Rejuvenated metallic glass strips produced via twin-roll casting

1. Introduction

Metallic glasses (MGs) generally form via fast quenching of alloy melts by avoiding crystallization [[1], [2], [3], [4]]. The long-range disorder structures of MGs render them exceptional mechanical or physical properties [[4], [5], [6]], including high strength, high elastic limit, and outstanding soft magnetic properties. Since the first discovery of the glassy Au-Si thin flake obtained by using the splat-quench method [7], thin MG ribbons in many alloy systems have been produced via single-roll melt spinning technique [4,5], as shown in Fig. 1. Fe-based MG ribbons, as a remarkable example, have been successfully and widely used in power distribution transformer cores due to their excellent soft magnetic properties and the low energy loss [8,9]. The single-roll melt spinning method can continuously produce thin MG ribbons with thickness ranging from tens of micrometers to ^∼150 μm [10,11], but this method is unable to produce MG strips with thickness larger than 150 μm. Twin-roll casting technique (see Fig. 1) is widely used in the industrial production of Al-based and Mg-based alloy strips with thickness ranging from one hundred micrometers to several millimeters [12,13], and several attempts have been tried to produce MG strips by using twin-roll casting technique [[14], [15], [16]]. The previous Fe-based [14] and Zr-based [15] MG strips produced via twin-roll casting imply that this technique is a feasible method for producing MG strips with thickness larger than 150 μm.

View original graphic| Download| PPT slide

Fig. 1.   Schematic illustration of single-roll melting spinning and twin-roll casting for producing metallic glass ribbons and strips, respectively.

The microstructure of MGs is strongly associated with cooling rates [1,17], which is different from that of crystalline alloys. During solidification of crystalline alloys, the cooling rate mainly alters the grain size, but the unit cell, i.e. the fashion of atomic packing, of the phase is generally independent of cooling rates. For comparison, the inherent structures of MGs, which are often depicted as configurational coordinates in the potential energy landscape, are susceptible to the external factors, including cooling rates, exerted force and plastic flow [1,17,18]. A higher cooling rate generally causes a higher content of free volume, which are “vacancies” in amorphous solids, frozen in MGs. An MG with a higher content of free volume has a higher energy state on the potential energy landscape, a higher enthalpy, a lower density, a lower hardness and better plastic deformability [1,[17], [18], [19], [20], [21], [22]]. The transition of an MG from a lower energy state to a higher energy state is known as “rejuvenation”, and the reverse process is called “relaxation” [1,17]. Many methods have been recently explored to rejuvenate MGs, including cyclic elastic loading [23,24], constrained loading [25], irradiation [26,27], and cryogenic thermal cycling [19].

As seen in Fig. 1, the main difference between single-roll melt spinning and twin-roll casting is the large viscous flow deformation caused by the squeezing force during solidification in the latter case, and in comparison, the solidification is external stress-free in the former case. In this paper, the ribbons and strips of Zr_41.2Ti_13.8Cu_12.5Ni₁₀Be_22.5 (Vit. 1) with different thicknesses were produced via single-roll melt spinning and twin-roll casting, respectively. It is found that for either of the methods, increasing the thickness of ribbons or strips, i.e. decreasing the cooling rate, causes a lower energy state of MG. However, although the cooling rate of the 320 μm-thick strip produced by twin-roll casting is almost one magnitude lower than that of the 110 μm-thick ribbon produced by single-roll melt spinning, the 320 μm-thick strip exhibits a larger relaxation enthalpy and a lower hardness as compared with that of the 110 μm-thick ribbon. This implies that the 320 μm-thick strip possesses a higher energy and rejuvenated state as compared to the 110 μm-thick ribbon. Molecular dynamics simulations reveal that the rejuvenation of the 320 μm-thick strip at a lower cooling rate arises from the flow deformation caused by the squeezing force during twin-roll casting.

