Sensitive Five-Fold Local Symmetry to Kinetic Energy of Depositing Atoms in Cu-Zr Thin Film Growth

We have investigated the glass formation ability of Cu-Zr alloy by molecular dynamics simulation of the deposition process. The atomistic structures of ZrxCu100−x metallic glass films have been systematically examined under the growth conditions of hypereutectic-eutectic, near-eutectic, and hypoeutectic regions by the radial distribution function and simulated X-ray diffraction. The structure analysis using Voronoi polyhedron index method demonstrates the variations of short-range order and five-fold local symmetry in ZrxCu100−x metallic glass films with respect to the growth conditions. We manifest that the five-fold local symmetry is sensitive to the kinetic energy of the depositing atoms. There is positive correlation between the degree of five-fold local symmetry and glass forming ability. Our results suggest that sputtering conditions greatly affect the local atomic structures and consequential properties. The glass forming ability could be scaled by the degree of five-fold local symmetry. Our study might be useful in optimizing sputtering conditions in real experiments, as well as promising implications in material design of advanced glassy materials.


Introduction
A solid metal in general is a crystalline where atoms have long range orders. When they are quenched with ultra-high cooling rate from melt, they form amorphous metals, namely metal glasses (MGs). Unlike common glasses, metallic glasses are metals, and therefore good in electrical conductivity and thermal conductivity. Without long-range orders or translational symmetry in structure, metallic glasses in general possess much higher strength, hardness, and corrosion resistance than their crystalline counterparts [1][2][3][4]. As a result, MGs have attracted extensive interest due to their exotic mechanical properties and promising applications [5,6]. However, the poor glass-forming ability (GFA) limits the broad engineering application of MGs, which restricts both the shape and the size of MGs. The key to the design of MGs with great GFA is to understand the structural origin of GFA, which, however, is a long-standing challenge.
Metallic glasses are usually obtained by rapid cooling of multicomponent alloy that the crystallization is avoided. The binary Zr-Cu alloy is a paradigm system to study the atomistic structures properties relationships and glass transition due to the outstanding glass formation ability which originated from the ideal atomic size ratio between the two constituents [7]. Recent studies have shown that binary Zr-Cu alloys have the ability to form a variety of stable metallic glass with different atom ratios [8,9], and the amorphous film formed by Zr-Cu alloys has prefect mechanical [10][11][12] and superconducting properties [13].  Figure 1 shows the initial deposition model used in this work. Each simulation box has periodic boundary conditions only in x and y directions. Si (1 0 0) with fixed bottom atoms was used as the substrate and the dimensions were (25 × 25 × 10) Å 3 . In the deposited process, two kind of deposited atoms were generated randomly. Both Zr and Cu atoms deposited from an initial position 5-7 Å above the deposition model.  Figure 1 shows the initial deposition model used in this work. Each simulation box has periodic boundary conditions only in x and y directions. Si (1 0 0) with fixed bottom atoms was used as the substrate and the dimensions were (25 × 25 × 10) Å 3 . In the deposited process, two kind of deposited atoms were generated randomly. Both Zr and Cu atoms deposited from an initial position 5-7 Å above the deposition model. For all simulations, the microcanonical ensemble (NVE) was employed with a Berendsen thermostat [34]. Berendsen thermostat can effectively dissipate the high energy coming from Zr-Cu atoms. The temperature of the simulation box was controlled to 300 K, which is similar to experimental temperature condition.

Modeling
In these simulations, embedded atom method (EAM) many-body potential [35] was applied to describe the atomic interactions of the Zr-Cu systems. The interactions between Si and Si atoms were described by Tersoff empirical potential [36]. Lennard Jones potential with Lorenz Berthelot mixing rules [37] was employed for the atomic interactions between Si and Zr-Cu systems.

