Microstructural and texture evolution of Copper-(Chromium, Molybdenum, Tungsten) composites deformed by high-pressure-torsion

Cu-refractory metal composites containing Cr, Mo or W were subjected to severe plastic deformation using room temperature high-pressure torsion (HPT). A lamellar microstructure developed in each of the composites at equivalent strains of ~75. The refractory metals developed {hkl}<111>fibre textures with a slight tilt to the tangential direction. This texture was stronger and more clearly defined in Mo and W than in Cr. By applying additional HPT deformation to these samples, perpendicular to the original shear strain, it was found that the lamellar structure of Cu30Mo70 and Cu20W80 (wt.%) composites could be retained at high equivalent strains and the refractory layer thickness could be reduced to 20-50 nm in Cu20W80 and 10-20 nm in Cu30Mo70. Although neighbouring regions of the microstucture were aligned and there was evidence of local texture in both composites, the bulk texture of Cu30Mo70 became weaker during this second step of HPT deformation. This was attributed to the refractory metal lamellae being discontinuous and imperfectly aligned. This work shows that it is possible to form ultrafine composites of Cu-group VI refractory metals via high-pressure torsion, with namolamellar structures being possible where there is a sufficient volume fraction of Mo or W.


Introduction
Copper-refractory composites are important industrial materials consisting of varying proportions of copper and one or more of the refractory metals, namely Nb, Ta, Cr, Mo, W and Re. These composites are employed in high temperature, high current applications such as heat sinks [1], electrodes in thermoelectric devices [1][2][3] and in aerospace components [4] which demand high thermal and/or electrical conductivity of Cu and the capability to operate at higher temperatures than copper alone can sustain. The high melting points (from 1907 • C(Cr)-3422 • C (W)) [5] and low thermal expansion coefficients (4.5-7.3 µm m −1 K −1 ) [6] of the refractory elements provide the required high temperature strength and dimensional stability.
The lamellar microstructure endows these composites with excellent thermal stability [12,15] by restricting diffusion through the alternating layers. Fracture toughness studies on lamellar composites have also shown a strong anisotropy due to crack deflection along the lamellae [16]. Thus, the ability to develop nanolamellar composites of refractory elements in Cu matrix will open-up the possibility for developing high strength, tough and thermally stable components for high temperature applications.
The production of such nanolaminates relies on the high ductility of the group V refractory metals, with ductile to brittle transformation temperatures (DBTTs) of ∼77 K [17,18] and ∼ 4 K [19] for Nb and Ta, respectively. Unfortunately the group VI refractory metals (Cr,Mo,W) are more brittle, with DBTTs in the range of 553-573 K for 99.96% purity Cr [20], ≈ 373 K [21] for Mo and 400 K for W [22].
High-pressure torsion (HPT) is a technique for applying high strains under quasi-hydrostatic conditions, thus permitting the deformation of brittle materials. This has been applied to Cu-Cr [23][24][25] Cu-W [26][27][28] and Cu-Mo [29] composites. Extensive deformation of Cu-Cr [24] results in the formation of equiaxed, nanometer scale grains. A similar microstructure was reported after two-step deformation of Cu50Mo50 [29]. In contrast, a Cu30Mo70 (wt.%) composite retained a lamellar structure with layer widths of ∼5 nm for Cu and ∼10-20 nm for Mo [29]. It appears that Cu-10wt.%W and Cu-25wt.%W composites also retained lamellar structures for strains up to at least 900 [28]. However, each of these studies were conducted using different sample sizes, pressure and applied strain, making it difficult to compare various Cu-refractory metal composites.
This work sets out to provide a comparison of the microstructural evolution of copper composites containing a group VI refractory element (Cr, Mo,W) during equivalent roomtemperature HPT deformation. In particular it examines the extent of microstructural refinement, the formation and decomposition of lamellar microstructures and the development of crystallographic texture during deformation.

