Evolution of dislocation substructures in metals via high-strain-rate nanoindentation Authors: Yuwei Zhang, Benjamin L. Hackett, Jiaqi Dong, Kelvin Y. Xie, George M. Pharr Published: 2023 ABSTRACT SignificanceDevelopment of impact-resistant materials is crucial for numerous applications, including space exploration and the advancement of weapon and armor systems. We utilize a unique nanoindentation system capable of generating high strain rates along with advanced electron microscopy techniques to investigate the microstructural evolution in metals. This unlocks the potential of employing high-strain-rate nanoindentation as a valuable tool for investigating the high-strain-rate responses of small features (thin films, heterogeneous structured material, interfaces, etc.) that are challenging to test using conventional high-strain-rate platforms but also offers a promising approach toward accelerating high-strain-rate testing and significantly reducing the associated costs. To that end, we also discuss some of the conditions under which micro- and macro-scale tests may and may not be comparable. Material scientists commonly consider design of materials based on their performance under quasistatic loading at strain rates in the range 10−4 to 101 s−1. However, it is well established that material behavior can change as the deformation rates approach ballistic impact conditions, i.e., strain rates >103 s−1 (1). This presents a challenge in designing impact-resistant materials, which are needed for a variety of important applications such as space exploration and improved weapon and armor systems. As such, there is a need to better understand the behavior of these materials in order to engineer the next generation of materials that will fulfill these applications. While machine-learning aided materials informatics promises to dramatically accelerate materials design and discovery (2), validating materials characterization and performance through standard high-strain-rate testing techniques, i.e., plate impact (3), split Hopkinson bar (4–6), and gas gun impact (7), still remains a bottleneck in the high-strain-rate materials design cycle due to the low throughput nature and high cost of these testing techniques. In this regard, high-strain-rate nanoindentation has recently emerged as a potentially useful high-throughput testing technique (8), but it remains unclear whether small-scale measurements adequately capture material deformation behavior at macroscopic scales (9–12). Many aspects of the mechanisms of deformation at small scales have not been explored, especially at high strain rates (13–21). At bulk scales, lattice defects such as dislocations and twins accommodate the majority of the plastic response in crystalline metals (22). These defects and resulting substructures generally evolve from pre-existing defects in the material (23). At high strain rates, plastic deformation is often associated with an increase in dislocation density for FCC (face-centered cubic) metals (e.g., Cu, Al, Ni) (24–32) and an increase in dislocation velocity for BCC (body-centered cubic) metals (e.g., Nb, Ta) (24, 33, 34). However, at scales which are small compared to the defect density, the mechanisms enabling high-strain-rate deformation have not been established and could be different from those at the macroscale due to, for example, the absence of strength-determining lattice defects such as dislocations in the small volume of material probed and the influences of grain size in polycrystalline materials. In this work, we use nanoindentation and transmission electron microscopy (TEM) with precession electron beam diffraction (PED) to reveal the microstructural evolution in materials deformed at small scales by nanoindentation from very low to very high strain rates, i.e., 10−2 s−1 to >104 s−1. Specifically, we use aluminum and molybdenum single crystals as model FCC and BCC metals with well-known differences in deformation behavior. Based on these results, we compare the conditions under which small- and large-scale testing are similar and different and develop a nanoindentation-based method to quantify the effects of high-strain-rate impact-induced microstructural evolution on subsequent quasistatic re-loading. The nanoindentation-based methods promise to accelerate high-strain-rate testing times and substantially reduce costs. Results and Discussion Instrumentation. Arrays of Berkovich indentations were produced using a specially designed high-strain-rate nanoindentation system based on a modified InForce 1000 nanoindenter actuator head (KLA Instruments, Oak Ridge, TN) (8). As shown in Fig. 1A, the capacitive gauge normally used to measure the displacement was replaced by a laser interferometer with a 10-pm displacement resolution and capable of acquiring data at a rate of 1.25 MHz. The enhanced data acquisition rate increases the time resolution needed to measure indentation depths and loads during high-strain-rate experiments (8). With this modification, the system is capable of operating over a wide range of indentation strain rates ( ε˙=h˙/h , where h and h˙ are the depth and depth rate) ranging from quasistatic tests ( ε˙ ~ 10−4 -10−2 s−1) up to impact tests ( ε˙ > 104 s−1) (8, 9, 35). Quasistatic tests were conducted at constant indentation strain rate ε˙ through feedback control of the loading rate (36). (A) Schematic of the high-strain-rate nanoindentation actuator. (B) Schematic of the high-strain-rate nanoindentation testing procedures. (C) Scanning electron microscope image of a 7-by-7 indent matrix produced through high-strain-rate nanoindentation within 1 h. For the impact tests, the indenter was backed off from the surface of the specimen and then accelerated toward it to achieve maximum velocity (Fig. 1B). As such, the strain rate varied during the course of the test, starting at high values at the beginning, decreasing gradually with an increase in indentation depth, and dropping sharply at the end of the test. A schematic of the procedures used in the impact