Simulation Process
We use the modified Thompson formula (detailed in Supplementary) to calculate the mean kinetic energy [38] of the incoming Zr and Cu atoms. We examine three different experimental conditions ( Table 1). The velocities of the incoming atoms are randomly sampled from a Maxwell Boltzmann distribution. The aim is to simulate the three different conditions in metallic glass growth, and to gain insights of the effect of both the incoming atom kinetic energy and the composition on the expected phase formation of Zr-Cu system. The deposited atoms are randomly generated. The initial velocity and direction are given to make them continuously deposited to the substrate. The time step is chosen to be 0.001 ps for the compromise of computation speed and numerical instability. Every 2 ps, 5 atom are released toward For all simulations, the microcanonical ensemble (NVE) was employed with a Berendsen thermostat [34]. Berendsen thermostat can effectively dissipate the high energy coming from Zr-Cu atoms. The temperature of the simulation box was controlled to 300 K, which is similar to experimental temperature condition.
In these simulations, embedded atom method (EAM) many-body potential [35] was applied to describe the atomic interactions of the Zr-Cu systems. The interactions between Si and Si atoms were described by Tersoff empirical potential [36]. Lennard Jones potential with Lorenz Berthelot mixing rules [37] was employed for the atomic interactions between Si and Zr-Cu systems.

Simulation Process
We use the modified Thompson formula (detailed in Supplementary) to calculate the mean kinetic energy [38] of the incoming Zr and Cu atoms. We examine three different experimental conditions ( Table 1). The velocities of the incoming atoms are randomly sampled from a Maxwell Boltzmann distribution. The aim is to simulate the three different conditions in metallic glass growth, and to gain insights of the effect of both the incoming atom kinetic energy and the composition on the expected phase formation of Zr-Cu system. The deposited atoms are randomly generated. The initial velocity and direction are given to make them continuously deposited to the substrate. The time step is chosen to be 0.001 ps for the compromise of computation speed and numerical instability. Every 2 ps, 5 atom are released toward the surface. This rate of 0.5 atom/ps is carefully selected so that there is enough time for thermal relaxation to take place with Berendsen thermostat.
For post-simulation analysis, we use the radial distribution functions (RDF) to determine the correlation between the atoms. The simulated X-ray diffraction (XRD) was employed to study the phase of Zr-Cu film and the Voronoi polyhedron index method (VPIM) [39] was employed to analyze the local atomic structure and FFLS of the system.

Results and Discussion
In order to study the deposition process and structural evolution of Zr-Cu alloy, MD simulations of the growth of Zr x Cu 100−x alloy and x is percentage of the atomic concentrations. We investigate x = 50, 70, 90, which is response to the sub-eutectic, near-eutectic, and hypoeutectic composition, respectively. The growth of the alloys was performed in three different energies, as listed in Table 1, with initial conditions similar to the experimental operating ones.

Zr 50 Cu 50
The Zr 50 Cu 50 deposition processes with different conditions on the Si substrate are shown in Figure 2. The results show that as the mean kinetic energies increases, the morphology goes through three different stages: cluster-film-film. For Zr 50 Cu 50 the Zr and Cu atoms grow as a cluster with height 7.0 nm under low energy. Under middle and high energies, the deposited atoms grow as clusters before 5 ns. Finally, two 5.5 nm thickness films were formed at middle and high energies. The mean kinetic energy of the deposited atoms is relatively lower at low energy (E Zr = 0.13 eV, E Cu = 0.34 eV), which is less than the activation energy for atomic diffusion. Therefore, the atoms are easier to aggregate with each other to form clusters. the surface. This rate of 0.5 atom/ps is carefully selected so that there is enough time for thermal relaxation to take place with Berendsen thermostat. For post-simulation analysis, we use the radial distribution functions (RDF) to determine the correlation between the atoms. The simulated X-ray diffraction (XRD) was employed to study the phase of Zr-Cu film and the Voronoi polyhedron index method (VPIM) [39] was employed to analyze the local atomic structure and FFLS of the system.

Results and Discussion
In order to study the deposition process and structural evolution of Zr-Cu alloy, MD simulations of the growth of ZrxCu100−x alloy and x is percentage of the atomic concentrations. We investigate x = 50, 70, 90, which is response to the sub-eutectic, near-eutectic, and hypoeutectic composition, respectively. The growth of the alloys was performed in three different energies, as listed in Table 1, with initial conditions similar to the experimental operating ones.