Experimental details
The materials used in this work were provided by Plansee SE, Austria. Cu-Mo and Cu-W composites took the form of liquid-metal infiltrated (LMI) plates, with compositions of Cu30Mo70, Cu50Mo50 and Cu20W80 (All compositions are given in wt.%). Cu-Cr was in the form of 4 mm slices sectioned from directionally-solidified Cu57Cr43 billets.
Step 1 HPT was carried out at room temperature on a 4000 kN apparatus rotating at 0.0625 revolutions per minute (rpm) under an applied pressure of 4.5 GPa. The sample thickness, t and number of revolutions, N, was adjusted to achieve a nominal von Mises' equivalent strain, ǫ eq of ∼75 (equal to the shear strain, γ, divided by √ 3) at a given radius, r, based on the equation [30]: Sample were constrained within tool steel anvils of diameter 30 mm and anvil gap 3.5 mm (Cu-Mo, Cu-W) or 1.5 mm (Cu-Cr), corresponding to a nominal sample thickness of 7.5 mm and 3.5 mm, respectively. Table 1 lists the deformation conditions for each composite, and gives the sampling radius and equivalent strain used for XRD analysis.  Vickers microindentation hardness testing was conducted on cross-sectioned samples of each HPT-deformed composites, using 300 g or 500 g loads.
The as-received and post-deformation microstructures of each composite were examined using a Gemini 1525 scanning electron microscope (SEM). Sample preparation was via mechanical polishing and grinding, with final polishing using a Buehler vibromet polisher.
Composites were then subjected to higher-strain HPT deformation to further refine the Because the shear plane in step 2 is perpendicular to that in step 1 this method has been shown to be particularly effective for microstructural refinement [31]. Unfortunately, the 4 mm thickness of the Cu-Cr starting material did not permit this procedure. Therefore, 8 mm diameter discs of the as-received material with a nominal height of 1 mm were deformed to 100 rotations (This was considered acceptable since previous studies on this material showed minimal change in hardness and grain size between deformation of 100 and to 1000 rotations [25]). Hardness testing was conducted to determine appropriate strain levels for further analysis. The sample size and deformation conditions for step 2 HPT samples are set out in Table 2.
Suitably-deformed HPT discs were mechanically ground and polished, dimpled and thinned to perforation using a Gatan 691 precision ion polisher. The electron transparent region corresponded with a radial distance of ∼3 mm. The corresponding strains (ǫ 1 :ǫ 2 ) were ∼50:1000 for Cu-Mo , ∼100:700 for Cu-W and ∼1500 for Cu-Cr(ǫ 2 only) composites. These foils were examined using a JEOL 2100F transmission electron microscope, operating at 200 kV using high-angle annular dark field scanning TEM (HAADF-STEM). Inner and outer collection angles of 65.51 mrad and 174.9 mrad, respectively, were used to provide atomic contrast imaging.
