From powders to bulk metallic glass composites

One way to adjust the properties of materials is by changing its microstructure. This concept is not easily applicable on bulk metallic glasses (BMGs), because they do not consist of grains or different phases and so their microstructure is very homogeneous. One obvious way to integrate inhomogeneities is to produce bulk metallic glass composites (BMGCs). Here we show how to generate BMGCs via high-pressure torsion (HPT) starting from powders (amorphous Zr-MG and crystalline Cu). Using this approach, the composition can be varied and by changing the applied shear strains, the refinement of the microstructure is adjustable. This process permits to produce amorphous/crystalline composites where the scale of the phases can be varied from the micro- to the nanometer regime. Even mixing of the two phases and the generation of new metallic glasses can be achieved. The refinement of microstructure increases the hardness and a hardness higher than the initial BMG can be obtained.


Microstructural evolution as a function of strain
In Fig. 1a, the microstructure of Zr-MG + 20 wt% Cu as a function of the applied strain is presented (the shear direction is indicated with an arrow). All micrographs are backscatter-detector images to distinguish easily between Cu and the Zr-based MG. Images in the lower row have a significant higher magnification. The two phases in the coarse state differ in their mechanical properties, which strongly influences the deformation. In the beginning, Cu sustains most of the deformation as it has a lower hardness and yield strength. Zr-MG is more difficult to deform and the initial particles are easily distinguished as they change their initial shape and size only marginally (one of those particle is marked with a white arrow in Fig. 1a). Cu works as a glue and holds the amorphous particles together. Due to the heavy plastic deformation, its grain size is refined down to the range of several hundreds of nanometers. Higher strains (see micrographs for γ = 50) force the Zr-MG to deform and bond together. Therefore, the amorphous phase forms long elongated regions with a length of about 100 µm in the shear direction and widths up to 10 µm, where some features from the initial particles are still detectable (area marked by the black arrow). The Cu bands become thinner and more evenly distributed, but large Cu-containing areas can still be found. The Cu grains refine and become less than 100 nm in size. Micrographs at γ = 80 show a further refinement of the microstructure. Cu and Zr-MG lamellae decrease in length and in width: All of the Cu lamellas are thinner than 1 µm and most of them exhibit a thickness of 20 nm. Cu grains are not detectable in the SEM micrographs anymore, but the thickness of the thinner lamellas can be seen as an upper limit for the grain size in this bands. The elongated regions of the amorphous phase shrinks below 100 µm in length and 10 µm in width. From γ = 80 to γ = 165 a pronounced increase of the number of the thin Zr-MG lamellas occurs, so that the Cu network in the images at low magnification, is in reality a Cu rich region consisting of Cu lamellas with lengths below 1 µm and widths below 100 nm separated by about equally sized amorphous lamellas. Additionally, the boundaries between the two phases change from sharp and easily detectable to more blurry interfaces as the Cu starts to mix with the amorphous phase at higher strains. This process starts already at γ = 80, but is clearly recognizable at γ = 165. The mixing of the two phases progresses with increasing applied deformation. At γ = 250 elongated Cu rich and Zr-MG rich regions can still be detected with smooth transitions in between, but at γ = 390 Cu the SEM micrographs seems to show a single phase metallic glass.
The ratio of the amorphous and crystalline phase can be easily modified by varying the ratio of the two initial powders. To investigate the influence of the ratio of the two phases, compositions with higher contents of Cu were produced and SEM micrographs of the cross sections of samples with 40 wt% Cu at different applied strains are shown in Fig. 1b. At first glance the major difference visible in the composition with less Cu in Fig. 1a, is the higher amount of the crystalline phase, but the deformation mechanism does not differ. The softer Cu phase starts to deform first, but with higher amount of applied strain, the harder Zr-MG starts to form long elongated bands, which shrink with increasing deformation. At higher strains, the two phases start to mix again until a homogeneous material is formed. However, by comparing the micrographs for the two compositions, it can be clearly seen that a higher content of Cu requires a higher strains to reach the same refinement of the microstructure. To obtain the saturation microstructure (i.e. the monolithic metallic glass), the applied strain for Zr-MG + 40 wt% Cu has to be doubled compared to the Zr-MG + 20 wt% Cu.