Zr50Cu50
The Zr50Cu50 deposition processes with different conditions on the Si substrate are shown in Figure 2. The results show that as the mean kinetic energies increases, the morphology goes through three different stages: cluster-film-film. For Zr50Cu50 the Zr and Cu atoms grow as a cluster with height 7.0 nm under low energy. Under middle and high energies, the deposited atoms grow as clusters before 5 ns. Finally, two 5.5 nm thickness films were formed at middle and high energies. The mean kinetic energy of the deposited atoms is relatively lower at low energy (EZr = 0.13 eV, ECu = 0.34 eV), which is less than the activation energy for atomic diffusion. Therefore, the atoms are easier to aggregate with each other to form clusters. The total RDFs for Zr50Cu50 under three energies are presented in Figure 3a. For MGs, the RDF peak positions are the inherited characteristic constant sequence [40] and the peaks show the nearest neighbor distance of atoms, which can be compared with the experimental values of the two elementals (Supplementary Table S1). The RDF shows an asymmetrical first peak and a second peak with no spilt in three energies, implying that there exists a strong short-range order. While the shoulder peak on the right side of the first peak in deposited Zr50Cu50 films are not observed as The total RDFs for Zr 50 Cu 50 under three energies are presented in Figure 3a. For MGs, the RDF peak positions are the inherited characteristic constant sequence [40] and the peaks show the nearest neighbor distance of atoms, which can be compared with the experimental values of the two elementals (Supplementary Table S1). The RDF shows an asymmetrical first peak and a second peak with no spilt in three energies, implying that there exists a strong short-range order. While the shoulder peak on the right side of the first peak in deposited Zr 50 Cu 50 films are not observed as literature [41]. The three energies in our simulations show that the RDF distributions are similar, and no effect of the mean kinetic energy is observed for Zr 50 Cu 50 . literature [41]. The three energies in our simulations show that the RDF distributions are similar, and no effect of the mean kinetic energy is observed for Zr50Cu50. The XRD patterns for Zr50Cu50 are illustrated in Figure 3b as a function of the mean kinetic energy. All of these patterns show broad diffraction peaks without any detectable sharp Bragg peaks, implying a low order structure. Zr50Cu50 films show an amorphous XRD pattern which is consistent with the calculated RDF in Figure 3a. As the mean kinetic energies increases, the value of the first peak in the XRD pattern gradually decreases. The diffraction angles of the first XRD peaks for Zr50Cu50 under three energies are between 38 and 39 degrees.

Zr70Cu30
Deposition processes of Zr70Cu30 on Si (1 0 0) at 2 ns, 5 ns and 10 ns under three different energies are presented in Figure 4. Initially, small nanoclusters formed on the Si (1 0 0) surface and formed thin films gradually as the number of incoming atoms increases. One can see from Figure 4 that as increasing the mean kinetic energies, the morphology goes through three stages: cluster-cluster-film at 5 ns. At high energy, the thin film was formed before the other two energies. Finally, Zr70Cu30 films have been formed with thickness ranging from 6.0 to 6.4 nm. The XRD patterns for Zr 50 Cu 50 are illustrated in Figure 3b as a function of the mean kinetic energy. All of these patterns show broad diffraction peaks without any detectable sharp Bragg peaks, implying a low order structure. Zr 50 Cu 50 films show an amorphous XRD pattern which is consistent with the calculated RDF in Figure 3a. As the mean kinetic energies increases, the value of the first peak in the XRD pattern gradually decreases. The diffraction angles of the first XRD peaks for Zr 50 Cu 50 under three energies are between 38 and 39 degrees.

Zr 70 Cu 30
Deposition processes of Zr 70 Cu 30 on Si (1 0 0) at 2 ns, 5 ns and 10 ns under three different energies are presented in Figure 4. Initially, small nanoclusters formed on the Si (1 0 0) surface and formed thin films gradually as the number of incoming atoms increases. One can see from Figure 4 that as increasing the mean kinetic energies, the morphology goes through three stages: cluster-cluster-film at 5 ns. At high energy, the thin film was formed before the other two energies. Finally, Zr 70 Cu 30 films have been formed with thickness ranging from 6.0 to 6.4 nm. literature [41]. The three energies in our simulations show that the RDF distributions are similar, and no effect of the mean kinetic energy is observed for Zr50Cu50. The XRD patterns for Zr50Cu50 are illustrated in Figure 3b as a function of the mean kinetic energy. All of these patterns show broad diffraction peaks without any detectable sharp Bragg peaks, implying a low order structure. Zr50Cu50 films show an amorphous XRD pattern which is consistent with the calculated RDF in Figure 3a. As the mean kinetic energies increases, the value of the first peak in the XRD pattern gradually decreases. The diffraction angles of the first XRD peaks for Zr50Cu50 under three energies are between 38 and 39 degrees.

