In-situ study of crystallization kinetics in ternary bulk metallic glass alloys with different glass forming abilities Free

The glass forming ability (GFA) and the phase stability of glassy alloys, and their liquids, have been one of the most important issues in the design and development of bulk metallic glass (BMG) with superior properties. Several empirical criteria for evaluating the GFA have been proposed, such as the reduced glass transition temperature,¹  $T r g = T g T l$ where T_l and T_g are the liquidus and glass transition temperatures, respectively. Another popular criteria is the γ parameter,² defined as $γ = T x T g + T x$⁠, where T_x is the crystallization temperature.

Heating metallic glasses into the supercooled liquid regime induces crystallization. Many factors can affect the crystallization behaviors. The growth rate of nanometer sized crystalline precipitates is controlled by long-range diffusion, which is known to be slow in BMGs. From thermodynamics point of view, the nucleation rate³ for continuous nucleation is determined by the difference in chemical potentials (Δg) and interfacial energy (σ) between the crystalline precipitates and the glass matrix as follows:

where W* is known as the nucleation barrier, which can be expressed as $W * = 16 π σ 3 3 ( Δ g ) 2$⁠, and D is the diffusion coefficient. Recently, Tang and Harrowell^4,5 reported an important molecular dynamics simulation study of crystal growth rates in metallic glasses with different GFA. Their results revealed, however, that even after the difference in diffusion and all other factors are taken into account, the crystal growth rates for a good glass former, ZrCu, are still much slower than that in a poor glass former, NiAl, which they suggest was due to the discrepancy in local structure ordering at the liquid-crystal interface. Thus, in spite of extensive studies so far, crystallization behaviors in metallic glasses, and their relationship to the GFA, are still far from being settled. In this paper, we report an experimental study of the crystallization kinetics of two typical Zr-based ternary metallic glasses using in-situ transmission electron microcopy (TEM) and time-resolved neutron diffraction and correlate the experimental results with their GFAs.

Amorphous alloy ingots with compositions of Zr₅₆Cu₃₆Al₈ and Zr₄₆Cu₄₆Al₈ (atomic %) were, respectively, prepared by arc melting a mixture of Zr (99.99%), Cu (99.99%), and Al (99.99%) in appropriate amounts under a Ti-gettered argon atmosphere. Each ingot was re-melted six times to ensure compositional homogeneity and quenched into cooper-mold by suction casting under high purity Ar atmosphere. The critical casting diameters, D_c, for Zr₅₆Cu₃₆Al₈ and Zr₄₆Cu₄₆Al₈ are 3 mm and 5 mm, respectively, which were determined using X-ray diffraction. To maintain consistency, 3 mm diameter samples were chosen for subsequent experimental studies. Thermal analysis was carried out using a Perkin-Elmer Diamond Differential Scanning Calorimetry (DSC) with a scanning rate of 10 K/min. In-situ neutron scattering experiments were conducted using the Nanoscale-Ordered Materials Diffractometer (NOMAD),⁶ at the Spallation Neutron Source, Oak Ridge National Laboratory.⁷ The as-cast amorphous samples of 3 mm diameter and 30 mm in length were isothermally annealed in a high-vacuum (∼10⁻⁵mbar) furnace. In-situ neutron scattering data were collected with a time resolution of 2 min during isothermal annealing until the samples reached a crystallization equilibrium. The structure factors S(Q), where Q is the momentum transfer, were deduced using NOMAD data analysis software. In-situ TEM images were taken by a Philips CM20 every 2–3 min. Thin foils with 3 mm diameter and 10 μm thickness were polished for TEM. The TEM specimens were final thinned by ion milling. For the convenience of discussion, Zr₅₆Cu₃₆Al₈ and Zr₄₆Cu₄₆Al₈ are abbreviated as Zr56 and Zr46, respectively, in the following text.

Figure 1 shows the DSC scans of Zr56 and Zr46 alloys with a heating rate of 10 K/min. Upon heating, both alloys exhibit a clear glass transition event followed by exothermic events characteristic of crystallization. For the Zr56 sample, T_g is around 683 K and only one crystallization event with an onset temperature T_x of 751 K was observed during continuous heating. For the Zr46 sample, T_g is about 709 K but there are two crystallization peaks instead, at T_x ∼ 776 K and ∼779 K, respectively. These two exothermal peaks suggest a two-step crystallization process.⁸ Table I summarizes the T_g and T_x values for these two glassy alloys, along with their T_l obtained from literatures.^9,10 Based on these data, the GFA indicators T_rg and γ can be estimated, as shown in Table I. The values of T_rg and γ for Zr46 are both larger than those for Zr56, indicating that Zr46 is a better glass former, which it is. The T_rg and γ values are consistent with experimentally determined critical diameters, D_c, which are also listed in Table I, along with the estimated critical cooling rates, R_c.

