Strength Degradation in AA5083 Armor Plate After Exposure to Elevated Temperatures

Since as far back as the Vietnam War, weight-sensitive combat vehicles have sought ballistic protection using high strength aluminum alloys [1]. While there are a number of aluminum alloys qualified for use in military armor plate applications, the 5xxx series are particularly widespread. This class of non-heat treatable alloys has a combination of favorable properties that continue to make them appealing for armor applications: high strength (among the lightweight alloys), wrought plate availability, weldability, and good corrosion resistance.

Aluminum alloys in the 5xxx series contain appreciable amounts of substitutional magnesium in its crystal structure which, in combination with dislocation structures introduced during cold working, impart its high strength. Those alloys with greater than 3 wt% Mg, such as 5083, 5059, and 5456, are supersaturated at room temperature so that, over time, excursion above temperatures as low as 60 °C has shown to provide enough thermal energy for the diffusion of Mg atoms and nucleation of β-phase as Al₃Mg₂. The loss of solutionized Mg and attendant decrease in the lattice strain energy diminishes strength and work hardening mechanisms.

Upon separation of the Mg solute from solution, the location of β-phase nucleation is determined by a number of factors, including the thermo-mechanical processing pathway used during production. Much research attention has been given to grain boundary β-phase, referred to as “sensitization,” because it is the precursor to intergranular corrosion and intergranular stress corrosion cracking. This is particularly important for applications in corrosive environments where fatigue loading may be the dominating factor, as is the case in marine shipbuilding applications. Precipitation of β-phase away from the grain boundaries, so-called matrix β-phase, has received less attention yet its influence in strength-dominated applications, such as armor protection, may be critically important.

Here we explore the evolution of β-phase morphology and its impact on the strength of AA5083-H131 rolled plate (armor grade aluminum) from three different commercial aluminum producers. As-received material is subjected to a 100 °C aging heat treatment over a total period of 30 days, a commonly practiced laboratory method for accelerating the effects of naturally occurring solar radiation and ambient temperature exposure. At periodic intervals, the β-phase morphology on the plate’s intended impact surface (i.e., plate “face”) is micrographed and analyzed using computational image processing techniques which is then correlated to strength measurements. Emergent trends are elucidated and discussed for new insight into the potential effectiveness of fielded armor.

Alloy AA5083-H131 rolled plate with thickness of 57.15 mm was acquired from three different aluminum producers (hereinafter referred to as producers A, B, and C). Because it has been previously reported that microstructure varies considerably among suppliers [4], material from multiple mills were sought for this study. Material certificates attesting conformance to MIL-DTL-46027 K [5] for armor grade aluminum via ballistic testing were obtained for each plate. The chemical composition measured via optical emission spectrometry (OES) is reported in Table 1 for producers A, B, and C. Test specimens were harvested according to the locations and orientations shown in Fig. 1.

All specimens underwent isothermal aging in a laboratory oven held at 100 ± 2 °C (with the exception of an “as-received” set retained for analysis) and a time-temperature log was maintained throughout the treatment period. The samples were removed at intervals of 1 day (24 ± 0.5 h), 3 days (72 ± 1 h), 7 days (168 ± 1 h), 15 days (360 ± 24 h), and 30 days (720 ± 24 h) and allowed to naturally cool to room temperature.

Round tension test specimens with a gauge diameter of 12.7 ± 0.254 mm and gauge length of 50.8 ± 0.127 mm were tested in quasi-static tension using a screw-driven testing machine at an applied nominal strain rate of 10⁻⁴ s⁻¹ until failure. Specimens were harvested from the through-thickness center of each plate with their longitudinal axis parallel to the rolling direction, cf. Fig. 1. One specimen was tested at each aging interval. The measured load cell force was used to calculate the nominal stress (using the measured cross-sectional area of the specimen), and a nominal strain was obtained from a specimen strain gauge. Yield strength was obtained using the 0.2% offset method in accordance with ASTM B557-15 [6].

Microscopy specimen LT and LS surfaces were wet-ground successively from 60 grit to a surface finish of 1200 grit with SiC paper, and subsequently polished with polycrystalline diamond suspension successively from 3 μm to a mirror finish of 1 μm. The polished specimens were rinsed with de-ionized water, and dried with clean compressed air.

To study the β-phase microstructure, an etchant was prepared by adding 40 parts by volume phosphoric acid (85% concentration reagent grade) to 60 parts by volume de-ionized water and each specimen was immersed for 3 min, rinsed with de-ionized water, and dried with clean compressed air. The specimens were then imaged with a Nikon Diaphot^™ inverted microscope at 500× magnification (10× magnification eyepiece lens multiplied by 50× objective lens). Images were taken at three different locations on the LT surface.