Zr70Cu30
Deposition processes of Zr70Cu30 on Si (1 0 0) at 2 ns, 5 ns and 10 ns under three different energies are presented in Figure 4. Initially, small nanoclusters formed on the Si (1 0 0) surface and formed thin films gradually as the number of incoming atoms increases. One can see from Figure 4 that as increasing the mean kinetic energies, the morphology goes through three stages: cluster-cluster-film at 5 ns. At high energy, the thin film was formed before the other two energies. Finally, Zr70Cu30 films have been formed with thickness ranging from 6.0 to 6.4 nm. The total RDF distributions of Zr 70 Cu 30 deposited films are displayed in Figure 5 for three deposition energies. The first and second peaks broaden with no splits which shows a typical amorphous characteristic. The amorphous nature of Zr 70 Cu 30 films under three energies is clearly shown in the snapshots presented in Figure 4, which means the near-eutectic alloy Zr 70 Cu 30 has a strong GFA.  The total RDF distributions of Zr70Cu30 deposited films are displayed in Figure 5 for three deposition energies. The first and second peaks broaden with no splits which shows a typical amorphous characteristic. The amorphous nature of Zr70Cu30 films under three energies is clearly shown in the snapshots presented in Figure 4, which means the near-eutectic alloy Zr70Cu30 has a strong GFA. The simulated XRD intensities versus diffraction angle 2θ for Zr70Cu30 are illustrated in Figure 5b. Broad diffraction peaks with no split show an amorphous Zr70Cu30 structure. As the mean kinetic energies increases, the first peak decreases successively and the corresponding diffraction angle 2θ decreases. The main peak diffraction angles for Zr70Cu30 under three energies are around 37 degrees.

Zr90Cu10
The deposition processes of the Zr90Cu10 are displayed in Figure 6. Zr90Cu10 is deposited as a cluster at 5 ns under low energy, and a film is formed with thickness about 7.1 nm at the end. Under middle and high energies, the deposited atoms grow as films with height around 7.0 nm. The simulated XRD intensities versus diffraction angle 2θ for Zr 70 Cu 30 are illustrated in Figure 5b. Broad diffraction peaks with no split show an amorphous Zr 70 Cu 30 structure. As the mean kinetic energies increases, the first peak decreases successively and the corresponding diffraction angle 2θ decreases. The main peak diffraction angles for Zr 70 Cu 30 under three energies are around 37 degrees.

