- Email: [email protected]
Microstructural evolution of nanolayered Cu–Nb composites subjected to high-pressure torsion
Available online at www.sciencedirect.com
ScienceDirect Acta Materialia 72 (2014) 178–191 www.elsevier.com/locate/actamat
Microstructural evolution of nanolayered Cu–Nb composites subjected to high-pressure torsion E.H. Ekiz a, T.G. Lach a, R.S. Averback a, N.A. Mara b, I.J. Beyerlein b, M. Pouryazdan c,d, H. Hahn c,d, P. Bellon a,⇑ a
Department of Materials Science and Engineering, University of Illinois Urbana Champaign, 1304 West Green St, Urbana, IL 61801, USA b Los Alamos National Laboratory, Los Alamos, NM 87545, USA c Institute of Nanotechnology, Karlsruhe Institute of Technology (KIT), 76021 Karlsruhe, Germany d Joint Research Laboratory Nanomaterials at Technische Universita¨t Darmstadt, 64287 Darmstadt, Germany Received 3 February 2014; received in revised form 15 March 2014; accepted 17 March 2014 Available online 20 April 2014
Abstract Bulk nanolayered Cu/Nb composites fabricated by accumulative roll bonding (ARB), leading to a nominal layer thickness of 18 nm, were subjected to large shear deformation by high-pressure torsion at room temperature. The evolution of the microstructure was characterized using X-ray diffraction, transmission electron microscopy and atom probe tomography. At shear strains of 4, the crystallographic texture started to change from the one stabilized by ARB, with a Kurdjumov–Sachs orientation relationship and a dominant {1 1 2}Cu||{1 1 2}Nb interface plane, toward textures unlike the shear texture of monolithic Cu and Nb. At larger strains, exceeding 10, the initial layered structure was progressively replaced by a three-dimensional Cu–Nb nanocomposite. This structure remained stable with respect to grain size, morphology and global texture from strains of 290 to the largest ones used in this study, 5900. The threedimensional self-organized nanocomposites comprised biconnected Cu-rich and Nb-rich regions, with a remarkably small coexistence length scale, 10 nm. The results are discussed in the context of the effect of severe plastic deformation and strain path on microstructure and texture stability in highly immiscible alloy systems. Ó 2014 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Keywords: Copper alloys; Niobium alloys; Nanocomposite; High-pressure torsion; Severe plastic deformation
1. Introduction Nanostructures can profoundly alter the properties of materials [1]. In metals it is mostly the mechanical properties that are affected, owing to the large density of interfaces and their interactions with dislocations and other defects [2–4]. While earlier synthesis relied on inert gas condensation and powder processing [2], the demonstration that bulk nanostructured materials could be fabricated inexpensively and in bulk form by applying severe plastic ⇑ Corresponding author.
E-mail address: [email protected] (P. Bellon).
deformation (SPD) has led to a growing interest in understanding and controlling the microstructural evolutions in materials subjected to SPD [5]. Several SPD techniques are commonly employed, including equal channel angular extrusion (ECAE), accumulative roll bonding (ARB) [6] and high-pressure torsion (HPT) – see Refs. [7,8] for recent reviews on these techniques. These processing routes differ in terms of the strain path of the material, e.g., nearly simple shear in ECAE and HPT but pure shear in ARB, as well as in the accessible strain values, ranging from strains of order 1 per pass in ECAE and rolling to strains of several thousand in HPT. From a fundamental as well as a practical perspective, it is therefore interesting to
http://dx.doi.org/10.1016/j.actamat.2014.03.040 1359-6454/Ó 2014 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
investigate the stability of nanostructures fabricated by each of these methods, as well as the stability of nanostructures produced by one processing scheme and subsequently subjected to a different SPD route. The properties of traditional materials are often optimized by including second phases and inclusions, and this approach is very likely to provide similar benefits for nanostructured materials, although work in this area is limited. Accumulative roll bonding and wire drawing of dissimilar materials, however, have been used to successfully synthesize bulk nanocomposite materials. One notable application of this approach has been the synthesis of nanolayered Cu/Nb composites [9,10], with individual layer thicknesses as small as 9 nm. These nanocomposites, originally developed for high conductivity and high strength applications, also hold great promise for applications requiring high resistance to energetic particle irradiation [11–13], to SPD [14] and to elevated temperature exposure [15]. One remarkable characteristic of these Cu/ Nb nanocomposites is that below a layer thickness of 1 lm, rolling stabilizes a Kurdjumov–Sachs orientation relationship between adjacent Cu and Nb grains, where h1 1 1iCu||h1 1 0iNb||RD, h1 1 0iCu||h1 1 1iNb||TD and h1 1 2iCu ||h1 1 2iNb||ND, with RD, TD and ND being the rolling direction, the transverse direction and the direction normal to the rolling plane, respectively, and h1 1 2iCu||h1 1 2iNb||ND forming a preferred interface between Cu and Nb. The stabilization of this unusual interface orientation, which is referred to as {1 1 2} Kurdjumov–Sachs, or simply {1 1 2}KS, has been rationalized by the favorable conditions for slip transfer across the Cu/Nb interfaces for these interfaces [14]. It is noteworthy that this interface plane is distinct from the one observed when Cu/Nb multilayers are grown by physical vapor deposition, for which the dominant interface habit plane is {1 1 1}Cu||{1 1 0}Nb. While these {1 1 1}Cu||{1 1 0}Nb KS interfaces have low shear strength [11], computer simulations have indicated that {1 1 2}KS interfaces possess high shear strength [16], and the stabilization of these interfaces may be useful for achieving optimal properties. The purpose of the present work is to further investigate the evolutions of microstructures in Cu–Nb subjected to SPD. Specifically, we examine the stability and evolution of ARB Cu/Nb nanocomposites subjected to a change in strain path from pure shear (ARB) to nearly simple shear using HPT. In addition we explore how the microstructure evolves at very high strains when it goes from two-dimensional (2-D) layers to threedimensional (3-D) interconnected phases. An additional motivation for the present work concerns the atomic mixing of Cu and Nb under SPD. These metals are highly immiscible, as indicated by the very large miscibility gap in the solid state as well as in the liquid state in the Cu–Nb equilibrium phase diagram. The evolution of mixtures of such highly immiscible elements during SPD is in fact not well understood. This contrasts with the case of moderately immiscible elements such as Cu–Ag [17], Cu–Fe [18,19] and Cu–Co [20], where it has been shown
179
that plastic deformation at low homologous temperatures can force the formation of solid solutions over the entire range of compositions (for a review see Refs. [21,22]). Solubility limits in highly immiscible alloys, on the other hand, are only slightly enhanced, even during SPD at very low temperatures. What is particularly interesting about these systems is that they undergo self-organization during low-temperature SPD; i.e., they select a length scale for phase separation that in the steady state depends only on the intensity (applied stresses) employed for SPD. For ball-milled Ni50Ag50 powders [23], a length scale of 20–40 nm has been reported while in Cu80Nb10Ag10 and Cu80Nb20, Nb precipitates reach a stable steady-state size of 20 nm after prolonged deformation by HPT or lowtemperature ball milling [24,25]. If ARB nanolayered Cu/Nb materials are destabilized by large HPT deformation, it is therefore intriguing to investigate whether they would self-organize, and if so at what length scale and by what mechanism. In this study, we have subjected ARB nanolayered Cu/Nb material to large HPT strains and characterized the evolution of the microstructure as a function of the applied strain, employing X-ray diffraction (XRD), transmission electron microscopy (TEM), scanning TEM (STEM), and atom probe tomography (APT). In Section 2, we detail the experimental procedures. In Section 3, we first report on the evolution of the two-phase microstructure, and then on the evolution of the crystallographic texture. We then discuss the implications of the experimental results in Section 4. 2. Experimental procedure 2.1. Fabrication of Cu–Nb nanolayered composites Nanolayered Cu/Nb composites were prepared by a combination of ARB, annealing and rolling steps, following the method detailed in Ref. [26]. Samples were composed of 50/50 volume fraction Cu/Nb, using starting materials of 99.99% and 99.97% purity, respectively, corresponding to a 51/49 ratio of weight fractions. The material selected for the present work was processed to a rolling strain of 10.93, resulting in a nominal thickness of 18 nm for the Cu and Nb layers. The final thickness of the as-fabricated ARB material was 380 lm. 2.2. SPD by high-pressure torsion HPT was applied to the above ARB nanocomposite to investigate its evolution when subjected to large shear strains. Discs of radius R = 5 mm were cut from the ARB sheet using electrical discharge machining. Samples then were subjected to HPT at room temperature, as described in Ref. [27]. An unconstrained anvil geometry was selected to minimize deformation heterogeneity, and the anvils were sand-blasted prior to torsion to prevent slipping [28]. The lower anvil was continuously rotated at
180
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
a rate of 1.2 rpm against the stationary upper anvil, with a confining hydrostatic pressure of 4.5 GPa. In order to cover a wide range of shear strains, four different values of rotation were used, 50°, 210°, 5 turns and 20 turns. During HPT deformation, if one assumes that the material flow is uniform and that the sample thickness is constant, the total shear strain c at a given distance r from the center of a disc of thickness t is given by: c¼
2pNr t
ð1Þ
where N is the number of turns applied by anvils [5,29]. In the present work, however, the sample thickness was observed to decrease significantly as N increased, in part because of the unconstrained anvil geometry. The thickness of the samples, measured by SEM imaging after HPT on cross sections, was found to decrease faster as the radial distance to the center of the disc increased. Four radial distances were selected for microstructural characterization, r = 1, 2, 3 and 4 mm, and thus the thickness evolution, t(N) for these four radial distances were measured and fitted using an exponential relaxation, as shown in Fig. 1. The shear strain evolution of the function was then deduced by integrating a differential form of Eq. (1), which includes the thickness evolution with HPT turns, t(N): dc ¼
2pr dN tðN Þ
ð2Þ
The resulting shear strains are reported in Table 1 for the four radial distances and the four HPT rotations N. As seen from Table 1, the combination of four samples and four radial distances covers a large spectrum of shear strains, from 2.0 to 6000. We note that the strain values obtained using the differential expression Eq. (2) can be significantly lower than the ones given by the standard Eq. (1) when using the final thickness, leading here to a reduction of 50% in the worst case (5 turns, r = 3 mm). For large
Fig. 1. Fitted thickness profiles for positions of HPT samples measured at r = 1, 2, 3 and 4 mm from the center of the disc after various HPT rotations. The horizontal black dashed line corresponds to the initial ARB sheet thickness.
Table 1 HPT shear strains calculated using Eq. (2) for the four samples and the four radial distances selected for the present work. For each r, the volume sampled was associated with a spot size of 100 lm. Two samples were subjected to 20 turns, the second one being only characterized at 2 mm and 3 mm, to provide supplemental data for the texture analysis.
50° 210° 5 turns 20 turns
1 mm
2 mm
3 mm
4 mm
2.0 8.8 107 960
4.2 18.8 286 1990 2490
6.5 30.3 560 3500 4500
10.8 68.5 1360 5970
strains, however, the sample thickness reached near steady-state values, and the differences did not exceed 20%. In practical terms, therefore, these errors have little effect on the results. In the following, the analyzed regions will be referred to using the strain values obtained from Eq. (2). This allows for a unique identification of sample and analyzed region since each strain value corresponds to a unique combination of rotation N and radial distance r, see Table 1. 2.3. Characterization methods For microstructural analysis, axes attached to the samples were defined as the normal direction (ND), the rolling direction (RD) and the transverse direction (TD), as illustrated in Fig. 2. The RD and TD directions refer to directions on the top surface of the samples during rolling, but these directions become less meaningful during HPT deformation since they are not invariant during torsion around the ND. For instance, for a given HPT rotation, the RD is rotated by an amount that increases with the depth of the analyzed region within the thickness of the sample. The samples were first characterized by XRD with a Panalytical Philips X’Pert X-ray diffractometer using Cu Ka radiation equipped with a monocapillary system. The top surface of the samples was polished to remove 20 lm to suppress contribution from surface artifacts. Local XRD scans and texture measurements were performed for four radial distances, r = 1, 2, 3 and 4 mm,
Fig. 2. Schematic illustration showing the axes convention and the radial distances used for analysis. The ND was common during rolling (ARB) and shearing (HPT).
