2025 Volume 34 Issue 3
Article Contents

Shan Zhang, Qingan Li, Yong Yang, Pengfei Guan. Atomic origin of minor alloying element effect on glass forming ability of metallic glass[J]. Chinese Physics B, 2025, 34(3): 036105. doi: 10.1088/1674-1056/ada756
Citation: Shan Zhang, Qingan Li, Yong Yang, Pengfei Guan. Atomic origin of minor alloying element effect on glass forming ability of metallic glass[J]. Chinese Physics B, 2025, 34(3): 036105. doi: 10.1088/1674-1056/ada756

Atomic origin of minor alloying element effect on glass forming ability of metallic glass

  • Received Date: 17/10/2024
    Accepted Date: 25/12/2024
    Available Online: 01/03/2025
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Figures(7)

Article Metrics

Article views(190) PDF downloads(3) Cited by(0)

Access History

Atomic origin of minor alloying element effect on glass forming ability of metallic glass

Abstract: The glass-forming ability (GFA) of metallic glasses is a key scientific challenge in their development and application, with compositional dependence playing a crucial role. Experimental studies have demonstrated that the addition of specific minor elements can greatly enhance the GFA of parent alloys, yet the underlying mechanism remains unclear. In this study, we use the ZrCuAl system as a model to explore how the addition of minor Al influences the crystallization rate by modulating the properties of the crystal–liquid interface, thereby affecting the GFA. The results reveal that the minor addition of Al significantly reduces the crystal growth rate, a phenomenon not governed by particle density fluctuations at the interface. The impact of minor element additions extends beyond a modest increase in crystal-unfavorable motifs in the bulk supercooled liquid. More importantly, it leads to a significant enrichment of these motifs at the crystal-supercooled liquid interface, forming a dense topological network of crystal-unfavorable structures that effectively prevent the growth of the crystalline interface and enhance GFA. Our results provide valuable insights for the design and development of high-performance metallic glasses.

1.   Introduction
  • Metallic glasses (MGs) are non-equilibrium metastable systems that have found significant industrial applications due to their excellent strength and high hardness.[1–4] They also exhibit excellent resistance against wear and corrosion,[5,6] remarkable irradiation resistance,[7] and demonstrate superior performance in catalysis, magnetism, and electronics.[8–10] In the field of materials science, understanding the glass forming ability (GFA) of materials holds great importance and has consistently been a focal point of research interest, driven by its significance. The GFA of a material is primarily influenced by its chemical composition, and theoretical models have been developed to predict and explain GFA,[11–15] providing a fundamental understanding of the underlying mechanisms.[16–18]

    However, despite notable advancements, accurately explaining the GFA of materials continues to be a challenging task due to the complex nature of glass transition. Recent studies have highlighted the significant influence of minor alloying elements on GFA.[19] Experimental investigations have demonstrated that the addition of a small percentage of minor elements to various MGs often leads to substantial improvements in both their GFA and properties.[20–22] For example, the GFA of the CuZr-based bulk metallic glasses (BMGs) can be dramatically enhanced by incorporating minor element additions, resulting in large diameters of up to 15 mm.[23,24] Similarly, the critical diameter of Ce–Al–Cu glassy rods can be significantly increased from 2 mm to at least 3 mm–10 mm by introducing minor amounts of elements, such as Fe, Co, Ni, Nb, Zn, and Si.[25] In the case of Gd55Co20Al25 MG, the critical diameter increased from 2 mm to 7 mm through the addition of 0.5-at.% Si, while maintaining the magnetocaloric effect.[26] Furthermore, the inclusion of minor elements in MGs has been found to enhance plasticity[27] and saturation magnetization.[28] Despite the progress made in understanding the impact of minor elements on GFA,[26,29–31] the atomic origin of the effect of minor alloying element remains elusive.

    In this work, we employed molecular dynamics (MD) simulations to elucidate the atomic-scale mechanisms behind the enhancement of GFA through the addition of minor Al elements in CuZr-based MGs. As the Al content increases from 0 to 8%, the crystallization rate declines rapidly from 0.2 m/s to 0.02 m/s. Through the analysis of the atomic structure at the liquid–crystal interface during crystallization, we discovered that the addition of minor Al to CuZr-based MGs alters the local topological structure order while almost not affecting the overall atomic density distribution. This alteration leads to the swift disappearance of the crystal-like structure and the emergence of the local atomic clusters of high fivefold symmetry, resulting in a network of such atomic clusters at the solid–liquid interface that disrupts the continuous pathways necessary for crystal structure formation.

