Sequence of Stages in the Microstructure Evolution in Copper under

Research Article www.acsami.org

Sequence of Stages in the Microstructure Evolution in Copper under Mild Reciprocating Tribological Loading Christian Greiner,*,† Zhilong Liu,† Luis Strassberger,† and Peter Gumbsch†,‡ †

Institute for Applied Materials (IAM), Karlsruhe Institute of Technology (KIT), Kaiserstrasse 12, 76131 Karlsruhe, Germany Fraunhofer IWM, Woehlerstrasse 11, 79108 Freiburg, Germany

‡

S Supporting Information *

ABSTRACT: Tailoring the surface properties of a material for low friction and little wear has long been a goal of tribological research. Since the microstructure of the material under the contact strongly influences tribological performance, the ability to control this microstructure is thereby of key importance. However, there is a significant lack of knowledge about the elementary mechanisms of microstructure evolution under tribological load. To cover different stages of this microstructure evolution, high-purity copper was investigated after increasing numbers of sliding cycles of a sapphire sphere in reciprocating motion. Scanning electron and focused ion beam (FIB) microscopy were applied to monitor the microstructure changes. A thin tribologically deformed layer which grew from tens of nanometers to several micrometers with increasing number of cycles was observed in cross-sections. By analyzing dislocation structures and local orientation changes in the cross-sectional areas, dislocation activity, the occurrence of a distinct dislocation trace line, and the emergence of new subgrain boundaries could be observed at different depths. These results strongly suggest that dislocation self-organization is a key elementary mechanism for the microstructure evolution under a tribological load. The distinct elementary processes at different stages of sliding identified here will be essential for the future modeling of the microstructure evolution in tribological contacts. KEYWORDS: tribology, copper, sapphire, STEM, EBSD, microstructure

1. INTRODUCTION

structure evolution under the sliding contact is therefore key for tailoring materials that combine low friction and wear. In the literature, there are two general directions of microstructure research under a tribological load: the first starts with an annealed, large-grained bulk material. Applying a tribological load then leads to modifications in the microstructure that result in a (ultra-)fine-crystalline microstructure near the surface.11−15 The second approach starts with samples exhibiting a fine or even nanocrystalline layer or thin film on the surface. The published results demonstrate that recrystallization on the surface and grain boundary motion are dominant mechanisms for making the nanocrystalline grains grow bigger.16−18 Microstructure evolution and changes in wear properties under a tribological load, in engineering applications often referred to as the “running-in” of a tribosystem, were investigated and explained in both directions for different materials systems such as copper,11−13,19−21 steels,22 and cobalt-based alloys for artificial hip joints.23 The consensus is that there is a change in the microstructure of ductile metals. Deformed layers, commonly named as “tribo-layers” or “third

The study of interacting surfaces in relative motioncalled tribologyis of great importance in modern life,1 particularly for metallic materials which are ubiquitously used in engineering components such as combustion engines,2 or artificial limbs and joints.3 The energy necessary to overcome unwanted friction in the powertrain system of a passenger car for example is still above 25% of the total fuel energy.4 Being able to tailor materials that combine low friction and little wear will allow an increase in the energy efficiency of many engineering systems. Early on it was highlighted by Bowden and Tabor that junctions are formed and sheared during a metallic sliding contact.5 If one of the sliding partners is harder than the other, the softer surface will be ploughed to an appreciable depth by the hard surface’s asperities, and hence the bulk properties of the softer surface determine the friction and wear properties of the entire tribosystem.5 The coefficient of friction (COF) in a sliding system is therefore expressed as the ratio of the shear strength to the yield pressure of the softer metal.5 The microstructure of the subsurface material, of course, has a strong influence on the tribological properties of the surface,6−8 and at the same time, the subsurface microstructure is strongly affected by plasticity and the nature of the corresponding dislocation activity under a tribological load.9−11 The micro© 2016 American Chemical Society

Received: April 5, 2016 Accepted: June 1, 2016 Published: June 1, 2016 15809

DOI: 10.1021/acsami.6b04035 ACS Appl. Mater. Interfaces 2016, 8, 15809−15819

Research Article

ACS Applied Materials & Interfaces

Figure 1. Scanning electron microscopy (SEM) images of OFHC copper under reciprocating tribological loading after different cycle numbers: (a) original material before tribological testing; (b) after 10 cycles; (c) after 100 cycles; (d) after 5000 cycles; (e) thickness of the deformed layer, plotted against cycle number. The cross-sections were performed at the middle of the wear track, perpendicular to the sliding surface and parallel to the sliding direction. The contrast at the top of the images is from the two protective platinum layers, the copper surface is marked by arrows.

results,27,28 and large-scale atomistic simulations have demonstrated the self-organization of nanocrystalline grains until the optimal orientation for the slip systems is established.27 Based on these experimental and numerical results, different approaches have been taken to explain the mechanisms behind changes in the subsurface microstructure. The principle processes for the microstructure evolution that in general are considered in the literature are plastic deformation, influence of environment (including counter body) and mechanical mixing.29,30 However, due to the complexity of any sliding contact, the experiments and models in the existing studies did not yield a satisfactory picture for the elementary mechanisms in play, especially for the initial stages of tribological loading which are key for all further changes in microstructure. Some

body”, have been observed in many types of contacts and for different loads. Numerical studies for different friction systems have also been applied using molecular dynamic simulations to model the microstructure changes under a tribological load. Sliding speed,24−26 crystallographic orientation of the surface,25,26 and existing lattice defects25 are considered as variables during these simulations. Corresponding results of an amorphous adlayer on the surface and dislocation activity have been reported.25,26 With non-equilibrium conditions, the dependence of friction force on sliding speed has been simulated, and at a high sliding speed, stacking faults and perfect dislocations loops have been observed in both Cu and Ag contact bodies.24 Crystallographic orientation has also shown its effect on the 15810