The bulk crystallographic texture of samples deformed till ǫ eq ∼ 75 (i.e, first step of HPT) were measured using a Rigaku Smartlab system with Cu Kα radiation. An approximate ∼4 mm×4 mm samples normal to the radial direction were used for these measurements. The defocusing corrections due to sample tilts were made using powders or powder compacts of the same size and composition. The texture measurements on the second step HPT deformed samples were performed with synchrotron radiation at the high energy materials science beamline P07 (operated by the Helmholtz-Zentrum Geesthacht) of the PETRA III synchrotron facility at DESY in Hamburg with photon energies of 101 keV. For this experiment, radial slices with a nominal width of 1 mm were cut from the HPT discs (see Figure 1(c)). The transmitted diffraction rings were measured with a beam size of 0.8 mm (high) x 1 mm (wide). The samples were rotated about their radial axis by 180 • with an interval of 5 • in order to obtain threedimensional texture information. Further, the intensity of the diffracted beams were corrected for the attenuation. A LaB 6 NIST standard was used for calibrating the detector distance and orientation with respect to the sample. For both techniques, the measured pole figures were analysed with Labosoft software. For W and Mo (110)

Hardness
The hardness of the composites after HPT deformation is plotted against the applied equivalent strain (ǫ eq ) in Figure 2. The change in hardness during the first step of HPT deformation is shown in Figure 2(a). Each of the composites underwent a rapid increase in hardness commencing at strains of ǫ 1 ∼20, with the greatest increases in hardness occurring in the Cu-Mo composites. The arrow indicates the strain value of ǫ 1 ∼75 at which the SEM and XRD investigations described hereafter (Sections 3.3.1 and 3.3.2, respectively) were performed. Figure 2(b) shows the hardness of the composites after step 2 deformation. The hardness of Cu57Cr43 in Fig. 2(b) increases gradually and approximately linearly with log(ǫ) while that of the Cu50Mo50 composite reaches a saturation in hardness. The hardness of the Cu30Mo70 and Cu20W80 composites continues to increase substantially, indicating that these composites have not reached a saturated state. Although the Cu20W80 composite was harder than Cu30Mo70 in the asreceived condition, the latter composite became substantially harder after HPT deformation, owing to more extensive microstructural refinement, as is shown in the following section.

Initial microstructure
Backscattered electron micrographs of the as-received composites are provided in Figure 3. The Cu-Cr composite ( Fig. 3(a)) had a coarse structure consisting of irregular Cr particles within a Cu matrix. The mean projected area of the Cr particles was 1340 µm 2 (equivalent in projected area to a spheroidal particle of diameter 41 µm) although particles with Feret diameters from 5-100 µm were observed. Micron-scale pores are evident, primarily at the interfaces between the two components. The Cu-Mo (Figure 3(c),3(b)) and Cu-W (Figure 3(d)) composites contained substantially finer, equiaxed particles of the refractory phase. The average diameter of the refractory metal particles was 3.8 µm for Cu30Mo70, 5.0 µm for Cu50Mo50 and 3.6 µm for Cu20W80.  3.3 HPT deformed (ǫ ∼ 75)

Microstructure
All composites adopted a lamellar morphology after HPT deformation to a strain of ∼75. Figure 4 shows representative backscattered electron images of each composite, with the shear plane horizontal in each micrograph. The refractory particles have undergone considerable plastic deformation, elongating in the shear direction, although the orientation of individual particles can vary substantially. The level of microstructural refinement is clearly greater at the rim of the sample (Figures 4(b,d,f,h)) than at the centre (a,c,e,g), due to the radial dependence of the equivalent strain. The microstructures of the Cu-W and Cu-Mo alloys are more uniform than that of Cu-Cr, in which much coarser Cr particles can be identified (See Fig 4(b)).
The refractory particle size in Cu57Cr43 decreased substantially from 41 µm in the asreceived condition to 1.3 µm (normal to the shear plane) after deformation (See Table 3). The HPT-deformed Cu50Mo50 and Cu30Mo70 had similar layer thickness values of 450 nm and 400 nm, respectively. Although the initial particle size in Cu20W80 was smaller than in the Mocontaining composites the HPT-deformed microstructure was coarser, with a layer thickness of 900 nm.

Texture Analysis
At an imposed equivalent strain of ∼75 the refractory components in Cu-Mo and Cu-W composites showed {hkl} 111 fibre textures. Recalculated (110) and (222) pole figures (PFs) for the refractory metal in each composite are provided in Figure 5, with the ideal shear texture for a body-centred cubic (BCC) phase overlaid on the experimental data. In the (110) pole figures the fibre texture was most pronounced in Cu50Mo50 and Cu30Mo70 with a maximum intensity of 1.8 and 1.7 respectively. (All texture intensities are expressed as multiples of the intensity for a random distribution (MRD)). The Cu57Cr43 composite had a maximum intensity of 1.4 on the 110 PF, similar to Cu-W, but with a less well-defined texture. The fibres were tilted away from the ideal shear texture and measurements on the (110) PF yielded tilt angles of 8 • in Cu30Mo70, 9 • in Cu20W80 and 12 • in Cu50Mo50 (The texture data is summarised in Table S Figure 5(e-h)). The tilt angles calculated from the (222) pole figures differed from those for (110) by less than the 5 • step size used for the texture scans, except for Cu57Cr43, where the tilt was estimated as 11 • from (110) and 19 • from (222).