2.   Computational details
  • In our MD simulations, we focused on the (100) and (110) faces of (ZrCu)100−xAlx alloys with varying Al content (x = 0, 2, and 8) to investigate the influence of minor element on crystal growth rate. To carry out these simulations, we utilized the widely used package of Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS),[32] employing the embedded atom method potentials,[33] specifically designed for the Cu–Zr–Al alloy system. Our simulation cells had 11520 atoms and the dimensions of approximately 4 nm×4 nm×24 nm, and periodic boundary conditions were applied in all three directions. As depicted in Fig. 1(a), the substrate consisted of a Zr50Cu50 (x = 0) crystalline solid layer with a thickness of about 4 nm, while the growth layer comprised a supercooled liquid layer of (ZrCu)100−xAlx (x = 0, 2, and 8) with a thickness of 10 nm. To isolate the impact of the crystal/liquid interface on the crystal growth rate, we introduced a vacuum layer of 10-nm along the long axis of the cell. The simulations proceeded as follows: (i) The samples were heated from 900 K to 1200 K using the constant number, pressure, and temperature (NPT) ensemble. (ii) A vacuum layer of 10 nm was established along the z-axis, and the supercooled liquid region (z = 4 nm– 14 nm) was melted. The entire simulation cell was then under the constant number, volume, and temperature (NVT) condition. At each temperature, the atoms in the bottom 2 nm of the crystalline layer were kept fixed, while the remaining 2 nm served as a buffer-layer. The amorphous layer beyond the 4-nm region was allowed to relax at various temperatures under NVT conditions. After this relaxation, the 2-nm buffer-layer and the amorphous layer relaxed under NVT conditions at the target temperature point. (iii) The samples were relaxed at the melting temperatures to achieve a well-equilibrated crystal/liquid system. (iv) Finally, the crystal growth rates were measured at various temperatures.

3.   Results and discussion
  • According to Ref. [34], the crystal growth rate can be calculated using the equation v=V˙dl/Nl(Vc−Vl), where V˙ represents the rate of change in the volume of the simulation cell, dl is the distance of the liquid layer to crystal layer along with the interface normal, Nl is the average number of atoms from the liquid to crystal phase, and Vc and Vl denote the volumes of the crystal and liquid layer, respectively. Figure 1(b) illustrates the crystal growth rates as a function of temperature for different compositions of (ZrCu)100−xAlx (x = 0, 2, and 8) with (110) and (100) surfaces. Notably, the growth rates of two surfaces with the same composition were almost identical. Moreover, the temperature at which the maximum crystal growth rate occurs remains consistent regardless of changes in chemical composition. The addition of Al to the (ZrCu)100−xAlx alloy significantly reduces the crystal growth rate. Specifically, when the Al content reaches 8%, the crystal growth rate decreases from 0.2 m/s to 0.02 m/s. Additionally, the temperature range for crystallization (v (T) > 0) narrows, indicating that (ZrCu)92Al8 exhibits enhanced GFA. In theory,[34] the crystal growth rate can be further described by the following equation:

    where Δμ(T) represents the difference in chemical potential between the liquid and crystal phases; kB is the Boltzmann constant; and k(T) is a prefactor that can be determined using the Wilson–Frenkel theory.[35,36] Finally, the crystal growth rate can be formulated as

    In this formulation, D represents the diffusion constant, σ is obtained from the position of the first peak in the radial distribution function, and L is the lattice spacing normal to the surface. The term v(T)σ2/(DL) signifies the overall number of crystal layers deposited per diffusion time. As shown in Fig. 1(b), the crystallization growth rates of different crystal planes were roughly the same, with similar overall trends. However, as shown in Fig. 1(c), the intrinsic growth rates of the {100} and {110} planes are different. According to Eq. (2), the different intrinsic growth rate of each plane is mainly due to the change in different lattice spacing, related to L. The constant C, which can be extracted from the slope of the fitting lines in Fig. 1(c) with its value depicted in Fig. 2(b), represents the rate of liquid layer addition per diffusion time. Notably, as the Al content increases to 8%, C approaches zero at deeply supercooled temperatures.