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
using a 100 lm spot size, from spots along the initial TD, as illustrated in Fig. 2. Data were acquired for three poles for each phase: h1 1 1i, h2 0 0i and h2 2 0i for Cu, and h1 1 0i, h2 0 0i and h2 1 1i for Nb. Orientation distribution functions (ODFs) were then calculated after background subtraction. Minor sample misalignments due to sample mounting were corrected by applying cubic (m 3m) symmetry operations for the crystal symmetry and triclinic ( 1) symmetry operations for the sample symmetry. Pole figures (PFs) and inverse pole figures (IPFs) were then reconstructed from the ODFs for both Cu and Nb phases. MTex 3.1 software was used for texture plot and analyses [30,31]. The texture results were resolved and displayed in the (SD–ND) shear plane. Selected regions of the samples subjected to HPT were also analyzed by TEM, at radial distances of r = 2 and 4 mm. Thin foils were prepared using a focused ion beam (FIB) (FEI Strata Dual Beam 235 and Helios NanoLabe 600i) and conventional lift-out methods. Operating voltages from 2 kV to 30 kV and currents from 20 nA to 1 pA were selected to optimize sample quality while minimizing gallium implantation. The majority of these TEM samples were taken directly from the top surface of the samples, which had been previously polished for XRD data acquisition, as noted above. The electron transparent volumes probed in these samples were therefore 20–30 lm below the original sample surfaces. In order to assess the homogeneity of the microstructures, additional TEM samples were prepared for c = 286 from a cross-section obtained by cutting the 10 mm disc along a diameter parallel to the RD. TEM samples were then lifted out from this cross-section at various depths. The microstructures were analyzed by combining bright field TEM (BF-TEM), dark field TEM (DF-TEM), selected-area electron diffraction (SAED) and high-angle annular dark field in scanning transmission electron microscopy (HAADF-STEM) mode, using JEOL 2100LaB6 and JEOL 2010F microscopes, both operated at an accelerating voltage of 200 kV.
181
APT was employed to characterize the Cu and Nb spatial distribution in the sample deformed by HPT to 20 turns. The corresponding disc was cut in half and wedge-polished to a thickness of 30 lm using an Allied Multiprep System; posts with thickness 30 lm, length 200 lm and spacing 300 lm were then cut into the edge of the discs using a low-speed diamond saw. This enabled fabrication of multiple tips along the radius of a single HPT sample. APT tips were shaped to a radius of curvature of 50 nm on the posts using a Ga-ion beam in a FIB, FEI Strata DB235 or FEI Helios NanoLabe 600i. The APT measurements were performed at the Northwestern University Center for Atom Probe Tomography (NUCAPT). Analysis of the APT measurements was performed using the IVAS 3.6 (Imago Visualization and Analysis Software). 3. Results 3.1. Initial microstructure of ARB material Electron microscopy confirmed the previously reported nanolayering of the as-fabricated ARB Cu–Nb composite, as illustrated in Fig. 3. The average thickness of the layers, calculated from over 100 measurements, was 16.8 ± 8 nm. This value is consistent with the nominal value of 18 nm. Sharp Cu/Nb interfaces were observed, and SAED patterns, see Fig. 3b, indicate a strong preferred orientation for the Cu and Nb crystals, with [1 1 1]Cu// [1 1 0]Nb//RD. This is consistent with the {1 1 2}KS texture reported by Carpenter et al. [32] using precession electron diffraction. 3.2. Microstructural evolution with HPT deformation The evolution of phases in the ARB material with HPT strain was first evaluated using XRD, as summarized in Fig. 4. Bragg reflections of the body centered cubic (bcc)
Fig. 3. As-fabricated ARB Cu–Nb nanocomposite: (a) bright field TEM micrograph and (b) corresponding SAED pattern; the main reflections are indexed in (c). The rolling and transverse directions are shown in (a).
182
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
and face centered cubic (fcc) phases remained present, even at the largest strain of 5970, indicating the persistence of phase coexistence between Cu-rich and Nb-rich phases. The relative intensities of Cu peaks and Nb peaks, however, were significantly modified by HPT strain, pointing to the development of a new preferred texture. This point will be further detailed in Section 3.3. In addition, Bragg reflections were shifted and became broader, indicating that HPT deformation led to some amount of forced chemical mixing and to a reduction in grain size. The forced chemical mixing between Cu and Nb was analyzed using lattice parameter measurements obtained from the {1 1 0} Nb and {1 1 1} Cu reflections. Additional peaks could not be used as their intensities became very small at large HPT strains, as seen in Fig. 4. For the as-fabricated ARB material, however, four and five reflections could be detected for the Cu and the Nb phases, respectively, and the lattice parameters were taken as averages over these reflections. Vegard’s law could not be used to quantify the degree of chemical mixing from these lattice parameter measurements since the Cu–Nb system is highly immiscible, with heats of solution of 98.7 kJ mol1 for Nb in Cu, and 46.0 kJ mol1 for Cu in Nb [33]. Instead, we used the atomic volumes reported by Ashkenazy et al. [34] for Cu–Nb dilute solid solutions [34] obtained by molecular statics simulations. Expressions for the concentrations of Nb in Cu and Cu in Nb as a function of the solid solution lattice parameters were obtained using the following equation: a3 1 a3 A CB ¼ 0 ð3Þ A ðDV =V ÞB
Fig. 4. XRD spectra from the ARB and HPT-processed samples. For clarity only the spectra for radial distances r = 2 and 4 mm are displayed, and the spectra are shifted in intensity by an arbitrary amount. Corresponding strain values are indicated on the graph, see also Table 1. Background has been removed from each spectrum. Dashed lines correspond to Bragg reflections expected for pure Nb and pure Cu.
where C AB is the concentration of B species in matrix A, a the corresponding lattice parameter, a0 the lattice parameA ter of the pure A phase and ðDV =V ÞB the relative volume change resulting from the introduction of one B solute atom in a pure A crystal. These relative volume changes measured in the above atomistic simulations were 0.53 and 0.28 for Nb in Cu, and Cu in Nb, respectively [34]. The evolution of the solubility with strain is plotted in Fig. 5, together with the evolution of the grain size, obtained from {1 1 1} Cu and {1 1 0} Nb peak broadening using Scherrer’s equation. It should be noted that the initial Cu grain size is very close to the nominal ARB layer thickness, indicating that layers contain just a single grain, or at most two, along the ND direction; this is in agreement with our TEM observations. One notices also that the initial Nb grain size is slightly smaller than that of Cu, 13.5 nm compared to 17 nm. Similar trends have been reported for the Cu and Nb layer thickness values in similarly prepared samples [14]. Lastly, we note that the solubilities reached after ARB were small, typically less than 1 at.%. HPT deformation to strains less than 10 did not result in significant changes in composition or grain size, as seen in Fig. 5. Upon increasing the strain further, however, the grain size of the Cu and Nb phases were progressively
Fig. 5. (a) Chemical mixing vs. strain and (b) grain size vs. strain for ARB Cu–Nb nanocomposite subjected to HPT deformation.
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
reduced and reached steady-state values of 6.2 and 4.7 nm, respectively, at a strain of 286. The evolution of solubility within the two phases was less symmetric, as the Cu concentration in Nb started to increase continuously when the strain exceeded 10, reaching a steady-state value of 10 at.% for strains exceeding 1000. In contrast, the Nb concentration in Cu started only to increase at a strain of 68.5, and reached a maximum of 1.7 at.% for a strain of 286, before decreasing slightly to 1.5 at.%, at larger strains. The detailed evolution of the microstructures was investigated using TEM. For low HPT strains, c = 4.2, the microstructure was very similar to that of the as-fabricated ARB sample, with well-defined layers and sharp interfaces, as seen in Fig. 6a and b. One noticeable difference at this low strain, however, was a slight waviness of the interfaces, which is clearly visible in HAADF-STEM images, see Fig. 6b. At a strain of c = 10.8, the interfacial waviness became more pronounced, and, while the layered structure was still preserved in most regions, strain localization was observed in other areas, leading to the presence of folds and vortex-like features, see Fig. 6c and d. This strain localization will be discussed in Section 4. Further increasing the HPT strain resulted in a progressive destabilization of the nanolayered structure of the as-fabricated ARB material. After a strain of c = 68.5, the microstructure had entered a “mixed” state with nanolayers coexisting with equiaxed grains, as shown in Fig. 7a
183
and b. This microstructural transformation was completed at a strain of 286, leading to complete destabilization of the 2-D Cu/Nb and the formation of a 3-D nanocomposite, with very small grain sizes. Grain sizes were estimated to be 5 nm from Z-contrast STEM images, see Fig. 8, in agreement with the XRD data. TEM analysis performed on specimens prepared from different depths for c = 286 confirmed that this microstructure was present throughout the thickness of the sample, as seen by comparing Fig. 8a, taken close to the disc surface, and Fig. 8b taken near the mid-plane of the disc. This microstructure remained stable upon increasing the HPT strain to the largest value used in this study, 5970, as illustrated in Fig. 9a and b. It is thus concluded that this microstructure is stable under HPT deformation. In addition, selected area diffraction, Fig. 9b, indicates that this 3-D steady-state microstructure had a texture quite distinct from that of the as-fabricated ARB sample, as already suggested by the XRD spectra, see Fig. 4. This point will be discussed in Section 3.3. The 3-D structure of the fine scale nanocomposite stabilized by large HPT strains could not be easily accessed by TEM owing to overlapping effects through the thickness of the TEM foils. We thus employed APT to investigate this microstructure for the sample deformed by HPT to a shear stain of 1990. Reconstruction maps were built from the APT data by binning the detected atoms into contiguous 1 nm 1 nm 1 nm cubes, hereafter referred to as voxels, forming a simple cubic lattice. The total number
Fig. 6. Microstructure for an HPT strain of (a and b) 4.2 and (c and d) 10.8. Bright field TEM images are shown in (a and c) and Z-contrast STEM images in (b and d). In the Z-contrast images, Nb-rich layers appear brighter than Cu-rich ones.
184
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
Fig. 7. Microstructure for an HPT strain of 68.5: (a) bright field TEM micrograph, (b) Z-contrast STEM image, (c) SAED pattern of area shown in (a and d) indexing of the main reflections in (c).
Fig. 8. Microstructure for an HPT strain of 286. Z-contrast STEM images taken from (a) 25 lm below the surface, (b) 50 lm below the surface. The specimen thickness after HPT was 106 lm. The sample (b) has been obtained by FIB from a cross-section as detailed in Section 2.3.
of atoms detected in each voxel and the local composition were then obtained. The voxel composition is denoted in the reconstruction by its color, ranging from red (pure Cu) to blue (pure Nb) in 15 steps. The voxel concentration maps confirmed the coexistence of Cu-rich and Nb-rich regions, as illustrated in Fig. 10a. The difference in evaporation fields between Cu and Nb [35], however, led to local magnification effects that resulted in inhomogeneous distributions of atoms detected in each 1 nm3 voxel. These aberrations therefore limited the chemical resolution and prevented us from determining accurately the composition of the Cu-rich and Nb-rich phases. The connectivity of these regions was nevertheless assessed by using isoconcentration surfaces defined by specified threshold values of the Cu concentration. An example is given in Fig. 10b using a Cu concentration threshold of 40 at.%. While the size of the largest connected interface varied somewhat with the
Fig. 9. Microstructure for an HPT strain of 5970: (a) bright field TEM micrograph, (b) dark field TEM micrograph formed using a Nb (1 1 0) reflection; (c) SAED pattern of area shown in (a); (d) indexing of the main reflections visible in (c).
threshold value, for reasonable values of the threshold, ranging from 30 to 70 at.% Cu, there always exists one interface percolating through the sample. This analysis therefore indicates that the Cu-rich and the Nb-rich regions are also percolating through the sample. The length scale of the decomposition was determined from the APT data by defining cylinders along the z-axis (the tip axis), and plotting the Nb and Cu concentration profiles as a function of z. This axis was chosen because it is less sensitive to the local magnification aberrations [36,37]. Notice in Fig. 10a, for instance, that the Cu-rich and Nb-rich regions in the x and y directions are considerably more extended. The cylinder diameter was varied from 5 to 25 nm, and for diameters between 5 and 15 nm welldefined composition modulations were observed with an average period of 10 nm, independent of the cylinder diameter, as seen in Fig. 11. This length scale is in good agreement with the one deduced from Z-contrast TEM micrographs, see for instance Figs. 8 and 9. 3.3. Texture evolution The XRD 2h scans in Fig. 4, and TEM diffraction patterns in Fig. 9c, suggested significant crystallographic texture evolution during HPT of the ARB Cu/Nb nanolayered material. To further investigate texture evolution, we analyzed the XRD texture measurements, using IPFs and ODFs, with grain orientations represented by the three Euler angles (u1, U, u2) in Bunge’s notation. Prior to HPT straining, the textures of the as-fabricated ARB were consistent with the texture measurements
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
185
Fig. 10. APT reconstruction maps for a Cu/Nb sample deformed by HPT to a shear strain of 1990. (a) Concentration map, with color scheme varying from red (pure Cu) to blue (pure Nb). (b) Isoconcentration surfaces obtained by applying a Cu concentration threshold of 40 at.%. The green surface displayed is the largest connected set of these surfaces. Axes are in nm and apply to both figures. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 11. One-dimensional composition profiles generated from the data shown in Fig. 10, using a cylinder along the tip axis, (a) 5 nm and (b) 15 nm in diameter.