tests is given in Fig. 1B, where the circled region represents the first loading and unloading of the indenter used to analyze the mechanical properties. In addition, a variation on the impact test referred to here as an “interrupted impact test” was used in some of the aluminum tests to extend the range of strain rates that could be achieved. Details on how the “impact tests” and “interrupted impact tests” were conducted are included in the “Materials and Methods” section. Fig. 1C shows a 7-by-7 matrix of impact indentations that were all produced in less than 1 h in a region of ~0.05 mm2. The load–depth curves and hardness values of the corresponding indents were calculated following procedures developed by Hackett et al. (8). The testing system can probe mechanical properties over 9 orders of magnitude of indentation strain rate in a single testing platform, which helps to strengthen the understanding of deformation mechanisms over a wide range of strain rate. Using focused ion beam sectioning techniques, cross-sections of many of the nanoindentations produced in Al and Mo were extracted for examination of the deformation microstructure by TEM and precession electron diffraction. Mechanical Properties. Arrays of Berkovich indentations were produced with the modified nanoindenter on a (111) single crystal of Al and a (110) single crystal of Mo, with the results summarized in Fig. 2. Fig. 2 A and D display the load and depth curves for the constant strain rate (CSR) tests, high-strain-rate impact tests, and interrupted impact tests (only performed in aluminum to probe the mechanical response larger than a strain rate of 104 s−1). The indents produced by the CSR method maintained a relatively constant indentation strain rate h˙/h across the entire depth range, as shown in Fig. 2 B and E. However, for the impact tests, the strain rate was highest at the beginning of the test but then decayed gradually and dropped sharply near the maximum depth of penetration (8). For purposes of analysis, an average strain rate for the impact tests was estimated through:[1]ε˙time=∫0tloadingε˙dttloading, [2]ε˙depth=∫0hloadingε˙dhhloading,Fig. 2.Strain-rate-dependent mechanical behavior observed in nanoindentation testing of single crystal aluminum (A–C) and single crystal molybdenum (D–F). The data include load–depth curves during loading (A and D), the depth dependence of the indentation strain rate (B and E), and the strain rate dependence of the hardness (C and F). Each testing condition consists of at least 4 separate measurements to confirm the repeatability.where tloading is the total time during the loading period from contact of the surface to the peak load, hloading is the maximum penetration depth with respect to the sample surface, and ε˙ is the instantaneous indentation strain rate. Note that Eq. 1 represents a time-averaged strain rate, while Eq. 2 is depth averaged. For the tests conducted in this work, both equations yielded similar values (details are given in SI Appendix, Fig. S1), so for the purposes of analysis in this manuscript, only the time-averaged value of Eq. 1 is used. Strain-rate-dependent mechanical behavior observed in nanoindentation testing of single crystal aluminum (A–C) and single crystal molybdenum (D–F). The data include load–depth curves during loading (A and D), the depth dependence of the indentation strain rate (B and E), and the strain rate dependence of the hardness (C and F). Each testing condition consists of at least 4 separate measurements to confirm the repeatability. With these definitions, Fig. 2 C and F display the dependencies of the measured hardness, H, on the indentation strain rate, ε˙ , where the hardnesses H = P/Ac were determined by measuring the projected area of the contact impressions Ac by atomic force microscopy (AFM). To minimize the influences of indentation size effects (ISE) on the measurements (37), all tests were terminated at essentially the same total indentation depth (the variation of the total indentation depth was within ±/-25 nm). For the Mo data shown in Fig. 2F, the H vs. ε˙ behavior is very linear on a double logarithmic plot, indicating a relatively CSR sensitivity m=∂ lnH/∂ lnε˙ of about 0.03. This is very similar to the strain rate sensitivity measured by conventional testing methods (38) and indentation strain rate jump tests (39). The strong linearity across all indentation strain rates suggests that the rate-controlling deformation mechanism remains very much the same over nearly 6 orders of magnitude of strain rate (10−2 s−1 to 104 s−1). On the other hand, the Al data in Fig. 2C are relatively linear in the strain rate range 10−2 s−1 to 102 s−1 with a strain rate sensitivity of 0.015 [similar to that observed in conventional testing (40, 41)], but at higher rates (>103 s−1), there is a distinct upturn in strength, especially for the strain rate achieved through the interrupted impact test. This behavior is similar to that observed at higher strain rates in split Hopkinson bar testing (24, 42, 43), thus suggesting a transition in the deformation mechanism that is possibly related to enhanced dislocation-dislocation interactions at higher strain rates, as observed in conventional high-strain-rate testing (24). Clues to the origins of these behaviors were obtained from microstructural examination of focused ion beam cross-sections (FIB) of the indentation hardness impressions, which are now discussed. Effects of Indentation Strain Rate and Depth on Dislocation Substructure. Beginning with the behavior of aluminum, deformation substructures formed beneath nanoindentations at low ( ε˙ = 10−2 s−1), medium ( ε˙ = 102 s−1), and high ( ε˙ ~ 4,000 s−1) strain rates were obtained with scanning TEM of thin cross-sections extracted from the centers of indents using FIB milling. During imaging, the samples were tilted to the [110] zone axis so that the majority of perfect dislocations (Burger’s vector a/2 <011>) would be visible. The dislocation microstructures beneath the indents at three different indentation