Zr 90 Cu 10
The deposition processes of the Zr 90 Cu 10 are displayed in Figure 6. Zr 90 Cu 10 is deposited as a cluster at 5 ns under low energy, and a film is formed with thickness about 7.1 nm at the end. Under middle and high energies, the deposited atoms grow as films with height around 7.0 nm.
The RDF of Zr 90 Cu 10 under three energies was calculated and is plotted in Figure 7a. For Zr 90 Cu 10 deposition RDF, four distinct main peaks with a split of second peak arise, showing the well-crystallized structure. The crystallinity under the low energy was better than that of the other two energies according to the peak size and the split degree of second peak, that means the Zr 90 Cu 10 deposited film shows a better crystal structure under the low energy.
The relationship between XRD intensity and diffraction angle 2θ of Zr 90 Cu 10 alloy films is presented in Figure 7b for different deposition energies. The XRD patterns exhibit three main peaks at all three cases, and the second peak at the middle energy shows a split. The position of the main XRD peak for Zr 90 Cu 10 under three energies are between 35 and 36 degrees. Similar to the RDF pattern, the XRD pattern also shows that the three Zr 90 Cu 10 alloy films have a perfect crystal structure. Among them, the peak of XRD pattern under low energy is bigger, but overall, the difference of XRD pattern under three conditions is not obvious. The RDF of Zr90Cu10 under three energies was calculated and is plotted in Figure 7a. For Zr90Cu10 deposition RDF, four distinct main peaks with a split of second peak arise, showing the wellcrystallized structure. The crystallinity under the low energy was better than that of the other two energies according to the peak size and the split degree of second peak, that means the Zr90Cu10 deposited film shows a better crystal structure under the low energy.
The relationship between XRD intensity and diffraction angle 2θ of Zr90Cu10 alloy films is presented in Figure 7b for different deposition energies. The XRD patterns exhibit three main peaks at all three cases, and the second peak at the middle energy shows a split. The position of the main XRD peak for Zr90Cu10 under three energies are between 35 and 36 degrees. Similar to the RDF pattern, the XRD pattern also shows that the three Zr90Cu10 alloy films have a perfect crystal structure. Among them, the peak of XRD pattern under low energy is bigger, but overall, the difference of XRD pattern under three conditions is not obvious.   The RDF of Zr90Cu10 under three energies was calculated and is plotted in Figure 7a. For Zr90Cu10 deposition RDF, four distinct main peaks with a split of second peak arise, showing the wellcrystallized structure. The crystallinity under the low energy was better than that of the other two energies according to the peak size and the split degree of second peak, that means the Zr90Cu10 deposited film shows a better crystal structure under the low energy.
The relationship between XRD intensity and diffraction angle 2θ of Zr90Cu10 alloy films is presented in Figure 7b for different deposition energies. The XRD patterns exhibit three main peaks at all three cases, and the second peak at the middle energy shows a split. The position of the main XRD peak for Zr90Cu10 under three energies are between 35 and 36 degrees. Similar to the RDF pattern, the XRD pattern also shows that the three Zr90Cu10 alloy films have a perfect crystal structure. Among them, the peak of XRD pattern under low energy is bigger, but overall, the difference of XRD pattern under three conditions is not obvious.  For deeper understanding of the film growth mechanism, snapshots of Zr 90 Cu 10 deposited films under different energies are shown in Figure 8. The deposited film is a polycrystalline structure with both A and B crystal orientations under low energy (Figure 8a). At the beginning of the deposition process, small clusters are formed with a random orientation distribution. With time goes on the film growth rates is almost the same, but the growth directions are different under different kinetic energies. This is because that the mean kinetic energy of the deposited atoms is less than the activation energy used for diffusion under low energy, thus forming two cluster structures with different growth directions. Nanostructures with different lattice orientations were formed as the deposition proceeds, that shows that the alloy thin film can be controlled toward a predetermined single crystal or poly-crystal direction by changing the energy condition of Zr 90 Cu 10 . The deposited film has a single crystal structure under middle energy (Figure 8b). While Figure 8c shows a crystalline structure on the bottom and an amorphous structure on the top. The thickness of the amorphous structure is about 1.5 nm.
or poly-crystal direction by changing the energy condition of Zr90Cu10. The deposited film has a single crystal structure under middle energy (Figure 8b). While Figure 8c shows a crystalline structure on the bottom and an amorphous structure on the top. The thickness of the amorphous structure is about 1.5 nm. Based on these simulations, the thickness of films increases when increasing Zr concentrations. This caused by the difference of atomic radius, since the radius of Zr atoms is larger than that of Cu atoms. Comparing the RDF and XRD patterns of ZrxCu100−x (x = 50, 70, 90) deposited films under three energies, it is easy to find that the impact of the elemental composition is bigger than deposition conditions (mean kinetic energy) for GFA of Zr-Cu system. Meanwhile, Zr50Cu50 and Zr70Cu30 have stronger ability to form metallic glass, while Zr90Cu10 tends to form crystalline structure and high energy is good for forming thin films.
In the relationship diagram of the Ω and δ (Figure 9), Zr-Cu alloys are located in the area for bulk metallic glass (B). Meanwhile, the Ω and δ values of Zr50Cu50 and Zr70Cu30 correspond to the central of BMGs, indicating a strong amorphous forming ability. As increasing the Zr concentration, alloys trend to enter the area of BMGs to solid-solutions. This is consistent with the results of our simulations, that the system prefers to form crystalline structure with high concentration of Zr. For Zr90Cu10, the thin film structure of polycrystalline, single crystal and crystalline-amorphous Based on these simulations, the thickness of films increases when increasing Zr concentrations. This caused by the difference of atomic radius, since the radius of Zr atoms is larger than that of Cu atoms. Comparing the RDF and XRD patterns of Zr x Cu 100−x (x = 50, 70, 90) deposited films under three energies, it is easy to find that the impact of the elemental composition is bigger than deposition conditions (mean kinetic energy) for GFA of Zr-Cu system. Meanwhile, Zr 50 Cu 50 and Zr 70 Cu 30 have stronger ability to form metallic glass, while Zr 90 Cu 10 tends to form crystalline structure and high energy is good for forming thin films.
In the relationship diagram of the Ω and δ (Figure 9), Zr-Cu alloys are located in the area for bulk metallic glass (B). Meanwhile, the Ω and δ values of Zr 50 Cu 50 and Zr 70 Cu 30 correspond to the central of BMGs, indicating a strong amorphous forming ability. As increasing the Zr concentration, alloys trend to enter the area of BMGs to solid-solutions. This is consistent with the results of our simulations, that the system prefers to form crystalline structure with high concentration of Zr. For Zr 90 Cu 10 , the thin film structure of polycrystalline, single crystal and crystalline-amorphous composite layer were obtained under low, middle and high energy conditions, which shows that the alloy with high Zr concentration is less likely to form metallic glass under the deposition. Zr 90 Cu 10 located at the edge of the B area, that is closer to the area of solid-solutions (S) than the others. That theoretically explains our simulation results well.
Materials 2018, 11, x FOR PEER REVIEW 9 of 13 composite layer were obtained under low, middle and high energy conditions, which shows that the alloy with high Zr concentration is less likely to form metallic glass under the deposition. Zr90Cu10 located at the edge of the B area, that is closer to the area of solid-solutions (S) than the others. That theoretically explains our simulation results well. . Ω is the ratio of the average melting temperature timing the mixing entropy and the mixing enthalpy, and δ is the mean square deviation of the atomic size (reproduced from reference [29]).
Three groups of MD simulations revealed the formation of metallic glass microscopically, and the relationship between GFA and parameters (Ω and δ) values. The simulation results which confirmed the correctness of the theory and demonstrated the reliability of MD simulation, is of great significance to predict the thin film structure using magnetron sputtering deposition in MD simulation.