Research Article


sample. Figure S2a shows the friction force in each experiment as a function of the cycle number. The tribological conditions were chosen particularly mild, not to plough the surface and not to cause any material loss from the surface (depth profiles of each wear track are presented in Figure S2b). No loose wear particles were observed after any of our tests. The microstructure was examined using a focused ion beam/ scanning electron dual beam microscope (FIB/SEM; Helios NanoLab DualBeam 650 from FEI, Hillsboro, OR, USA). The surface was protected from ion beam damage by depositing two platinum layers, the first one employing the electron beam only and the second thicker layer using the ion beam. Cross-sectional SEM images were taken to observe the thickness of the tribologically deformed layer. Scanning transmission electron microscopy (STEM) images were taken with an acceleration voltage of 30 kV and beam currents of 50 and 100 pA. Transmission electron microscopy (TEM) diffraction was performed to examine the nature of the material in the cross-sectional area and orientation changes along the normal direction (using a Philips CM 30 TEM, Amsterdam, The Netherlands). A state of the art FIB lift-out technique with little ion beam damage was applied to prepare the (S)TEM lamellae for this study.37 EBSD on the cross-sectional area was performed using a lift-out technique similar to that for STEM sample preparation38 with a sample thickness of 1 μm. This can remarkably increase the pattern quality for the cross-sectional area compared to EBSD on volume material. The cross-sectional EBSD scans were performed on a 70° pretilted surface with acceleration voltages of 25 and 30 kV and beam currents of 3.2 and 6.4 nA. The step size was 200 nm for cross-sectional EBSD, a commonly used step size in recent years.39 The Kikuchi patterns were acquired by the detector NordlysMax2 and were indexed by the software AZtecHKL (both, Oxford Instruments, Oxfordshire, U.K.). EBSD mapping analysis was carried out with MTEX, a well-developed Matlab toolbox for texture analysis.39,40 From the cross-sectional area EBSD data, the local orientation changes can be interpreted as an estimate for the density of geometrically necessary dislocations (GNDs). The GND density analysis can expose microstructural features which are introduced by a nonuniform deformation during sliding41,42 like the beginning of plasticity related changes in the microstructure. Following Gao et al.,43,44 the GND density, ρGND, was calculated directly from the misorientation angle, ϑ, irrespective of grain orientation:

studies divided the deformed area into different types of layers13,31 and summarized the behavior. Grains are found to be severely reduced in size in the subsurface area.11,14,15 Subgrain rotation, which may last for the whole sliding process, has been widely agreed on as one explanation for the microstructure change in the subsurface area.13 It has been reported around axes perpendicular but also along the sliding direction.13,32 A plastic flow model has been used to describe the behavior of the near surface material.18,33 Recently, the slip system and the path in which the subgrains are reoriented in the sliding contact of a copper single crystal was investigated establishing that the size and localization of the tribo-layer depends on the crystallographic orientation with respect to loading.31 In similar studies, the generation of wear particles was correlated with the parallelization of the sliding plane to a (111) plane.28,34 In summary, the majority of the existing literature focuses on late stages of microstructural changes under a tribological load and there is a significant lack of investigations of the early phases. Therefore, we are missing fundamental understanding of the elementary mechanisms acting at different stages of this microstructural evolution. This is where our research starts with a series of systematic experiments investigating the early stages of third body formation by monitoring the microstructure evolution underneath the tribologically loaded surface. Highpurity copper is chosen as a model material and tested against sapphire spheres which are so much harder and more inert that they do not participate in any mechanical mixing or microstructure evolution process. A reciprocating motion was chosen to investigate potential differences between the dead centers and the middle of the wear track.

2. MATERIAL AND METHODS Oxygen-free high-conductivity (OFHC) copper plates (Goodfellow, Friedberg, Germany) with a purity higher than 99.95% were annealed in a vacuum of 1.5 × 10−6 mbar for 2 h at a temperature of 500 °C. The plates were then left inside the oven under vacuum to cool to room temperature (with an average cooling rate of 80 K/h). The annealing process was followed by grinding with SiC paper of #800 down to #4000 grid. Mechanical polishing was carried out with a 3 μm diamond suspension for 5 min and with a 1 μm diamond suspension for 8 min (both DP-suspension M products were purchased from Struers, Willich, Germany). The samples were then electropolished in D2 electrolyte (Struers). Right before testing, the samples were sonicated in isopropanol for 15 min. This sample preparation process was developed in order to yield a reproducible surface chemistry and a starting microstructure with the least initial defect density possible (as verified by FIB and TEM; see Figure 1a). Sapphire spheres with a diameter of 10 mm were provided by SWIP (Bruegg, Switzerland). Even though it is associated with a complex 3D stress field,35 a spherical counter body was chosen as it allows for a defined contact area. We also expect that in engineering applications involving tribological loading complex stress fields will be encountered, making a sphere-on-plate contact a good model system. A reciprocating linear tribometer36 was used for the dry sliding experiments (Supporting Information Figure S1). The only variable in the tests was the cycle number, which was increased from 1, 10, 100, 500, 1000, and up to 5000 cycles. All other parameters were kept constant in order to systematically investigate the effect of the cycle number on the microstructure evolution. A normal load of 2 N was applied to the sapphire sphere. The corresponding Hertzian contact pressure35 was 530 MPa, as calculated with a Young’s modulus of copper (117 GPa) and sapphire (345 GPa) and a Poisson ratio of copper (0.34) and sapphire (0.3). The sliding speed was 500 μm/s and the stroke length 12 mm. The tests were conducted at room temperature and in 50% relative humidity. Each test was performed with a new sapphire sphere, and at a new location on the copper

ρGND =

2ϑ ub

(1)

where u is the distance between the two points used to calculate the 1 misorientation and b the length of the Burgers vector of a 2 ⟨110⟩ dislocation, which is 0.255 nm for copper. This method assumes a cube with an edge length of u and each (edge type) GND in this cube will stretch one edge with the length of one Burger’s vector. The sodefined GND density can then be correlated with a misorientation tilt angle, regardless of slip system.39 Even though this method might appear like a simple estimation, its results are very similar to a more thorough analysis.44 Because the angular resolution limit of our EBSD measurements is 0.08°, the lowest GND density which could be detected with this method is thus 5 × 1013 m−2.45 This limit also describes the minimum error of the GND analysis.45