depths (400, 600, and 800 nm) and three different indentation strain rates ( ε˙ = 10−2, 102, and ~ 4,000 s−1) are displayed in Fig. 3. The figure clearly indicates that the dislocation structure depends on both the depth of penetration and the strain rate, with increasing depth and increasing strain rate transitioning from relatively diffusely spread individual dislocations to well-deformed dislocation cell and subgrain structures. Considering first the 400-nm deep indents in Fig. 3 A, F, and K, it is clear that the dislocations produced at low strain rates ( ε˙ = 10−2 s−1) and medium strain rates ( ε˙ = 102 s−1) are diffusely spread over a finite volume beneath the hardness impression, while distinct cell and subgrain structures form for the high-strain-rate indents ( ε˙ ~ 4,000 s−1), mostly near the hardness impression but to a lesser extent beneath the surface. As the indentation depth increases to 600 nm, the cell/subgrain structure for the high-strain-rate indents becomes more distinct and widespread (Fig. 3L), extending to a depth of almost 5 times of indentation depth. On the other hand, no cell/subgrain structure is observed for the low-strain-rate indent, and cells/subgrains are just starting to appear near the hardness impression at medium strain rates. The formation of cell and subgrain boundaries represents a considerable dislocation accumulation locally and results in crystallographic polygonization to minimize the local strain energy (44). TEM and PED characterization of dislocation structures as a function of indentation depth and strain rate for single crystal aluminum. Panels (A–C), (F–H), and (K–M) are scanning transmission electron micrographs of the microstructure underneath indents produced at three different indentation depths and three different indentation strain rates. Panels (D), (I), and (N) are KAMs for the 600-nm indents depicting the local dislocation density acquired from precession electron diffraction (PED). Panels (E), (J), and (O) are point-to-origin lattice misorientations of the corresponding line profiles for the 600-nm indents marked in B, G, and L, highlighting the lattice rotation of the deformed region. To further characterize the effects of strain rate on the dislocation microstructure, the kernel average misorientation (KAM) of the local crystal lattice as measured with precession electron diffraction (PED) was mapped for the indentations made to a depth of 600 nm. Results are shown in Fig. 3 D, I, and N, where the measured misorientations along the horizontal lines just beneath the hardness impressions included in Fig. 3 B, G, and L are shown in Fig. 3 E, J, and O. Fig. 3 N and O show that the misorientation across the boundaries of the long vertical subgrain in the high-strain-rate indent (Fig. 3L) is large and sharp at about 5°, indicating the formation of a well-defined subgrain. On the other hand, the lattice rotations for the low- and medium-strain rate indents are much smaller and much more diffuse. At an indentation depth of 800 nm, the subgrain boundaries that have been formed under the indent in the low- (Fig. 3C) and medium- (Fig. 3H) strain rate indents begin to resemble the substructural features in the high-strain-rate indent (Fig. 3M). However, the point to origin misorientation of the high-strain-rate indent displays larger lattice misorientation values than the low and medium indents in the same region, as demonstrated in SI Appendix, Fig. S2. Collectively, these results suggest that high-strain-rate impact deformation is accompanied by the rapid development of well-defined subgrains, whereas subgrain formation at the lower rates occurs in a much more gradual way. Assuming that well-formed subgrain boundaries provide significant strengthening, the observed behavior could explain the high rate increase in strain rate sensitivity and deviation in linearity of the log-log hardness vs. strain rate plot in Fig. 2C. In a similar vein, single crystal molybdenum was examined using the same procedures, with the resulting dislocation substructures shown in Fig. 4. Interestingly, the dislocation configurations in the high-strain-rate ( ε˙ ≈ 104 s−1) and low-strain-rate ( ε˙ = 10−2 s−1) indents are very similar for the 400-nm and 800-nm deep indents, at least to first order. For both (Fig. 4 A and E), a diffuse distribution of dislocations extends beneath the contact impression to a distance at least 5 times the total indentation depth, although a cell or subgrain appears to be forming on the indenter face side of the hardness impression (the right side in the figure). To further quantify the low-strain-rate and high-strain-rate indents at a depth of 400 nm, the lattice orientation of the same areas was mapped, as shown in Fig. 4 C and G, with the point-to-origin misorientation profiles along the lines drawn in Fig. 4 A and E provided in Fig. 4 D and H. The lattice misorientations across the cell display a similar profile with no indication of a change in behavior related to the indentation strain rates. In addition, in comparison to the sharp subgrain boundaries formed in aluminum, the dislocation structure in Mo is more cell-like than subgrain-like, with the cell misorientations showing a gradual rather than sharp transition across the boundaries and through the cell. Overall, in single crystal Mo, the similarities of the dislocation configurations and lattice misorientations between the low-strain-rate and high-strain-rate indents at two different indentation depths appear closely related to the linearity of the hardness and strain rate relationship across the over 6 orders of strain rates as shown in Fig. 2F. Once again, similar deformation behavior has been observed in BCC metals and alloys deformed with conventional high-strain-rate testing platforms (24, 33). TEM and PED characterization of dislocation structures as a function of indentation strain rate and indentation depth in the single crystal molybdenum. Panels (A and B) and (E and F) are scanning transmission electron micrographs of the