2. Experimental and simulation procedures

The Zr_41.2Ti_13.8Cu_12.5Ni₁₀Be_22.5 (at.%, Vit. 1) ingots were prepared by arc melting high-purity Zr, Ti, Cu, Ni and Be metals (with purities ≥ 99.8%) in a Ti-gettered high-purity argon atmosphere. All ingots were re-melted four times to ensure chemical homogeneity. Small pieces in weight of 5-10 g of the Vit. 1 ingot were put inside the quartz tube, and were re-melted in the induction coil. Then the melt was spouted on a copper single-roll or copper twin-roll to form a ribbon or a strip, respectively, as schematically shown in Fig. 1. Ribbons in thickness of 40 μm and 110 μm were prepared by single-roll spinning, and strips in thickness of 320 μm and 490 μm were prepared by twin-roll casting. The diameter of the single-roll is 210 mm. The rotation speed and the nozzle-roll distance are 1500 r min^-1 and 0.8 mm for the 40 μm-thick ribbon, and 600 r min^-1 and 0.8 mm for the 110 μm-thick ribbon. The diameter of the twin-roll is 80 mm, the total spring force is ˜46.8 N, and the rotation speed for the 320 μm-thick and 490 μm-thick is 180 r min^-1 and 150 r min^-1, respectively. In order to compare the cooling rates of the 110 μm-thick ribbon and the 320 μm-thick strip, eutectic Al-33 wt% Cu alloy was chosen to prepare a ribbon and a strip by single-roll melt spinning and twin-roll casting, respectively, under the same fabrication conditions as that of the 110 μm-thick ribbon and the 320 μm-thick strip. The cooling rates then can be experimentally obtained by measuring the average eutectic lamellar spacing.

Specimens cut from both of the stripes were ground on both sides down to ˜100 μm, and the as-produced ribbons were characterized by means of X-ray diffraction (XRD, Philips PW 1050, Cu-K_α), differential scanning calorimetry (DSC, Netzsch 204F1), transmission electron microscopy (TEM, FEI Tecnai F20), and the nanoindentation measurement (Agilent Technologies Nano Indenter G20) with a Berkovich tip. DSC was performed at a heating rate of 0.33 K s^-1. The specimens for conducting TEM investigations were prepared by twin-jet electrolytic thinning in a solution of 30 ml perchloric acid +175 ml 1-butanol +295 ml methanol at 248 K and a constant voltage of 20 V. The selected area electron diffraction (SAED) patterns were integrated into 1D intensity profile by following the standard operations in the PASAD tools [28]. All indentations were performed with a displacement rate of 1 nm s^-1 and a maximum depth of 300 nm.

Molecular dynamics simulations were performed in LAMMPS [29,30] by using the embedded atom method (EAM) potential developed by Cheng et al. [29]. The Zr-Cu-Al system was selected because of its reliable empirical potential, and the random Zr₄₆Cu₄₆Al₈ configuration was composed of 4600 Zr atoms, 4600 Cu atoms and 800 Al atoms. At 2000 K and zero pressure, the model was fully equilibrated for 1 ns under periodic boundary conditions (PBCs), within an NPT ensemble (constant number, constant pressure, and constant temperature) [30]. The Nose-Hoover thermostat and Parrinello-Rahman technique were adopted to control the temperature and pressure [31,32]. Two models were performed. One was subjected to quenching to 50 K at a cooling rate of 10¹² K s^-1 under hydrostatic pressure 0 GPa, and the other one was subjected to quenching to 50 K at a cooling rate of 5 × 10¹¹ K s^-1 under 1 GPa pressure along x-direction.

3. Results

Fig. 2(a) and (b) shows the 40 μm-thick and 110 μm-thick ribbons, respectively, produced by single-roll melt spinning. The 320 μm-thick and 490 μm-thick strips produced by twin-roll casting are presented in Fig. 2(c) and (d), respectively. The XRD spectra of the ribbons and strips are shown in Fig. 2(e), and only broad diffraction humps were detected, which suggests the full amorphous nature of the prepared ribbons and strips. It is noteworthy that for a relatively constant cooling rate, which has a vital influence on the structural evolution of MGs, the strips were deliberately prepared with a length less than the circumference of the copper roll, as shown in Fig. 2(c) and (d). But the twin-roll casting itself is capable for producing very long MG strips, and a twin-roll prepared long Vit. 1 strip is shown in Fig. S1 in the Supplementary Materials.

View original graphic| Download| PPT slide

Fig. 2.   Appearance of Vit. 1 ribbons prepared by single-roll melt spinning (a, b), and strips prepared by twin-roll casting (c, d). The respective thickness of ribbons/strips is given. (e) The XRD spectra of the Vit. 1 ribbons and strips.