The content of Cu was further increased to investigate the limits of mixing and to see any resulting change of the deformation behavior. In Fig. 1c, SEM micrographs of Zr-MG 60 wt% Cu can be seen. Compared to the samples with 20 and 40 wt% Cu, the deformation behavior in the beginning does not change and a lamellar structure is formed (see γ = 550-1600). At higher applied strains (γ = 3300-5000), the micrographs exhibits a deviation from the deformation behavior from the samples with low Cu content. The lamellas shrink in thickness with higher applied strain, but mainly a break-down of the lamellar configuration occurs. A transition microstructure develops with residual lamellar blocks embedded (see circled areas in Fig. 1c) in a nanocrystalline (see black arrow) and amorphous structure (see white arrow). At γ = 11000, both phases seem to be fully mixed and a single phase microstructure is obtained. The amount of strain needed to reach this saturation microstructure is about 10 times higher than for Zr-MG 40 wt% Cu and about 20 times higher than for Zr-MG 20 wt% Cu.
The composition with the highest Cu content investigated was Zr-MG 80 wt% Cu. SEM micrographs of the cross sections are displayed in Fig. 2. Again, the Zr-MG particles elongate and a lamellar structure is obtained in the beginning, but at an applied strain of γ = 5000 and higher, the lamellas start to break and the microstructure does not change significantly anymore even after applying extremely high strains (up to γ = 18900). Since the microstructure does not change significantly even after this enormous increase of the applied strain, it is assumed that a saturation is reached. The saturation microstructure seems to consist of a nanocrystalline Cu-rich matrix (in Fig. 2 at higher magnification crystals indicated with black arrows can be seen as small freckles in the light grey area) in which elongated Zr-MG bands (some are indicated with a white arrow in Fig. 2) with a length of several micrometers are embedded.
The evolution of the microstructure is reflected also in XRD measurements. XRD profiles of samples from Zr-MG (undeformed and deformed), crystalline Cu, and the four compositions investigated in a near saturated condition are displayed in Fig. 3. The measured range is concentrated on the first (and strongest) peak of Zr-MG, which is very broad and has low intensity compared to the more distinctive peaks of crystalline Cu. The XRD results of a HPT deformed pure Cu powder are also depicted in Fig. 3, which has a grain size of about 100 nm 24 . The XRD peaks of the undeformed and deformed Zr-MG coincide exactly, which indicates that the HPT process does not have an influence on the single phase BMG. The influence of the applied strain can be clearly seen at the three samples of Zr-MG 60 wt% Cu as the number of rotations in the HPT process was varied from 50 to 200 and 500. The two Cu-peaks are very dominant for the specimen with 50 rotations, but shrink as the applied strain increases. After 200 rotations, the first Cu peak has a similar height as the amorphous peak and both Cu peaks broadened significantly, which can be caused by two effects: smaller crystal sizes and higher defect densities in the crystals. The crystalline peaks disappear after 500 rotations and only the amorphous peak remains.   crystalline Cu (indicated with arrows in Fig. 3). On the other hand, even after 500 rotations Zr-MG 80 wt% Cu shows a dominant Cu peak and a very weak amorphous peak (indicated with an arrow in Fig. 3). Comparing the crystalline peak of Zr-MG 60 wt% Cu and Zr-MG 80 wt% Cu to the pure Cu sample, the Cu peak is shifted to lower angles due to partial solution (or formation of a supersaturated solid solution) and broadens mainly due to the small crystalline size. Furthermore, mixing Cu into the amorphous phase shifts its peak positions to higher 2-Theta values and the difference between Zr-MG and Zr-MG 60 wt% Cu is approximately 5°. In order to demonstrate the possibility of analyzing the alloying region for generation of new metallic glasses, a sample with Zr-MG 20 wt% Cu 10 wt% Ni was investigated and after 150 rotations a fully amorphous structure is obtained.

