., Calvino, J.J., Marcos, M., and Rodriguez-Chacon, M.A., Corros. Sci., 40, 1803 (1998).

(28) Fahrenholtz, W.G., O'Keefe, M.J., Zhou, H., and Grant, J.T., Surf. Coat. Tech., 155, 208 (2002).

(29) Rivera, B.F., Johnson, B.Y., O'Keefe, M.J., and Fahrenholtz, W.G., Surf. Coat. Technol., 176, 349 (2004).

(30) Wang, C., Jiang, F., and Wang, F., Corrosion, 60, 237 (2004).

(31) Campestrini, P., Terryn, H., Hovestad, A., and de Wit, H., Surf. Coat. Technol., 176, 365 (2004).

(32) Schmitt Handsberg, Th. and Schubach, O., ATB Metallurgie, 43 (1,2), 9 (2003).

(33) Hughes, A.E., Hardin, S.G., Harvey, T.G., Nikour, T., Hinton, B., Galassi, A., McAdam, G., Stonham, A., Harris, S.J., Church, S., Figgures, C., Dixon, D., Bowden, C., Morgan, P., Toh, S.K., McCulloch, D., and Du Plessis, J., ATB Metallurgie, 43 (1,2), 264 (2003).

(34) Hughes, A.E., Hardin, S.G., Wittel, K., and Miller, P.R., "Accelerated Cerium-Based Conversion Coatings," NACE 2000, Orlando, FL, March 27-31, 2000; Corrosion 2000 Research Topical Symposium--"Surface Conversions of Aluminum and Ferrous Alloys for Corrosion Resistance," NACE International (Houston, TX), 47-66, 2000.

(35) Hughes, A.E., Taylor, R.J., and Hinton, B.R.W., Surf. Interface Anal., 23, 540 (1995).

(36) Hughes, A.E., Gorman, J.D., Miller, P.R., Sexton, B.A., Paterson, P.J.K., and Taylor, R.J., Surf. Interface Anal., 36, 290 (2004).

(37) Hughes, A.E., Mol, A.J., Hinton, B.R.W., and van der Zwaag, S., Corros. Sci., 47, 107 (2005).

(38) Hughes, A.E., Nelson, K.J.H., Taylor, R.J., Hinton, B.R.W., Henderson, M.J., Wilson, L., and Nugent, S.A., "Metal Treatment with Acidic, Rare Earth Ion Containing Cleaning Solution," EU Patent No. 0719350.

(39) Hardin, S.G., Nelson, K.J.H., Wittel, K., and Hughes, A.E., "Process and Solution for Providing a Conversion Coating on a Metallic Surface." U.S. Patent 6,503,565.

(40) Mol, A., "Filiform Corrosion of Aluminum Alloys. The Effect of Microstructural Variations in the Substrate," Ph.D. Thesis, Laboratory of Materials Science of the Delft University of Technology, The Netherlands, 2000.

(41) Leth-Olsen, H. and Nisancioglu, K., Corrosion, 53 (9), 705 (1997).

(42) McMurray, H.N., Williams, G., and O'Driscoll, S., J. Electrochem. Soc., 151 (7): B406-B414 (2004).

(43) McMurray, H.N. and Williams, G., Corrosion, 60 (3), 219 (2004).

(44) Hughes, A.E., Nelson, K.J.H., and Turney, T.W., "Process and Solution for Providing a Conversion Coating on a Metal Surface," EU Patent No. 0804633.

(45) LeBozec, N., Persson, D., Thierry, D., and Axelsen, S.B., Corrosion, 60, 584 (2004).

(46) Scamans, G.M., Asfeth, A., Thompson, G.E., and Zhou, X., Proc. 3rd International Symposium of the Aluminum Surface Science and Technology, ATB Metall., 43 (1,2), 90 (2003).

(47) Leth-Olsen, H. and Nisancioglu, K., Corros. Sci., 40 (7), 1179 (1998).

(48) Slabaugh, W.H., Dejager, W., Hoover, S.E., and Hutchinson, L.I., "Filiform Corrosion of Aluminum," J. PAINT TECHNOL., 44, No. 566, 76 (1972).

(49) Nelson, K.J.H., Hughes, A.E., Taylor, R.J., Hinton, B.R.W., Wilson, L., and Henderson, M., Mater. Sci. Technol., 17 (10), 1211 (2001).

(50) Caicedo-Martinez, C.E., Koroleva, E.V., Thompson, G.E., Skeldon, P., Shimizu, K., Habazaki, H., Hoellrigl, G., Smith, G., Flukes, G., and Foord, D.T., Proc. 3rd International Symposium of the Aluminum Surface Science and Technology, ATB Metall., 43 (1,2), 301 (2003).

(51) Martyak, N.M., McCaskie, J.E., Hulsmann, T., and Schroer, D., Interfinish 96, 14th World Congress, Institute of Metal Finishing, Vol. 3, p. 229, Birmingham, September 12-14, 1996.