3. RESULTS 3.1. Cross-Sectional Scanning Electron Microscopy. FIB cross-sections were prepared in sliding direction in the middle of the wear tracks after all sliding cycles were investigated (Figures 1 and S3) and also at the two dead centers of each wear track (Figure S4) to investigate the evolution of the deformed layer in the subsurface area. In the unloaded material (Figure 1a), there is no contrast change in the vicinity of the surface and no deformed layer observed in the subsurface area. After 10 cycles (Figure 1b), we observe a 15811


Research Article


Figure 2. Geometrically necessary dislocation (GND) density analysis in the cross-sectional area of wear tracks of OFHC copper under tribological loading after different cycle numbers: (a) after 10 cycles; (b) after 100 cycles; (c) after 500 cycles; (d) after 5000 cycles. The cross-sections were milled perpendicular to the sliding surface and parallel to the sliding direction. The samples’ surfaces are marked by black arrows. The color bar in the GND density mapping represents the number of GNDs per unit area (where, for example, 6e+15 represents 6 × 1015/m2).

remainder of this work, only the middle of the wear track is therefore analyzed further. The observation of the tribologically deformed layer, the change of its thickness with increasing cycle number and an apparent orientation dependence, required further investigation of the microstructure. Therefore, cross-sectional electron backscatter diffraction (EBSD) measurements were performed. 3.2. Cross-Sectional Electron Backscatter Diffraction Mapping. Cross-sectional EBSD measurements were performed in the middle of the wear tracks. The aim of these measurements was to further characterize the crystallographic orientation and the GND density. The results for the orientation mapping and the GND density analysis of the cross-sectional areas are shown in Figure S5 and Figure 2, respectively. Orientation gradients are observed in the crosssections for higher cycle numbers (Figure S5c,d). In some cases, not the entire subsurface area could be fully indexed due to heavy plastic deformation. The amorphous protective platinum layers are also not indexed. From the sample after 10 cycles to the one after 5000 cycles (Figure 2a−d), the average GND density continuously increases with cycle number. In all samples, the area from the surface to a depth of 1 μm has a much higher GND density than deeper inside the material, indicating heavy plastic deformation near the surface. In the sample after 10 cycles (Figure 2a), a grain boundary is covered in the mapping. The average GND density in the left grain with a surface normal

subsurface area with a slightly changed microstructure. A tribologically deformed layer with a thickness of 0.52 ± 0.19 μm is found. This thickness is defined as the distance from the surfacealong the sample’s normal directionto the point where the contrast no longer changes after enhancing contract and brightness of each SEM image. This distance is determined by measuring its extension from the surface into the bulk of the material at five different positions on each image and then calculating the average mean value as well as its standard deviation. This deformed layer increases in thickness with cycle number (Figure 1b−d). The thickness of the deformed layer can be very different in grains with different orientation (Figure 1c) and at grain boundaries (Figure 1c), where it often is higher than inside a grain. For the highest cycle number of 5000 (Figure 1d), the deformed layer has grown as thick as 15.73 ± 1.12 μm. Except very near the surface, this deformed layer appears homogeneous in the normal direction of the sample in classical SEM secondary electron contrast. In Figure 1e, the thickness of this layer is plotted against the cycle number in a double-logarithmic fashion. The same cross-sectional investigation was also performed at both dead centers of all wear tracks (Figure S4). No systematic difference for the thickness of the deformed layer can be found between the middle of the wear track and the two dead centers. The appearance of the deformed layer (thickness and the orientation dependence) is very similar to that in the middle of the wear tracks. For the 15812


Research Article


Figure 3. Scanning transmission electron microscopy (STEM) images of OFHC copper under tribological loading after different cycle numbers: (a) original material before tribological testing; (b) after 1 cycles; (c) after 10 cycles; (d) after 100 cycles [small-angle grain boundary (SAGB) and clusters are marked by white arrows]; (e) after 500 cycles. The cross-sections were cut at the middle of the wear track, perpendicular to the sliding surface and parallel to the sliding direction. The contrast at the top of the images is from the two protective platinum layers, and the sample surface is marked by white arrows.

no features near the surface, demonstrating the high quality of our sample preparation technique. After only one cycle, a distinct line is visible underneath the surface (highlighted by an arrow in Figure 3b) which is associated with a contrast change above and below the line. This feature is observed at a mean depth of 137 ± 5 nm beneath the surface and is also found after 10 sliding cycles (at a depth of 152 ± 18 nm, Figure 3c). These depths are taken as the mean value of five different measurements in each STEM image. After 100 cycles (Figure 3d), a line-shape feature that shows bright contrast from one side and dark contrast from the other (highlighted by an arrow and labeled SAGB) is observed at a depth of 200 nm. At the same time, some hemispherical clusters with brighter contrast begin to appear on the surface (also highlighted by an arrow in Figure 3d,e). With the cycle number increasing to 500 the

orientation close to [001] is higher than that in the right grain with an orientation close to [101]. In the mapping after 100 cycles (Figure 2b), only one grain is covered. The initiation of a network of lines of high GND density appears at depths of more than 1 μm. After 500 cycles (Figure 2c), a high GND density network is clearly visible, and after 5000 cycles (Figure 2d), the network has developed further. 3.3. Cross-Sectional Transmission Electron Microscopy. Dislocation networks formed at several micrometers depth appear to evolve underneath a very highly deformed surface layer which cannot be resolved well in EBSD. STEM was thus applied to overcome the limited resolution of EBSD. Figure 3 presents STEM images for the original, undeformed sample (Figure 3a), as well as for those after tribological loading for one to 500 cycles (Figure 3b−e). Figure 3a shows 15813


Research Article


Figure 4. Transmission election microscopy (TEM) selected area diffraction images of OFHC copper under tribological loading of 10 cycles. A line scan is performed from the surface of the sample along the sample’s normal direction: (a) TEM image of the lamella [circles and the arrow mark the direction and the position of the line scan]; (b−d) diffraction patterns of the line scan; (e) plot of the relative orientation change against the distance from the surface; (f) schema showing the diffraction patterns rotation and the formation of the high GND density position. 15814