microstructure underneath indents produced at two different indentation strain rates and depths of 400 nm and 800 nm. Panels (C) and (G) are PED lattice orientation maps for the 400-nm indents revealing the formation of a cell structure by color contrast viewed from the horizontal in-plane direction of the FIB-liftout foil. Panels (D) and (H) are PED point-to-origin lattice misorientations of the corresponding line profiles for the 400-nm indents marked in (A and E), highlighting the lattice rotation at two different indentation strain rates. The differences in dislocation substructure evolution in Al and Mo may be linked to fundamental differences in dislocation mobility resulting from the different crystal structures of the FCC aluminum and the BCC molybdenum. In pure FCC metals, dislocations are generally quite mobile, with hardening produced by dislocation–dislocation interactions that become increasingly stronger as dislocations multiply during plastic deformation. As illustrated in the micrographs of Fig. 3, at all indentation strain rates, this process begins with the development of diffusely scattered dislocations and progresses to the development of cells and subgrains as straining continues. In aluminum, cell wall formation is presumably aided by the relatively high stacking fault energy of this material in the ways it promotes dislocation cross-slip. However, it is clear from the evidence in Fig. 3 that the cell formation process in aluminum is also influenced by the indentation stain rate, with higher strain rates accelerating cell formation and leading to well-defined cell and subgrain structures at smaller indentation depths. Since well-formed cell and subgrain boundaries can lead to further strengthening, this tendency of greater cell and subgrain formation at higher strain rates may explain the origin of the upswing in hardness observed at high strain rates in Fig. 2C. For the case of Mo, the similarity of dislocation configurations between two extreme strain rates and the lack of strong cell and subgrain formation is likely related to the high lattice friction observed in pure BCC metals, which limits dislocation mobility and thus contributes strongly to the strength in a very different way compared to aluminum (33, 45). One important consequence is that the extra hardening associated with cell formation is not observed, and there is no upswing in the hardness vs. strain rate behavior observed in Fig. 2F. Quasistatic Reloading. In order to further document the influences of strain rate on hardness in aluminum and molybdenum, an indentation testing method was developed to probe the relationships between hardness, indentation strain rate, and microstructural evolution at low and high strain rates. Referred to as “quasistatic reloading”, the test involves first deforming the specimen at a high strain rate either by impact testing ( ε˙ > 103 s−1) or by high CSR testing ( ε˙ = 102 s−1) and then quasistatically reloading the newly formed hardness impression to a deeper depth at a much slower CSR ( ε˙ = 10−2 s−1). As such, the test is essentially an indentation strain rate change test in which the strain rate begins very high and transitions to very low. The test then probes how deformation microstructures produced by pre-straining at high strain rates affects subsequent deformation and hardness at lower strain rates. Load–depth and hardness–depth curves for Al and Mo for the quasistatic reloading experiments are given in Fig. 5. In each of the plots, the gray curves and discrete data points were acquired from quasistatic loading at ε˙ = 0.01 s−1, the blue curves and discrete data points from impact loading at ε˙ ~ 4,000 s−1 for Al and 11,000 s−1 for Mo, the red curves from quasistatic reloading at ε˙ = 0.01 s−1 after impact loading, and the green curves from quasistatic reloading at ε˙ = 0.01 s−1 after CSR loading at the relatively high rate of ε˙ = 100 s−1. The discrete data points appearing in Fig. 5 B and D are hardnesses determined from the known indentation loads and direct measurement of the contact areas by AFM, whereas the continuous curves are the hardnesses determined by assuming that the ratio of contact depth to the total indentation depth remains constant, as discussed in the Materials and Methods section (8). Note that the quasistatic (gray) and high rate impact (blue) curves all start at zero depth, whereas the quasistatic reloading curves (red and green) all start from a finite depth on the order of 1,000 nm. This is because the quasistatic reloading tests begin not at the sample surface but rather at the depth of the hardness impression formed at high strain rates prior to the quasistatic reloading. The initial sharp rise in load and hardness in the reloading tests is due to the elastic reloading of the hardness impression, after which plasticity commences and there is a transition in behavior. Evaluation of high-strain-rate pre-straining induced microstructural changes by means of quasistatic reloading tests. (A) Load–depth relationship and (B) hardness–depth relationship for impact tests, quasistatic tests, and quasistatic reloading tests in the single crystal aluminum specimen. (C) Load–depth relationship and (D) hardness–depth relationship for impact tests, quasistatic tests, and quasistatic reloading of the single crystal molybdenum specimen. Each testing condition consists of at least 3 separate measurements to confirm the repeatability. For Al (Fig. 5 A and B), it is evident that the high rate pre-straining markedly increases the hardness as compared to that in constant low-rate quasistatic testing, with the increases in hardness increasing with the strain rate during pre-straining. This suggests that high-strain-rate deformation produces dislocation microstructures with higher strength and hardness. However, with further straining the hardnesses of the quasistatic reloaded indents tend back toward the values observed in constant-rate quasi-static tests, thus suggesting that the hardened microstructures relax back to their