On nanostructured molybdenum–copper composites produced by high-pressure torsion

Molybdenum–copper composites are used in high-power electrical applications such as thermoelectrics where a combination of dimensional stability, high mechanical strength and high thermal and electrical conductivity are required [1‐3]. Molybdenum has the high melting point (\(2610\,^\circ \hbox {C}\) [4]) and low thermal expansion coefficient [5] characteristic of refractory metals such as Nb, Ta, W and Cr. While molybdenum has similar electrical and thermal conductivity to tungsten, the lower density of Mo makes it preferred for weight-critical applications [5] and its lower ductile-to-brittle transformation temperature (DBTT) of \(\sim 373\,\hbox {K}\) [6] makes forming Mo–Cu composites easier than equivalent Cu–W composites. In addition to providing the composites with thermal and electrical conductivity, copper also increases the ductility of the composites [5, 7] which is useful both during production and in service.

Mo and Cu have a positive enthalpy of mixing of +18 Kj/mol and have negligible solid solubility even at the melting temperature of copper [8]. The lack of mutual solubility is advantageous for retaining high conductivity [3, 9]; however, it also results in a weak interface and hence poor mechanical bonding between the phases [10].

Mo–Cu composites are conventionally produced via liquid-metal infiltration [5, 11]. This involves sintering Mo powders to make a porous preform which can then be infiltrated with molten Cu. This process is only viable in a limited composition range, and the rate of densification is low [12].

Efforts to accelerate densification have mainly involved using ball milling to produce nanoscale Mo and Cu powders with high defect concentrations. These can then be densified more rapidly via powder compaction and sintering at 1000–1200\(\,^\circ \hbox {C}\) [12‐14]. Although the ball-milled powders have grain sizes in the range of 10–20 nm, grain growth during sintering is rapid and the densified compacts have grain diameters of several microns [3, 11, 13]. Complete densification remains unachievable and intergranular porosity is still observed after sintering [3, 13].

High-pressure torsion (HPT) has been used for grain refinement and mechanical alloying of a wide range of Cu-based systems including Cu–Fe [15‐17], Cu–Cr [18, 19], Cu–Nb [20, 21] and Cu–W [22, 23]. Deformation is carried out under quasi-hydrostatic conditions, allowing brittle materials such as refractory metals to be extensively strained without the need for external heating. This, in turn, avoids the grain growth which accompanies high-temperature sintering. Previous studies on pure Mo found more extensive grain refinement could be achieved with HPT (with an average grain size of \(0.2\,\mu \hbox {m}\)) than with either equal channel angular extrusion (ECAE) or multi-stage forming (average grain sizes of 2.5 and \(1.8\,\mu \hbox {m}\), respectively) due to the lower processing temperature [24]. HPT imposes a strain that is proportional to the radial distance at a given point, and thus, a single sample can provide information covering a wide range of strain conditions.

The objective of this investigation was to use HPT to produce bulk, full-dense, nanograined Mo–Cu composites and to examine their microstructures at different applied strains. Although there are number of reports on the HPT deformation of the related Cu–Cr (see for example, [18, 19, 25]) and several on Cu–W [22, 23] systems, to the best of the authors’ knowledge this work provides the first examination of HPT-deformed Cu–Mo composites.

This investigation was carried out using two liquid-metal infiltrated Mo–Cu composites provided by Plansee SE, Austria. These composites had compositions of Mo-30 wt% Cu (40 at.% Cu) and Mo-50 wt% Cu (60 at.% Cu). These are hereafter described as Cu30Mo70 and Cu50Mo50, respectively.

Deformation was applied using multi-step high-pressure torsion (HPT) [17]. Discs (height, \(h\sim 12\,\hbox {mm}\), diameter, \(\phi =30\,\hbox {mm}\)) were cut from the raw ingots by electrical discharge machining. These were deformed at room temperature in a 4000 kN HPT apparatus, using tool steel anvils with an anvil gap of 3.5 mm. A total of either 5 or 10 complete rotations (hereafter described as \(N_1=5\) or 10, with the subscript indicating the first step of deformation) were performed at an angular velocity (\(\omega \)) of 0.0625 revolutions per minute (rpm) under an applied pressure of \(\sim 4.5\,\hbox {GPa}\).

Cylindrical samples (\({h}\sim 20\,\hbox {mm}\), \(\phi =7.5\,\hbox {mm}\)) were cut from the discs, perpendicular to the shear plane of this disc (see Fig. 1). These cylinders were sectioned into samples with height \(\sim \)1.1 mm. The average strain in these discs was calculated as 25 for \(N_1=5\) and 50 for \(N_1=10\). Step 2 deformation was conducted using a 400 kN HPT apparatus at an applied pressure of 7.4 GPa and \(\omega =0.4\,\hbox {rpm}\). The second step of deformation consisted of \(N_2=10-50\) rotations of deformation with a 0.25 mm anvil gap and cavity diameter of 8 mm.