The glassy nature of all the prepared ribbons and strips was further confirmed in TEM. Since the cooling rate of 490 μm-thick strip is the lowest, only the TEM results of 490 μm-thick strip are selected to be shown in Fig. 3. Fig. 3(a) displays a featureless microstructure of the 490 μm-thick strip, and only diffuse rings were detected in its selective area electron diffraction (SAED) pattern (inset in Fig. 3(a)). The high-resolution TEM image in Fig. 3(b) shows maze-like patterns, and only a diffuse ring is observed in the corresponding fast Fourier transformation (FFT) image, which confirms the lack of long-rang order in the 490 μm-thick strip.

View original graphic| Download| PPT slide

Fig. 3.   (a) TEM micrograph with an inset of SAED pattern and (b) HRTEM image with an inset of the FFT pattern of the twin-roll produced 490 μm-thick strip.

The heating-up DSC traces of the ribbons and strips are shown in Fig. 4(a). All the ribbons and strips exhibit very similar glass transition and crystallization behaviors, and the glass transition temperature, T_g, is 627 ± 3 K, and the onset crystallization temperature, T_x, is 703 ± 4 K. The DSC traces in the range from 350 K to 650 K are magnified and plotted in Fig. 4(b). The endothermic relaxation events are clearly shown. The relaxation enthalpies,ΔH_rel, were quantitatively calculated to be 1253 ± 30 J mol^-1 for the 40 μm-thick ribbon, 976 ± 12 J mol^-1 for the 110 μm-thick ribbon, 1050 ± 24 J mol^-1 for the 320 μm-thick strip, and 917 ± 12 J mol^-1 for the 490 μm-thick strip, respectively, which are also presented in Fig. 5 for a better readability. With increasing the ribbon thickness, the relaxation enthalpy of the single-roll produced 110 μm-thick ribbon is much smaller than that of the 40 μm-thick ribbon. Similarly, the twin-roll casting prepared 490 μm-thick ribbon possesses a smaller relaxation enthalpy as compared to the 320 μm-thick ribbon. However, it is interesting that the relaxation enthalpy of the twin-roll prepared 320 μm-thick strip exhibits a higher relaxation enthalpy than that of the single-roll prepared 110 μm-thick ribbon, which implies that the structure of 320 μm-thick strip is at a higher energy state as compared with that of the 110 μm-thick ribbon.

View original graphic| Download| PPT slide

Fig. 4.   (a) DSC traces and (b) the enlarged DSC trances before T_g of Vit. 1 ribbons and strips.

View original graphic| Download| PPT slide

Fig. 5.   DSC-measured relaxation enthalpies of the Vit. 1 ribbons and strips.

The nanoindentation measurements were carefully performed on the freely-solidified surfaces of the single-roll prepared ribbons as well as on the center regions of the twin-roll prepared strips (see details in the experimental section). The measured average hardness is 6.17 ± 0.45 GPa for the 40 μm-thick ribbon, 7.48 ± 0.20 GPa for the 110 μm-thick ribbon, 7.29 ± 0.13 GPa for the 320 μm-thick strip, and 7.56 ± 0.23 GPa for the 490 μm-thick strip, respectively, which are plotted in Fig. 6. With increasing the ribbon thickness, the hardness value of the single-roll produced 110 μm-thick ribbon is much higher than that of the single-roll produced the 40 μm-thick ribbon. Similarly, the twin-roll casting prepared 490 μm-thick ribbon exhibits a high hardness value than the twin-roll casting prepared 320 μm-thick ribbon. Opposite to the measured relaxation enthalpies, the twin-roll produced 320 μm-thick strip has a slightly lower hardness value than the single-roll prepared 110 μm-thick ribbon. A lower hardness value of an MG often implies a weaker deformation resistance arising from a higher content of free volume in the amorphous structure [24,25].

View original graphic| Download| PPT slide

Fig. 6.   Nanoindentation-measured hardness of the Vit. 1 ribbons and strips.