reported by Carpenter et al. [26,32] on similarly prepared samples. HPT deformation led to a rapid change from
the initial ARB texture. For strains as small as 4.2, ODFs indicate that the initial textures have been largely erased, as seen in the ODF sections u2 = 0° and u2 = 45°, shown in Figs. 12 and 13 for the Cu and the Nb phases, respectively. It is noteworthy that, at that strain value, the layered structure has not yet been destabilized, see Fig. 6a and b. At larger strains, a new steady-state texture evolves, as we now describe. For the Cu phase, as strain increases a preferred texture develops progressively, with a strong preference for ND// {1 1 1}, i.e., the so-called A-fiber, see the ODFs in Fig. 12 as well as the PFs and IPFs, Figs. S.1 and S.2 in Supplementary information. The main ideal texture components present at large strains were the A=A, corresponding to ð1 1 1Þ½1 1 0 and ð1 1 1Þ½1 1 0, and the A1 =A2 components, corresponding to ð1 1 1Þ½1 1 2 and ð1 1 1Þ½1 1 2. Strains in excess of 2500 were needed for the complete stabilization of this texture. The texture evolution of the Nb-rich phase is more complex, as illustrated in Fig. 13. While low HPT strains greatly suppressed the initial ARB texture, transient textures developed for strains ranging from 18.8 all the way up to 1360, and at larger strains the Nb texture became nearly random. Similarly to the Cu phase, strains greater than 2500 were necessary to fully stabilize this texture. In order to investigate whether the local texture variations could account for the unusual texture development, texture analysis was performed using a larger X-ray beam, 2 mm 2 mm, which we refer to as macro XRD, instead of the 100 lm beam size used in the results reported in the previous paragraphs. A comparison of the ODFs obtained with these two different beam sizes indicates that there is a good agreement between the textures at large strains, see Fig. 14, with an A-fiber texture in Cu and a near random texture in Nb. In the macro XRD Cu results, the intensities of the twin-symmetric A and A components are very similar, as expected from symmetry. Note that the comparison between micro and macro XRD results is performed here
186
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
Fig. 12. U2 = 0° and U2 = 45° sections of Cu phase ODFs for selected HPT strains. Scale bar shown is in multiples of random distribution (min = 0, max = 4). Locations of common simple shear ideal texture components are given at the bottom.
only for strains large enough for the texture be in steady state since the large X-ray beam covered regions subjected to a wide range of strain, e.g., 5100–12,550 when the X-ray beam is centered on r = 4 mm, as in Fig. 14. 4. Discussion HPT has been employed to evaluate the stability of ARB Cu/Nb nanolayered composites to large shear strains. We found that HPT led to profound morphological, microstructural and chemical evolutions of the Cu/Nb nanolayered material, eventually leading to the destabilization of the 2-D nanocomposite in favor of an ultrafine 3-D nanocomposite. For the ARB material investigated here, with an initial nominal layer thickness of 18 nm, TEM revealed
that the layered structure remained largely stable for strains up to 10, although interfaces became increasingly wavy. This observation is consistent with prior reports of wavy layered patterns forming during the mechanical alloying of ductile metals [21], and during HPT deformation of two-phase materials [38]. Atomistic simulations and modeling have also shown that the co-deformation of layered structures induces a kinetic roughening of interfaces [27], resulting in wavy interfaces. When increasing the HPT strain to 68.5, the layered structure became severely disrupted and partly replaced by a 3-D structure, with coexisting Cu phases and Nb-rich phases. For a strain of 286, this morphological transformation was complete and resulted in the stabilization of a 3-D nanocomposite, which remained present for the largest shear strains used in this
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
187
Fig. 13. U2 = 0° and U2 = 45° sections of Nb phase ODFs for selected HPT strains. Scale bar shown is in multiples of random distribution (min = 0, max = 4). Locations of common simple shear ideal texture components are given at the bottom.
study, c 6000. Analysis of the XRD data indicates that the grain size of the Cu and Nb phases reached a steady state at a strain of 286, whereas the chemical composition of these phases appear to continue to evolve slightly, finally reaching a steady state for strains exceeding 3500. In contrast to moderately immiscible alloy systems such as Cu–Ag [27,39,40] or Cu–Fe [41], a large HPT strain could thus not force the highly immiscible Cu and Nb elements into a solid solution. This is consistent with the stabilization of nanocomposites in Cu–Nb mixtures subjected to SPD by other techniques such as ball milling [42] and wire drawing [43,44], as well as in other highly immiscible alloy systems such as Ni–Ag [45,46], Fe–Ag [47,48], Cu–Mo [49] and Cu–W. It should be noted here that the report of complete alloying of Cu90Nb10 mixtures by ball milling
[24,50] is questionable since the Cu lattice parameter never exceeded 0.364 nm (indicating a Nb solubility of 1.6 at.%); however, ball milling was performed at liquid N2 temperatures, and partial amorphization at the interfaces is a possibility. Our analysis of atomic mixing in Section 3.1 suggests that the Nb content was at most 2 at.%. Such a low solubility is in fact consistent with the APT data obtained from wire drawing alloys [44], HPT [25] of Cu–Nb–Ag alloys and indeed the ball milling result of Refs. [24,49]. Excess Nb atoms can form very small precipitates, which would then be very difficult to detect by XRD. It is interesting to note that the solubility of Cu in Nb was significantly larger, saturating at 10 at.% at large strains, as shown in Fig. 5. Similar values of 10–15 at.% have in fact been measured by APT in wire-drawn Cu/Nb
188
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
Fig. 14. Comparison between the ODFs obtained from XRD data collected with a 100 lm beam size (micro XRD), and those obtained with a 2 mm beam size (macro XRD), for the ARB material and for the samples deformed to large strains by HPT. Top row for Cu, and bottom row for Nb.
composites [43]. This larger nonequilibrium solubility of Cu in Nb compared to Nb in Cu can be rationalized by the lower heat of solution of Cu in Nb, 0.6 eV, compared to 1.3 eV for Nb in Cu [34]. The evolution of textures in the Cu and Nb phases with strain also revealed new insights about SPD in strongly immiscible alloys. First, the ARB texture was already largely destabilized after a strain of 4.2, see Figs. 12 and 13. This is consistent with molecular dynamics (MD) simulation results indicating that the {1 1 2}KS orientation relationship is not stable under simple shear deformation [51,52]. Second, the textures stabilized at very large strains, an A-fiber texture for Cu and a nearly random texture for Nb, appear to be quite distinct from the textures reported for nominally pure Cu deformed by HPT [53,54] and the ones expected for bcc elements, such as Fe [55], deformed in simple shear. In particular in the case of Cu, the dominant texture components identified by Enikeev et al. [53] were B=B, with some significant C component as well. None of these components is present in the Cu/Nb nanocomposites at large HPT strains, as shown in Fig. 12. One should be careful in comparing these results, however, since the results reported for monolithic Cu have been obtained using quasi-constrained HPT, for strains not exceeding 300, and on pure Cu. Third, the Cu and Nb phase textures required unusually large strains to reach steady state, and furthermore the Nb phase presents strong transient textures for strains ranging from 18.8 to 1360. These results suggest that the texture evolution and stabilization in these Cu/Nb nanocomposites is strongly influenced by the constraints imposed by 3-D nanoscale phase co-existence. In that context, we note that the transient textures in Nb started to develop at a strain of
18.8–68.5, which coincides with the destabilization of the layered morphology. Moreover the same large strains required for the stabilization of textures are also the ones required for the Cu and Nb solubilities to reach steady state, see Fig. 5. It is remarkable that at large strains the Cu phase developed a fairly strong texture, with maximum intensities in the ODFs of approximately four times that of a random distribution, while the Nb phase reached a near random texture. This result suggests that, at steady state, plastic deformation is mainly accommodated by deformation of the Cu phase. One intriguing aspect of the transformation of the ARB nanolayered structure to a 3-D nanocomposite structure pertains to the mechanisms driving this morphological evolution. It is often observed that layered structures can unravel by a pinch-off mechanism. The kinetic roughening of interfaces imposed by plastic deformation could certainly lead to the pinching off of layers, especially considering the small initial layer thickness used in the present study. On the other hand, we observed the presence of S-shape or Z-shape folds and vortex-like features, see Fig. 6c and d and the low magnification STEM image Fig. 15. These features suggest that another plastic instability might also be taking place. Macroscopic swirls and vortex-like features have been reported for duplex stainless steels [56] processed by HPT, and ascribed to the development of a Kelvin–Helmholtz instability. The scale of the swirls in those studies, however, were typically hundreds of microns, and thus much larger than the folds observed here, and the geometry of those swirls seems to be directly derived from the geometry of the torsion test. The folds that we observed are, in fact, more similar to those recently reported in hypoeutectic Cu–Ag alloys subjected to HPT [39]. These
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
Cu–Ag folds have been linked to shear banding, which does not appear to be present in our Cu/Nb samples. It is also possible that the folds in Cu/Nb are related to the folds forming in layered geological materials subjected to shear deformation by plate tectonics [57–59]. Despite the vast difference in length scale, the physical processes involved in both cases are similar. Further work is needed to elucidate the origin of folds in nanolayered materials subjected to HPT. The 3-D nanocomposites stabilized at large HPT strains possess remarkable features. First the scale of phase co-existence is very small. Z-contrast STEM imaging indicates that this scale is 5–10 nm, see Fig. 8. A complete characterization of this structure is very challenging because the scale is finer than the typical TEM sample thickness, leading to significant superposition for the near equiatomic composition of the ARB samples. APT was thus used to further evaluate these nanostructures. Although APT lacks atomic resolution because of the large difference in evaporation fields for Cu and Nb, atom reconstruction maps confirmed the 3-D coexistence of Cu-rich and Nb-rich phases, with a characteristic length scale of 10 nm, as illustrated in Fig. 11. We note also that isoconcentration surfaces generated in this study establish that both phases are connected, which is not surprising for mixtures with nearly equiatomic compositions. Moreover, TEM dark field imaging and XRD peak broadening showed that the Cu and Nb grain sizes are also of the order of 5 nm. It is therefore concluded that each Cu-rich or Nb-rich ligament comprises 1–2 grains. As indicated in Section 1, one of the interests in developing Cu/Nb nanolaminates and nanocomposites resides in their
Fig. 15. Low magnification Z-contrast STEM image for a strain of 10.8. Note the presence of folds and vortex-like features. The near-vertical dark bands result from foil thickness variations introduced by FIB sample preparation. Nb-rich layers appear brighter than Cu-rich ones.