lower-strain-rate structural states with further deformation and straining. A second important observation in the aluminum data is that the hardnesses immediately after quasistatic reloading do not reach the values observed in the impact tests at the same indentation depths, implying that there are other contributions to the strength and hardness at the higher strain rates. In sharp contrast, the quasistatic reloading tests for molybdenum in Fig. 5 C and D, exhibit a very different behavior. Specifically, immediately after quasi-static reloading, there is no increase in indentation load or hardness; rather, the reloaded indentation load and hardness values converge immediately to those observed in the constant low-rate quasistatic tests, thus suggesting there is essentially no effect of high-strain-rate pre-straining on the deformation microstructure. In other words, the strain-rate-dependent hardening mechanism in BCC-Mo is essentially independent of the pre-strain condition. This is conceivably explained by a larger contribution of the inherent lattice resistance to the strength and hardness of pure BCC metals. The lack of significant subgrain or cell structures during the indentation of molybdenum and the fact that the low-rate and high-rate structures are so similar, as illustrated in Fig. 4, further support this observation. The Behavior of Aluminum at Very Small Indentation Depths. The aluminum results in Figs. 3 and 5 show that the development of dislocation cell and subgrain structures during small-scale indentation depends not only on strain rate but also on the depth of indentation, presumably in the way the depth determines the number of geometrically necessary dislocations required to accommodate the increase in volume of the hardness impression. To further explore this, a quasistatic reloading test was performed at the small indentation depth of ~150 nm where there was no obvious formation of cell or subgrain substructures. Fig. 6A shows that at this depth, only a few randomly scattered dislocations are observed beneath the indent. No prismatic loops were observed (46), in contrast to recent nanoindentation experiments conducted in situ in a TEM (47) and in molecular dynamics simulations (48, 49). This is possibly because the loops are observed only at very small indenter depths, i.e., approximately 20 nm or less (47). (A) Scanning transmission electron micrograph of the microstructure underneath an indent produced by high-strain-rate nanoindentation to a depth of 150 nm in the single crystal aluminum specimen. (B) Comparison of load–depth curves for quasistatic tests and quasistatic reloading tests for the single crystal aluminum specimen. The Inset in Fig. 6B shows that quasistatic reloading of the high-strain-rate indent (the blue curve starting at ~150 nm) does not display any obvious enhancement of the load relative to a test conducted entirely under quasistatic conditions. On the other hand, as also shown in Fig. 6B, for a high-strain-rate indent made to a depth of 800 nm and then quasistatically reloaded, there is a detectable increase in the load above the quasistatic value. Collectively, these results indicate that at very small indentation depths, there are very few obstacles to hinder the motion of newly created dislocations, and the hardness is thus likely determined largely by the stresses needed to nucleate and/or multiply new dislocations from near surface sources. Only later in the indentation process when the dislocation density grows to a critical value do cells and subgrains begin to form. This then leads to increases in the indentation load and hardness caused by the formation of well-developed dislocation substructures. From this perspective, the way in which the dislocation substructure develops and evolves is likely to play an important role in producing the ISE, at least in some materials. The ISE is characterized by an increase in measured hardness with decreasing depth of penetration when the indentation depths fall into the micrometer and sub-micrometer range (50–52). The classic explanation for the ISE is the mechanism proposed by Nix and Gao (37), which assumes that the geometrically necessary dislocations created to accommodate the formation of the hardness impression are confined to reside within the hemispherical zone directly beneath the indent, with the size of the zone determined by the radius of the indent in the surface. To the contrary, the observations here for aluminum and molybdenum, like those reported in previous studies of the nanoindentation of nickel (53) and aluminum (47), suggest that during the initial stages of indentation, there is no such zone in these well-annealed materials, but rather that the dislocations randomly move unimpeded into the hinterland far away from the indentation and/or escape from the free surface. Only when the indentation depth reaches a critical value do dislocations begin to accumulate beneath the indent and form a plastic zone that resembles something like that in the Nix–Gao model. From this perspective, the Nix–Gao model may apply at slightly greater indentation depths, but at the smallest depths in well-annealed materials, the origin of the ISE appears to be more related to the operation of dislocation sources, sometimes referred to as “dislocation source starvation”. Interestingly, this may explain the experimental observation that there often two distinct regions of slope in log-log plots of hardness vs. the inverse of indentation depth at small depths (53–55). The Nix–Gao model predicts such a linearity, but with only one value of the slope. We also wish to note that the magnitude and nature of the ISE likely depends on many factors other than crystal structure such as homologous temperature, stacking fault energy, shear modulus, and initial dislocation density (and configuration), albeit in complex ways that we do not yet fully understand. Microstructural Comparison to Standard High-Strain-Rate Testing Methods. An important question remains as to whether the substructural