Initiation and short crack growth behaviour of environmentally induced cracks in AA5083 H131 investigated across time and length scales

2.1 Materials

Commercially available H131 temper AA5083 alloy plates of 29-mm thickness having the composition summarised in Table 1 were used in this study. The composition of the as-received plates was measured using optical emission spectroscopy. The susceptibility to IGC was evaluated using the Nitric Acid Mass Loss Test (NAMLT) according to ASTM G67 (ASTM International, 2018). The DoS value obtained for the as-received samples was 8.6 mg/cm² (Holroyd et al., 2017a). The as-received material had typical pancake-shaped elongated grains along the rolling direction and with an average length (L) of 150 μm and widths of 35 μm (S) and 80 μm (T). The measured short transverse tensile yield stress (YS), ultimate tensile stress (UTS) and plane-strain fracture toughness (KIC) properties were 260 MPa, 375 MPa and 31 M Nm^−3/2, respectively (Seifi et al., 2016).

Cylindrical tensile test samples of 12.7 mm gauge length and 3.2 mm diameter were machined in the short-transverse orientation from the as-received plates for SSRT (Seifi et al., 2016). The samples were then mechanically polished using an SiC abrasive grinding paper of P4000 grade and then ultrasonically cleaned with ethanol followed by warm air drying. In order to enhance the susceptibility of the as-received samples to IGC, an additional sensitization treatment was performed at 80°C for 225 h. This resulted in an increase in DoS (NAMLT value) to ~40 mg/cm². The sensitized samples were further subjected to pre-exposure in a 0.6 M NaCl solution at RT for a period of 180 h in order to introduce surface corrosion attack, e.g. IGC, as well as charging the sample with hydrogen. The surface corrosion features are expected to act as realistic stress raising sites during SSRT (Holroyd et al., 2017a). Following pre-exposure, the gauge lengths of tensile specimens were rinsed, carefully dried and immediately subjected to SSRT. The regions of the sample gauge length on either side of the central region were masked with Parafilm™ (Bemis Company, Inc.). This helped to prevent failure in the shoulder region of the SSRT samples and also provided a window of environmental exposure commensurate with the field of view (FOV) of the X-ray imaging.

2.2 SSRT

SSRT of the samples were performed at a nominal strain rate of 4×10⁻⁵ s⁻¹ using a Deben CT5000 5kN mechanical test rig (Deben UK). The test was performed by straining the sample up to UTS at this strain rate, after which the displacement was fixed to provide constant strain until final failure of the sample occurred. Within the constant strain rate regime of the SSRT prior to achieving UTS, the straining was paused intermittently to allow XCT data acquisition taking ~90 s to avoid image blurring. This results in a stepwise increment in the applied load on the sample. A humid atmosphere around the sample during testing was achieved by placing wet sponges at the base (~20 mm away from the sample gauge length) of the vitreous glassy carbon test chamber, giving a humidity level of ~70% relative humidity at RT.

2.3 Synchrotron 3D XCT

Synchrotron XCT was conducted at the I13-2 imaging beamline at Diamond Light Source, UK. A “pink beam” (Rivers, 2016) was used for imaging to maximize the flux around a nominal energy of 28 keV. The detector provided a 4.2 mm×3.5 mm FOV with a (1.63 μm)³ voxel size. The exposure time was 0.035 s for each of the 1200 projections, totalling up to ~90 s per tomogram acquisition. A total of 43 tomograms were collected in situ over a period of 13,260 s (3 h, 41 min) of testing in addition to those collected prior to and after testing. After an initial low-noise XCT scan with a 0.09s exposure time, scans were collected at regular intervals until the yield strength was exceeded and then recorded continuously once the UTS was reached. Further, an additional tomogram with an exposure time of 0.09 s was acquired after testing. Reconstruction was performed using the Gridrec algorithm (Dowd et al., 1999) of the TomoPy software (Gürsoy et al., 2014) module within a Savu pipeline (TomoPy – TomoPy 1.1.3 documentation, n.d.).