189
potential for absorbing point defects and dislocations, owing to their high interfacial area. In that respect, the 3D nanocomposites stabilized by large HPT strains offer a high potential. The specific interfacial area is estimated from APT isoconcentration surfaces to be 0.30 nm1, compared to 0.056 nm1 for the as-fabricated ARB material. Prior studies on Cu/Nb and Cu/V layered structures have indicated, however, that the efficiency for absorbing point defects or He atoms is also a function of the crystallographic orientation of the interfaces [16,60]. It would thus be interesting to determine the He storage capacity of these 3-D nanocomposites, and to compare it with that measured on Cu/Nb nanolayers with well-defined interfaces [60]. The steady-state grain size of the 3-D Cu/Nb nanocomposites is much smaller than the ones obtained after room temperature HPT deformation of the pure metals, typically 150 nm for pure Cu [61,62] and 100–150 nm for pure Nb [63]. In the pure Cu samples deformed by HPT, however, grain size distribution and microtexture developments provided clear evidence for recrystallization [54]. In contrast, Lim and Rollett [10] showed that, in Cu–Nb layered composites, recrystallization in the Cu layer was suppressed during rolling when the layer thickness was below 1 lm. Our analysis of the microstructure and texture of 3-D nanocomposites stabilized by HPT is consistent with this result, leading to very small grain sizes. We also note that the steady-state grain size stabilized by ball milling is also quite small, typically 5 nm [24], probably for similar reasons. One interesting fundamental question raised by the stabilization of these 3-D nanocomposites concerns the selection of the length scale of phase coexistence. In the case of moderately immiscible alloys, it has been proposed that nanoscale phase separation results from the competition between thermally activated phase separation and forced mixing [27,40,64]. Recent atomistic simulations support this simple picture and the corresponding analysis of the length scale selection resulting from this dynamical competition [65]. Such an understanding is missing in the case of highly immiscible systems. In a recent investigation of phase evolution in Cu80Nb10Ag10 subjected to low temperature SPD by HPT and ball milling, Wang et al. [25] showed that, in the absence of thermally activated diffusion, Nb precipitates reached a well defined steady-state size, 20 nm, and that this size results from plasticity-driven mechanisms promoting precipitate shrinkage as well as re-precipitation. More work is needed to determine whether a similar rationalization could also apply to the bi-connected Cu–Nb 3-D composites stabilized by HPT. The starting material for the present investigation was an ARB Cu/Nb nanolaminate with (1 1 2)KS orientation, thus with high shear strength interfaces. It would be interesting to investigate the evolution of Cu/Nb nanolaminates with low interface shear strength, such as PVD Cu/Nb nanolayers where interfaces are predominantly (1 1 1)KS [11]. Hoagland et al. [66] calculated that dislocations would be attracted by these weak interfaces, and these authors
190
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
predicted that “weaker boundaries afford a greater barrier to the transport of slip across the boundary than very strong boundaries”. This prediction was in fact recently confirmed by MD simulations [51,52] showing that, in simple shear deformation, (1 1 1)KS and NW Cu/Nb interfaces are much more resistant to chemical mixing than (1 1 2)KS interfaces. At large enough shear strains, however, the activation of multiple slip systems should eventually lead to the formation of interface steps, leading to interface roughening, and to the progressive destabilization of the initial layered structure. We thus anticipate that the threshold strain required for destabilizing layered structures with low shear strength interfaces would be larger than the one reported here for (1 1 2)KS interfaces, but, in the limit of very large strains, a 3-D nanocomposite would nevertheless be stabilized. Evaluating these predictions is left for future work. Lastly, it is interesting to note that wire drawing experiments and atomistic simulations indicated that the forced mixing of Cu and Nb could lead to the formation of an amorphous phase [34,43,67]. No amorphous phase was identified in the present work by either TEM or XRD, but we note that the detection of such a phase in the 3-D nanocomposites stabilized at large HPT strains would be very difficult owing to the fine scale of phase coexistence compared to typical TEM foil thickness. 5. Conclusion Nanolayered Cu/Nb composites fabricated by ARB, with a nominal layer thickness of 18 nm, were subjected to large shear strains using HPT at room temperature. The evolution of the microstructure proceeded in successive steps. First, at strains of 4.2 and beyond, XRD characterization indicated that the {1 1 2}KS orientation relationship and texture stabilized by ARB were destabilized. TEM showed that the layered morphology resisted HPT deformation to strains up to 10.8. As the HPT strain was further increased, a 3-D microstructure replaced the initial ARB layered structure, with a reduction in the length scale of phase coexistence. The average grain size of both Cu and Nb was also reduced, down to 5 nm at a strain of 286, while strains as large as 3500 were required for the chemical composition of these phases to reach steady state. The steady-state solubility of Nb in Cu and Cu in Nb were found to reach 1.5 at.% and 10 at.%, respectively. This significant difference in solubility was assigned to the large asymmetry of the thermal solubility between the Cu-rich side and the Nb-rich side of the equilibrium Cu–Nb phase diagram. At large strains, the textures of the Cu and Nb phases stabilized around an A-fiber texture and a near random texture, respectively. These textures appear thus to be distinct from those reported in monolithic Cu and Nb subjected to HPT, suggesting a significant effect of the 3-D Cu–Nb nanoscale phase coexistence on texture. The length scale of these 3-D nanocomposites was very fine, and determined by
TEM and APT to be 5–10 nm. The present findings suggest that HPT can be used to stabilize interconnected nanocomposites at very small scales and with a very high density of interfaces. In addition to its stability under HPT deformation, this morphology thus offers attractive features for trapping point defects and rare gas atoms under irradiation for the design of radiation-resistant materials. Acknowledgements This work was supported as part of the Center for Materials at Irradiation and Mechanical Extremes, an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences under Award Number 2008LANL1026. APT was performed at the Northwestern University Center for Atom-Probe Tomography (NUCAPT) whose localelectrode atom-probe (LEAP) tomograph was purchased and upgraded with funding from NSF-MRI (DMR0420532) and ONR-DURIP (N00014-0400798, N000140610539, N00014-0910781) Grants. Instrumentation at NUCAPT was supported by the Initiative for Sustainability and Energy at Northwestern (ISEN). NUCAPT is a Shared Facility at the Materials Research Center of Northwestern University, supported by the National Science Foundation’s MRSEC program (DMR-1121262. Stimulating discussions with Drs. A. Misra and J. Carpenter (LANL), and Prof. A. Rollett (CMU), are gratefully acknowledged. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.actamat.2014.03.040. References [1] Vollath D. Nanomaterials: an introduction to synthesis, properties and application. Weinheim: Wiley-VCH; 2008. [2] Gleiter H. Prog Mater Sci 1989;33:223. [3] Birringer R. Mater Sci Eng A 1989;117:33. [4] Meyers MA, Mishra A, Benson DJ. Prog Mater Sci 2006;51:427. [5] Valiev RZ, Islamgaliev RK, Alexandrov IV. Prog Mater Sci 2000;45:103. [6] Saito Y, Utsunomiya H, Tsuji N, Sakai T. Acta Mater 1999;47:579. [7] Valiev R. Int J Mater Res 2009;100:757. [8] Zhilyaev AP, Langdon TG. Prog Mater Sci 2008;53:893. [9] Jha SC, Delagi RG, Forster JA, Krotz PD. Metall Trans A 1993;24:15. [10] Lim SCV, Rollett AD. Mater Sci Eng A 2009;520:189. [11] Demkowicz MJ, Hoagland RG, Hirth JP. Phys Rev Lett 2008;100:136102. [12] Misra A, Demkowicz MJ, Zhang X, Hoagland RG. JOM 2007;59:62. [13] Zhang L, Demkowicz MJ. Appl Phys Lett 2013;103:061604. [14] Beyerlein IJ, Mara NA, Carpenter JS, Nizolek T, Mook WM, Wynn TA, et al. J Mater Res 2013;28:1799. [15] Zheng S, Beyerlein IJ, Carpenter JS, Kang K, Wang J, Han W, et al. Nat Commun 2013;4:1696.
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191 [16] Demkowicz MJ, Thilly L. Acta Mater 2011;59:7744. [17] Klassen T, Herr U, Averback RS. Acta Mater 1997;45:2921. [18] Eckert J, Holzer JC, Krill CE, Johnson WL. J Appl Phys 1993;73:2794. [19] Huang JY, Yu YD, Wu YK, Li DX, Ye HQ. Acta Mater 1997;45:113. [20] Gente C, Oehring M, Bormann R. Phys Rev B 1993;48:13244. [21] Suryanarayana C. Prog Mater Sci 2001;46:1. [22] Ma E. Prog Mater Sci 2005;50:413. [23] Zghal S, Bhattacharya P, Twesten R, Wu F, Bellon P. Metastable, mechanically alloyed and nanocrystalline materials 2002;386–3:165. [24] Botcharova E, Heilmaier M, Freudenberger J, Drew G, Kudashow D, Martin U, et al. J Alloys Compd 2003;351:119. [25] Wang M, Averback RS, Bellon P, Dillon SJ. Acta Mater 2014;62:276. [26] Carpenter JS, Vogel SC, LeDonne JE, Hammon DL, Beyerlein IJ, Mara NA. Acta Mater 2012;60:1576. [27] Pouryazdan M, Schwen D, Wang D, Scherer T, Hahn H, Averback RS, et al. Phys Rev B 2012;86:144302. [28] Pippan R, Scheriau S, Hohenwarter A, Hafok M. Mater Sci Forum 2008:584–6. Part 1:16. [29] Vorhauer A, Pippan R. Scripta Mater 2004;51:921. [30] Hielscher R, Schaeben H. J Appl Crystallogr 2008;41:1024. [31] Bachmann F, Hielscher R, Schaeben H. Solid State Phenomena 2010;160:63. [32] Carpenter JS, Liu X, Darbal A, Nuhfer NT, McCabe RJ, Vogel SC, et al. Scripta Mater 2012;67:336. [33] Chakrabarti DJ, Laughlin DE. Bull Alloy Phase Diagr 1982;2:455. [34] Ashkenazy Y, Vo NQ, Schwen D, Averback RS, Bellon P. Acta Mater 2012;60:984. [35] Miller MK. Atom probe field ion microscopy. Oxford: Clarendon Press; 1996. [36] Vurpillot F, Bostel A, Blavette D. Appl Phys Lett 2000;76:3127. [37] Isheim D, Hellman OC, Seidman DN, Danoix F, Blavette D. Scripta Mater 2000;42:645. [38] Cao Y, Wang YB, Alhajeri SN, Liao XZ, Zheng WL, Ringer SP, et al. J Mater Sci 2010;45:765. [39] Tian YZ, Wu SD, Zhang ZF, Figueiredo RB, Gao N, Langdon TG. Acta Mater 2011;59:2783. [40] Arshad SN, Lach TG, Pouryazdan M, Hahn H, Bellon P, Dillon SJ, et al. Scripta Mater 2013;68:215. [41] Quelennec X, Menand A, Le Breton JM, Pippan R, Sauvage X. Philos Mag 2010;90:1179. [42] Benghalem A, Morris DG. Scripta Metall Mater 1992;27:739.
191
[43] Sauvage X, Renaud L, Deconihout B, Blavette D, Ping DH, Hono K. Acta Mater 2001;49:389. [44] Raabe D, Ohsaki S, Hono K. Acta Mater 2009;57:5254. [45] Xu J, Herr U, Klasse T, Averback RS. J Appl Phys 1996;79:3935. [46] He JH, Sheng HW, Schilling PJ, Chien CL, Ma E. Phys Rev Lett 2001;86:2826. [47] Angiolini M, Krasnowski M, Mazzone G, Montone A, Urchulutegui M, VittoriAntisari M. Nanophase Mater 1995;195:13. [48] Shingu PH, Huang B, Kuyama J, Ishihara KN, Nasu S. In: Artz A, editor. New materials by mechanical alloying techniques. Deutsche Gesellschaft fu¨r Metallkunde; 1989. p. 319. [49] Botcharova E, Freudenberger J, Schultz L. J Mater Sci 2004;39:5287. [50] Botcharova E, Freudenberger J, Schultz L. J Alloys Compd 2004;365:157. [51] Vo NQ, Zhou J, Ashkenazy Y, Schwen D, Averback RS, Bellon P. JOM 2013;65:382. [52] Zhou J, Averback RS, Bellon P. Acta Mater 2014, in press. [53] Enikeev NA, Schafler E, Zehetbauer M, Alexandrov IV, Valiev RZ. Mater Sci Forum 2008:584–6. Part 1:367. [54] Al-Fadhalah KJ, Alhajeri SN, Almazrouee AI, Langdon TG. J Mater Sci 2013;48:4563. [55] Li SY, Gazder AA, Beyerlein IJ, Davies CHJ, Pereloma EV. Acta Mater 2007;55:1017. [56] Cao Y, Kawasaki M, Wang YB, Alhajeri SN, Liao XZ, Zheng WL, et al. J Mater Sci 2010;45:4545. [57] Carreras J, Druguet E, Griera A. J Struct Geol 2005;27:1229. [58] Schmalholz SM, Schmid DW. Philos Trans R Soc A 2012;370:1798. [59] Llorens MG, Bons PD, Griera A, Gomez-Rivas E, Evans LA. J Struct Geol 2013;50:209. [60] Demkowicz MJ, Bhattacharyya D, Usov I, Wang YQ, Nastasi M, Misra A. Appl Phys Lett 2010;97:161903. [61] Jiang HG, Zhu YT, Butt DP, Alexandrov IV, Lowe TC. Mater Sci Eng A 2000;290:128. [62] Horita Z, Langdon TG. Mater Sci Eng A 2005;410:422. [63] Popov VV, Popova EN, Stolbovskiy AV. Mater Sci Eng A 2012;539:22. [64] Odunuga S, Li Y, Krasnochtchekov P, Bellon P, Averback RS. Phys Rev Lett 2005;95:045901. [65] Schwen D, Wang M, Averback RS, Bellon P. J Mater Res 2013;28:2687. [66] Hoagland RG, Hirth JP, Misra A. Philos Mag 2006;86:3537. [67] Vo NQ, Averback RS, Ashkenazy Y, Bellon P, Wang J. J Mater Res 2012;27:1621.