Evaluating the texture of copper proved to be impractical. The low diffracted intensity of Cu compared to W made it impossible to determine the texture for Cu in Cu-W. Although there was sufficient diffracted intensity in the two Cu-Mo composites, measurements revealed only weak, and poorly defined textures. It was only in Cu57Cr43 that a moderate (max=2. 3) {hkl} 110 texture [32,33] was observed. Further analysis of the texture was therefore restricted to the refractory metal texture, however, the (111) pole figure for Cu in Cu57Cr43 have been provided in Figure S-1 of the supplementary material.

Microstructure
Significant differences between the composites became evident after second-step HPT deformation. Figure 6 presents typical high-angle annular dark field scanning TEM (HAADF-STEM) micrographs, in which the Cu and refractory components can be readily identified due to atomic number contrast. The lamellae decomposed in Cu57Cr43 and Cu50Mo50 leaving a microstructure consisting of equiaxed nanoscale grains, reported for each of these materials [24,29]. The refractory particle size was 14 nm in Cu57Cr43 and 17 nm in Cu50Mo50. In contrast, Cu20W80 and Cu30Mo70 retained lamellar microstructures after 50 and 35 rotations of step 2 deformation, respectively, during which the refractory layer width was reduced to tens of nanometers. Lamellar spacings of 10-20 nm and 20-50 nm were measured for Mo and W, respectively. Cu was present between the refractory lamellae, often as layers of ∼ 5 nm thickness. This type of microstructure has been shown for Cu30Mo70 [29] but to the best of the authors knowledge is the first report of such nanolamellar Cu-W composites. Selected area diffraction (SAD) of the Cu57Cr43 (Fig. 7(a)) and Cu50Mo50 samples ( Fig. 7(b)) yielded ring patterns with minimal radial variation in intensity. Composites which retained a lamellar microstructure displayed substantial radial variations in diffracted intensity. This phenomenon indicates that the persistence of the lamellar structure is also associated with texture at a local scale, and is more pronounced in Cu20W80 (Fig. 7(d)) than in Cu30Mo70 (Fig. 7(c)). Each of the SAD patterns was obtained close to the edge of the foil, using an identical aperture   Figure 7: Selected area diffraction patterns obtained from each composites after second step HPT deformation. An identical aperture diameter of 120 µm was used in each instance, and images were obtained close to the edge of the foil. The persistence of the lamellar microstructure is associated with local texture, as can be seen in (c,d).
decreased to 0 degrees, as has been predicted for an ideal shear texture at high strains [34], although the fibres here are not well defined.