To reveal grain morphology, specimens were exposed to Barker’s etchant which was prepared by adding 5 mL of fluoroboric acid (50% concentration reagent grade) to 200 mL de-ionized water. Each specimen was immersed in the etchant with an applied potential of 30 V from a DC supply for 2–4 min (30 s at a time after an initial 2 min until the grain structure was visible) and were thereafter rinsed with de-ionized water and dried with clean compressed air. The surfaces were then imaged with a Hirox RH-8800^™ microscope equipped with polarized lighting at 200–400× magnification.

Digital image analysis of the phosphoric acid-etched specimen micrographs was performed using the Mathematica^® programming environment. The approach taken here was to measure and partition, on first order, β-phase precipitation into two distinct groups: matrix β dispersoids and grain boundary β-phase networks. To do so, we relied on an automated computer vision algorithm to count the β dispersoids in each image for quantifying matrix β while a human driven process was used to measure the length of β-phase networks on the grain boundaries. We acknowledge that neither process is without some error: dispersoids counted as matrix β may in fact be discontinuous and evolving β-phase at sub-grains or parent grain edges, or human bias may impact the grain boundary measurement, as a few examples. However, the methods developed here show consistency and correlation (as described in Sec. 3) which sufficiently supports a reasonable first-order approximation for the respective matrix and grain boundary β-phase amounts.

To count the number of matrix β precipitates, each 2560 × 1920 pixel field was first transformed to grayscale with pixel intensities scaled over real numbers of the set [0,1] and then converted to a binary image using a threshold of 0.45. The resulting binary was then culled of background noise by removing artifacts with an equivalent disk radius (radius of a disk that has the same area of the subject artifact) of less than 2 pixels and large precipitates with an equivalent disk radius greater than 20 pixels. To confirm that these larger precipitates are not β-phase and therefore excluded from the matrix tally, a scanning electron microscope (SEM) equipped with an energy dispersive X-ray spectrometer was used to acquire composition maps of the LT surface, see Fig. 2. The maps positively identify larger precipitates (10–100 times larger than β dispersoids) as Al(Mn,Fe) intermetallics and void of any significant amount of Mg. The centroids of remaining artifacts were then identified and counted. The resulting image from each process step is shown in Fig. 3. The mean and standard deviation about the mean for three images, each taken at a different location on the LT surface, were computed.

To measure the length of grain boundary β in the image field, an interactive software interface was built to trace user-identified β-phase at the grain boundaries, picking points along the β-phase pathway and tallying the pixel length along straight lines between the points. The cumulative total was then converted to a total measurable length in microns using the scale marker for calibration. An example of a traced image of grain boundary β is shown in Fig. 4. As before, a mean and standard deviation about the mean for three images, each taken at a different location on the LT surface, were computed.

Micrographs of the phosphoric-etched LT surface on specimens from producer B plate are shown in Fig. 5. The aging period is listed in the upper left corner of each image. β-Phase dispersoids are evident in the as-received condition but there is no readily identifiable β-phase network. Large Al(Mn, Fe) precipitates can be seen in the as-received condition indicating that they formed during thermo-mechanical processing of the material at the mill as also reported by others [7]. After just a single day of aging, there is a significant jump in the population of observed spherical precipitates. It is only after a 7-day exposure that there is clear evidence of discontinuous grain boundary β-phase which grows in volume and continuity as seen in the 30-day specimen. A similar microstructural evolution was observed in the micrographs from producers A and C, which is captured in Figs. 7 and 8 (to be discussed).

The stress–strain behavior for aged material is shown in Fig. 6. Test specimens exhibit stress–strain curves with well-defined yield behavior, some degree of strain hardening, and constant material stiffness and elongation throughout aging. The plots also reveal that all three products share an immediate loss in yield and tensile strength after the initial 24 h. Through the duration of the aging period, the stress–strain curves are densely packed, indicating little change in mechanical behavior with the exception of the 30 day sample from producer B.

The immediate loss in yield strength is made clear in Fig. 7 where it is plotted as a function of aging time. After 24 h, losses of 7.7%, 9.7%, and 9.0% occur for producers A, B, and C, respectively. Superimposed on these same figures is the computational tally for the β-phase matrix dispersoids (right hand ordinate axis). The rapid population growth of β-phase after short periods of thermal exposure strongly correlates with the sudden loss of alloy strength. At periods beyond 3 days, all three products once again follow the same general trend: a plateau in both β-phase nucleation and yield strength (with the exception of the 30 day sample from producer B). The majority loss of strength among all three producers is presumably linked to precipitating Mg atoms from the supersatured crystal lattice which otherwise serve as the primary strengthening mechanism for 5xxx series aluminum.