Research Article

ACS Applied Materials & Interfaces clusters on the surface grow deeper into the material and cover more area on the surface (Figure 3e). In Figure 3e, one can also observe a subgrain forming in the subsurface area. To further understand the initiation as well as the formation of the aforementioned high GND density networks, a transmission electron microscopy selected area diffraction line scan was performed on the wear track after 10 loading cycles. This allowed quantification of the degree of the crystallographic orientation changes in the cross-sectional area along the normal direction of the sample (Figure 4a). The orientation of this lamella is the same as for the lamella in Figure 3c. The circles and the arrow mark the selected area where the diffraction patterns were collected (Figure 4b−d). The diffraction pattern in Figure 4c (taken around 600 nm away from the surface) is indexed in order to determine the zone axis (see Figure S6). Two possible zone axes, [013] (Figure S6a) and [001] (Figure S6b), both fully matching the diffraction spots, are found. Measuring the rotation angles of the diffraction patterns and plotting them against the depth from the surface results in Figure 4e. The area deeper than 1 μm is considered as not rotated (0°) when measuring the rotation angles. This plot demonstrates that the rotation along the axis parallel to the surface and perpendicular to the sliding direction increases from −2.8° at the surface to 1° at a depth of 800 nm. When the diffraction spot moves deeper along the normal direction, the crystallographic rotation reduces to 0° at a depth of 1 μm and reaches a steady value from here on deeper into the material. The rotation angle in the area with a depth between 500 nm and 1 μm is positive, and the rotation angle in the area from the surface to a depth of 500 nm is negative. In addition, the change of the diffraction spots’ intensity (see Figure 4b−d) indicates that the area is also rotated along the axis parallel to both the surface and the sliding direction. The spots in a depth of 600 nm (Figure 4c) are brighter than in depths of 0 nm (Figure 4b) and 800 nm (Figure 4d). The intensity does not change much in the area deeper than 1 μm. The lamella shown in Figure 3e was also observed in the TEM (see Figure 5a). At a depth of roughly 350 nm, we can see a contrast change which was not observed in the STEM image. TEM images were also taken from the lamella shown in Figure 3d, and a selected area diffraction analysis was performed on one of the clusters, in order to further determine the nature of these clusters. A TEM image and the diffraction pattern are presented in Figure 5b. The diffraction pattern shows a halo which appears to be slightly spotty.

Figure 5. Transmission electron microscopy (TEM) images of OFHC copper under tribological loading: (a) lamella prepared in the wear track after 500 cycles loading; [surface is marked as the solid black line in the image and the position of the dislocation trace line is marked as the dashed black line]; (b) lamella prepared in the wear track after 100 cycles of loading [surface is marked as the black line in the image, the diffraction area is marked by the white arrow, and the diffraction pattern is inserted at the bottom right corner of the TEM image]. Both lamellae were prepared in the middle of the wear tracks, parallel to the sliding direction.

4. DISCUSSION 4.1. Tribologically Deformed Layer. In our results, no systematic difference between the middle of the wear track and the two dead centers was found. This is contrary to some literature, e.g., in the case of combustion engines, where wear at the two dead centers of the cylinder has been reported to be higher than that for other in-cylinder parts.46 In combustion engines, of course, the sliding speed difference between the dead centers and the middle of the track is much larger than in our experiments. Since one of the main purposes of this work was to study the early stages of microstructural changes and to investigate the elementary mechanisms for these changes, our loading conditionsnormal load and sliding speedwere chosen to be very mild. Therefore, a possible explanation for the difference between the published results and ours with respect to the dead centers is most likely found in the small

normal load (Hertzian pressure) and the low sliding speed (strain rate). The observation and the significance of a tribologically deformed layer has been widely reported in the literature.6,11,16,17 What we observed and now interpret is the contrast change in the images in Figure 1, which can be correlated to the plastic deformation taking place in the 15815


Research Article


dislocation trace line, which appears to be thereby pushed slightly deeper into the material. The depth of the line in Figure 3d (after 100 cycles) is only somewhat deeper at about 200 nm, indicating that the dislocation trace line, which may already be interpreted as a small-angle boundary of very small misorientation, may evolve into an SAGB with a larger misorientation with increasing cycle number. This finding is of special significance for the understanding of the elementary mechanisms of third body formation because the thickness of the deformed layer is on the micrometer scale and significantly increases with cycle number (Figure 1e). One could thus expect the formation of subgrain boundaries to occur deeper inside the material with increasing cycle number. The relatively constant depth of the dislocation trace line/ SAGB observed in Figure 3b,c demonstrates that the depth of the trace line has a different origin than the emission of the leading dislocations. Also the trace line is only weakly dependent on the cycle number and once formed appears to mainly increase in misorientation and not move significantly. The difference in depth and in response to the sliding cycle number confirms that the formation of the dislocation trace line/SAGB and the evolution of the deformed layer are two distinct processes. While dislocation injection may lead to deformation micrometers deep into the surface, the reorganization of dislocations into a trace line near the surface may first require a reasonably high dislocation density and may also occur as a distinct second step after injection. From this we hypothesize that dislocations keep accumulating in the dislocation trace line at a depth of around 150−200 nm for up to 100 cycles, before the formation of subgrains starts for higher cycle numbers. Subgrains may then be pushed deeper into the material. When the cycle number reached 500, dislocation cells were formed and led to other new subgrains at greater depth (Figure 3e). This being said, in the TEM image in Figure 5a for a lamella after 500 cycles, the contrast change which could correspond to a SAGB from the original dislocation trace line can still be seen at a depth of about 350 nm. Interestingly, this phenomenon of the microstructural changes occurring from the surface to a certain depth has been found elsewhere in the literature with a different friction system (Ni−W alloy running against a WC sphere) and even with a different initial microstructure (nanocrystalline).16 It has been reported that a surface layer showing significant grain growth (from 3 to 20 nm) always appears from the surface to a depth of only a few hundred nanometers while the maximum Hertzian pressure of 991 MPa is only reached at a depth of 26 μm.16 Even though this study started with a nanocrystalline material, the depth of this grain growth layer agrees well with our results for the depth of the dislocation trace line, which further indicates that dislocation activity from the surface to this depth dominates the deformation mechanisms in this area. Of course, the central role of dislocations for microstructural changes under tribological loading was previously reported by other authors.11,14,15,48 However, since most of these other studies were performed at significantly higher contact stresses they quickly resulted in the generation wear particles,11,14,15 in mechanical mixing with the counter body,15 and the beginning of recrystallization.14,15 The earliest stage of dislocation selforganization and the sequence of subsequent stages of microstructure evolution, which are identified in the present study, have not been deciphered previously. The relatively small depth at which the trace line as the first indication for a