    Figure 2(a) demonstrates that the addition of Al has minimal impact on the diffusion constant D values for Cu and Zr atoms, except for slight variations at extremely high temperatures. Moreover, the introduction of minor elements has negligible effects on the values of σ and L on both the (110) and (100) surfaces. However, the crucial factor, C, plays a significant role in distinguishing the crystal growth rates. As depicted in Fig. 1(c), the ZrCu alloys reveal a temperature-independent nature of C, which aligns with previous research.[34] In contrast, when Al is added, C exhibits a decrease at lower temperatures on the same surface. Particularly, when Al comprises 8% of the composition, a notable reduction in the C parameter is observed on the (110) surface at lower temperatures. This substantial deviation in C for (ZrCu)92Al8 suggests unusually sluggish dynamics at lower temperatures.

  • Based on previous studies,[34,37,38] it is clear that the crystal growth rate is influenced by the structure and chemical composition of the crystal/liquid interface. Figures 3(a) and 3(c) depict the normalized atomic population as a function of distance from the solid/liquid interface, accompanied by the corresponding top views of atom positions shown in Figs. 3(b) and 3(d). In Figs. 3(a) and 3(c), it can be observed that the peak concentration of Cu/Zr decreases, approaching that of the supercooled liquid over approximately five atomic layers. This behavior allows us to extract a correlation length of density peak for Cu/Zr density distribution. The introduction of Al into the system does not result in significant variations in the correlation length of Cu/Zr density distribution.

    Our attention now turns to the structure of local atomic packing. Figure 4(a) presents the distributions of normalized density ρ as a function of distance from the crystal layers in the (ZrCu)100−xAlx alloys (x = 0 and 8), revealing nearly identical curves. These density distributions were obtained using the iso-configurational averaging method. The introduction of trace elements has no discernable impact on the macroscopic order. The presence of an icosahedral structure, characterized by five-fold symmetry, creates frustration and competition with crystallization. To quantify the local atomic symmetry, we employed the parameter f5 defined as f5=n5/∑i=36ni, where ni represents the number of i-edged faces of the Voronoi polyhedron. Figure 4(b) illustrates the normalized sum-f5 values for each layer as well as the sum atomic f5 values. Significantly, the normalized sum-f5 value of (ZrCu)92Al8 increases rapidly in the three closest interface layers, particularly in the third layer. This indicates a larger number of icosahedron-like structures within the interface layers. As the distance from the solid–liquid interface increases (beyond the 3rd layer), the probability of five-fold symmetry clusters gradually approaches the liquid level of Zr50Cu50. The addition of a small amount of Al rapidly enhances the presence of five-fold symmetry clusters at the solid–liquid interface, a phenomenon that diminishes with increasing distance from the interface. In Figs. 4(c)–4(f), it can be observed that the addition of Al leads to larger f5 values for Al and its neighboring atoms. Overall, the minor addition of Al does not alter the density order, but it does influence the local packing order.

  • To analyze the microstructure details of the solid–liquid surface, we employed bond-orientational order (BOO) parameters (qlm)[39] to measure the local orientational order between a particle and its neighboring atoms, which can be expressed as

    where rij represents the bond between particles i and j, Ylm(θ(rij),ϕ(rij)) denotes the spherical harmonics of bond rij with l degree and m order, and Nb(i) represents the number of the first nearest neighbor of particle i.

    It is common to utilize multiple BOO to characterize a particle’s local environment. The coarse-grained Qlm(i) is used to differentiate between different ordered environments and reduce the overlap between two different BOO parameter maps.[40] The plot of Q4 against Q6 is commonly employed to distinguish between disordered (liquid-like) and ordered (crystal-like) environments. Figure 5(a) demonstrates that Zr50Cu50 exhibits a connected structure in the interface layers, indicating a cohesive atomic environment surrounding the solid–liquid interface in Zr50Cu50. However, in the case of (ZrCu)92Al8, a distinct boundary appears in the same region, as depicted in Fig. 5(b). The addition of Al induces a significant alteration in the atomic packing structure, leading to the emergence of a new structure, as evidenced by an additional boundary line in the Q4–Q6 interface plot.