A similar analysis is conducted for grain boundary β. Figure 8 shows the same yield strength data for the three products (left hand ordinate axis) but this time overlaid with the total length of the grain boundary β network per unit area (right hand ordinate axis). Over the same period, the increase in grain boundary β is only gradual, suggesting that the sudden reduction in strength is principally triggered by the nucleation of the matrix dispersoids alone. This is further reinforced by the continued growth of the grain boundary network throughout the entire exposure period even while the strength has plateaued. Again the behavior is consistent in material from all three aluminum producers. The implication is that strength degradation occurs before the alloy develops a sensitized microstructure.

Generally, the higher interfacial energy at grain boundaries should favor nucleation of grain boundary β-phase over precipitation in the matrix. This has been observed in numerous studies on AA5456 and AA5083 with the H116 temper [8–10]. Yet here with the H-131 temper, the opposite appears true: β-phase is seen at both intragranular and intergranular sites, and the rapid nucleation of matrix β is favored over the more slowly evolving grain boundary network. One cause for the discrepant observations between H116 and H131 tempers is the number of potential nucleation sites. Full hard tempers correlate to approximately 75% cold reduction after annealing [11], while the H131 temper refers to 3/8 hard (28% cold reduction) and the H116 temper refers to 1/8 hard (only 9.4% cold reduction). The hardening effect imposed by production is the result of an increase in dislocation density and their mutual interactions [12]. Yi et al. [11] reported a 62% increase in the dislocation density in an H131 sample over an H116 sample.

Dislocation centers can contribute in three ways to β-phase nucleation. First, dislocations serve as sites for heterogeneous nucleation in the matrix. Despite grain, subgrain, and phase boundaries having higher defect energies, classic nucleation theory accounts for the relative influence of these defect sites as well, including the number of sites per unit volume [13]. As a result, the high dislocation density of H131 thermodynamically favors matrix dislocation over that in H116. Second, the high dislocation concentration supports significant pipe diffusion [14] which has been observed by others to accelerate β-phase growth in the matrix of 5xxx series aluminum [15]. Pipe diffusion occurs along diffusivity pathways created by dislocations (and other linear defects) that can be orders of magnitude higher when compared to lattice (or bulk) diffusion [16]. Third, the stored elastic strain energy in the heavily cold-worked material increases bulk diffusion rates of substitutional Mg atoms as well. The combination of enhanced pipe diffusion and lattice diffusion accelerate nucleation and growth at the dislocation structures in the matrix.

Figure 9 shows micrographs taken after Barker’s etching of the LT and LS surfaces. The LS surface reveals elongated grains in the rolling direction with no indications of a recrystallized microstructure near the L-edge which can sometimes be formed by the heat and pressure (stress) of the hot rollers during production. A comparison of the images taken near the L-edge and the through-thickness center shows no significant gradient in grain structure which provides some assurance of the strong correlation between β-phase morphology examined on the LT surface and the behavior of the tensile specimens which were harvested from the material’s center plane. This opens the possibility of determining the condition of a plate from an interrogation at its surface.

The findings here suggest that field exposure to diurnal solar radiation cycles, ambient temperature fluctuations, or vehicle engine/exhaust heat could lead to a loss in strength and therefore ballistic protection without the development of a sensitized condition. Future work should focus on sampling fielded armor to determine the degree of similitude with the accelerated laboratory conditions used here.

It has been found that:

1. AA5083-H131 armor plates exhibit a significant (7–10%) loss of yield strength over the first 24 h period of a 100 °C aging treatment. After 3 days, two of the three plates exhibited a yield strength plateau in which no considerable strength loss was observed over the full 30 day period.

2. During the aging treatments, the growth of matrix β-phase dispersoids was rapid and generally plateaued after 3 days, while the growth of grain boundary β networks occurred gradually over the 30 day period.

3. A loss of yield strength strongly correlates with the rapid emergence of matrix β-phase and not the slowly evolving grain boundary β networks. A sensitized microstructure is not necessary for strength degradation.

4. These trends were observed in the armor plate made by all three of the aluminum producers investigated here.

The authors wish to acknowledge Utibe-Eno Charles-Granville (Ph.D. candidate, University of Virginia) for the sample preparation and metallography.

There are no conflicts of interest.

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.