2.4 Analysis methodology

Avizo™ software (ThermoFisher Scientific; Thermo Fisher Scientific, 2018) was employed to visualise and analyse the reconstructed XCT data. The data from the time step just prior to failure were first utilized to identify the full extent of cracking that had developed during SSRT within the FOV. The individual cracks were then numbered and tracked back through the time sequence using the time-lapse tomograms to identify the location and time step of origin. The point of initiation in time was identified by checking for the presence of the crack in a specific tomogram and confirmed only if it was measurable to at least three voxels in dimension, i.e. at least 5 μm and clearly visible as a change from the previous time step. The temporal resolution of identifying the time of initiation is limited in accuracy by the time elapsed between the two tomograms in which the crack had initiated and was large enough to be confidently measured. The crack length was always measured as the depth which the crack penetrates radially inward toward the sample centre from the surface initiation site. Topographical features of the crack (i.e. variations in height) are excluded in these measurements as the cracks are generally very flat. The length of all the identified cracks was measured for each of the collected tomograms. This measurement protocol for crack lengths was semi-automated using the scripting capability within the Avizo™ software module.

2.5 Post-mortem characterization

The morphology of the fracture surface of the sample after SSRT was investigated using an FEI Quanta 650 field emission gun scanning electron microscope coupled with an energy-dispersive X-ray spectroscopy capability. Through correlation with the XCT data, a specific crack was identified in the scanning electron microscopy (SEM) for further examination using focused ion beam serial section tomography. The volume of interest for this analysis was chosen from the XCT data volume rendering and was identified in the SEM by correlating the corresponding surface corrosion features and distance from the fracture surface on the sample gauge length. A volume with approximate dimensions 175×175×200 μm³ containing the majority of the EIC crack of interest was lifted out using an FEI Helios Xe⁺ plasma focused ion beam (PFIB); further details of this technique can be found elsewhere (Burnett et al., 2017; Slater et al., 2017; Gillen et al., 2018). After extraction, the block was attached to a Cu transmission electron microscopy grid. The grid holding the block was subsequently mounted onto a pre-tilted holder, and serial sectioning was performed using the PFIB operated at 30 keV with a current of 59 nA with the FEI Auto Slice and View 4 software. The slice thickness was chosen to be 300 nm in order to capture the finer details of the crack path and to enable chemical mapping of small precipitates. A rocking mill of 9° was also used to reduce curtaining effects during milling, which can be exaggerated by the presence of the crack in the milling direction. After each slice, simultaneous electron backscatter diffraction (EBSD) and energy-dispersive spectroscopy (EDS) acquisitions were performed with a step size of 250 nm using the electron beam operated at 20 keV and 22 nA using the Oxford Instruments Aztec software. In addition, at each slice, a secondary electron (SE) image was collected in order to image the crack path and correlate to the crystallographic and chemical data. The total volume of analysed material was 175×175×80 μm³ captured in ~300 slices.

3.1 SSRT

The engineering stress vs. the relative strain from the SSRT of the sample tested is shown in Figure 1A. As can be seen, at around 325 MPa (close to UTS), the test was switched to constant strain or fixed displacement mode from the SSRT mode. Initially, the stress increases linearly with strain followed by yielding of the sample at ~261 MPa. After switching to fixed displacement mode at ~2.2% plastic strain, the stress across the sample drops from approximately 325–225 MPa over a period of just over 4500 s as the EIC cracks grow, leading to final failure. The stress vs. time during the fixed displacement regime of the SSRT is shown in Figure 1B. The drop in stress prior to failure was not immediate but proceeded in a slow downward trend coupled with several small load drops.

Figure 1:

(A) Engineering stress vs. relative strain response of the AA5083 sample, showing the region of switch from SSRT to fixed displacement testing, and the final failure. From 0 to ~3% relative strain, the plot displays a region showing the slack of the SSRT system. (B) Engineering stress and relative strain vs. test time showing the region of constant strain or fixed displacement highlighting the load drops observed until final failure.