subsurface area. The thickness of the deformed layer increases with cycle number (Figure 1e), indicating that the plastic deformation is getting more and more severe. The thickness of the deformed layer is different in grains with different crystallographic orientations due to the elastic and plastic anisotropy of copper. The thickness of the deformed layer is larger at grain boundaries, indicating that the plastic deformation is locally higher around grain boundaries compared to inside a grain. One explanation for this behavior is that the grain boundaries can be considered as dislocation obstacles where dislocations are accumulated. Consequently it is expected to find more dislocations in the vicinity of grain boundaries. When plotting the depth of the deformed layer versus the number of sliding cycles (Figure 1e), the slope of the resulting curves reflects the sensitivity of the material’s microstructure to the sliding distance (or time). Thus, the slope might be used as a parameter to characterize the dynamics of the microstructure evolution under reciprocating tribological loading. For our experiments, the slopes of the curves in Figure 1e are about 0.5 (in this a double-logarithmic plot). This corresponds to a square root growth law which agrees well with observations in the literature for the formation of an amorphous layer between two sliding diamond surfaces.26 This might be an indication that this growth law is of a more general nature: a hypothesis that has to be investigated further. 4.2. Dislocation Trace Line and Small-Angle Grain Boundary. Our STEM results (Figure 3) reveal dislocation structures and the formation of subgrain boundaries in a much smaller depth than the general dislocation activity that we detected by EBSD (Figure 2). The pronounced contrast change above and below the distinct line in Figure 3b,c (after one and 10 cycles) at a depth of approximately 150 nm clearly indicates a distinct orientation change in the subsurface area. This feature is interpreted as a trace line of dislocations with an accumulated Burgers vector density that leads to the observed contrast change. As it is also a trace of the tribological loading and consists of dislocations, we introduce the term “dislocation trace line” when referring to this feature. The line-shaped feature in Figure 3d (after 100 cycles, marked by the white arrow) at a depth of approximately 200 nm shows a local orientation gradient which is compatible with a small-angle grain boundary (SAGB). The formation of the dislocation trace line and of a SAGB can be rationalized as follows: First, the generation and emission of dislocations from the surface of the sliding contact is expected. Dislocations are probably continuously emitted below the sliding sphere and pushed into the material to a depth that depends on the normal load and the number of passes (Figure 1e). The passing of the sliding sphere leads to sign changes in the stress field as, for example, shown in Hamilton’s solution for the stresses beneath a sliding sphere.47 This apparently leads to the formation of a planar network (or layer) of dislocations that appears as a trace line in STEM. The area above the dislocation trace line appears clearer, because dislocations that have been generated at the surface are either pushed into the trace line or leave again through the nearby surface. The area beneath the trace line contains many more dislocations and thus appears darker (Figure 3b,c). Our GND density analysis for the cross-sectional EBSD scans also gives strong evidence for a high concentration of dislocations at a larger depth of at least 0.5 μm (Figure 2b,c). With the continuing emission of dislocations (more sliding cycles), increasingly more dislocations are stored in the 15816


Research Article


sample after 5000 cycles of loading, presented in Figure S7. If dynamic recrystallization occurred, one would not expect a continuous increase in hardness toward the sample surface as seen in our data. 4.4. Amorphous or Nanocrystalline Clusters. The slightly spotty halo observed in the diffraction pattern in Figure 5b suggests that the hemispherical clusters observed in Figure 3d,e could be nanocrystalline or of amorphous nature. In Figure 3d,e, the microstructure inside the hemispherical clusters cannot be resolved in the STEM images, suggesting that the size of crystallized grains inside these clusters has to be smaller than 10 nm (if they are crystalline at all). A detailed analysis of these clusters in terms of their chemical composition, crystallinity, and especially their nucleation and growth has not been performed here. Yet, preliminary STEM-EDS analysis indicates a somewhat increased oxygen content in the clusters. Careful quantification of the oxygen content and a detailed phase analysis clearly requires further investigation. However, explaining the observation of the nanocrystalline clusters as the continuation of the grain refinement processes described above until nanocrystalline clusters are formed, as proposed in the literature,19 appears unlikely for two reasons: First, it would be very difficult to explain a sharp boundary, clearly seen in our micrographs (Figure 3d,e), between the nanocrystalline and the coarse-grained regions; second, it would be very difficult to reconcile this with the observed dislocation structure formation and grain refinement discussed above. It seems much more reasonable to assume that amorphization occurs due to the severe plastic deformation under tribological loading. This would imply a mechanically driven phase transformation from a crystalline to an amorphous phase at the sharp interface between the clusters and the coarser grained material. An indication of such a phase transformation is the sharp transition between the clusters and the deformed material itself. Although amorphization of pure copper has not been found in triboexperiments or severe plastic deformation, it is reported in MD simulations.25 While such amorphization can only indirectly be argued for here and remains somewhat speculative, mechanically driven amorphization has clearly been found for example in tribological experiments on diamond.49,50 Further experiments and simulations will be necessary to fully understand the nature of and the mechanisms behind the formation of these amorphous/nanocrystalline clusters. 4.5. Sequence of Processes. In summary and directly on the basis of our observations, we propose the following sequence of elementary processes governing the microstructural changes in copper under a mild tribological load: (1) The sliding contact between the sapphire sphere and the copper plate first generates and emits dislocations into the material. Significant amounts of dislocations are emitted several micrometers deep into the material forming a deformed surface layer, and this layer extends deeper with continued reciprocating loading. (2) Simultaneously a straight dislocation trace line forms in the very first cycle 150 nm beneath the surface where dislocations reorganize, leaving the area above the dislocation trace line with fewer dislocations (giving this area brighter contrast in Figure 3b,c, compared to below the trace line). We speculate that under the reciprocating tribological loading, this dislocation trace line accumulates successive dislocations and an increase in misorientation with increasing cycle number. (3) After 100 cycles, the dislocation trace line evolves into a less flat SAGB at a somewhat larger depth of around 200 nm