ScienceDirect Acta Materialia 72 (2014) 178–191 www.elsevier.com/locate/actamat
Microstructural evolution of nanolayered Cu–Nb composites subjected to high-pressure torsion E.H. Ekiz a, T.G. Lach a, R.S. Averback a, N.A. Mara b, I.J. Beyerlein b, M. Pouryazdan c,d, H. Hahn c,d, P. Bellon a,⇑ a
Department of Materials Science and Engineering, University of Illinois Urbana Champaign, 1304 West Green St, Urbana, IL 61801, USA b Los Alamos National Laboratory, Los Alamos, NM 87545, USA c Institute of Nanotechnology, Karlsruhe Institute of Technology (KIT), 76021 Karlsruhe, Germany d Joint Research Laboratory Nanomaterials at Technische Universita¨t Darmstadt, 64287 Darmstadt, Germany Received 3 February 2014; received in revised form 15 March 2014; accepted 17 March 2014 Available online 20 April 2014
Abstract Bulk nanolayered Cu/Nb composites fabricated by accumulative roll bonding (ARB), leading to a nominal layer thickness of 18 nm, were subjected to large shear deformation by high-pressure torsion at room temperature. The evolution of the microstructure was characterized using X-ray diffraction, transmission electron microscopy and atom probe tomography. At shear strains of 4, the crystallographic texture started to change from the one stabilized by ARB, with a Kurdjumov–Sachs orientation relationship and a dominant {1 1 2}Cu||{1 1 2}Nb interface plane, toward textures unlike the shear texture of monolithic Cu and Nb. At larger strains, exceeding 10, the initial layered structure was progressively replaced by a three-dimensional Cu–Nb nanocomposite. This structure remained stable with respect to grain size, morphology and global texture from strains of 290 to the largest ones used in this study, 5900. The threedimensional self-organized nanocomposites comprised biconnected Cu-rich and Nb-rich regions, with a remarkably small coexistence length scale, 10 nm. The results are discussed in the context of the effect of severe plastic deformation and strain path on microstructure and texture stability in highly immiscible alloy systems. Ó 2014 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Keywords: Copper alloys; Niobium alloys; Nanocomposite; High-pressure torsion; Severe plastic deformation
1. Introduction Nanostructures can profoundly alter the properties of materials [1]. In metals it is mostly the mechanical properties that are affected, owing to the large density of interfaces and their interactions with dislocations and other defects [2–4]. While earlier synthesis relied on inert gas condensation and powder processing [2], the demonstration that bulk nanostructured materials could be fabricated inexpensively and in bulk form by applying severe plastic ⇑ Corresponding author.
E-mail address: [email protected] (P. Bellon).
deformation (SPD) has led to a growing interest in understanding and controlling the microstructural evolutions in materials subjected to SPD [5]. Several SPD techniques are commonly employed, including equal channel angular extrusion (ECAE), accumulative roll bonding (ARB) [6] and high-pressure torsion (HPT) – see Refs. [7,8] for recent reviews on these techniques. These processing routes differ in terms of the strain path of the material, e.g., nearly simple shear in ECAE and HPT but pure shear in ARB, as well as in the accessible strain values, ranging from strains of order 1 per pass in ECAE and rolling to strains of several thousand in HPT. From a fundamental as well as a practical perspective, it is therefore interesting to
http://dx.doi.org/10.1016/j.actamat.2014.03.040 1359-6454/Ó 2014 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
investigate the stability of nanostructures fabricated by each of these methods, as well as the stability of nanostructures produced by one processing scheme and subsequently subjected to a different SPD route. The properties of traditional materials are often optimized by including second phases and inclusions, and this approach is very likely to provide similar benefits for nanostructured materials, although work in this area is limited. Accumulative roll bonding and wire drawing of dissimilar materials, however, have been used to successfully synthesize bulk nanocomposite materials. One notable application of this approach has been the synthesis of nanolayered Cu/Nb composites [9,10], with individual layer thicknesses as small as 9 nm. These nanocomposites, originally developed for high conductivity and high strength applications, also hold great promise for applications requiring high resistance to energetic particle irradiation [11–13], to SPD [14] and to elevated temperature exposure [15]. One remarkable characteristic of these Cu/ Nb nanocomposites is that below a layer thickness of 1 lm, rolling stabilizes a Kurdjumov–Sachs orientation relationship between adjacent Cu and Nb grains, where h1 1 1iCu||h1 1 0iNb||RD, h1 1 0iCu||h1 1 1iNb||TD and h1 1 2iCu ||h1 1 2iNb||ND, with RD, TD and ND being the rolling direction, the transverse direction and the direction normal to the rolling plane, respectively, and h1 1 2iCu||h1 1 2iNb||ND forming a preferred interface between Cu and Nb. The stabilization of this unusual interface orientation, which is referred to as {1 1 2} Kurdjumov–Sachs, or simply {1 1 2}KS, has been rationalized by the favorable conditions for slip transfer across the Cu/Nb interfaces for these interfaces [14]. It is noteworthy that this interface plane is distinct from the one observed when Cu/Nb multilayers are grown by physical vapor deposition, for which the dominant interface habit plane is {1 1 1}Cu||{1 1 0}Nb. While these {1 1 1}Cu||{1 1 0}Nb KS interfaces have low shear strength [11], computer simulations have indicated that {1 1 2}KS interfaces possess high shear strength [16], and the stabilization of these interfaces may be useful for achieving optimal properties. The purpose of the present work is to further investigate the evolutions of microstructures in Cu–Nb subjected to SPD. Specifically, we examine the stability and evolution of ARB Cu/Nb nanocomposites subjected to a change in strain path from pure shear (ARB) to nearly simple shear using HPT. In addition we explore how the microstructure evolves at very high strains when it goes from two-dimensional (2-D) layers to threedimensional (3-D) interconnected phases. An additional motivation for the present work concerns the atomic mixing of Cu and Nb under SPD. These metals are highly immiscible, as indicated by the very large miscibility gap in the solid state as well as in the liquid state in the Cu–Nb equilibrium phase diagram. The evolution of mixtures of such highly immiscible elements during SPD is in fact not well understood. This contrasts with the case of moderately immiscible elements such as Cu–Ag [17], Cu–Fe [18,19] and Cu–Co [20], where it has been shown
179
that plastic deformation at low homologous temperatures can force the formation of solid solutions over the entire range of compositions (for a review see Refs. [21,22]). Solubility limits in highly immiscible alloys, on the other hand, are only slightly enhanced, even during SPD at very low temperatures. What is particularly interesting about these systems is that they undergo self-organization during low-temperature SPD; i.e., they select a length scale for phase separation that in the steady state depends only on the intensity (applied stresses) employed for SPD. For ball-milled Ni50Ag50 powders [23], a length scale of 20–40 nm has been reported while in Cu80Nb10Ag10 and Cu80Nb20, Nb precipitates reach a stable steady-state size of 20 nm after prolonged deformation by HPT or lowtemperature ball milling [24,25]. If ARB nanolayered Cu/Nb materials are destabilized by large HPT deformation, it is therefore intriguing to investigate whether they would self-organize, and if so at what length scale and by what mechanism. In this study, we have subjected ARB nanolayered Cu/Nb material to large HPT strains and characterized the evolution of the microstructure as a function of the applied strain, employing X-ray diffraction (XRD), transmission electron microscopy (TEM), scanning TEM (STEM), and atom probe tomography (APT). In Section 2, we detail the experimental procedures. In Section 3, we first report on the evolution of the two-phase microstructure, and then on the evolution of the crystallographic texture. We then discuss the implications of the experimental results in Section 4. 2. Experimental procedure 2.1. Fabrication of Cu–Nb nanolayered composites Nanolayered Cu/Nb composites were prepared by a combination of ARB, annealing and rolling steps, following the method detailed in Ref. [26]. Samples were composed of 50/50 volume fraction Cu/Nb, using starting materials of 99.99% and 99.97% purity, respectively, corresponding to a 51/49 ratio of weight fractions. The material selected for the present work was processed to a rolling strain of 10.93, resulting in a nominal thickness of 18 nm for the Cu and Nb layers. The final thickness of the as-fabricated ARB material was 380 lm. 2.2. SPD by high-pressure torsion HPT was applied to the above ARB nanocomposite to investigate its evolution when subjected to large shear strains. Discs of radius R = 5 mm were cut from the ARB sheet using electrical discharge machining. Samples then were subjected to HPT at room temperature, as described in Ref. [27]. An unconstrained anvil geometry was selected to minimize deformation heterogeneity, and the anvils were sand-blasted prior to torsion to prevent slipping [28]. The lower anvil was continuously rotated at
180
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
a rate of 1.2 rpm against the stationary upper anvil, with a confining hydrostatic pressure of 4.5 GPa. In order to cover a wide range of shear strains, four different values of rotation were used, 50°, 210°, 5 turns and 20 turns. During HPT deformation, if one assumes that the material flow is uniform and that the sample thickness is constant, the total shear strain c at a given distance r from the center of a disc of thickness t is given by: c¼
2pNr t
ð1Þ
where N is the number of turns applied by anvils [5,29]. In the present work, however, the sample thickness was observed to decrease significantly as N increased, in part because of the unconstrained anvil geometry. The thickness of the samples, measured by SEM imaging after HPT on cross sections, was found to decrease faster as the radial distance to the center of the disc increased. Four radial distances were selected for microstructural characterization, r = 1, 2, 3 and 4 mm, and thus the thickness evolution, t(N) for these four radial distances were measured and fitted using an exponential relaxation, as shown in Fig. 1. The shear strain evolution of the function was then deduced by integrating a differential form of Eq. (1), which includes the thickness evolution with HPT turns, t(N): dc ¼
2pr dN tðN Þ
ð2Þ
The resulting shear strains are reported in Table 1 for the four radial distances and the four HPT rotations N. As seen from Table 1, the combination of four samples and four radial distances covers a large spectrum of shear strains, from 2.0 to 6000. We note that the strain values obtained using the differential expression Eq. (2) can be significantly lower than the ones given by the standard Eq. (1) when using the final thickness, leading here to a reduction of 50% in the worst case (5 turns, r = 3 mm). For large
Fig. 1. Fitted thickness profiles for positions of HPT samples measured at r = 1, 2, 3 and 4 mm from the center of the disc after various HPT rotations. The horizontal black dashed line corresponds to the initial ARB sheet thickness.
Table 1 HPT shear strains calculated using Eq. (2) for the four samples and the four radial distances selected for the present work. For each r, the volume sampled was associated with a spot size of 100 lm. Two samples were subjected to 20 turns, the second one being only characterized at 2 mm and 3 mm, to provide supplemental data for the texture analysis.
50° 210° 5 turns 20 turns
1 mm
2 mm
3 mm
4 mm
2.0 8.8 107 960
4.2 18.8 286 1990 2490
6.5 30.3 560 3500 4500
10.8 68.5 1360 5970
strains, however, the sample thickness reached near steady-state values, and the differences did not exceed 20%. In practical terms, therefore, these errors have little effect on the results. In the following, the analyzed regions will be referred to using the strain values obtained from Eq. (2). This allows for a unique identification of sample and analyzed region since each strain value corresponds to a unique combination of rotation N and radial distance r, see Table 1. 2.3. Characterization methods For microstructural analysis, axes attached to the samples were defined as the normal direction (ND), the rolling direction (RD) and the transverse direction (TD), as illustrated in Fig. 2. The RD and TD directions refer to directions on the top surface of the samples during rolling, but these directions become less meaningful during HPT deformation since they are not invariant during torsion around the ND. For instance, for a given HPT rotation, the RD is rotated by an amount that increases with the depth of the analyzed region within the thickness of the sample. The samples were first characterized by XRD with a Panalytical Philips X’Pert X-ray diffractometer using Cu Ka radiation equipped with a monocapillary system. The top surface of the samples was polished to remove 20 lm to suppress contribution from surface artifacts. Local XRD scans and texture measurements were performed for four radial distances, r = 1, 2, 3 and 4 mm,
Fig. 2. Schematic illustration showing the axes convention and the radial distances used for analysis. The ND was common during rolling (ARB) and shearing (HPT).