Figure 2A–D shows the sample just prior to failure and shows features such as the corrosion sites due to pre-exposure (covered by corrosion products in some locations) as well as the catalogued (numbered) EIC cracks, while the route of the final failure is shown in Figure 2E.

Figure 2:

(A–D) Volume renderings of the gauge length of the sample at four different viewing angles (θ) obtained at a time step just prior to failure showing all the 44 EIC cracks that were numbered and investigated, and (E) volume rendering of the sample gauge length after failure showing the relative location of the specific EIC crack marked “3” on the sample gauge length that was subject to further characterisation. The protective Parafilm™ used to prevent cracking outside the XCT FOV is also visible at the bottom end of the image.

The number of EIC cracks measured from the XCT data acquired just prior to failure is shown in Figure 3A as a function of the crack length. The engineering stress and plastic strain on the sample at this stage are ~245 MPa (post UTS) and ~2.2%, respectively. The number of EIC crack initiation events and the total number of cracks as a function of SSRT test time obtained from the time-lapse XCT data are presented in Figure 3B. Crack initiation only starts at around 215 MPa (i.e. below yield) or after about 3700 s from the start of straining. The number of initiation events recorded gradually increases beyond 0.2% yield strength of the sample and drops sharply when the loading is changed to fixed displacement mode. Indeed, only four additional initiation events were observed after the switch to fixed displacement mode (as seen in Figure 3B), although it is possible these cracks also initiated in the constant strain rate region and only grew large enough to be resolved by the XCT imaging slightly later. A total of 44 cracks were indexed and tracked through the test time, and their size, location and time of initiation were recorded within the spatial and temporal resolution limits. Of the 44 cracks that were indexed using the time-lapse XCT data, five minor cracks (<50 μm) and six major cracks (>50 μm, can be identified on the fracture surface in Figure 4A) contributed to the final failure. Further, it was found that of all the 44 cracks that were indexed, 42 of them initiated from a surface feature corresponding to intermetallic phases or corrosion sites based on the proximity to the associated crack.

Figure 3:

(A) Distribution of crack lengths in the sample measured from XCT data acquired just prior to overload failure (bin size=50 μm), and (B) EIC crack initiation events per time step, and total number of cracks as a function of test time. All data measured from the reconstructed XCT images.

Figure 4:

SE SEM images showing (A) overview of the fracture surface with IGSCC (type 1, dotted red lines), type 2 cracking (solid red lines) and micro-void coalescence (MVC; region within the dotted orange lines) along with the specific identified EIC crack numbers correlated from XCT data. (B) A closer look at the IGSCC (crack 15) fracture surface showing the region of initiation at a corrosion site (dotted green lines). (C) Backscattered SEM image of region shown in (B) showing the bright-appearing cluster of intermetallic phases at the surface corrosion site.

The fracture surface in Figure 4A shows three major distinct regions displaying features corresponding to the IGSCC region (type 1 fracture), type 2 EIC and micro-void coalescence (MVC) in addition to the surface corrosion sites. Thirty-five percent to 40% of the fracture surface is identified as IGSCC, which displays a flat morphology and appears “shiny” when viewed optically. Further, it is observed that the IGSCC regions are followed by type 2 EIC cracking features characterized by their stepped topography and very large and shallow dimple-like features containing patches of particles (Holroyd et al., 2017a). It is also observed that in all the instances, type 2 EIC cracking is always preceded by type 1 IGSCC. The total IGSCC area (type 1) and type 2 EIC area comprise approximately 60% of the fracture surface when viewed in plan view. The remaining 40% of the fracture surface is attributed to MVC that occurred during the final stages of fast fracture and is characterized by the steep topography. Most of the IGSCC sites found on the fracture surface could be tracked back to features corresponding to surface corrosion sites. An example of this is presented in Figure 4B (SE image) and Figure 4C (corresponding backscattered electron image), where a cluster of intermetallic particles (confirmed by SEM EDS to be AlFeMnSi based) at the surface of the sample gauge length have formed a corrosion site during the pre-exposure treatment in the NaCl solution. This corrosion site is followed by IGSCC (type 1 EIC cracking).