Discussion
Each of the refractory metal composites developed a lamellar structure during step 1 HPT deformation, with the refractory layer spacing ranging from 400 nm (Cu30Mo70) to 1.3 µm (Cu57Cr43) at ǫ ∼75 (See Table 3). This microstructure is comprised of elongated refractory particles within the Cu matrix and bears a closer structural resemblance to the HPT deformed Cu-Ag alloys [35] than the more regular, planar composites produced by ARB [8][9][10]. Deformation temperature has a strong influence on microstructural refinement. It has been shown that refinement of a Cu-25wt.%W composite was noticeably more rapid during HPT deformation at room temperature than at 200-400 • C [27]. Thus, the deformation temperature can have a strong influence on the particle refinement process. In the present study, microstructure refinement was most extensive in the Cu57Cr43 alloy. The Cr has the highest DBTT of the three refractory elements (Cr, Mo, W) and thus it appear that the DBTT affects the ductility of the particles and hence their susceptibility for fragmentation by brittle fracture. For coarse grained refractory materials, the propensity for brittle fracture is more along the high angle grain boundaries. Thus, during the first-step of HPT deformation the coarser Cr particles breaks down at a faster rate. It has also been shown that introduction of dislocation sources by plastic deformation in tungsten can improve the ductility [36]. The presence of elongated particles ( Figure 4) indicates that the quasi hydrostatic state along with the applied shear strain can provide an appropriate stress state for ductilisation of refractory particles. However, this effect is most likely to be operating more effectively for Mo and W particles than for Cr due to the higher DBTT of the latter. Such a tendency was clearly observed during second-step HPT deformation. In Cu30Mo70 and Cu20W80, containing 67 and 65 vol.% percentage of refractory particles, the lamellar structure with elongated refractory particles was maintained during second-step HPT deformation. However, for Cu57Cr43 and Cu50Mo50 composites, with refractory volume fractions of only 49% and 46% respectively, the elongated particle structure of the first-step HPT broke down into isolated and equiaxed hard particles embedded in the softer Cu matrix. This showed that apart from the DBTT of the particles the propensity for plastic deformation is also influenced by the volume fraction of the hard phase [37,38]. For a lower volume fraction of refractory particles the local deformation state acting on the particles possibly deviates significantly from the applied macroscopic deformation state leading to fragmentation of the lamellae.
The signature of such differences in the macroscopic and microscopic deformation state is evident from the crystallographic pole figures. The composites in which the lamellar structure was maintained developed a clear {hkl} 111 fibre texture (Figures 5 and 8). However, in the composites where the lamellar structure was decomposed, no prominent texture was observed. In this case, the equiaxed refractory particles are uniformly dispersed in the Cu matrix and rotate freely in response to the local deformation. This microstructure is also associated with the saturated hardness for Cu50Mo50, indicating that further refinement of the microstructure is significantly slowed down due to the unobstructed motion of hard refractory particles in Cu matrix. On the other hand, in Cu30Mo70 and Cu20W80, where a lamellar structure is maintained, further hardening was observed with increasing deformation (Figure 2). It is not clear till what extent of deformation the lamellar microstructure will be retained, as several equiaxed Mo and W particles started to form at higher strains ( Figure 6(c)-6(d)). Although, at microscopic scale the lamellae still show signature of prominent texture, as observed by the SAD patterns (Figure 7(c)-7(d)), the bulk texture measured by synchrotron radiation shows weakening or spread about the ideal positions ( Figure 8). With a 120 µm aperture and assuming a reasonable foil thickness of 100 nm, the SAD pattern sampled a volume of 1100 µm 3 ; five orders of magnitude less than the synchrotron X-ray beam. The variability in the shape and orientation of the refractory particles and their gradual fragmentation at higher strains, as observed in SEM ( Figure 4) and STEM (Figure 6), leads to broadening and weakening of the texture due to averaging over the larger volume. A quantitative examination of the processes of plastic deformation and fragmentation of the refractory particles would require detailed stereological analysis which is beyond the scope of the present work although such a study is currently underway.
Although it was possible to produce nanostructured refractory metal composites via HPT, it should be noted that the refinement of the refractory particle size was substantially less than would be expected from the applied shear deformation. In an ideal case the thickness of the lamellae, d, is given by: for shear strains of γ 1 and γ 2 where γ 2 is applied perpendicular to γ 1 and where the initial particle thickness is d 0 [31]. The expected particle size and layer thicknesses are given in Table 3, together with the experimentally measured values. For completeness the refractory particle size is provided for both lamellar and isotropic composites, as the layer thickness or particle diameter, respectively. After step 1 HPT deformation of Cu-Mo and Cu-W composites, the measured particle size is an order of magnitude larger than predicted, even though there was no indication of macroscopic slippage. The discrepancy is even greater during step 2 deformation where the strain imposed would theoretically reduce the layer thickness to (physically unrealistic) subnanometer scale, but where the measured values are actually two to three orders of magnitude greater.