microstructural change due to tribological loading occurs and its independence from the specific microstructural state of the material, e.g., grain orientation, suggests that the mechanism behind its formation must be a general property of the tribological experiment, e.g., of the stress field near the sample surface or of the surface geometry. 4.3. GND Density Networks. The cross-sectional EBSD experiments cannot provide statistically reliable results due to the limited area in the scans. The direct information in the EBSD results is the crystallographic orientation of each pixel, which can be used to calculate a crystallographic orientation gradient. The misorientation between neighboring pixels can be interpreted as the GND density. The GND density analysis confirms that the primary mechanism behind the microstructure evolution in our experiments is dislocation activity. The GND density inside a grain becomes higher with increasing cycle number as the nonuniform plastic deformation is increasing with progressing tribological loading. A weak orientation dependence of the GND density can be seen in the two grains in Figure 2a (and Figure S5a), with the left (100) grain having a higher GND density than the right grain which is close to a (111) orientation. This conforms with both elastic and plastic anisotropy for compression normal to these surfaces. The GND density after 100 cycles (Figure 2b) does not appear much higher than that after 10 cycles (Figure 2a) which may be because the grain picked by chance for 100 cycles (Figure 2b) is close to a (111) orientation (Figure S5b). This orientation facilitates glide parallel to the surface and has unfavorably oriented slip systems for dislocation motion into the depth of the material. Besides the observation of the high GND density network, a net grain rotation can also be found in the cross-sectional area after high cycle number loading. Three rotations across the lines in the network (R1, R2, and R3 in Figure 2d) are analyzed in terms of misorientation between pairs of points at both sides of them (shown as the two ends of short black arrows in Figure 2d) in the cross-sectional area after 5000 loading cycles. The rotation angles around an out of plane axis are 2.5°, 2.7°, and 2.2°, respectively, indicating mostly a rotation around the transverse direction. This is another direct observation of the formation of small-angle boundaries near the surface and a further indication of (sub)grain rotation induced by dislocation activity. Subgrain formation is further substantiated by the selected area diffraction study on a lamella cut after 10 sliding cycles presented in Figure 4. From the diffraction patterns presented in Figure 4b−d, one can see that, with the diffraction area moving deeper into the material, the diffraction patterns rotate and the intensity of the spots in the diffraction patterns varies. This indicates that the crystallographic orientation changes (subgrain rotation) along two axes. With increasing depth, the rotation of the subsurface area along each axis can be in two opposite directions (see Figure 4f). This will make the crystallographic orientations at different depths (for example from 0 to 500 nm and from 500 nm to 1 μm) vary further for higher cycle number tests. The absolute degree of the relative change in orientation as measured through analyzing the diffraction pattern rotation (approximately 2.8° from the surface to a depth greater than 1 μm) is very similar to the rotation angles determined through EBSD (2.2°−2.7°). Our results do not give evidence for the occurrence of dynamic recrystallization. This is additionally supported by a cross-sectional microhardness measurement performed on a 15817


Research Article

ACS Applied Materials & Interfaces (Figure 3d). We also see indications of the nucleation of a new trace line further to the left in Figure 3d. (4) With an additional increase in cycle number, GND networks begin to appear in the deformed layer to a depth of several micrometers. They evolve (Figure 2c,d) into SAGBs which eventually form subgrains (Figure 3e). It appears important to note here that the formation of the dislocation trace line and the evolution of the GND density networks at greater depth underneath the surface must clearly be considered as two distinct processes. (5) Yet a separate process is the formation of amorphous/ nanocrystalline clusters on the surface after 100 sliding cycles (Figure 3d). The exact mechanism of their formation as well as their importance for friction and wear of copper remains to be determined. It appears that any modeling aiming at predicting the performance of a tribosystem with respect to the evolution of friction and wear, especially during the early stages of sliding, must consider these different processes since each of them changes the local properties in a very significant way. The increase of dislocation density in the deformed layer leads to an increase in hardness. The local orientation changes may soften the surface above the trace line, and the amorphous clusters introduce a yet different phase in the contact area of as yet unknown (but probably softer) mechanical properties. Although our investigation still leaves many open questions as, for example, to the dependence of the different processes on normal load and sliding speed, to the origin of the trace line, and to the nucleation and chemistry of the amorphous/ nanocrystalline clusters, we believe that the microstructural evolution of the material underneath contacting surfaces holds the key to understanding and rationally designing tribosystems.

evolution of tribological properties and rational design of tribosystems.

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsami.6b04035. Schematic of the experimental setup, results for the friction force and profilometry of the wear tracks after different cycle numbers of reciprocating tribological loading, additional SEM images of OFHC copper under reciprocating tribological loading after different cycle numbers, at the middle of the wear track, SEM images of OFHC copper under tribological loading after different cycle numbers at both dead centers, additional cross-sectional EBSD mapping of OFHC copper under tribological loading after different cycle numbers, indexing of the diffraction pattern presented in Figure 4c, and microhardness mapping for a cross-sectional area of a wear track after 5000 cycles loading (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was funded by the German Research Foundation (DFG) under Projects GR 4174/1 and Gu 367/30. We thank M. Wenk for help with GND analysis software development. Discussions with Dr. Johannes Schneider are gratefully acknowledged.