In order to further clarify the subtle changes in structures of the prepared glassy ribbons and strips, their 2D SAED patterns, as shown in Fig. S2, were carefully centered and integrated into 1D profiles with intensity I(Q) versus Q, where Q is the momentum transfer, as presented in Fig. 7(a). The positions of the first maximum, Q₁, were determined from the 1D profiles, which are 25.49 ± 0.22 nm^-1 for the 40 μm-thick ribbon, 26.07 ± 0.10 nm^-1 for the 110 μm-thick ribbon, 25.92 ± 0.11 nm^-1 for the 320 μm-thick strip, and 26.10 ± 0.08 nm^-1 for the 490 μm-thick strip, respectively, as shown in Fig. 7(b). With increasing the ribbon thickness, the Q₁ value of the single-roll prepared 110 μm-thick ribbon is higher than that of the single-roll prepared the 40 μm-thick ribbon, and the twin-roll casting prepared 490 μm-thick ribbon has higher Q₁ than the twin-roll casting prepared the 320 μm-thick ribbon. Similar to the measured hardness values, the Q₁ value of the twin-roll prepared 320 μm-thick strip possesses a slightly lower than that of single-roll prepared 110 μm-thick ribbon. A lower Q₁ value of an MG generally means a looser atomic packing with a higher content of free volume [33,34].

View original graphic| Download| PPT slide

Fig. 7.   (a) The integrated 1D diffraction profiles from their SAED patterns, and (b) the measured position of the first diffraction maximum (Q₁).

4. Discussion

For either single-roll melt-spinning or twin-roll casting, the cooling rate decreases with increasing the thickness of ribbons or strips. As a result, the 40 μm-thick ribbon is in a higher energy state as compared with the 110 μm-thick ribbon, and similarly, the 320 μm-thick strip possesses a higher energy state as compared with the 490 μm-thick strip. A higher energy state of an MG means a higher content of free volume, a larger relaxation enthalpy, a lower hardness, and a lower Q₁ value [17,20,[34], [35], [36], [37], [38], [39]]. The different cooling rates cause the measured different relaxation enthalpies, hardness values, and the Q₁ values of single-roll prepared ribbons and twin-roll prepared strips, respectively. However, the twin-roll prepared 320 μm-thick strip is much thicker than the single-roll prepared 110 μm-thick strip, but the relaxation enthalpy of the former is lower, and the hardness and the Q₁ value of the former are noticeably higher than those of the latter, suggesting a higher energy state of the 320 μm-thick strip than the 110 μm-thick strip. Could the cooling rate of the 320 μm-thick strip be higher than that of the 110 μm-thick ribbon?

4.1. Determination of cooling rates

In order to experimentally determine the cooling rates of the 110 μm-thick ribbon and the 320 μm-thick strip, the eutectic Al-33 wt% Cu alloy was adopted to make a ribbon and a strip under the same conditions as that of the 110 μm-thick ribbon and the 320 μm-thick strip. The eutectic Al-33 wt%Cu alloy has been widely used to estimate the cooling rates of copper mold casting [40], water quenching [41], or even the 3D printing [42] by measuring the eutectic lamellar spacing of the as-solidified microstructure. The eutectic temperature of the Al-33 wt% Cu alloy is 821 K [40], which is very close to the average temperature (826 K) of the liquids temperature (1025 K [43]) and the glass transition temperature (627 K) of Vit. 1. Therefore, it is rational to use the Al-33 wt%Cu alloy to estimate the cooling rate of Vit. 1.

The thickness of the single-roll prepared ribbon and the twin-roll prepared strip is 115 ± 4 μm and 325 ± 8 μm, respectively. The slight difference in thickness between the Vit. 1 and the Al-33 wt%Cu ribbon/strip could be caused by the different melt viscosity of two alloys. The microstructures of the Al-33 wt% Cu ribbon and strip are shown in Fig. 8, and the average eutectic lamellar spacing, λ, is measured as 0.16 μm in the ribbon and 0.27 μm in the strip. The cooling rate can be calculated by [41,42]:

$\frac{dT}{dt}=\frac{ΔH_f}{V_mC_p}\frac{3K}{tλ^2}$ (1)

where K is the coefficient of thermarl conductivity, ΔH_f the melting enthalpy, V_m the molar volume, C_p the heat capacity under constant pressure, t the thickness of the ribbon or the half thickness of the strip. K = 2.75 × 10^-8 mm³ s^-1 was measured for an Al-Cu alloy and $\frac{ΔH_f}{V_mC_p}$=400K [40]. Then the cooling rate can be calculated from Eq. (1) to be 1.2 × 10⁴ K s^-1 for the ribbon and 3.1 × 10³ K s^-1 for the strip.