The ideal co-deformation of the two components assumed in Equation 2 clearly does not occur in practice for Cu-Mo or Cu-W composites and it appears that strain tends to partition to the softer Cu component. Co-deformation would result in a uniform reorientation and elongation Table 3: A summary of the sample morphology and refractory metal particle size. "AR" indicates the as-received condition. The measured refractory metal size is given as either the average Feret diameter for isotropic samples or the refractory layer thickness for lamellar samples. For 2-step HPT deformation, γ 1 was calculated for the centre of the cylinder cut from the step 1 HPT disc. This was multiplied by the shear strain imposed during step 2 HPT. The expected particle size for ideal co-deformation is also given (See Equation 2). show that the elongated particles are non-uniform in shape and have a spread of orientations, which becomes more pronounced after step 2 HPT. This is reflected in the development of texture in Cu30Mo70 and Cu20W80 during step 1 deformation (Fig. 5) and its subsequent spread and weakening during step 2 HPT (Fig. 8) as the lamellae become more convoluted and misoriented. The isotropic (step 2 HPT) Cu50Mo50 composite also had a much larger Mo particle size than expected. The microstructural refinement of the Cu-Cr composite was noticeable different to that of the Mo and W-based composites. The difference between ideal and measured layer spacing was roughly a factor of 2 during step 1 deformation and the particle size after step 2 deformation was remarkably close that expected from the shear strain. As noted above, Cr has the highest DBTT of the refractory metals examined in this work, and this factor, combined with the larger initial particle size would allow brittle fracture to play a large role during the early stages of deformation. In addition, forced mixing can dissolve up to 32wt.% Cu in Cr [25] and at higher strains this is likely to influence not only the volume fraction of the phases, but also their mechanical behaviour. Although it is possible to identify the DBTT and solubility as key parameters, the deformation behaviour of such lamellar composites is not well explained and this topic warrants further investigation. Strong, well-defined orientation relationships between the co-deformed phases are a wellknown characteristic of ARB Cu-Nb composites [32,33] and influence the mechanical properties and radiation damage tolerance. However, the difficulty in obtaining the bulk texture data for the Cu phase makes it impossible to determine the orientation relationship between Cu and the refractory phases in present study and further local microscopic studies, such as the SAD data (Figure 7), are warranted.

Conclusions
In summary, the present study showed that it is possible to circumvent the poor ductility of the group VI refractory elements (Cr, Mo, W) to produce ultrafine-grained composites in a Cu matrix by HPT deformation. Cu and the refractory phases undergo limited co-deformation at room temperature and develop a discontinuous lamellar microstructure at an equivalent strain of ∼75. The refractory particles show considerable irregularity in their shape, size and orientation with respect to the shear plane. The refractory metals develop typical BCC {hkl} 111 fibre textures, with a tilt to the tangential direction. The texture is stronger and more clearly defined in Cu-Mo and Cu-W than in Cu-Cr.
Further HPT deformation lead to the breakdown of the lamellar structure in Cu57Cr43 and Cu50Mo50, resulting in a saturated or near-saturated microstructure consisting of equiaxed, essentially randomly-oriented, nanometer-scale grains of the two phases. The grain sizes of Cr and Mo were measured as 14 nm and 17 nm, respectively. Cu30Mo70 and Cu20W80 (wt.%) composites retained a nearly lamellar structure to extremely high applied strains and the refractory metal layer thickness was reduced to 10-20 nm (Cu-Mo) or 20-50 nm (Cu-W). Despite the microstructure being textured at a local scale, the irregularity of the refractory particle orientation did not allow for the development of strong, global texture.
The differences in the microstructure evolution of Cu57Cr43 and Cu50Mo50 from Cu30Mo70 and Cu20W80 composites can be attributed to the variation in DBTT and the volume fraction of the refractory metal. Thus, a wide range of microstructures can be generated by tuning the fundamental deformation behaviour and the volume fraction of the refractory elements.