Figure 2(a) compares the S(Q) of the two alloys at room temperature. The S(Q) data as a function of annealing time are shown in three dimension (3D) plots in Figures 2(b) and 2(c). The annealing temperatures were set at 45 K above their respective T_g, and the precision for temperature control was 1–2 K. For both samples, there were incubation periods, within which the diffraction patterns showed no discernable changes. From Figures 2(b) and 2(c), we estimated that the incubation time is about 90 and 98 min, respectively, for Zr56 and Zr46. The appearance of the Bragg peaks marked the end of the incubation period, after which the intensities of diffraction peaks started to grow while the scattering intensities from the amorphous matrix were receding, indicative of continued growth of the crystalline phases by consuming the amorphous matrix.¹¹ 

The inset of Figure 2(a) shows that, at the crystallization equilibrium, there is a clear distinction between the diffraction patterns of these two samples. For the Zr56 sample, the diffraction peaks are sharp and strong. By performing Rietveld refinement using GSAS, the crystalline phase was identified to be the tetragonal Zr₂Cu. In contrast, the crystalline peaks for the Zr46 sample are shallow and broad, indicating that the crystalline phase of the Zr46 glass is poorly ordered. Rietveld analysis indicated that the phase in Zr46 is the orthorhombic Zr₇Cu₁₀.

Figure 2(d) shows the temporal evolution of the crystalline phases in both samples. The intensity was obtained by integrating over the Q-range where prominent crystalline peaks appear. The initial flat curves represent the incubation periods. For Zr56, the slope turned steep and the growth rate became quite large around ∼90 min. Even after ∼240 min of annealing, the crystalline phase in the Zr56 alloy was still growing.

For the Zr46 sample, the scattering intensity increased slowly after ∼98 min. In the early stage of growth, the growth rate in Zr46 was significantly slower than that in Zr56. At ∼150 min, the scattering intensity in the Zr46 sample reached a maximum. After this point, the intensity curve turned sharply into a plateau, indicating no more growth.

To gain further insights into the crystallization behaviors of both alloys, in-situ TEM studies were performed (for details, see supplementary material¹²). Figures 3(a) and 3(b) display the bright field images of the Zr56 sample isothermally annealed at an effective temperature of 731 ± 1 K (T_g + 48 K) for 90 and 240 min, respectively. Figures 3(c) and 3(d) are the equivalent for Zr46 annealed at 759 ± 1 K (T_g + 50 K) for 90 and 240 min, respectively. From Figure 3(a), in Zr56, a small amount of crystals emerged from the amorphous matrix after 90 min of annealing, and diffraction spots can be seen superimposed on the amorphous diffuse halos in the selected area diffraction (SAED) (inset). In contrast, in Zr46, the image in Figure 3(c) shows many black dots scattered in the matrix. After ∼90 min annealing, the SAED pattern (inset) shows that the first amorphous halo became shaper and two more outer-diffraction rings appeared, but no diffraction spots yet. The large number of those black dots indicates a high density of nucleation.

With increasing annealing time, the average size and volume fraction of crystals increased for both alloys. The bright field image in Figure 3(b) shows uniformly distributed islands in the amorphous matrix of the Zr56 sample. In Figure 3(d), however, the bright field image reveals a different microstructure: a high density of nanoparticles pinned to each other and formed a “flower-like” structure. The corresponding diffraction pattern shows that diffraction spots distributed in the diffuse rings, which indicates that the residual matrix was dotted with a large amount of nano-size particles.

The in-situ experimental data, from both neutron diffraction and TEM, demonstrated clear differences in crystallization behaviors between the two samples under study. The Zr56 sample, an average glass former, is relatively easy to crystallize. The DSC scans exhibited one crystallization temperature, and the crystalline phase was identified to be Zr₂Cu. The crystalline peaks are sharp, indicating a well-ordered Zr₂Cu phase. The kinetics of crystallization is characterized by continuous nucleation and growth, with crystalline particles reaching 125 nm as measured by TEM.

The Zr46 sample, a better glass former, showed much more resistance against crystallization. The DSC scans showed two crystallization events, and the final crystalline phase was identified to be Zr₇Cu₁₀. The neutron diffraction peaks from the Zr₇Cu₁₀ particles are, however, broad, indicting significant local disorder, both chemical and topological. In-situ TEM showed a high-density of crystalline nucleus during the incubation period, as evidenced by the large number of black dots in, e.g., Figure 3(c). The growth of the nanometer sized crystalline particles is slower in Zr46, which can be seen from the time-resolved neutron diffraction data in Figure 2(c). Furthermore, at ∼150 min, the intensity of the crystalline peaks stopped growing altogether, suggesting soft impingement of nanometer sized crystalline particles. This observation is corroborated by TEM images shown in Figure 3(d), which showed particles pinned to each other. The particle size at equilibrium, as estimated from TEM, is about 55 nm. Overall, the crystallization kinetics of Zr46 can be reconciled with high-density site-saturated nucleation, followed by slow growth.