Effect of strain on the mechanical properties
Microstructure influences strongly the mechanical properties, which were investigated in this study by using hardness measurement. In Fig. 4, the Vicker's hardness is plotted versus the applied strain (see methods). Since the strain varies over about three orders of magnitudes, several specimens have been used. The fitted lines are for guiding the eye. The hardness of the single phase Zr-MG scatters strongly at low applied strain, because the particles are not welded together completely in this early state of deformation. To eliminate an effect from the HPT deformation on the hardness beside consolidation, undeformed powder and a Zr-MG sample with 30 rotations were investigated via nanoindentation. Both conditions show the same hardness with 6.12 ± 0.33 GPa for the powder and 6.11 ± 0.07 GPa for the deformed specimen. The influence of the applied strain is more pronounced for the BMGCs, because not only the particles are welded together but also a refining and mixing of the two phases occur. Three compositions (with 20, 40 and 60 wt% Cu) show a similar behavior. At low strains, the hardness is lower than the single phase Zr-MG but increases strongly with deformation. The slope decreases with higher deformation and the curves level off at a hardness higher than the Zr-MG. Exceeding the hardness of both initial powders can only be explained by mixing the two phases, shifting the chemical composition and so forming a new metallic glass with higher hardness. Higher contents of Cu shift the curve to the right, which means that welding, refining and mixing require more strain if the fraction of the softer phase is higher. This behavior of the hardness corresponds to the evolution of the microstructure where the influence of Cu can also be seen clearly by comparing micrographs of the different composition. Increasing the content of Cu in the amorphous phase also leads to higher hardness (see Zr-MG 20 wt% Cu and Zr-MG 40 wt% Cu), but it requires more deformation. Zr-MG 80 wt% Cu also show an increase in hardness with larger applied strains, but the slope is less steep compared to the other compositions and the hardness of Zr-MG is not reached even after γ = 10 4 . The material gets harder, but the rate is so slow that a real mixing (as it is the case for the other three compositions) is not practical, as days are needed to achieve sufficient deformation. Not only the shape and position of the hardness curves depend on the composition, but also the hardness at saturation. In Fig. 5, the approximated hardness for the saturated microstructure depending on the composition is shown and the highest value is reached for the composition with 40 wt% Cu (which corresponds to 32at% Zr and 55.1at% Cu). The hardness drops by approximately 15% for Zr-MG compared to Zr-MG 40 wt% Cu.

Discussion
The microstructure of the HPT deformed Cu + Zr-MG mixture and their mechanical properties depend strongly on the composition and on the applied strain. At low strains, a bulk composite is formed, in which the elongated phases start to refine and the hardness increases strongly. Increasing the applied strain, continuous thinning of the lamellae can be found for 20 and 40 wt% of Cu, whereas, the lamellas tend to break down at higher contents of Cu. For 60 and 80 wt% Cu, the length of the lamellae decreases more than their thickness. Additional, a mixing of the two phases takes place and at very high strains, a single phase BMG is obtained for all composition except Zr-MG 80 wt% Cu. However, compared to the beginning the microstructure changes slower as the applied strain increases, which is also reflected in the hardness curves as all level off at higher degrees of deformation. The strain necessary to reach saturation increases significantly with increasing content of the softer crystalline phase (e.g.: more than 20 times if the content of Cu is increased from 20 wt% to 60 wt%). This can be seen in SEM, XRD and the hardness measurements (more Cu also shifts the hardness curves to higher strains). The position of the amorphous peak shifts to higher degrees as more Cu is mixed into the MG (the peak position of Zr-MG and Zr-MG 60 wt% Cu differs with approximately 5°), but it can also been shown how the crystalline Cu in the composites is affected. As the applied strain increases, the crystalline peak broadens due to grain refinement. The grain size becomes smaller during HPT deformation in the presence of a second phase than in a single phase crystalline material. Mixing of the two phases takes place also for Zr-MG 80 wt% Cu. In this case, a part of Zr-MG is dissolved in the crystalline Cu, it remains crystalline (nanocrystalline) and a supersaturated solid solution is formed. The exact quantity is hard to obtain, because the Zr-MG contains four other elements beside Cu (if only Zr was considered, the peak shift would indicate dissolving 0.16 at% into the crystalline Cu 46 ). Two effects could simultaneously prevent a full mixing of the two phases