5. CONCLUSIONS In this contribution, we systematically investigated the microstructure evolution of high-purity copper in mild normal load tribological model experiments of a sapphire sphere sliding against a copper plate in a dry sliding linear reciprocating fashion. The microstructure evolution in the copper plate was monitored with STEM/FIB dual beam microscopy, EBSD, and TEM. We find that: • a tribologically deformed layer is formed which increases in thickness with cycle number following a square root growth law, reaching a depth of 10 μm after 5000 cycles; • networks of geometrically necessary dislocations (GNDs) are formed within the deformed layer after only a few hundred cycles. In later stages these GND networks lead to the formation of subgrains and further grain refinement in the tribologically deformed layer; • within this tribologically deformed layer, a distinct dislocation trace line already forms in the very first cycle parallel to the surface at a depth of 150 nm. This dislocation trace line most probably evolves into a small-angle grain boundary and regenerates again within the subgrain; • finally, nanocrystalline or possibly amorphous clusters form at the surface after only 100 sliding cycles. We speculate that the formation of these clusters introduces a completely new phase at the tribologically loaded surface which eventually dominates friction and wear of the contact. Our study reveals three distinct elementary mechanisms of the microstructure evolution at different stages of reciprocating sliding. It thereby provides the basis for future modeling of the

■

REFERENCES

(1) Bhushan, B. Introduction to Tribology, 1st ed.; John Wiley & Sons: New York, NY, USA, 2002; pp 18−21. (2) Gresham, R. M. Who Cares? Tribol. Lubr. Technol. 2011, 67 (10), 22−23. (3) Buescher, R.; Taeger, G.; Dudzinski, W.; Gleising, B.; Wimmer, M. A.; Fischer, A. Subsurface Microstructure of Metal-on-Metal Hip Joints and Its Relationship to Wear Particle Generation. J. Biomed. Mater. Res., Part B 2005, 72B (1), 206−214. (4) Holmberg, K.; Andersson, P.; Erdemir, A. Global Energy Consumption Due to Friction in Passenger Cars. Tribol. Int. 2012, 47, 221−234. (5) Bowden, F. P.; Tabor, D. The Friction and Lubrication of Solids, 1st ed.; Clarendon Press: Oxford, U.K., 1950; pp 20−35. (6) Rigney, D. A.; Glaeser, W. A. Significance of Near-Surface Microstructure in Wear Process. Wear 1978, 46 (1), 241−250. (7) Rigney, D. A.; Hirth, J. P. Plastic Deformation and Sliding Friction of Metals. Wear 1979, 53 (2), 345−370. (8) zum Gahr, K.-H. Formation of Wear Debris by the Abrasion of Ductile Metals. Wear 1981, 74 (2), 353−373. (9) Hattori, T.; Kaneko, Y.; Hashimoto, S. Wear-Induced Microstructure in Ni/Cu Nano-Multilayers. J. Mater. Sci. 2008, 43 (11), 3923−3930. (10) Stoyanov, P.; Romero, P. A.; Merz, R.; Kopnarski, M.; Stricker, M.; Stemmer, P.; Dienwiebel, M.; Moseler, M. Nano Scale Sliding Friction Phenomena at the Interface of Diamond-Like Carbon and Tungsten. Acta Mater. 2014, 67, 395−408. (11) Hughes, D.; Hansen, N. Graded Nanostructures Produced by Sliding and Exhibiting Universal Behavior. Phys. Rev. Lett. 2001, 87 (13), 135503.

15818


Research Article

ACS Applied Materials & Interfaces (12) Emge, A.; Karthikeyan, S.; Kim, H. J.; Rigney, D. A. The Effect of Sliding Velocity on the Tribological Behavior of Copper. Wear 2007, 263, 614−618. (13) Heilmann, P.; Clark, W. A. T.; Rigney, D. A. Orientation Determination of Subsurface Cells Generated by Sliding. Acta Metall. 1983, 31 (8), 1293−1305. (14) Turley, D.; Samuels, L. The Nature of Mechanically Polished Surfaces of Copper. Metallography 1981, 14 (4), 275−294. (15) Hughes, D. A.; Dawson, D. B.; Korellis, J. S.; Weingarten, L. I. A Microstructurally Based Method for Stress Estimates. Wear 1995, 181183, 458−468. (16) Rupert, T. J.; Schuh, C. A. Sliding Wear of Nanocrystalline NiW: Structural Evolution and the Apparent Breakdown of Archard Scaling. Acta Mater. 2010, 58 (12), 4137−4148. (17) Li, W. L.; Tao, N. R.; Han, Z.; Lu, K. Comparisons of Dry Sliding Tribological Behaviors between Coarse-Grained and Nanocrystalline Copper. Wear 2012, 274-275, 306−312. (18) Luo, Z.-P.; Zhang, G.-P.; Schwaiger, R. Microstructural Vortex Formation During Cyclic Sliding of Cu/Au Multilayers. Scr. Mater. 2015, 107 (0), 67−70. (19) Emge, A.; Karthikeyan, S.; Rigney, D. A. The Effects of Sliding Velocity and Sliding Time on Nanocrystalline Tribolayer Development and Properties in Copper. Wear 2009, 267 (1−4), 562−567. (20) Bednar, M. S.; Kuhlmann-Wilsdorf, D. Amorphous and Alloy Film Formation in Sliding of Silver on Copper. Wear 1995, 181-183, 922−937. (21) Shakhvorostov, D.; Jian, L.; Nold, E.; Beuchle, G.; Scherge, M. Influence of Cu Grain Size on Running-In Related Phenomena. Tribol. Lett. 2007, 28 (3), 307−318. (22) Hsu, K. L.; Ahn, T. M.; Rigney, D. A. Friction, Wear and Microstructure of Unlubricated Austenitic Stainless Steels. Wear 1980, 60 (1), 13−37. (23) Pourzal, R.; Theissmann, R.; Gleising, B.; Williams, S.; Fischer, A. Micro-Structural Alterations in MoM Hip Implants. Mater. Sci. Forum 2010, 638−642, 1872−1877. (24) Hammerberg, J. E.; Germann, T. C.; Holian, B. L.; Ravelo, R. Nonequilibrium Molecular Dynamics Simulations of Metallic Friction at Ta/Al and Cu/Ag Interfaces. Metall. Mater. Trans. A 2004, 35 (9), 2741−2745. (25) Karthikeyan, S.; Agrawal, A.; Rigney, D. A. Molecular Dynamics Simulations of Sliding in an Fe-Cu Tribopair System. Wear 2009, 267 (5−8), 1166−1176. (26) Pastewka, L.; Moser, S.; Gumbsch, P.; Moseler, M. Anisotropic Mechanical Amorphization Drives Wear in Diamond. Nat. Mater. 2011, 10 (1), 34−38. (27) Romero, P. A.; Jarvi, T. T.; Beckmann, N.; Mrovec, M.; Moseler, M. Coarse Graining and Localized Plasticity Between Sliding Nanocrystalline Metals. Phys. Rev. Lett. 2014, 113 (3), 036101. (28) Beckmann, N.; Romero, P. A.; Linsler, D.; Dienwiebel, M.; Stolz, U.; Moseler, M.; Gumbsch, P. Origins of Folding Instabilities on Polycrystalline Metal Surfaces. Phys. Rev. Appl. 2014, 2 (6), 064004. (29) Rigney, D. A.; Karthikeyan, S. The Evolution of Tribomaterial During Sliding: A Brief Introduction. Tribol. Lett. 2010, 39 (1), 3−7. (30) Young, J. L.; Kuhlmann-Wilsdorf, D.; Hull, R. The Generation of Mechanically Mixed Layers (Mmls) During Sliding Contact and the Effects of Lubricant Thereon. Wear 2000, 246 (1−2), 74−90. (31) Chumaevsky, A. V.; Lychagin, D. V.; Tarasov, S. Y.; Melnikov, A. G. Fragmentation, Texturing and Plastic Flow in the Subsurface of Friction-Processed Copper Single Crystal. Adv. Mater. Res. 2013, 872, 30−35. (32) Ohno, Y.; Inotani, J.; Kaneko, Y.; Hashimoto, S. Orientation Dependence of High-Angle Grain Boundary Formation During Sliding Wear in Copper Single Crystals. Nippon Kinzoku Gakkaishi 2010, 74 (6), 384−391. (33) Karthikeyan, S.; Kim, H. J.; Rigney, D. A. Velocity and StrainRate Profiles in Materials Subjected to Unlubricated Sliding. Phys. Rev. Lett. 2005, 95 (10), 106001.