phenomena observed at small scales in this study are similar to those in more conventional tests such as split Hopkinson bar or plate impact testing. Specifically, at very small indentation depths (~150 nm), where the nanoindentation contact dimension approaches the pre-existing dislocation spacing, the deformation behavior observed in this study may be very different from that in macroscopic tests due to the small number of pre-existing dislocations in the tested volume. In this regard, it is interesting to compare the results obtained in this study to a general set of observations on the macroscopic behavior of FCC and BCC metals. In BCC metals, the dislocation configurations and proposed deformation mechanisms during low- and high-strain-rate nanoindentation observed here are indeed very similar to those observed in conventional macroscopic testing. Specifically, Gray et al. (33) conducted investigations on a variety of BCC metals and alloys (e.g., Nb, Ta, and Ta-10W alloy) and reported that the dislocation substructures formed in samples deformed at both high and low strain rates are much the same, consisting predominantly of long and straight screw dislocations with little tendency to regroup into subgrain and cell structures. This is consistent with the nanoindentation observations reported here for Mo, since the dislocation structures in Fig. 4 show very little indication of cell/subgrain formation but consist mostly of diffuse networks of individual dislocations. In addition, Gray et al. found that shock-prestrained BCC metals and alloys do not show an increase in strength when subsequently tested under quasistatic reloading conditions, which they attribute to the significant contribution of lattice friction to the flow stress (24, 33). This is essentially the same as what was observed here with nanoindentation in the quasistatic reloading tests of molybdenum (33). Thus, in numerous respects, the dislocation configurations and deformation mechanisms observed during high-strain-rate nanoindentation of BCC Mo closely align with those observed with traditional high-strain-rate testing methods. In FCC materials, there is also similarity of high-strain-rate nanoindentation and traditional high-strain-rate testing in that subgrains and cells form during testing and that the subgrain/cell structures produced during high-strain-rate loading yield stronger structures than those formed under quasistatic conditions (29, 30, 33, 56, 57). However, an important difference concerns the nature of the subgrains and cells formed by the two different testing techniques. For conventional high-rate testing methods, well-defined subgrains form mostly only at lower strain rates, whereas high-strain-rate testing produces less well-defined and loosely entangled dislocation cells (24, 29, 30). In this study, it was found that high strain rates during nanoindentation testing enhance the formation of well-defined subgrains and that these form at smaller indentation depths (Fig. 3L). Although the exact reason for the difference is not clear, some possibilities are a difference in the absolute values of strains produced by the two testing methods, the spatially non-uniform straining produced by nanoindentation, and/or the size of the plastically deformed region relative to the length scales of the predominant microstructural features influencing strength. Another possibility is the testing speed. In the high-rate nanoindentation testing system, the maximum velocity of the indenter during a high rate impact test is only ~0.3 m/s, so that unlike in most conventional high-rate testing systems, the influences of stress wave propagation on the deformation processes may not be as important. Another point worth noting that may lead to important differences in behavior between small-scale and large-scale high-strain-rate testing concerns the size of the plastically deformed region relative to important strength-controlling microstructural features such as the grain size. Clearly, in macroscopic testing, where the size of the specimen is much larger than the average grain size, enhancements in strength due to grain size effects are to be expected, but they may not be observed in nanoindentation since the plasticity can take place in single grains. Another scale-dependent factor of this type is the dislocation density. Specifically, in highly annealed materials, the average spacing between dislocations can sometimes be greater that the linear dimension of plastic volume sampled by the nanoindentation, in which case plasticity must be initiated in dislocation source starved regions. This effect has been well documented as the origin of “pop-in” at the theoretical strength of the material in well-annealed metals and ceramics (58). Comparison of Nanoindentation Data to Uniaxial Testing Data. The predicted uniaxial strain rates and stresses calculated from the nanoindentation data are shown in Fig. 7 and compared with results for aluminum alloy 1100 obtained elsewhere from uniaxial compression tests and direct disc impact tests (59). This alloy was chosen for comparison because it is nearly pure aluminum (99% Al), albeit in polycrystalline rather than a single crystalline form. The uniaxial results estimated for a strain of 10% are plotted in Fig. 7. The strain of 10% was chosen to facilitate direct comparison to the Berkovich indentation results, for which the characteristic strain as defined by Tabor is usually in the range 8 to 10%. Note that the two sets of results are similar and follow the same trends, indicating that the high-strain-rate nanoindentation methods have the potential to capture the important features of the macroscopic behavior, as has been observed in other forms of high-strain-rate indentation testing (60, 61). A better comparison might be obtained if the two methods were applied to exactly the same material. Comparison of equivalent uniaxial strain rate and stress of single crystal aluminum calculated from the nanoindentation data at a characteristic strain of 8% (assumed