Vickers microindentation hardness testing was conducted on as-received material and following each step of HPT deformation. Scanning electron microscopy was used to characterise the microstructure in the as-received condition and following HPT deformation. Samples for SEM examination were ground and polished using standard techniques, within final polishing taking place on a Buehler Vibromet polisher. SEM examinations were carried out using a Zeiss Leo 1525 microscope equipped with field emission gun and Gemini lens system. Micrographs were recorded using the backscattered electron detector (BSD) with source voltages of 15–20 kV. Stereological measurements were taken using ImageJ software (Version 1.50g). The Mo particle size in the as-received condition was measured by a processing consisting of (i) background subtraction, (ii) thresholding, (iii) separation of connected particles via the watershed algorithm and (iv) measurement of the projected particle area. Determining the lamellar spacing required similar preprocessing steps of background subtraction and thresholding. The spacing was then measured by counting the number of intercepts using a circular grid to give the true mean spacing [26]. This follows the established methodology developed for measuring lamellar spacings in pearlitic steels.

Cross-sectional and plan-section TEM foils prepared by dimple grinding to a thickness of 10–15\(\,\mu \hbox {m}\), followed by precision ion polishing (PIPS) to perforation using a Gatan 691 instrument. For planar samples (see Fig. 2), the electron beam is oriented normal to the shear plane, whereas for cross-sectional samples, the electron beam is oriented parallel to this plane. Note that in both cases, electron beam is normal to the shear direction and elongation of the components will be visible. PIPS made use of a liquid nitrogen-cooled sample stage to minimise specimen heating during foil preparation. The foils examined were extracted from the outer section of the HPT disc, and the hole position corresponds to a radial distance of approximately 3 mm. TEM and scanning TEM (STEM) observations were made using CM12 and JEOL Cs-corrected 2100F microscopes operating at 120 and 200 kV, respectively. High-angle annular dark-field STEM (HAADF-STEM) images were obtained on the JEOL microscope with inner and outer collection angles of 65.51 and 174.9 mrad, respectively. Lamellar spacings were determined from the STEM images using the same procedure as for the SEM micrographs. It was not possible to obtain reliable data from heavily deformed Cu50Mo50 composites due to the decomposition of the lamellar structure, as is described in the following section.

Liquid-metal infiltrated Mo–Cu proved to be amenable to room-temperature deformation via HPT, making it possible to produce nanometre-scale composites. No radial cracking was observed after step 1 deformation, where the maximum equivalent strain was \(\sim 72\). Although radial cracks were sometimes evident after step 2 deformation, it was generally possible to deform to 50 revolutions. Deformation-induced heating during HPT is only expected to raise the temperature to around 373K during step 1 deformation and 300K during step 2 deformation. This is much lower than the temperatures of 573–673 K [6] usually required to ensure sufficient ductility in Mo; however, the high imposed pressure and quasi-hydrostatic conditions allowed for deformation of both Cu and Mo components.

Due to the fine scale of the HPT material and the absence of intergranular porosity, the hardness values in the strain-saturated state compare very favourably with densified ball-milled materials. The hardness values reported in Mo–Cu with\(\hbox{Mo} < 20\) wt% alloys were in the range of 200–259 \(H_V\) [3, 28, 29], and ball-milled and annealed pure Mo was reported to have a hardness of 250 \(H_V\). These values are substantially below the values of 475 and \(600\,H_V\) measured herein for HPT-deformed Cu50Mo50 and Cu30Mo70, respectively.

Both of the Cu–Mo composites developed similar lamellar microstructures during the first step of deformation (Fig. 6), with the microstructure being more refined at higher strains (c,d) and in the Cu30Mo70 composite for equal strain (b, d). However, in the Cu50Mo50 composite, this microstructure decomposes during step 2 deformation in Cu50Mo50 (Fig. 9) unlike Cu30Mo70 (Fig. 8). Previous investigations on binary systems containing phases with substantial differences in hardness have reported the lamellar structure formed during HPT break-up at higher strains and is replaced by a microstructure consisting of equiaxed grains [18]. It has been argued that once such microstructures develop, additional deformation is localised in the softer phase and does not further refine the microstructure [18]. The microstructure then enters a steady state, with the mechanical properties, including hardness, remaining unchanged even if further strain is applied. Strain localisation in the soft phase occurs more readily for small volume fractions of the hard phase, where there is limited contact between particles of the hard phase during deformation. Therefore, the higher volume fraction of Mo in Cu30Mo70 (67% compared to 47% in Cu50Mo50) is thought to be responsible for the retention of a lamellar structure even after extensive two-step deformation.