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
using a 100 lm spot size, from spots along the initial TD, as illustrated in Fig. 2. Data were acquired for three poles for each phase: h1 1 1i, h2 0 0i and h2 2 0i for Cu, and h1 1 0i, h2 0 0i and h2 1 1i for Nb. Orientation distribution functions (ODFs) were then calculated after background subtraction. Minor sample misalignments due to sample mounting were corrected by applying cubic (m 3m) symmetry operations for the crystal symmetry and triclinic ( 1) symmetry operations for the sample symmetry. Pole figures (PFs) and inverse pole figures (IPFs) were then reconstructed from the ODFs for both Cu and Nb phases. MTex 3.1 software was used for texture plot and analyses [30,31]. The texture results were resolved and displayed in the (SD–ND) shear plane. Selected regions of the samples subjected to HPT were also analyzed by TEM, at radial distances of r = 2 and 4 mm. Thin foils were prepared using a focused ion beam (FIB) (FEI Strata Dual Beam 235 and Helios NanoLabe 600i) and conventional lift-out methods. Operating voltages from 2 kV to 30 kV and currents from 20 nA to 1 pA were selected to optimize sample quality while minimizing gallium implantation. The majority of these TEM samples were taken directly from the top surface of the samples, which had been previously polished for XRD data acquisition, as noted above. The electron transparent volumes probed in these samples were therefore 20–30 lm below the original sample surfaces. In order to assess the homogeneity of the microstructures, additional TEM samples were prepared for c = 286 from a cross-section obtained by cutting the 10 mm disc along a diameter parallel to the RD. TEM samples were then lifted out from this cross-section at various depths. The microstructures were analyzed by combining bright field TEM (BF-TEM), dark field TEM (DF-TEM), selected-area electron diffraction (SAED) and high-angle annular dark field in scanning transmission electron microscopy (HAADF-STEM) mode, using JEOL 2100LaB6 and JEOL 2010F microscopes, both operated at an accelerating voltage of 200 kV.
181
APT was employed to characterize the Cu and Nb spatial distribution in the sample deformed by HPT to 20 turns. The corresponding disc was cut in half and wedge-polished to a thickness of 30 lm using an Allied Multiprep System; posts with thickness 30 lm, length 200 lm and spacing 300 lm were then cut into the edge of the discs using a low-speed diamond saw. This enabled fabrication of multiple tips along the radius of a single HPT sample. APT tips were shaped to a radius of curvature of 50 nm on the posts using a Ga-ion beam in a FIB, FEI Strata DB235 or FEI Helios NanoLabe 600i. The APT measurements were performed at the Northwestern University Center for Atom Probe Tomography (NUCAPT). Analysis of the APT measurements was performed using the IVAS 3.6 (Imago Visualization and Analysis Software). 3. Results 3.1. Initial microstructure of ARB material Electron microscopy confirmed the previously reported nanolayering of the as-fabricated ARB Cu–Nb composite, as illustrated in Fig. 3. The average thickness of the layers, calculated from over 100 measurements, was 16.8 ± 8 nm. This value is consistent with the nominal value of 18 nm. Sharp Cu/Nb interfaces were observed, and SAED patterns, see Fig. 3b, indicate a strong preferred orientation for the Cu and Nb crystals, with [1 1 1]Cu// [1 1 0]Nb//RD. This is consistent with the {1 1 2}KS texture reported by Carpenter et al. [32] using precession electron diffraction. 3.2. Microstructural evolution with HPT deformation The evolution of phases in the ARB material with HPT strain was first evaluated using XRD, as summarized in Fig. 4. Bragg reflections of the body centered cubic (bcc)
Fig. 3. As-fabricated ARB Cu–Nb nanocomposite: (a) bright field TEM micrograph and (b) corresponding SAED pattern; the main reflections are indexed in (c). The rolling and transverse directions are shown in (a).
182
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
and face centered cubic (fcc) phases remained present, even at the largest strain of 5970, indicating the persistence of phase coexistence between Cu-rich and Nb-rich phases. The relative intensities of Cu peaks and Nb peaks, however, were significantly modified by HPT strain, pointing to the development of a new preferred texture. This point will be further detailed in Section 3.3. In addition, Bragg reflections were shifted and became broader, indicating that HPT deformation led to some amount of forced chemical mixing and to a reduction in grain size. The forced chemical mixing between Cu and Nb was analyzed using lattice parameter measurements obtained from the {1 1 0} Nb and {1 1 1} Cu reflections. Additional peaks could not be used as their intensities became very small at large HPT strains, as seen in Fig. 4. For the as-fabricated ARB material, however, four and five reflections could be detected for the Cu and the Nb phases, respectively, and the lattice parameters were taken as averages over these reflections. Vegard’s law could not be used to quantify the degree of chemical mixing from these lattice parameter measurements since the Cu–Nb system is highly immiscible, with heats of solution of 98.7 kJ mol1 for Nb in Cu, and 46.0 kJ mol1 for Cu in Nb [33]. Instead, we used the atomic volumes reported by Ashkenazy et al. [34] for Cu–Nb dilute solid solutions [34] obtained by molecular statics simulations. Expressions for the concentrations of Nb in Cu and Cu in Nb as a function of the solid solution lattice parameters were obtained using the following equation: a3 1 a3 A CB ¼ 0 ð3Þ A ðDV =V ÞB
Fig. 4. XRD spectra from the ARB and HPT-processed samples. For clarity only the spectra for radial distances r = 2 and 4 mm are displayed, and the spectra are shifted in intensity by an arbitrary amount. Corresponding strain values are indicated on the graph, see also Table 1. Background has been removed from each spectrum. Dashed lines correspond to Bragg reflections expected for pure Nb and pure Cu.
where C AB is the concentration of B species in matrix A, a the corresponding lattice parameter, a0 the lattice parameA ter of the pure A phase and ðDV =V ÞB the relative volume change resulting from the introduction of one B solute atom in a pure A crystal. These relative volume changes measured in the above atomistic simulations were 0.53 and 0.28 for Nb in Cu, and Cu in Nb, respectively [34]. The evolution of the solubility with strain is plotted in Fig. 5, together with the evolution of the grain size, obtained from {1 1 1} Cu and {1 1 0} Nb peak broadening using Scherrer’s equation. It should be noted that the initial Cu grain size is very close to the nominal ARB layer thickness, indicating that layers contain just a single grain, or at most two, along the ND direction; this is in agreement with our TEM observations. One notices also that the initial Nb grain size is slightly smaller than that of Cu, 13.5 nm compared to 17 nm. Similar trends have been reported for the Cu and Nb layer thickness values in similarly prepared samples [14]. Lastly, we note that the solubilities reached after ARB were small, typically less than 1 at.%. HPT deformation to strains less than 10 did not result in significant changes in composition or grain size, as seen in Fig. 5. Upon increasing the strain further, however, the grain size of the Cu and Nb phases were progressively
Fig. 5. (a) Chemical mixing vs. strain and (b) grain size vs. strain for ARB Cu–Nb nanocomposite subjected to HPT deformation.
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
reduced and reached steady-state values of 6.2 and 4.7 nm, respectively, at a strain of 286. The evolution of solubility within the two phases was less symmetric, as the Cu concentration in Nb started to increase continuously when the strain exceeded 10, reaching a steady-state value of 10 at.% for strains exceeding 1000. In contrast, the Nb concentration in Cu started only to increase at a strain of 68.5, and reached a maximum of 1.7 at.% for a strain of 286, before decreasing slightly to 1.5 at.%, at larger strains. The detailed evolution of the microstructures was investigated using TEM. For low HPT strains, c = 4.2, the microstructure was very similar to that of the as-fabricated ARB sample, with well-defined layers and sharp interfaces, as seen in Fig. 6a and b. One noticeable difference at this low strain, however, was a slight waviness of the interfaces, which is clearly visible in HAADF-STEM images, see Fig. 6b. At a strain of c = 10.8, the interfacial waviness became more pronounced, and, while the layered structure was still preserved in most regions, strain localization was observed in other areas, leading to the presence of folds and vortex-like features, see Fig. 6c and d. This strain localization will be discussed in Section 4. Further increasing the HPT strain resulted in a progressive destabilization of the nanolayered structure of the as-fabricated ARB material. After a strain of c = 68.5, the microstructure had entered a “mixed” state with nanolayers coexisting with equiaxed grains, as shown in Fig. 7a
183
and b. This microstructural transformation was completed at a strain of 286, leading to complete destabilization of the 2-D Cu/Nb and the formation of a 3-D nanocomposite, with very small grain sizes. Grain sizes were estimated to be 5 nm from Z-contrast STEM images, see Fig. 8, in agreement with the XRD data. TEM analysis performed on specimens prepared from different depths for c = 286 confirmed that this microstructure was present throughout the thickness of the sample, as seen by comparing Fig. 8a, taken close to the disc surface, and Fig. 8b taken near the mid-plane of the disc. This microstructure remained stable upon increasing the HPT strain to the largest value used in this study, 5970, as illustrated in Fig. 9a and b. It is thus concluded that this microstructure is stable under HPT deformation. In addition, selected area diffraction, Fig. 9b, indicates that this 3-D steady-state microstructure had a texture quite distinct from that of the as-fabricated ARB sample, as already suggested by the XRD spectra, see Fig. 4. This point will be discussed in Section 3.3. The 3-D structure of the fine scale nanocomposite stabilized by large HPT strains could not be easily accessed by TEM owing to overlapping effects through the thickness of the TEM foils. We thus employed APT to investigate this microstructure for the sample deformed by HPT to a shear stain of 1990. Reconstruction maps were built from the APT data by binning the detected atoms into contiguous 1 nm 1 nm 1 nm cubes, hereafter referred to as voxels, forming a simple cubic lattice. The total number
Fig. 6. Microstructure for an HPT strain of (a and b) 4.2 and (c and d) 10.8. Bright field TEM images are shown in (a and c) and Z-contrast STEM images in (b and d). In the Z-contrast images, Nb-rich layers appear brighter than Cu-rich ones.
184
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
Fig. 7. Microstructure for an HPT strain of 68.5: (a) bright field TEM micrograph, (b) Z-contrast STEM image, (c) SAED pattern of area shown in (a and d) indexing of the main reflections in (c).
Fig. 8. Microstructure for an HPT strain of 286. Z-contrast STEM images taken from (a) 25 lm below the surface, (b) 50 lm below the surface. The specimen thickness after HPT was 106 lm. The sample (b) has been obtained by FIB from a cross-section as detailed in Section 2.3.
of atoms detected in each voxel and the local composition were then obtained. The voxel composition is denoted in the reconstruction by its color, ranging from red (pure Cu) to blue (pure Nb) in 15 steps. The voxel concentration maps confirmed the coexistence of Cu-rich and Nb-rich regions, as illustrated in Fig. 10a. The difference in evaporation fields between Cu and Nb [35], however, led to local magnification effects that resulted in inhomogeneous distributions of atoms detected in each 1 nm3 voxel. These aberrations therefore limited the chemical resolution and prevented us from determining accurately the composition of the Cu-rich and Nb-rich phases. The connectivity of these regions was nevertheless assessed by using isoconcentration surfaces defined by specified threshold values of the Cu concentration. An example is given in Fig. 10b using a Cu concentration threshold of 40 at.%. While the size of the largest connected interface varied somewhat with the
Fig. 9. Microstructure for an HPT strain of 5970: (a) bright field TEM micrograph, (b) dark field TEM micrograph formed using a Nb (1 1 0) reflection; (c) SAED pattern of area shown in (a); (d) indexing of the main reflections visible in (c).
threshold value, for reasonable values of the threshold, ranging from 30 to 70 at.% Cu, there always exists one interface percolating through the sample. This analysis therefore indicates that the Cu-rich and the Nb-rich regions are also percolating through the sample. The length scale of the decomposition was determined from the APT data by defining cylinders along the z-axis (the tip axis), and plotting the Nb and Cu concentration profiles as a function of z. This axis was chosen because it is less sensitive to the local magnification aberrations [36,37]. Notice in Fig. 10a, for instance, that the Cu-rich and Nb-rich regions in the x and y directions are considerably more extended. The cylinder diameter was varied from 5 to 25 nm, and for diameters between 5 and 15 nm welldefined composition modulations were observed with an average period of 10 nm, independent of the cylinder diameter, as seen in Fig. 11. This length scale is in good agreement with the one deduced from Z-contrast TEM micrographs, see for instance Figs. 8 and 9. 3.3. Texture evolution The XRD 2h scans in Fig. 4, and TEM diffraction patterns in Fig. 9c, suggested significant crystallographic texture evolution during HPT of the ARB Cu/Nb nanolayered material. To further investigate texture evolution, we analyzed the XRD texture measurements, using IPFs and ODFs, with grain orientations represented by the three Euler angles (u1, U, u2) in Bunge’s notation. Prior to HPT straining, the textures of the as-fabricated ARB were consistent with the texture measurements
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
185
Fig. 10. APT reconstruction maps for a Cu/Nb sample deformed by HPT to a shear strain of 1990. (a) Concentration map, with color scheme varying from red (pure Cu) to blue (pure Nb). (b) Isoconcentration surfaces obtained by applying a Cu concentration threshold of 40 at.%. The green surface displayed is the largest connected set of these surfaces. Axes are in nm and apply to both figures. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 11. One-dimensional composition profiles generated from the data shown in Fig. 10, using a cylinder along the tip axis, (a) 5 nm and (b) 15 nm in diameter.