View original graphic| Download| PPT slide

Fig. 8.   SEM micrographs of eutectic Al-33 wt% Cu alloy: (a) the ribbon prepared by single-roll melt spinning and (b) the strip prepared by single-roll casting. The characteristic spacing of the eutectic microstructure is used to estimate the cooling rate.

The above discussion implies that the cooling rate of the 320 μm-thick Vit. 1 strip is almost one magnitude lower than that of the 110 μm-thick Vit. 1 ribbon. The fact that the 320 μm-thick strip possesses a higher energy state as compared to the 110 μm-thick ribbon contradicts to the effect of cooling rates. Therefore, the 320 μm-thick strip with a lower cooling rate is rejuvenated as compared to the 110 μm-thick ribbon with a higher cooling rate. This rejuvenation process should be associated with the processing difference between the single-roll melt spinning and the twin-roll casting.

4.2. Molecular dynamics simulation

The main difference between single-roll melt spinning and twin-roll casting is that the supercooled liquid undergoes viscous flow deformation, known as rheology, caused by the squeezing force during twin-roll casting, and for comparison, the ribbons solidify stress-freely during the single-roll melt spinning. In order to clarify how the microstructure changes during the two different processing methods, molecular dynamics simulations were performed with two models: one is quenched at a rate of 1 × 10¹² K/s under hydrostatic pressure 0 GPa, and the other one is quenched at a lower rate of 5 × 10¹¹ K/s under 1 GPa pressure along x-direction, as shown in Fig. 9, in analogy to the single-roll melt spinning and the twin-roll casting, respectively.

View original graphic| Download| PPT slide

Fig. 9.   Schematic diagram of the model Cu₄₆Zr₄₆Al₈ MGs formed (a) at a cooling rate of 1 × 10¹² K/s under hydrostatic pressure 0 GPa and (b) at a cooling rate of 5 × 10¹¹ K/s under 1 GPa pressure along x-direction.

Fig. 10 shows the distributions of (a) Zr-centered, (b) Cu-centered and (c) Al-centered polyhedra in Cu₄₆Zr₄₆Al₈ MGs with and without pressure. All the polyhedral types were statistically analyzed, and only the ones with a percent higher than 5% are shown. With decreasing the cooling rate from 1 × 10¹² K/s to 5 × 10¹¹ K/s together with exerting 1 GPa pressure, the percent of most of Zr-centered and Cu-centered polyhedra increases, while only the percent of Zr-centered <0,0,10,3> and the Cu-centered <0,3,6,3> slightly decreases a bit. The percent of Al-centered <0,1,10,2> and <0,2,8,2> polyhedra increases, while the percent of most of Al-centered polyhedra, including <0,0,12,0 > , <0,2,8,1 > , <0,3,6,4> and <0,3,6,3 > , decreases. However, since Zr and Cu are the main constituents, it can be safely concluded that the population of polyhedra, i.e. the short-range order (SRO), increases with decreasing the cooling rate from 1 × 10¹² K/s to 5 × 10¹¹ K/s plus 1 GPa pressure. Under a lower cooling rate, the degree of SRO in an MG generally increases, which contributes to its smaller relaxation enthalpy and the lower energy state [1,17]. Therefore, the change of SRO caused by the two different conditions can explain the energy states of either ribbons or strips, but it cannot explain the rejuvenation of the 320 μm-thick strip as compared to the 110 μm-thick ribbon.

View original graphic| Download| PPT slide

Fig. 10.   Distributions of (a) Zr-centered, (b) Cu-centered and (c) Al-centered polyhedra in simulated Cu₄₆Zr₄₆Al₈ MGs.

The connection types of clusters, i.e. the mediate-range order (MRO), were further statistically analyzed. The connection types can be categorized by the numbers of shared atoms between two polyhedra. As shown in Fig. 11, the 1-atom, 2-atom, 3-atom connections refer to the connections by sharing a vertex, an edge and a face of polyhedra, while the 4-atom connection refers to sharing a distorted quadrilateral or squashed tetrahedron [44]. With decreasing the cooling rate from 1 × 10¹² K/s to 5 × 10¹¹ K/s plus 1 GPa pressure, the percent of 1-atom and 3-atom connections decreases from 45.1% to 38.7% and from 30.0% to 22.4%, respectively, while that of 2-atom and 4-atom connections increases from 17.6% to 28.3% and from 5.1% to 9.7%, respectively, as shown in Fig. 11.