By considering the crack length vs. time data during the fixed displacement regime of testing, it was observed that the velocity of some of the cracks reduced. One such EIC crack belonging to this subset (crack 3) was investigated in further detail. This specific crack initiated in the constant strain rate part of the test at a stress value of approximately 215–245 MPa, i.e. below the yield strength. The crack had a length of ~258 μm (the 5th largest) when the fixed displacement testing began and appeared to be still growing (the global average length of all the cracks at this stage was approximately 80 μm). At the end of the test, just prior to failure, the length of crack 3 increased by only 10% to ~283 μm (11th largest overall), while the average crack length had increased by 250% (to ~280 μm). By comparison, the longest crack had grown from 497 to 1447 μm. As it is clear that this specific crack had grown considerably in the constant strain rate mode of the test, but almost stopped growing completely in the fixed displacement regime, it was of interest to investigate this crack further in detail to try and understand how it initiated and grew and why it stopped. This crack did not contribute to the final failure as seen in Figure 2.

Figure 5 shows the XCT surface rendering as well as the corresponding SE image of the corrosion site from where the crack 3 initiated. In Figure 5A, cracking can be seen to originate from the edge of one of the surface corrosion sites, which can be seen more clearly in the SE image in Figure 5B. Four distinct surface corrosion sites are visible. In addition, the SE image also shows corrosion products on the sample surface and general corrosion of the surrounding Al matrix, which is beyond the spatial resolution of the XCT images. EIC cracking can be clearly observed connecting to the edge of one of the corrosion site in the SE image in Figure 5B.

Figure 5:

(A) Volume rendering of the sample gauge length surface obtained from the XCT data of the sample showing the region of interest where EIC crack 3 (marked with black arrow) initiated, and (B) the corresponding SEM SE image of the surface showing the corrosion sites associated with intermetallic particles and the EIC crack 3 (marked with white arrow) initiating from one of these sites.

In order to better visualise the growth of the crack, the 3D XCT volume of the crack was segmented and visualised at various stages of the test as shown in Figure 6. It is clear that there were no major changes to the crack length and morphology under fixed displacement conditions, indicative of minimal EIC propagation for this crack.

Figure 6:

3D volume renderings of the crack 3 generated by segmenting the XCT data of the crack volume during SSRT (blue crack t=8460 s) at the transition to constant displacement (green crack t=9480 s) and after a period of constant displacement just before the final failure of the specimen (orange crack t=13,080 s).

The crack length and the relative strain on the sample vs. the test time are shown in Figure 7 in addition to selected XCT virtual slices of crack 3 shown at different time steps. The initiation of crack 3 occurred from a 30 μm feature (to be discussed later) between t=3800 and 4500 s (Figure 7). The crack appears to grow visibly in the early stages before growing quite abruptly. The virtual slices of the cross section of the crack from the time-lapse XCT data shown in Figure 7B–J reveal the growth as well as the changes in path of the crack before appearing to become dormant in the fixed displacement regime. Last, a small slow increment of growth is observed just before sample failure. The highly branched nature of the crack is observed in Figure 7F–J. The stepwise increase in the strain is due to the intermittent loading of the sample that was employed prior to yielding to allow steady-state conditions during XCT acquisition. Beyond a plastic strain value of ~2% (t=~5500 s), the first instance of fixed displacement was applied. During this period, no crack growth was observed in the sample after over 1 h; therefore, it was decided that the load at fixed displacement was too low and displacement was increased in a further step (at t=~8000 s). Following this, crack growth occurred registering both in the mechanical response of the sample and the raw 2D X-ray projections. In the subsequent analysis of the 3D data, a rapid increase in the crack length is evident. A further increase in strain (at t=~9000 s) was followed by another increase in crack length. This step-up in load was performed to allow growth of the crack under subsequent fixed displacement mode in a reasonable amount of test time. During the fixed displacement stage (t>~9000 s), it was observed that the crack did not propagate for a period followed by a small, slow increment in length. The largest growth occurred during the reloading step at t=~8000 s. In addition, most of the instances of change in length of the crack can be correlated to the change in strain on the sample due to applied load change. The implications of this method of testing will be discussed later.