for the Berkovich indenter) with the uniaxial results of 1100 aluminum from conventional compression tests and direct disc impact tests (59). Conclusions High-strain-rate nanoindentation testing has been used in combination with TEM of indentation cross-sections to explore the dependence of hardness and dislocation microstructural evolution on indentation strain rate in the range 10−2 to >104 s−1 using single crystal aluminum (FCC) and molybdenum (BCC) as model materials. The primary conclusions are as follows: (1) In FCC aluminum, high-strain-rate indentation results in the rapid development of well-defined subgrains, while subgrain formation at lower rates occurs more gradually as straining proceeds. Prior to subgrain formation, the dislocation structure is relatively diffuse, with dislocations spread over a finite volume beneath the hardness impression. (2) At small depths of penetration and low strain rates, the behavior of BCC molybdenum is similar to that of aluminum but differs significantly at higher depths and strain rates in that there is not a strong tendency to form well-defined subgrains. There is some evidence that less well-defined cell structures may form, but they are fewer in number and do not have sharp lattice misorientations across their boundaries. (3) The observed differences in dislocation substructure for the two materials are likely linked to fundamental differences in dislocation mobility in FCC and BCC metals, as has been observed in conventional high-strain-rate testing. (4) Quasistatic reloading of hardness impressions after high-strain-rate impact testing revealed that high rate pre-straining increases the quasistatic strength of aluminum but not molybdenum. This is likely explained by the differences in dislocation substructures formed in the two materials, i.e., well-defined subgrains in aluminum but diffusely spread dislocation arrangements with some evidence of cell formation in molybdenum. (5) The observations suggest that the frequently observed increase in hardness at the smallest indentation depths is not due to the Nix–Gao mechanism but rather to dislocation source activation. At a later stage of deformation when a well-defined dislocation structure has formed in a zone beneath the hardness impression, the Nix–Gao mechanism may take over. (6) Although some aspects of the high-rate nanoindentation testing adequately capture the behavior observed in macroscopic testing, there are also some important differences that require further understanding and exploration before the technique can be adapted as a low-cost, high-throughput alternative to traditional macroscopic testing techniques. Materials and Methods Materials. A (111) single crystal of aluminum and a (110) single crystal of molybdenum were obtained from Surface Preparation Laboratory B.V. The Al single crystal was annealed in high purity argon at 673K for 12 h and then electropolished at room temperature using a solution consisting of 60% phosphoric acid, 20% sulfuric acid, and 10% chromic acid in distilled water. The single crystal of Mo was prepared for nanoindentation testing by mechanically polishing with diamond suspensions followed by vibratory polishing for 3 h with 50-nm colloidal silica. Prior to nanoindentation testing, both samples were further cleaned for 20 min using an ion beam milling system (Leica Microsystems GmbH). Electron backscatter diffraction was performed to measure the in-plane and out-of-plane crystal orientations. Nanoindentation. A special nanoindentation system described in detail elsewhere was used to perform all the nanoindentation experiments with a Berkovich diamond indenter (8). The samples were tested using two different methods: constant indentation strain rate (CSR) and high-strain-rate impact testing (8, 36). The indentation strain rate used here is conventionally defined as:[3]ε˙=h˙h,where h and h˙ are the depth of penetration relative to the surface of the specimen and indenter velocity, respectively (62). Constant indentation strain rate tests. For the constant indentation strain rate tests, a constant value of P˙/P was maintained during loading, where P and P˙ are the indentation load and loading rate, respectively (36). Five different rates were used: P˙/P = 0.02 s−1, 0.2 s−1, 2 s−1, 20 s−1, and 200 s−1. At the three lower rates, P˙/P = 0.02 s−1, 0.2 s−1, 2 s−1, the load values were calculated through:[4]P=F0-kh,where F0 is the force applied by the magnetic coil in the actuator, k is the stiffness of the springs in the actuator (k = 293.0 N/m), and h is the depth of penetration relative to the surface of the specimen measured by laser interferometry (8, 63). At higher loading rates, the dynamics of the loading system (inertia and damping) become significant at P˙/P = 20 s−1 and 200 s−1 (63, 64), requiring that the load values be calculated through:[5]P=F0-mh¨-bh˙-kh,where m is the mass of the moving indenter column (m = 7.883 g) and b is the damping of the system (b = 2.444 Ns/m). After the targeted peak load was achieved during the loading period, all tests were unloaded immediately without holding. The orientation of the triangular symmetry of the Berkovich indenter relative to the in-plane crystallographic directions is shown in SI Appendix, Fig. S3. High rate impact tests. The high rate impact testing procedure followed directly from that developed by Hackett et al. (8). The loading procedure is illustrated in Fig. 1B, where it is seen that the indenter starts well off the surface and is accelerated toward it to produce maximum velocity at impact. As shown in Fig. 1A, indenter displacements during impact loading were measured at a rate of 1.25 MHz using a laser interferometer (SmarAct GmbH, Oldenburg, Germany). Indenter loads were first calculated from the measured displacements through Newton’s laws of motion using low-pass Fourier filtering of the displacement data to remove high frequency noise (8). However, this procedure was often found to produce spurious undulations in the load–depth curves due to filtering