View original graphic| Download| PPT slide

Fig. 11.   Statistical connecting types of polyhedra with different shared atomic number(s).

Previously, Ding et al. [44] investigated how the cluster connection schemes evolve during cooling, and they found that a slower cooling rate promotes an increase in the population of 1-atom and 3-atom connections and a decrease of 2-atom and 4-atom connections. This means that the more relaxed amorphous structure at lower energy state obtained at a lower cooling rate is composed of more 1-atom and 3-atom connections and less 2-atom and 4-atom connections [44,45]. Molecular dynamics (MD) simulations [44,45] also indicate that the 1-atom and 3-atom connections exhibit higher deformation resistance as compared with the 2-atom and 4-atom connections.

In the current MD models with decreasing the cooling rate from 1 × 10¹² K/s to 5 × 10¹¹ K/s plus 1 GPa pressure, the percent of the 1-atom and 3-atom connections decreases and the percent of 2-atom and 4-atom connections increases, which is opposite to the evolution tendency of connection types by lowering the cooling rate [44,45]. This is because 1-atom and 3-atom connections exhibit higher deformation resistance as compared with the 2-atom and 4-atom connections, and thus the 1-atom and 3-atom connected clusters bears the major portion of the local stress. Consequently, the 1-atom and 3-atom connections are easily broken or altered to 2-atom or 4-atom connections for the accommodation of large deformation. Therefore, the exerted pressure and the accompanying deformation during twin-roll casting account for the special evolution of connection types, i.e. MRO, and bring the rejuvenation of the 320 μm-thick strip as compared to the 110 μm-thick ribbon, which solidifies stress-freely. It is noteworthy that the applying pressure during the transition from liquid to glass is an effective method to cause the rejuvenation, as revealed very recently by Feng et al [46]. The enrichment of 2-atom and 4-atom connections and the depletion of 1-atom and 3-atom in the current 320 μm-thick strip are the structural origin for its larger relaxation enthalpy, lower Q₁ value due to the higher content of free volume, and lower hardness value, as compared to that of the 110 μm-thick ribbon. In addition, the relaxation enthalpy (917 J mol^-1) of the 490 μm-thick strip is just slightly smaller than that (1050 J mol^-1) of the 320 μm-thick strip, and the smaller difference in hardness and Q₁ values of the strips are also highlighted, which might imply that the squeezing force together with the viscous flow deformation during twin-roll casting can retard the evolution of structural ordering and relaxation caused by lowering the cooling rate.

5. Conclusion

Ribbons with a thickness of 40 μm and 110 μm and strips with a thickness of 320 μm and 490 μm were produced by single-roll melt spinning and twin-roll casting, respectively. The increase in the thickness of either ribbons or strips causes a decrease in cooling rate and a lower energy state of the MG with a smaller relaxation enthalpy, a higher hardness and a higher Q₁ value. However, the twin-roll produced 320 μm-thick strip possesses a rejuvenated energy state as compared to the single-roll produced 110 μm-thick ribbon, although the cooling rate of the former is almost one magnitude lower than that of the latter. MD simulations reveal that the squeezing force and the accompanying flow deformation during twin-roll casting alter the connection types of the clusters in an MG, i.e. the population of 2-atom and 4-atom connections are prone to be increased during twin-roll casting, and such rejuvenation process can overwhelm the relaxation process caused by the lower cooling rate. Therefore, as compared with the single-roll melt spinning, the twin-roll casting is more likely to produce rejuvenated MG strips.

Acknowledgements

This work is supported financially by the National Natural Science Foundation of China (Nos. 51790484, 51701213 and 51801174), the National Key Research and Development Program of China (No. 2018YFB0703402), the Liaoning Revitalization Talents Program (Nos. XLYC1802078, XLYC1807062), the China Postdoctoral Science Foundation (No. 2018M633005), and the Dongguan Innovative Research Team Program (No. 2014607134).

Appendix A. Supplementary data

Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.jmst.2019.08.022.