Figure 7:

(A) Length of crack 3 measured from the time-lapse XCT data and calculated relative strain on the sample vs. test time showing – 0.2% yield strength – 275 MPa, initiation window of the crack 3, UTS – 325 MPa and the initial IGC site depth. (B–J) XCT virtual slices of the sample showing the crack at various stages of the test marked in (A). SD, straining direction.

Post-mortem characterization of crack 3 was conducted in order to better understand the interaction of this crack with the microstructure. EDS elemental and EBSD mapping were employed to relate the crack path to the local grain orientations and the nature of any local second phases. Selected SEM SE images corresponding EDS O K_α1 maps and EBSD inverse pole figure Z maps from this analysis are presented in Figure 8.

Figure 8:

(A) XCT virtual slice (filtered) showing initiation region and morphology of crack 3. (B) SEM SE surface view of the corrosion site from where the crack initiated showing the locations (marked in red) of the presented PFIB cross sections (C–F) from which SE images, EDS and EBSD maps were acquired. Slice (C) corresponds to the XCT virtual slice in (A).

The local microstructure and chemistry of the region neighbouring crack 3 are shown in Figure 8 via four PFIB sections. The SEM SE images in Figure 8C and D appear to show the crack originating from the bottom of the corrosion site, while that in Figure 8E shows that the crack has followed a GB which is at a ~45° angle to the surface. Some of the sharp changes of direction of the crack near the surface were initially thought to be tears through ligaments between different crack regions. However, the corresponding EBSD orientation maps show that in all instances, the crack is propagating along grain boundaries, as expected due to the IGSCC or type 1 nature of EIC. Close to the initiation site, the crack path is more tortuous as it follows the grain boundaries, but as the crack develops, it becomes more planar as it follows the boundary of the large (red) grain that extends almost throughout the entire volume to the right of the crack in Figure 8E and F. It appears that this large grain leads an interconnected high-angle GB with a number of neighbouring grains across the entire 3D volume, providing a direct route for crack propagation in the most mechanically favourable direction. The highly disjointed nature of these first increments of crack growth and the fact that further along the length of the crack (deeper into the material) the crack opens up lead us to believe that several very small cracks originally initiated from the particle cluster. It is surmised that with further growth, these small cracks then coalesced, leading to the opening of the crack and continuation as a single larger crack.

A total of 11 grain boundaries were analysed across these slices, out of which 10 are observed to be high-angle grain boundaries (misorientations ranging from 23° to 53°). However, one cracked GB was observed to be a low-angle boundary (~8° misorientation) despite the fact that low-angle grain boundaries (misorientation <15°) are generally considered as resistant to intergranular attack (Yuan et al., 2001; Davenport et al., 2006; Tan and Allen, 2010; Scotto D’Antuono et al., 2014; Seifi et al., 2015). Post-mortem analyses in another work (Seifi et al., 2015) also showed a propensity for EAC along high-angle grain boundaries in sensitized Al-Mg. The susceptibility of particular boundaries to attack can also be related to the propensity to induce β-phase precipitation, which determines whether precipitation along boundaries is continuous or discontinuous (Davenport et al., 2006). However, it was not possible to resolve any β-phase at grain boundaries during the serial sectioning experiment due to their estimated small size (<50 nm) (Scotto D’Antuono et al., 2014).