    To further characterize the microstructure, we employed the Voronoi index. Figures 5(c) and 5(d) present the proportions of the four dominant ordering Voronoi polyhedral in Zr50Cu50 and (ZrCu)92Al8, respectively, along the crystal growth direction at 1050 K. At the solid–liquid interface, the crystal order content of Zr50Cu50 gradually decreases while maintaining good connectivity in the subsequent layers. In contrast, the crystal order of (ZrCu)92Al8 sharply drops to less than 20% within one interface layer, resulting in a rapid accumulation of B2-unfavorable motifs. In CuZr-based MGs, these two competing structures, crystal order and B2-unfavorable motifs, have been observed to exhibit significantly different temperature response behaviors.[29,31,41] These changes impede further progress in the crystallization process. We performed a cluster-linking analysis on the icosahedral structures at the interface, as illustrated in Figs. 5(e) and 5(f). The results revealed that when x = 8, nearly all icosahedra at the interface are interconnected, forming predominantly uniform clusters of the same type. In contrast, when x = 0, the density of icosahedra is significantly lower, and they are spatially dispersed. According to Fig. 5(e), when the icosahedral-like network structure at the interface has a lower degree of connectivity, the connectivity between the crystalline surface layer and the liquid phase is not blocked, and the interface region contains a large number of B2-favorable motifs that exhibit characteristics of the liquid-to-crystal phase transition (as shown in the gray area of the 6 nm–8 nm region in Fig. 5(c)), which facilitates continued crystal growth. In contrast, when the network connectivity of the B2-unfavorable motifs, such as icosahedral motif, is higher, the crystalline surface layer is effectively separated from the liquid phase, leading to a significant reduction in B2-favorable motifs with liquid-to-crystal phase transition characteristics (as shown in the gray area of the 7.5 nm–8 nm region in Fig. 5(d)), thus effectively preventing the growth of the crystalline interface.

    This behavior is especially prominent at deep supercooled temperatures in (ZrCu)92Al8. Figure 6(a) illustrates the composition of the B2-unfavorable motifs: icosahedron 〈0, 0, 12, 0〉, icosahedral precursor 〈0, 2, 8, 2〉, and the two B2-favorable motifs: 〈0, 6, 0, 8〉, 〈0, 5, 2, 6〉 along the z-axis direction of (ZrCu)92Al8 at 950 K. The presence of the icosahedral structure and its precursors, which exhibit a high degree of five-fold symmetry, rapidly increases to 70% in the layer immediately following the solid–liquid interface. These structures can act as network modifiers, introducing structural disorder and promoting glass formation. As a result, the crystallization behavior becomes almost stagnant, leading to a substantial drop in the crystallization rate as depicted in Fig. 1(c) for (ZrCu)92Al8 at supercooled temperatures. In Fig. 6(b), we plotted the distributions of Q6 and w6 for the B2-unfavorable motifs and B2-favorable motifs in (ZrCu)92Al8 at 950 K.

    The diagonal distributions suggest that the two structures possess entirely different trends in the bond orientation BOO map.

    Figure 6(c) visualizes the spatial distribution of B2-unfavorable motif and B2-favorable motif. At a deep supercooled temperature, the solid–liquid interface of (ZrCu)92Al8 rapidly fills with B2-unfavorable motifs with near-total five-fold symmetry, blocking the continued growth of crystals and enhancing the GFA. The increase in icosahedral order near the interface can be attributed to two main factors: (i) the addition of Al significantly reduces the formation energy of 〈0, 0, 12, 0〉-like motif, as previously proposed; (ii) the wall-pinning effect induced by the interface leads to a significant slowdown in dynamics near the interface.[42] This slowdown indicates a lower effective temperature in the interfacial region, which provides more favorable conditions for the formation of icosahedral clusters. Additionally, the aggregation of icosahedral clusters further increases viscosity, promoting the formation of more icosahedral clusters. Ultimately, a dense network is woven near the interface, preventing the growth of the crystalline phase.

4.   Discussion
  • To achieve a more comprehensive understanding of minor alloying element effect of glass forming ability of metallic melts, we employed the forward flux sampling (FFS)[43,44] technique to study the nucleation processes of two glass-forming liquids with different compositions: Zr50Cu50 and Zr46Cu46Al8, in addition to the aforementioned growth processes. The simulated temperature was set at 1050 K, corresponding to the fastest growth rate shown in Fig. 1(b). Based on the bond orientational order (BOO) parameter Sl(i, j),[40,45] we propose using the averaged BOO parameter S¯l(i,j) to distinguish between solid and liquid:

    where Qlm(i) is defined by Eq. (5), and we set l = 6. Two particles i and j are defined to be connected if S¯6(i,j)>0.89, and particle i is considered solid-like if its number of connections is greater than or equal to 7. Ultimately, we count the number of atoms in the largest cluster composed of solid-like atoms as the parameter λ. To study the rate kAB of the system’s evolution from λA to λB (λA < λB), FFS inserts additional interfaces λA ≤ λ0 < λ1 < ⋯ < λn = λB between λA and λB. Consequently, the rate kAB can be expressed as[43]

    where Φ¯A,0 represents the average flux from λA to λ0, the factor h¯A represents the average time the system spends in the “attractive basin” A, and P(λi+1|λi) represents the probability of proceeding to λi+1 after reaching λi rather than returning to λA. Furthermore, we utilize the temporally coarse-grained FFS[44] and increase the sampling time to 1 ps to filter out undesirable high-frequency fluctuations.

    Figure 7 illustrates the simulated crystal nucleus growth probability for investigated compositions. Contrary to the growth process discussed earlier, Zr46Cu46Al8 exhibits a significantly higher nucleation rate compared to Zr50Cu50. Upon examining the specific morphology of the nuclei, it was found that both alloys formed the Laves phase rather than the B2 phase. This finding agrees with a previous study that reported the formation of the Laves phase in Zr46Cu46Al8 through extensive MD simulations[46] based on the same EAM potential.[33] This indicates that the addition of Al enhances local five-fold symmetry in the ZrCu melt, facilitating the formation of the Laves phase, which incorporates Cu-centered icosahedral motifs. The inhibition of B2 crystalline phase growth observed in interface growth simulations, attributed to the enhanced icosahedral network at the interface, may stem from the competition between the matrix B2 phase and the Laves phase, which tends to nucleate in the supercooled liquid. Regardless of the specific mechanism driving the increase in B2-unfavorable motifs at the interface, this phenomenon ultimately alters the structural evolution pathway and elevates the activation energy barrier,[47,48] thereby complicating the crystalline growth process. Consequently, regulating the topological stacking at the crystal–liquid interface through trace element addition emerges as an effective strategy for optimizing glass-forming ability. Achieving accurate representation of trace element effects in simulations, however, necessitates the use of high-precision interatomic potentials, such as machine-learning-based potentials.[49,50]

    In this study, we utilize the ZrCuAl system as a model to investigate how trace additions of Al influence the crystallization rate by modulating the properties of the crystal–liquid interface, thereby impacting the glass-forming ability (GFA). Unlike previous findings in various alloy systems, where differences in crystallization ability could be inferred from the decay of atomic density fluctuations, the addition of trace amounts of Al does not alter the density fluctuation decay in the interfacial region. Nevertheless, it leads to a substantial reduction in the crystallization rate, by several orders of magnitude compared to the ZrCu alloy system. Detailed analysis of the local structure reveals that even a small amount of Al modifies the motifs in the interfacial region. These motifs transition from crystal-phase-favorable motifs, which typically form under the influence of the crystalline field, to 〈0, 0, 12, 0〉 motifs, which are unfavorable for crystal growth. This shift increases the energy barrier for crystal growth, thereby reducing the growth rate. The underlying mechanism is that Al addition significantly lowers the formation energy of 〈0, 0, 12, 0〉 motifs. The formation role of motifs is governed by the cooperative interactions among elements. It underscores the inherent challenges and complexities of composition-dependent GFA in alloys, warranting systematic investigation using AI-associated large-scale atomic models in future studies.

    In summary, the introduction of minor elements led to a filling of the solid–liquid interface with crystal-unfavorable motifs, which is significantly higher than that in the bulk liquid under the same bath temperature, impeding the formation pathway of crystalline-like order. These structures with fivefold symmetry formed a network, hindering the formation and growth of crystalline order. This effect was particularly pronounced at deep supercooled temperatures, reminiscent of a pinning effect[51] that reduced structural rearrangement, increased the content of icosahedra, and ultimately improved the system’s GFA. Therefore, the effect of trace element addition is not limited to a modest increase in crystal-unfavorable motifs within the bulk supercooled liquid. More importantly, it leads to a significant enhancement of such motifs at the crystal–supercooled liquid interface, which provides critical insights that cannot be obtained solely from studies of bulk supercooled liquids. This finding offers a potential approach for the preliminary assessment of glass-forming ability (GFA) in simulations. By comparing the differences in motifs between the interfacial region and the crystal phase, a qualitative evaluation of GFA may be achievable.

Figure (7)  Reference (51)

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return