reported by Carpenter et al. [26,32] on similarly prepared samples. HPT deformation led to a rapid change from
the initial ARB texture. For strains as small as 4.2, ODFs indicate that the initial textures have been largely erased, as seen in the ODF sections u2 = 0° and u2 = 45°, shown in Figs. 12 and 13 for the Cu and the Nb phases, respectively. It is noteworthy that, at that strain value, the layered structure has not yet been destabilized, see Fig. 6a and b. At larger strains, a new steady-state texture evolves, as we now describe. For the Cu phase, as strain increases a preferred texture develops progressively, with a strong preference for ND// {1 1 1}, i.e., the so-called A-fiber, see the ODFs in Fig. 12 as well as the PFs and IPFs, Figs. S.1 and S.2 in Supplementary information. The main ideal texture components present at large strains were the A=A, corresponding to ð1 1 1Þ½1 1 0 and ð1 1 1Þ½1 1 0, and the A1 =A2 components, corresponding to ð1 1 1Þ½1 1 2 and ð1 1 1Þ½1 1 2. Strains in excess of 2500 were needed for the complete stabilization of this texture. The texture evolution of the Nb-rich phase is more complex, as illustrated in Fig. 13. While low HPT strains greatly suppressed the initial ARB texture, transient textures developed for strains ranging from 18.8 all the way up to 1360, and at larger strains the Nb texture became nearly random. Similarly to the Cu phase, strains greater than 2500 were necessary to fully stabilize this texture. In order to investigate whether the local texture variations could account for the unusual texture development, texture analysis was performed using a larger X-ray beam, 2 mm 2 mm, which we refer to as macro XRD, instead of the 100 lm beam size used in the results reported in the previous paragraphs. A comparison of the ODFs obtained with these two different beam sizes indicates that there is a good agreement between the textures at large strains, see Fig. 14, with an A-fiber texture in Cu and a near random texture in Nb. In the macro XRD Cu results, the intensities of the twin-symmetric A and A components are very similar, as expected from symmetry. Note that the comparison between micro and macro XRD results is performed here
186
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
Fig. 12. U2 = 0° and U2 = 45° sections of Cu phase ODFs for selected HPT strains. Scale bar shown is in multiples of random distribution (min = 0, max = 4). Locations of common simple shear ideal texture components are given at the bottom.
only for strains large enough for the texture be in steady state since the large X-ray beam covered regions subjected to a wide range of strain, e.g., 5100–12,550 when the X-ray beam is centered on r = 4 mm, as in Fig. 14. 4. Discussion HPT has been employed to evaluate the stability of ARB Cu/Nb nanolayered composites to large shear strains. We found that HPT led to profound morphological, microstructural and chemical evolutions of the Cu/Nb nanolayered material, eventually leading to the destabilization of the 2-D nanocomposite in favor of an ultrafine 3-D nanocomposite. For the ARB material investigated here, with an initial nominal layer thickness of 18 nm, TEM revealed
that the layered structure remained largely stable for strains up to 10, although interfaces became increasingly wavy. This observation is consistent with prior reports of wavy layered patterns forming during the mechanical alloying of ductile metals [21], and during HPT deformation of two-phase materials [38]. Atomistic simulations and modeling have also shown that the co-deformation of layered structures induces a kinetic roughening of interfaces [27], resulting in wavy interfaces. When increasing the HPT strain to 68.5, the layered structure became severely disrupted and partly replaced by a 3-D structure, with coexisting Cu phases and Nb-rich phases. For a strain of 286, this morphological transformation was complete and resulted in the stabilization of a 3-D nanocomposite, which remained present for the largest shear strains used in this
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
187
Fig. 13. U2 = 0° and U2 = 45° sections of Nb phase ODFs for selected HPT strains. Scale bar shown is in multiples of random distribution (min = 0, max = 4). Locations of common simple shear ideal texture components are given at the bottom.
study, c 6000. Analysis of the XRD data indicates that the grain size of the Cu and Nb phases reached a steady state at a strain of 286, whereas the chemical composition of these phases appear to continue to evolve slightly, finally reaching a steady state for strains exceeding 3500. In contrast to moderately immiscible alloy systems such as Cu–Ag [27,39,40] or Cu–Fe [41], a large HPT strain could thus not force the highly immiscible Cu and Nb elements into a solid solution. This is consistent with the stabilization of nanocomposites in Cu–Nb mixtures subjected to SPD by other techniques such as ball milling [42] and wire drawing [43,44], as well as in other highly immiscible alloy systems such as Ni–Ag [45,46], Fe–Ag [47,48], Cu–Mo [49] and Cu–W. It should be noted here that the report of complete alloying of Cu90Nb10 mixtures by ball milling
[24,50] is questionable since the Cu lattice parameter never exceeded 0.364 nm (indicating a Nb solubility of 1.6 at.%); however, ball milling was performed at liquid N2 temperatures, and partial amorphization at the interfaces is a possibility. Our analysis of atomic mixing in Section 3.1 suggests that the Nb content was at most 2 at.%. Such a low solubility is in fact consistent with the APT data obtained from wire drawing alloys [44], HPT [25] of Cu–Nb–Ag alloys and indeed the ball milling result of Refs. [24,49]. Excess Nb atoms can form very small precipitates, which would then be very difficult to detect by XRD. It is interesting to note that the solubility of Cu in Nb was significantly larger, saturating at 10 at.% at large strains, as shown in Fig. 5. Similar values of 10–15 at.% have in fact been measured by APT in wire-drawn Cu/Nb
188
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
Fig. 14. Comparison between the ODFs obtained from XRD data collected with a 100 lm beam size (micro XRD), and those obtained with a 2 mm beam size (macro XRD), for the ARB material and for the samples deformed to large strains by HPT. Top row for Cu, and bottom row for Nb.
composites [43]. This larger nonequilibrium solubility of Cu in Nb compared to Nb in Cu can be rationalized by the lower heat of solution of Cu in Nb, 0.6 eV, compared to 1.3 eV for Nb in Cu [34]. The evolution of textures in the Cu and Nb phases with strain also revealed new insights about SPD in strongly immiscible alloys. First, the ARB texture was already largely destabilized after a strain of 4.2, see Figs. 12 and 13. This is consistent with molecular dynamics (MD) simulation results indicating that the {1 1 2}KS orientation relationship is not stable under simple shear deformation [51,52]. Second, the textures stabilized at very large strains, an A-fiber texture for Cu and a nearly random texture for Nb, appear to be quite distinct from the textures reported for nominally pure Cu deformed by HPT [53,54] and the ones expected for bcc elements, such as Fe [55], deformed in simple shear. In particular in the case of Cu, the dominant texture components identified by Enikeev et al. [53] were B=B, with some significant C component as well. None of these components is present in the Cu/Nb nanocomposites at large HPT strains, as shown in Fig. 12. One should be careful in comparing these results, however, since the results reported for monolithic Cu have been obtained using quasi-constrained HPT, for strains not exceeding 300, and on pure Cu. Third, the Cu and Nb phase textures required unusually large strains to reach steady state, and furthermore the Nb phase presents strong transient textures for strains ranging from 18.8 to 1360. These results suggest that the texture evolution and stabilization in these Cu/Nb nanocomposites is strongly influenced by the constraints imposed by 3-D nanoscale phase co-existence. In that context, we note that the transient textures in Nb started to develop at a strain of
18.8–68.5, which coincides with the destabilization of the layered morphology. Moreover the same large strains required for the stabilization of textures are also the ones required for the Cu and Nb solubilities to reach steady state, see Fig. 5. It is remarkable that at large strains the Cu phase developed a fairly strong texture, with maximum intensities in the ODFs of approximately four times that of a random distribution, while the Nb phase reached a near random texture. This result suggests that, at steady state, plastic deformation is mainly accommodated by deformation of the Cu phase. One intriguing aspect of the transformation of the ARB nanolayered structure to a 3-D nanocomposite structure pertains to the mechanisms driving this morphological evolution. It is often observed that layered structures can unravel by a pinch-off mechanism. The kinetic roughening of interfaces imposed by plastic deformation could certainly lead to the pinching off of layers, especially considering the small initial layer thickness used in the present study. On the other hand, we observed the presence of S-shape or Z-shape folds and vortex-like features, see Fig. 6c and d and the low magnification STEM image Fig. 15. These features suggest that another plastic instability might also be taking place. Macroscopic swirls and vortex-like features have been reported for duplex stainless steels [56] processed by HPT, and ascribed to the development of a Kelvin–Helmholtz instability. The scale of the swirls in those studies, however, were typically hundreds of microns, and thus much larger than the folds observed here, and the geometry of those swirls seems to be directly derived from the geometry of the torsion test. The folds that we observed are, in fact, more similar to those recently reported in hypoeutectic Cu–Ag alloys subjected to HPT [39]. These
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
Cu–Ag folds have been linked to shear banding, which does not appear to be present in our Cu/Nb samples. It is also possible that the folds in Cu/Nb are related to the folds forming in layered geological materials subjected to shear deformation by plate tectonics [57–59]. Despite the vast difference in length scale, the physical processes involved in both cases are similar. Further work is needed to elucidate the origin of folds in nanolayered materials subjected to HPT. The 3-D nanocomposites stabilized at large HPT strains possess remarkable features. First the scale of phase co-existence is very small. Z-contrast STEM imaging indicates that this scale is 5–10 nm, see Fig. 8. A complete characterization of this structure is very challenging because the scale is finer than the typical TEM sample thickness, leading to significant superposition for the near equiatomic composition of the ARB samples. APT was thus used to further evaluate these nanostructures. Although APT lacks atomic resolution because of the large difference in evaporation fields for Cu and Nb, atom reconstruction maps confirmed the 3-D coexistence of Cu-rich and Nb-rich phases, with a characteristic length scale of 10 nm, as illustrated in Fig. 11. We note also that isoconcentration surfaces generated in this study establish that both phases are connected, which is not surprising for mixtures with nearly equiatomic compositions. Moreover, TEM dark field imaging and XRD peak broadening showed that the Cu and Nb grain sizes are also of the order of 5 nm. It is therefore concluded that each Cu-rich or Nb-rich ligament comprises 1–2 grains. As indicated in Section 1, one of the interests in developing Cu/Nb nanolaminates and nanocomposites resides in their
Fig. 15. Low magnification Z-contrast STEM image for a strain of 10.8. Note the presence of folds and vortex-like features. The near-vertical dark bands result from foil thickness variations introduced by FIB sample preparation. Nb-rich layers appear brighter than Cu-rich ones.