The observations made from the 3D volume renderings and the subsequent SEM characterization show that the pre-exposure treatment in NaCl solution has resulted in corrosion of the surface of the AA5083 sample. This material is known to contain phases such as α-AlFeMnSi and Mg₂Si in addition to β-Al₃Mg₂. Localised corrosion is focused around these particles, resulting in pitting or IGC (Oguocha et al., 2008). The α-AlFeMnSi phases are cathodic or electrochemically less active compared to the Al matrix, while the Mg₂Si and β-Al₃Mg₂ are anodic or electrochemically more active to the surrounding Al matrix (Buchheit, 1995; Birbilis & Buchheit, 2005). The cathodic intermetallic phases support the cathodic reactions and lead to the anodic dissolution of the surrounding matrix, leading to trenching and in some cases fall out of the intermetallic phase due to mechanical debonding. On the other hand, the anodic intermetallic phases in this material preferentially dissolve and cause pitting in the Al matrix and, if present at GBs, lead to preferential GB dissolution and intergranular attack. In the current investigation, most of the corrosion sites resolved on the sample surface could be identified to be associated with the cathodic intermetallic phases, and as mentioned earlier, 42 of the 44 EIC cracks initiated at these features, which are clearly visible as bright particles connected to the crack at the sample surface. As well as being active corrosion sites, it is clear that these sites are creating the conditions to initiate EIC cracks, although it is only during the subsequent exposure to the stress and humid air that these cracks actually initiate and start growing.

The observations from the PFIB cross sections of crack 3 show that this specific crack may have originated at the bottom of the corrosion site associated with the cathodic intermetallic phases or at the surface next to this corrosion site at the GB as seen in Figure 8E. Within the spatial and temporal resolution of the current XCT acquisition parameters, it is however not possible to determine at which location cracking first started. However, considering the sensitization treatment performed to the samples, the presence of the β-Al₃Mg₂ phase is expected at grain boundaries (Searles et al., 2001; Birbilis et al., 2013; Seifi et al., 2016; Zhang et al., 2016), which in the vicinity of the α-AlFeMnSi intermetallic phases in the corrosion site is expected to cause galvanic coupling, further enhancing corrosion propagation at the adjacent grain boundaries under humid conditions. This, under applied external load, leads to type 1 EIC cracking. Further, studies performed on environmentally assisted fatigue crack propagation show that in the majority of the instances, in the presence of a corrosion pit, EIC initiation occurs at the edge of the corrosion site on the surface of the sample and not at the bottom of the corrosion site (Turnbull, 2017). This is postulated to be due to the local increase in strain and also easier availability of oxygen from the atmosphere close to the sample surface, than at the bottom of the corrosion site. In short, the easy access to oxygen, combined with the enhanced galvanic coupling between the cathodic α-AlFeMnSi particles and the anodic β-Al₃Mg₂, provide a high probability for EIC initiation. In view of this, it seems most likely that this specific EIC crack, though observed to have originated from the bottom of the corrosion site from the XCT data resolution limits, could have actually initiated from the edge of the corrosion site on the sample surface and then propagated inward as seen in Figure 8 (Turnbull, 2017) to run along the entire outside edge of the intermetallic corrosion site. Once, one or more such EIC cracks have initiated (i.e. proto-cracks) (Staehle, 2011), they can grow and coalesce, thus providing sufficient length and driving force for the crack to enter the sustained propagation stage.

In summary, the pre-exposure of the sample has created small corrosion sites which provide preferential sites for EIC nucleation. Initiation starts below the yield strength and continues past the UTS but abruptly stops when constant strain rate switches to fixed displacement. However, growth of a single crack analysed in detail propagates at the highest velocity when the loading regime changes from fixed displacement to constant strain rate. Detailed analysis of a single crack revealed the site of initiation to be a α-AlFeMnSi particle cluster from which it appeared several small proto-cracks emerged, which later coalesced to form a single crack which then propagated to around 300 μm in length.