artifacts, so a piezoelectric load cell was incorporated into the load train beneath the specimen to measure the load directly. The load cell was calibrated using the load data derived from Newton’s laws applied to the high rate displacement data. SI Appendix, Fig. S4 presents the load depth curves obtained with the two different measures of the load, showing how the load cell smooths out the undulations in the load derived from the displacement data. It should be noted that the high rate impact testing procedure does not maintain a constant indentation strain rate, but rather, the strain rate begins high and then decelerates during the course of the test, as seen in Fig. 2 B and E. For this reason, an “average strain rate” has been used to characterize the rate at which the impact tests were performed. Interrupted impact tests. Interrupted impact tests were performed in a manner similar to the impact test described above but with additional testing steps to arrest and reverse the motion of the indenter prior to it being naturally stopped by the mechanical resistance offered by the impacted material. To achieve this, a large negative force relative to the initial driving force was applied to the indenter column via the force coil just after contact to stop the advance of the indenter at a desired depth. The interruption halts the test sooner, effectively retaining the highest-strain-rate conditions present at the beginning of initial contact. This results in the interrupted test having a higher average representative strain rate compared to a non-interrupted impact test sharing the same initially applied coil force magnitude. This testing method was only performed in single crystal Al, aiming to achieve a strain rate larger than 104 s−1. Quasistatic reloading tests. Quasistatic reloading tests began with the same procedure as the impact tests, starting with the indenter well above the surface and then accelerating it toward the specimen at a constant applied coil force F0 . After a short period of dynamic ringing, the indenter finally comes to rest on the hardness impression, at which time the indenter was held at a constant applied coil force F0 for 1 s and then loaded to a deeper depth using CSR loading to the target load at P˙/P = 0.02 s−1. The indentation strain rate during the quasistatic reloading maintains a relative constant value as shown in SI Appendix, Fig. S5. After achieving the target peak load, the indenter is fully unloaded. Hardness Calculation. The hardness values shown in Fig. 2 C and F and discrete symbols in Fig. 5 B and D were determined using (65):[6]H=PAc,where P is the peak load (including dynamic loading effects), which can be acquired from each individual load–depth curve, and Ac is the projected contact area, which was carefully measured through AFM. All the CSR and impact tests were terminated at approximately the same indentation depth (±/-25 nm) in order to mitigate hardness variation effects from the ISE (37). The continuous curves of hardness values displayed in Fig. 5 B and D were calculated using the constant hc / h method, as explained by Hackett et al. (8). The constant hc / h method assumes the ratio of contact depth to the total indentation depth remains constant. Its value is determined directly from the load–depth curve and the residual contact area measured through AFM. Conversion of Nanoindentation Data to Equivalent Uniaxial Reponses. In order to compare the nanoindentation indentation data obtained here for single crystal aluminum to data obtained by more conventional macroscopic uniaxial testing, the nanoindentation hardness data in Fig. 2C were converted to the equivalent uniaxial strain rate, ϵ˙ , and uniaxial stress, σ , using the method proposed by Su et al. (66). This is based on the relations:[7]ϵ˙=1c tanθ1hdhdt, and [8]σ=1Fc2Pπ(h tanθ)2,where θ is the equivalent half cone angle of the indenter (70.3° for a Berkovich indenter), c is the pile-up/sink-in parameter, and F is the reduced contact pressure. Noting that c=hc/h , the specific value of c for each strain rate from low to high-strain-rate nanoindentation testing was determined from AFM and scanning laser microscopy measurements of the projected contact areas, as detailed by Hackett et al. (8). The value F = 3 was assumed without using the polynomial fitting proposed by Su et al. (66), which follows from the observation that as the stress exponent (reciprocal of strain rate sensitivity) attains larger values, the reduced contact pressure asymptotically approaches a value of 3, which is equivalent to a Tabor constraint factor of 3 (67). Electron Microscopy. TEM samples were prepared using FIB milling and TEM specimen liftout using a Helios G4 DualBeam microscope (Thermo Fisher Scientific). Before depositing Pt to protect the region of interest, the indents were first covered with 200-nm amorphous carbon through e-beam deposition, aiming to prevent damage from FIB milling. Then, following the standard TEM sample preparation using FIB milling (68), a 2-μm-thick by 55-μm-long foil was lifted out which contained the center of several indents in a single strip. The foil was thinned down to about 200-nm thick for Al and 80-nm thick for Mo. A final low-kV cleaning process was performed on the Al sample at 2 kV with a beam current of 23 pA on both sides of the liftout foil in order to remove implantation damage from Ga+ ions. The S/TEM characterizations were performed using an FEI Technai G2 F20 ST FE TEM at 200 keV equipped with a NanoMEGAS ASTAR system to trace the orientation information by precession electron diffraction (PED). The diffraction pattern and misorientation information were recorded and analyzed at each pixel with a specific step size and a 0.3° precession angle. The Kernel-average-misorientation map was imaged by averaging the orientation difference between the first-neighboring pixel and a target pixel in the center. Supplementary Material Appendix 01 (PDF) Click here for additional data file. Dataset S01 (XLSX) Click here for additional data file.