The relationship between the hardness and lamellar spacing (s), of the composites is presented in Fig. 10, which shows a Hall–Petch dependence on the lamellar spacing for spacings as low as 160 nm. The gradient of the hardness values plotted against \(1/\sqrt{s}\) for lamellar spacing s, gave a Hall–Petch coefficient, K, of \(217\pm 7\,\hbox {MPa}/\mu \hbox {m}^{1/2}\). Similar relationships between lamellar spacing and hardness were demonstrated in eutectoid pearlitic steels [30] and have also been established for the widely studied Cu–Nb composites [31‐34]. This behaviour indicates that plastic deformation is interface controlled, due to the difficulty of transferring dislocations across the interface. However, further refinement of the microstructure had little effect, with a factor of 20 reduction in the spacing leading to a hardness increase of approximately 20%. This deviation from the Hall–Petch relationship for extremely fine structures has been reported in nanoscale Cu–Nb alloys [35]. This is thought to indicate a change in the deformation mechanism, with studies suggesting that interface sliding occurs [36].

The microstructural development of the Cu–Mo composites is initially similar to that reported for HPT deformation of other Group VI refractory metals. HPT deformation of a W-25 wt% Cu composite resulted in a lamellar microstructure for equivalent strains of at least 64 [37]. Another study using a two-step HPT procedure subjected W-25 wt% Cu to equivalent strains of 1000 without reaching a hardness plateau [23]. This would suggest that the material retains a lamellar microstructure under these conditions. HPT Cu–Cr composites have also been examined at high strains [18, 38], and it has been shown that the lamellar structure decomposed to form equiaxed, roughly spheroidal particles after strains of \(\sim 250\) [38]. These reached a grain size of \(\sim 15\,\hbox {nm}\) after strain of \({\approx}1400\) but showed no further reduction in grain size at a strain of \({\approx}4000\) [39]. In comparison, an average equivalent shear strain of \({\sim}50\) was imposed on the Cu–Mo composites during the first HPT deformation step and a further (multiplicative) strain of \(\sim 1000\) during the second deformation step, indicating the Cu30Mo70 layered microstructure is stable to extremely high strains.

The deformability of the refractory phase at a given temperature and hence the ductile-to-brittle transformation temperature are also critical to the retention of the lamellar microstructure. The DBTT of tungsten (370–470 K [40]) is only slightly higher than the value of \(\sim 373\,\hbox {K}\) [6] for Mo. However, chromium has a DBTT variously reported between 423 and 773 K [41‐43], and in Cu–Cr composites, the lamellar structure decomposes at much lower strains than either Cu–W or Cu–Mo.

Composites with similar layer thicknesses of \(\sim 10{-}20\,\hbox {nm}\) layers have also produced via accumulative roll bonding (ARB) of pure Cu and Nb [20, 44]. These layered structures were retained during subsequent HPT deformation of the ARB composite [20]. It is important to note, however, that molybdenum is significantly less ductile than niobium and that the quasi-hydrostatic conditions during HPT play a role in allowing for Mo/Cu co-deformation at room temperature. HPT therefore offers a means of generating ultrafine layered structures containing less ductile refractory elements, opening up the possibility of more detailed studies of their microstructure and properties. One area of interest concerns the texture of the composites, since the ARB Cu–Nb composites evolved atypical textures as the layer thickness approached the grain size due to the increasing influence of interfaces and the constraint imposed by neighbouring grains [44]. Further work is underway to examine the texture of Cu–Mo HPT composites.

Liquid-metal infiltrated Mo–Cu composites were found to be amenable to high-pressure torsion at room temperature. Both copper and molybdenum phases underwent deformation, giving a lamellar microstructure after the first step of HPT deformation, corresponding to an average equivalent shear strain of 50. In Cu50Mo50, the lamellar structure breaks up during the second step of HPT deformation and the microstructure consisted of equiaxed Mo nanoscale grains surrounded by Cu. This behaviour is similar to that of Cu–Cr composites and liquid-metal infiltrated Cu–W composites and is associated with a strain-saturated state in which the hardness reaches a plateau. However, in the Cu30Mo70 material, the lamellar structure was retained, and after stage-two deformation, the resulting composites had a fine lamellar structure with Cu and Mo forming 5 nm and 10–20 nm lamellae, respectively, with an overall lamellar spacing of 9 nm. Notwithstanding the poor room-temperature ductility of Mo, this microstructure was stable to equivalent strains of \(\sim 50000\) and the layer thicknesses became equivalent to those of accumulative roll-bonded Cu–Nb. Since no sintering was used, post-deformation grain growth was avoided. The fine microstructure and absence of porosity contributed to Vickers hardness values of 475 for Cu50Mo50 and 600 for Cu30Mo70. The combination of high-strength, and a fine, oriented microstructure would be well suited to thermoelectric materials and shows that HPT-deformed Cu–Mo composites warrant further investigation for such applications.