(34) Tarasov, S. Y.; Lychagin, D. V.; Chumaevskii, A. V. Orientation Dependence of Subsurface Deformation in Dry Sliding Wear of Cu Single Crystals. Appl. Surf. Sci. 2013, 274, 22−26. (35) Johnson, K. L. Contact Mechanics, 1rst ed; Cambridge University Press: Cambridge, UK, 1987; pp 65−92. (36) Greiner, C.; Schafer, M.; Popp, U.; Gumbsch, P. Contact Splitting and the Effect of Dimple Depth on Static Friction of Textured Surfaces. ACS Appl. Mater. Interfaces 2014, 6 (11), 7986− 7990. (37) Mayer, J.; Giannuzzi, L. A.; Kamino, T.; Michael, J. TEM Sample Preparation and FIB-Induced Damage. MRS Bull. 2007, 32 (5), 400−407. (38) Prasad, S. V.; Michael, J. K.; Christenson, T. R. EBSD Studies on Wear-Induced Subsurface Regions in LIGA Nickel. Scr. Mater. 2003, 48 (3), 255−260. (39) Schreijag, S.; Kaufmann, D.; Wenk, M.; Kraft, O.; Monig, R. Size and Microstructural Effects in the Mechanical Response of Alpha-Fe and Low Alloyed Steel. Acta Mater. 2015, 97, 94−104. (40) Hielscher, R.; Schaeben, H. A Novel Pole Figure Inversion Method: Specification of the MTEX Algorithm. J. Appl. Crystallogr. 2008, 41, 1024−1037. (41) Kubin, L. P.; Mortensen, A. Geometrically Necessary Dislocations and Strain-Gradient Plasticity: a Few Critical Issues. Scr. Mater. 2003, 48 (2), 119−125. (42) Gao, H. J.; Huang, Y. G. Geometrically Necessary Dislocation and Size-Dependent Plasticity. Scr. Mater. 2003, 48 (2), 113−118. (43) Gao, H.; Huang, Y.; Nix, W. D.; Hutchinson, J. W. MechanismBased Strain Gradient Plasticity - I. Theory. J. Mech. Phys. Solids 1999, 47 (6), 1239−1263. (44) Calcagnotto, M.; Ponge, D.; Demir, E.; Raabe, D. Orientation Gradients and Geometrically Necessary Dislocations in Ultrafine Grained Dual-Phase Steels Studied by 2D and 3D EBSD. Mater. Sci. Eng., A 2010, 527 (10−11), 2738−2746. (45) Wimmer, A.; Heinz, W.; Detzel, T.; Robl, W.; Nellessen, M.; Kirchlechner, C.; Dehm, G. Cyclic Bending Experiments on FreeStanding Cu Micron Lines Observed by Electron Backscatter Diffraction. Acta Mater. 2015, 83 (0), 460−469. (46) Dong, W. P.; Davis, E. J.; Butler, D. L.; Stout, K. J. Topographic Features of Cylinder Liners - An Application of Three-Dimensional Characterization Techniques. Tribol. Int. 1995, 28 (7), 453−463. (47) Hamilton, G. M. Explicit Equations for the Stresses Beneath a Sliding Spherical Contact. Proc. Inst. Mech. Eng., Part C 1983, 197 (1), 53−59. (48) Hughes, D. A.; Hansen, N. Deformation Structures Developing on Fine Scales. Philos. Mag. 2003, 83 (31−34), 3871−3893. (49) Zhang, X. Y.; Schneider, R.; Muller, E.; Mee, M.; Meier, S.; Gumbsch, P.; Gerthsen, D. Electron Microscopic Evidence for a Tribologically Induced Phase Transformation as the Origin of Wear in Diamond. J. Appl. Phys. 2014, 115 (6), 063508. (50) Bouchet, M. I. D. B.; Matta, C.; Vacher, B.; Le-Mogne, T.; Martin, J. M.; von Lautz, J.; Ma, T.; Pastewka, L.; Otschik, J.; Gumbsch, P.; Moseler, M. Energy Filtering Transmission Electron Microscopy and Atomistic Simulations of Tribo-Induced Hybridization Change of Nanocrystalline Diamond Coating. Carbon 2015, 87, 317−329.

15819


Sequence of Stages in the Microstructure Evolution in Copper under

Recommend Documents