189
potential for absorbing point defects and dislocations, owing to their high interfacial area. In that respect, the 3D nanocomposites stabilized by large HPT strains offer a high potential. The specific interfacial area is estimated from APT isoconcentration surfaces to be 0.30 nm1, compared to 0.056 nm1 for the as-fabricated ARB material. Prior studies on Cu/Nb and Cu/V layered structures have indicated, however, that the efficiency for absorbing point defects or He atoms is also a function of the crystallographic orientation of the interfaces [16,60]. It would thus be interesting to determine the He storage capacity of these 3-D nanocomposites, and to compare it with that measured on Cu/Nb nanolayers with well-defined interfaces [60]. The steady-state grain size of the 3-D Cu/Nb nanocomposites is much smaller than the ones obtained after room temperature HPT deformation of the pure metals, typically 150 nm for pure Cu [61,62] and 100–150 nm for pure Nb [63]. In the pure Cu samples deformed by HPT, however, grain size distribution and microtexture developments provided clear evidence for recrystallization [54]. In contrast, Lim and Rollett [10] showed that, in Cu–Nb layered composites, recrystallization in the Cu layer was suppressed during rolling when the layer thickness was below 1 lm. Our analysis of the microstructure and texture of 3-D nanocomposites stabilized by HPT is consistent with this result, leading to very small grain sizes. We also note that the steady-state grain size stabilized by ball milling is also quite small, typically 5 nm [24], probably for similar reasons. One interesting fundamental question raised by the stabilization of these 3-D nanocomposites concerns the selection of the length scale of phase coexistence. In the case of moderately immiscible alloys, it has been proposed that nanoscale phase separation results from the competition between thermally activated phase separation and forced mixing [27,40,64]. Recent atomistic simulations support this simple picture and the corresponding analysis of the length scale selection resulting from this dynamical competition [65]. Such an understanding is missing in the case of highly immiscible systems. In a recent investigation of phase evolution in Cu80Nb10Ag10 subjected to low temperature SPD by HPT and ball milling, Wang et al. [25] showed that, in the absence of thermally activated diffusion, Nb precipitates reached a well defined steady-state size, 20 nm, and that this size results from plasticity-driven mechanisms promoting precipitate shrinkage as well as re-precipitation. More work is needed to determine whether a similar rationalization could also apply to the bi-connected Cu–Nb 3-D composites stabilized by HPT. The starting material for the present investigation was an ARB Cu/Nb nanolaminate with (1 1 2)KS orientation, thus with high shear strength interfaces. It would be interesting to investigate the evolution of Cu/Nb nanolaminates with low interface shear strength, such as PVD Cu/Nb nanolayers where interfaces are predominantly (1 1 1)KS [11]. Hoagland et al. [66] calculated that dislocations would be attracted by these weak interfaces, and these authors
190
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191
predicted that “weaker boundaries afford a greater barrier to the transport of slip across the boundary than very strong boundaries”. This prediction was in fact recently confirmed by MD simulations [51,52] showing that, in simple shear deformation, (1 1 1)KS and NW Cu/Nb interfaces are much more resistant to chemical mixing than (1 1 2)KS interfaces. At large enough shear strains, however, the activation of multiple slip systems should eventually lead to the formation of interface steps, leading to interface roughening, and to the progressive destabilization of the initial layered structure. We thus anticipate that the threshold strain required for destabilizing layered structures with low shear strength interfaces would be larger than the one reported here for (1 1 2)KS interfaces, but, in the limit of very large strains, a 3-D nanocomposite would nevertheless be stabilized. Evaluating these predictions is left for future work. Lastly, it is interesting to note that wire drawing experiments and atomistic simulations indicated that the forced mixing of Cu and Nb could lead to the formation of an amorphous phase [34,43,67]. No amorphous phase was identified in the present work by either TEM or XRD, but we note that the detection of such a phase in the 3-D nanocomposites stabilized at large HPT strains would be very difficult owing to the fine scale of phase coexistence compared to typical TEM foil thickness. 5. Conclusion Nanolayered Cu/Nb composites fabricated by ARB, with a nominal layer thickness of 18 nm, were subjected to large shear strains using HPT at room temperature. The evolution of the microstructure proceeded in successive steps. First, at strains of 4.2 and beyond, XRD characterization indicated that the {1 1 2}KS orientation relationship and texture stabilized by ARB were destabilized. TEM showed that the layered morphology resisted HPT deformation to strains up to 10.8. As the HPT strain was further increased, a 3-D microstructure replaced the initial ARB layered structure, with a reduction in the length scale of phase coexistence. The average grain size of both Cu and Nb was also reduced, down to 5 nm at a strain of 286, while strains as large as 3500 were required for the chemical composition of these phases to reach steady state. The steady-state solubility of Nb in Cu and Cu in Nb were found to reach 1.5 at.% and 10 at.%, respectively. This significant difference in solubility was assigned to the large asymmetry of the thermal solubility between the Cu-rich side and the Nb-rich side of the equilibrium Cu–Nb phase diagram. At large strains, the textures of the Cu and Nb phases stabilized around an A-fiber texture and a near random texture, respectively. These textures appear thus to be distinct from those reported in monolithic Cu and Nb subjected to HPT, suggesting a significant effect of the 3-D Cu–Nb nanoscale phase coexistence on texture. The length scale of these 3-D nanocomposites was very fine, and determined by
TEM and APT to be 5–10 nm. The present findings suggest that HPT can be used to stabilize interconnected nanocomposites at very small scales and with a very high density of interfaces. In addition to its stability under HPT deformation, this morphology thus offers attractive features for trapping point defects and rare gas atoms under irradiation for the design of radiation-resistant materials. Acknowledgements This work was supported as part of the Center for Materials at Irradiation and Mechanical Extremes, an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences under Award Number 2008LANL1026. APT was performed at the Northwestern University Center for Atom-Probe Tomography (NUCAPT) whose localelectrode atom-probe (LEAP) tomograph was purchased and upgraded with funding from NSF-MRI (DMR0420532) and ONR-DURIP (N00014-0400798, N000140610539, N00014-0910781) Grants. Instrumentation at NUCAPT was supported by the Initiative for Sustainability and Energy at Northwestern (ISEN). NUCAPT is a Shared Facility at the Materials Research Center of Northwestern University, supported by the National Science Foundation’s MRSEC program (DMR-1121262. Stimulating discussions with Drs. A. Misra and J. Carpenter (LANL), and Prof. A. Rollett (CMU), are gratefully acknowledged. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.actamat.2014.03.040. References [1] Vollath D. Nanomaterials: an introduction to synthesis, properties and application. Weinheim: Wiley-VCH; 2008. [2] Gleiter H. Prog Mater Sci 1989;33:223. [3] Birringer R. Mater Sci Eng A 1989;117:33. [4] Meyers MA, Mishra A, Benson DJ. Prog Mater Sci 2006;51:427. [5] Valiev RZ, Islamgaliev RK, Alexandrov IV. Prog Mater Sci 2000;45:103. [6] Saito Y, Utsunomiya H, Tsuji N, Sakai T. Acta Mater 1999;47:579. [7] Valiev R. Int J Mater Res 2009;100:757. [8] Zhilyaev AP, Langdon TG. Prog Mater Sci 2008;53:893. [9] Jha SC, Delagi RG, Forster JA, Krotz PD. Metall Trans A 1993;24:15. [10] Lim SCV, Rollett AD. Mater Sci Eng A 2009;520:189. [11] Demkowicz MJ, Hoagland RG, Hirth JP. Phys Rev Lett 2008;100:136102. [12] Misra A, Demkowicz MJ, Zhang X, Hoagland RG. JOM 2007;59:62. [13] Zhang L, Demkowicz MJ. Appl Phys Lett 2013;103:061604. [14] Beyerlein IJ, Mara NA, Carpenter JS, Nizolek T, Mook WM, Wynn TA, et al. J Mater Res 2013;28:1799. [15] Zheng S, Beyerlein IJ, Carpenter JS, Kang K, Wang J, Han W, et al. Nat Commun 2013;4:1696.
E.H. Ekiz et al. / Acta Materialia 72 (2014) 178–191 [16] Demkowicz MJ, Thilly L. Acta Mater 2011;59:7744. [17] Klassen T, Herr U, Averback RS. Acta Mater 1997;45:2921. [18] Eckert J, Holzer JC, Krill CE, Johnson WL. J Appl Phys 1993;73:2794. [19] Huang JY, Yu YD, Wu YK, Li DX, Ye HQ. Acta Mater 1997;45:113. [20] Gente C, Oehring M, Bormann R. Phys Rev B 1993;48:13244. [21] Suryanarayana C. Prog Mater Sci 2001;46:1. [22] Ma E. Prog Mater Sci 2005;50:413. [23] Zghal S, Bhattacharya P, Twesten R, Wu F, Bellon P. Metastable, mechanically alloyed and nanocrystalline materials 2002;386–3:165. [24] Botcharova E, Heilmaier M, Freudenberger J, Drew G, Kudashow D, Martin U, et al. J Alloys Compd 2003;351:119. [25] Wang M, Averback RS, Bellon P, Dillon SJ. Acta Mater 2014;62:276. [26] Carpenter JS, Vogel SC, LeDonne JE, Hammon DL, Beyerlein IJ, Mara NA. Acta Mater 2012;60:1576. [27] Pouryazdan M, Schwen D, Wang D, Scherer T, Hahn H, Averback RS, et al. Phys Rev B 2012;86:144302. [28] Pippan R, Scheriau S, Hohenwarter A, Hafok M. Mater Sci Forum 2008:584–6. Part 1:16. [29] Vorhauer A, Pippan R. Scripta Mater 2004;51:921. [30] Hielscher R, Schaeben H. J Appl Crystallogr 2008;41:1024. [31] Bachmann F, Hielscher R, Schaeben H. Solid State Phenomena 2010;160:63. [32] Carpenter JS, Liu X, Darbal A, Nuhfer NT, McCabe RJ, Vogel SC, et al. Scripta Mater 2012;67:336. [33] Chakrabarti DJ, Laughlin DE. Bull Alloy Phase Diagr 1982;2:455. [34] Ashkenazy Y, Vo NQ, Schwen D, Averback RS, Bellon P. Acta Mater 2012;60:984. [35] Miller MK. Atom probe field ion microscopy. Oxford: Clarendon Press; 1996. [36] Vurpillot F, Bostel A, Blavette D. Appl Phys Lett 2000;76:3127. [37] Isheim D, Hellman OC, Seidman DN, Danoix F, Blavette D. Scripta Mater 2000;42:645. [38] Cao Y, Wang YB, Alhajeri SN, Liao XZ, Zheng WL, Ringer SP, et al. J Mater Sci 2010;45:765. [39] Tian YZ, Wu SD, Zhang ZF, Figueiredo RB, Gao N, Langdon TG. Acta Mater 2011;59:2783. [40] Arshad SN, Lach TG, Pouryazdan M, Hahn H, Bellon P, Dillon SJ, et al. Scripta Mater 2013;68:215. [41] Quelennec X, Menand A, Le Breton JM, Pippan R, Sauvage X. Philos Mag 2010;90:1179. [42] Benghalem A, Morris DG. Scripta Metall Mater 1992;27:739.
191
[43] Sauvage X, Renaud L, Deconihout B, Blavette D, Ping DH, Hono K. Acta Mater 2001;49:389. [44] Raabe D, Ohsaki S, Hono K. Acta Mater 2009;57:5254. [45] Xu J, Herr U, Klasse T, Averback RS. J Appl Phys 1996;79:3935. [46] He JH, Sheng HW, Schilling PJ, Chien CL, Ma E. Phys Rev Lett 2001;86:2826. [47] Angiolini M, Krasnowski M, Mazzone G, Montone A, Urchulutegui M, VittoriAntisari M. Nanophase Mater 1995;195:13. [48] Shingu PH, Huang B, Kuyama J, Ishihara KN, Nasu S. In: Artz A, editor. New materials by mechanical alloying techniques. Deutsche Gesellschaft fu¨r Metallkunde; 1989. p. 319. [49] Botcharova E, Freudenberger J, Schultz L. J Mater Sci 2004;39:5287. [50] Botcharova E, Freudenberger J, Schultz L. J Alloys Compd 2004;365:157. [51] Vo NQ, Zhou J, Ashkenazy Y, Schwen D, Averback RS, Bellon P. JOM 2013;65:382. [52] Zhou J, Averback RS, Bellon P. Acta Mater 2014, in press. [53] Enikeev NA, Schafler E, Zehetbauer M, Alexandrov IV, Valiev RZ. Mater Sci Forum 2008:584–6. Part 1:367. [54] Al-Fadhalah KJ, Alhajeri SN, Almazrouee AI, Langdon TG. J Mater Sci 2013;48:4563. [55] Li SY, Gazder AA, Beyerlein IJ, Davies CHJ, Pereloma EV. Acta Mater 2007;55:1017. [56] Cao Y, Kawasaki M, Wang YB, Alhajeri SN, Liao XZ, Zheng WL, et al. J Mater Sci 2010;45:4545. [57] Carreras J, Druguet E, Griera A. J Struct Geol 2005;27:1229. [58] Schmalholz SM, Schmid DW. Philos Trans R Soc A 2012;370:1798. [59] Llorens MG, Bons PD, Griera A, Gomez-Rivas E, Evans LA. J Struct Geol 2013;50:209. [60] Demkowicz MJ, Bhattacharyya D, Usov I, Wang YQ, Nastasi M, Misra A. Appl Phys Lett 2010;97:161903. [61] Jiang HG, Zhu YT, Butt DP, Alexandrov IV, Lowe TC. Mater Sci Eng A 2000;290:128. [62] Horita Z, Langdon TG. Mater Sci Eng A 2005;410:422. [63] Popov VV, Popova EN, Stolbovskiy AV. Mater Sci Eng A 2012;539:22. [64] Odunuga S, Li Y, Krasnochtchekov P, Bellon P, Averback RS. Phys Rev Lett 2005;95:045901. [65] Schwen D, Wang M, Averback RS, Bellon P. J Mater Res 2013;28:2687. [66] Hoagland RG, Hirth JP, Misra A. Philos Mag 2006;86:3537. [67] Vo NQ, Averback RS, Ashkenazy Y, Bellon P, Wang J. J Mater Res 2012;27:1621.