(52) Kendig, M.W. and Buchheit, R.G., Corrosion, 59 (5), 379 (2003).

(53) Kaesche, H., "Pitting Corrosion of Aluminum and Intergranular Corrosion of Al-Alloys," in Localized Corrosion, 3, NACE International, Houston, TX, p. 516, 1971.

(54) Sehgal, A., Frankel, G.S., Zoofan, B., and Rokhlin, S., J. Electrochem. Soc., 147, 140 (2000).

(55) Aldekwicz, A.J., Isaacs, H.S., and Davenport, A.J., J. Electrochem. Soc., 142, 3343 (1995).

(56) Asfeth, A., Nordlien, J.H., Scamens, G.M., and Nisancioglu, K., Corros. Sci., 44, 2529 (2002).

(57) Zhao, J., Frankel, G., and McCreery, R., J. Electrochem. Soc., 145, 2258 (1998).

(58) Poulain, V. and Petitjean, J.-P., Mater. Sci. Forum, 217-222, 1641 (1996).

(59) Aldykewicz, A.J., Davenport, A.J., and Isaacs, H.S., J. Electrochem. Soc., 143, 147 (1996).

APPENDIX

The area of filiform corrosion has been described as the area/unit length of scribe. In Figure A1(a), the maximum area/unit length of scribe can be described most generally by xl where x is the unit length of scribe and l is the length of the filament. Thus, the available area for corrosion to grow into increases in proportion to l.

[FIGURE A1 OMITTED]

For a filament growing from an arc, it can be seen in Figure A1(b) that there are additional wings adjacent to the central rectangle that are available for filament growth; hence, there is more available area for corrosion to develop.

To quantify the amount of additional area, the following approach was adopted. Assume that the distance [r.sub.1] is the radius from the center of the rivet to the scribe and the unit length of the arc at [r.sub.1] subtended by the angle [theta] is x. The area available for filament growth is the difference between the areas of the two sectors defined by [r.sub.1] and ([r.sub.1] + l).

A = [[theta]/2]{(1 + [r.sub.1])[.sup.2] - [r.sub.1.sup.2]}

= [[theta]/2]{[l.sup.2] + 2l[r.sub.1]} (A1)

Knowing that the length of the sector x = [r.sub.1][theta] and substituting this into equation (A1) and dividing by lx (area available for a linear scribe), the equation for the ratio of the area

[area x available x for x an x arc]/[area x available x for x line] = 1 + [l/2[r.sub.1]]

for a linear scribe to an arc of the same length is obtained.

S. Intem -- Ecole Polytechnique de l'universite de Nantes*

A.E. Hughes, A.K. Neufeld, and T. Markley -- CSIRO Manufacturing & Materials Technology ([dagger])

A.M. Glenn -- CSIRO Minerals**

* Rue Christian Pauc, BP 90604, 44306 Nantes, France.

([degger]) Corrosion Science and Surface Design, Locked Bag 33, Clayton Sth MDC, Clayton, Australia, 3169.

** Box 312, Clayton South, Australia, 3169.

Table 1 -- Conversion Coating Processing Conditions

Process                    Time (sec)  Temperature ([degrees]C)

Acetone wipe                --         --
Alkaline clean             600         60
Rinse 1 (distilled water)   60         15
Deoxidation                600         21
Rinse 2 (distilled water)   60         15
Conversion coating         120         21(CrCC), 45(CeCC)
Rinse 3                     60         15
Dry                         24+(hr)    21

Table 2 -- Surface Treatment and Exposure Time for Each Sample

                  Surface           Total Exposure  Photograph Frequency
Sample  Fastener  Treatment         Time (hr)       (days)

 1      Cu        Alkaline-cleaned  1000            2
 2      Cu        Deoxidized        1000            2
 3      Cu        CeCC              1000            2
 4      Cu        CrCC               250            2
 5      Cu        CrCC               500            2
 6      Cu        CrCC               750            2
 7      Cu        CrCC              1000            2
 8      Cu/Ni     Alkaline-cleaned  1000            2
 9      Cu/Ni     Deoxidized        1000            2
10      Cu/Ni     CeCC              1000            2
11      Cu/Ni     CrCC               250            2
12      Cu/Ni     CrCC               500            2
13      Cu/Ni     CrCC               750            2
14      Cu/Ni     CrCC              1000            2

Table 3 -- Initial Site Density, Corrosion Number, and Average Filament
Lengths

                                         Corrosion       Average
                Initial Site  Corrosion  Number          Filament
                Density       Number     1000 hr         Length (cm)
Treatment       (number/mm)   1000 hr    ([Mol.sup.16])  (this Study)

Alkaline clean  10.8          6.48       7.62            6
Deoxidation     11.9          2.50       --              --
CeCC            11.1          2.33       2.78            2.05
CrCC            11.1          2.55       2.19            2.3

                Average
                Filament
                Length (cm)
Treatment       ([Mol.sup.16])

Alkaline clean  5.40
Deoxidation     --
CeCC            3.33
CrCC            2.14