for this composition: The solubility of Cu in the amorphous phase is restricted and the limit lies between 71at% (Zr-MG 60 wt% Cu) and 85.9at% (Zr-MG 80 wt% Cu). Secondly, the amorphous particles are not forced to deform as strongly as in the other compositions, because the high content of soft Cu carries all of the deformation. The consequences of the second effect is the possibility of a further mixing of the two phases if the applied strain is pushed to higher values. However, this can only happen if the emerging amorphous phase has a higher strength and forces the initial MG to deform further. Comparing the approximated hardness at saturation, it can be seen that all compositions show a higher hardening than Zr-MG 80 wt% Cu. The highest value is obtained for Zr-MG 40 wt% Cu. Zr-MG 80 wt% Cu still consists considerably of a crystalline phase but has already a similar hardness to Zr-MG. Nevertheless, the composition can be continuously changed in a range from Zr 57 Cu 20 Al 10 Ni 8 Ti 5 (Zr-MG) to at least Zr 20 Cu 71 Al 3.6 Ni 2.9 Ti 1.8 (Zr-MG 60 wt% Cu). Additional to the compositions in this study, a fully amorphous sample with Zr-MG 20 wt% Cu 10 wt% Ni was produced. Compared to compositions from literature, the adjustable range is large (see Fig. 6) and new BMGs with a significant different chemical composition can be produced. Hence, the HPT can be used to fathom the limits of possible BMG's chemical compositions.
In summary, it is shown that BMGCs can be produced via HPT and their microstructure and mechanical properties can be adjusted by changing composition and the applied strain. The BMGCs can be forced to mix and the solubility range is highly extended, producing new BMGs. The chemical composition of this new BMGs can be adjusted over a wider range compared to conventional casting and by adding additional elements (e.g. Cu and Ni), the possibility of feasible BMG compositions increases drastically. Changing the composition influences also the mechanical properties and the hardness can be increased by approximately 15%. While the generation of new BMGs is important, even more interesting is the potential for tuning the properties of metal-BMG composites with the possibility to vary the second phase. The second phase can be freely changed in choosing different materials and compositions and its dimensions can be altered from the µm to the nanometer regime.

Methods
The metallic glass powder (Zr 57 Cu 20 Al 10 Ni 8 Ti 5 , spherical particle with diameters between 1 µm to 40 µm) was fabricated via high pressure gas atomization and was then mixed with crystalline Cu powder (spherical particle with diameters between 10 µm to 40 µm) in the respective compositions and blended by hand. The powder mixture was then filled into the gap between two grooved anvils and compacted by applying 4 GPa and 10° rotation in the HPT. For further deformation, the pressure was increased to 8 GPa (or 9 GPa) and samples with 2 to 500 rotations at room temperature were produced. The dimensions of the specimens were 6 mm in diameter and a height of approximately 450-600 µm. The applied shear strain γ can be estimated with where r is the radius, N is the numbers of rotations and t is the thickness after deformation. As the dimensions were approximately the same for all samples, the degree of deformation is adjusted by the numbers of applied rotations (more turns lead to higher deformation) and at the radius where the sample was investigated (the applied strain increases with the radial distance from the center of the disk). For SEM and hardness measurement, the coin-like HPT samples were cut in half, ground and polished to investigate the cross sections. In SEM the back-scatter detector was used to provide mass contrast so that the two phases are more easily distinguishable. Vickers hardness was measured along the diameter on the cross section with a load of 0.5 kg. An error of 2% for the measured hardness values is assumed and the standard deviation errors of the fitted curve were calculated and used for the error bars for Fig. 5. For nanoindentation testing of the powder and the deformed sample, a platform nanoindenter G200 (Keysight Tec) was used and the experiments were conducted under constant indentation strain rate (0.05 s −1 ) to a maximum indentation depth of 500 nm. The hardness was measured continuously over indentation depth and was averaged between an indentation depth between 400 to 450 nm. The non-deformed powder was embedded and then mechanical ground and polished like the surface of the deformed sample. Several particles were indented and for both materials, a mean value and the standard deviation were calculated from all indentations. A 5-circle X-ray diffractometer equipped with a source for Cu-K α radiation was used for XRD phase analysis of the specimens. Half samples were used and the surface was grounded to remove any impurities from the HPT process. The investigated 2-Theta range was concentrated around the first amorphous peak (the strongest).