The unique crystallization behavior observed for Zr46 shed some lights on the stability and hence better GFA of the alloy. Although a large number of crystalline nuclei formed during incubation, they grew little during further annealing. On the other hand, new crystalline particles are difficult to nucleate, which can be seen by examining the nucleation rate given by Eq. (1). To the first order of approximation, the σ values can be considered approximately the same for both alloys. Another parameter, Δg, is the Gibbs free energy difference between the original and new phase per unit volume, which depends on the magnitude of the supercooling, ΔT = T_l-T.³ Based on Table I, the supercooling of Zr56 is ∼484 K, and that of Zr46 is ∼405 K. Thus, the Zr46 alloy has a smaller Δg value, which leads to a larger W*, and hence a slower nucleation rate. Furthermore, the crystal growth rate is also smaller in Zr46, as can be seen from Figures 2(d) and 3(d), suggesting slower diffusion in Zr46. As the nucleation rate is proportional to the diffusion coefficient, see Eq. (1), this further reduces the nucleation rate of new crystalline particles in Zr46. Interestingly, a separate study by molecular dynamics simulation of (Cu_0.5Zr_0.5)_100−xAl_x (x ≤ 10) suggested that the diffusion coefficient reaches a minimum when x ∼ 7–8, close to the composition of Zr46.¹³ 

The slower diffusion in Zr46 is the main reason of the slow growth of the nucleated crystalline particles. At longer annealing times, the high-density of small crystalline particles began to overlap, further suppressing the growth and continuing crystallization. The slower diffusion in Zr46 is also accountable for the poorly ordered Zr₇Cu₁₀ phase observed in the Zr46 sample.

The exact source of the high-density of nucleation in Zr46 is not clear at the moment. One possibility is that in Zr46, there was a significant level of local density fluctuation in the quenched liquid state. Indeed, there is evidence by atom probe tomography that high-density Cu clustering occurs in as-cast nanostructured steel¹⁴ and in annealed Fe-based metallic glass ribbons.¹⁵ Another possibility lies in the fact that Zr46 follows a two-step crystallization, as its DSC exhibits two crystallization peaks,¹⁶ the final crystalline phase being Zr₇Cu₁₀. The first phase, yet to be identified, is metastable. It is possible that the chemical potential of this metastable phase is closer to that of the glass, which promotes the nucleation of the crystalline phases. The differences in the crystalline phases, Zr₂Cu vs. Zr₇Cu₁₀, particularly the local structure at the crystal/liquid interface, may also play an important role, as alluded to by Tang and Harrowell in Ref. 4. Further experiments are necessary to resolve which scenario is ultimately responsible.

Strictly speaking, the crystallization study reported here addresses the stability of the supercooled liquid during heating rather than the GFA, which are characterized by the critical heating and critical cooling rates (to avoid crystallization), respectively. For metallic glasses, the critical heating rate is usually much higher than the critical cooling rate, a phenomenon known as asymmetric heating/cooling.^17,18 However, as pointed out by Schroers et al.^17,18 and lately by Johnson et al.,¹⁹ for several well-known metallic glass alloys, the critical heating rate is about two orders of magnitude higher than the critical cool rate. In this regard, the stability of metallic glasses is related to their GFA, and the crystallization study such as ours provides an angle to look into the GFA of metallic glasses.

In summary, in-situ neutron diffraction and TEM studies revealed dramatically different crystallization behaviors in two Zr-based ternary metallic glass alloys. Zr₅₆Cu₃₆Al₈, an average glass former, crystallized to a well-ordered Zr₂Cu phase whose kinetics is characterized by continuous nucleation and growth. Crystallization in Zr₄₆Cu₄₆Al₈, a better glass former, is much more complicated, showing two crystallization peaks, and the kinetics is characterized by site-saturated nucleation followed by slow growth. These observations correlate well with the stability and GFA of the alloys under study and point to the differences in crystalline phases and/or nanoscale density fluctuation in the glassy alloys.

The neutron scattering experiments were carried out at the Spallation Neutron Source, which is sponsored by the Scientific User Facilities Division, Office of Basic Energy Sciences, U.S. Department of Energy, under Contract No. DE-AC05-00OR22725 with Oak Ridge National Laboratory. J.Z. and Z.P.L. thank financial support from National Natural Science Foundation of China (Grant Nos. 51010001, 51001009, and 51271212) and 111 Project (B07003).