Five-Fold Local Symmetry Analysis.
To gain insights on the GFA at atomic level, we do more atomic structural analysis. The amorphous structure of Zr-Cu alloys thin film was characterized by FFLS, which is one of the most popular methods to describe the nanostructure characteristics of MGs [43,44]. Different atomic clusters may contain different degree of FFLS, which can be expressed by The typical Voronoi polyhedrons (VPs) with three components are calculated and shown in Table S2 and Figure S1 in the supplementary information. According to the above-mentioned formula about FFLS, we calculate the average degree of FFLS. Solid line in Figure 10 shows the component dependence of the average degree of FFLS, which is denoted as W) under different energies obtained in MD simulations. With increasing Zr concentrations, W decreases gradually. The W of the first four i P Figure 9. The prediction of phase formation for alloys. The region (S) shows that the alloy can form a solid solution. The alloy in the region (I) mainly consists of intermetallic compounds. The region (S+I) indicate that the solid solution and the intermetallic compound coexist, and alloy forms an amorphous phase metallic glass at the region (B) for bulk metal glasses (BMGs). Ω is the ratio of the average melting temperature timing the mixing entropy and the mixing enthalpy, and δ is the mean square deviation of the atomic size (reproduced from reference [29]).
Three groups of MD simulations revealed the formation of metallic glass microscopically, and the relationship between GFA and parameters (Ω and δ) values. The simulation results which confirmed the correctness of the theory and demonstrated the reliability of MD simulation, is of great significance to predict the thin film structure using magnetron sputtering deposition in MD simulation.

Five-Fold Local Symmetry Analysis.
To gain insights on the GFA at atomic level, we do more atomic structural analysis. The amorphous structure of Zr-Cu alloys thin film was characterized by FFLS, which is one of the most popular methods to describe the nanostructure characteristics of MGs [43,44]. Different atomic clusters may contain different degree of FFLS, which can be expressed by f k  Table S2 and Figure S1 in the supplementary information. According to the above-mentioned formula about FFLS, we calculate the average degree of FFLS. Solid line in Figure 10 shows the component dependence of the average degree of FFLS, which is denoted as W) under different energies obtained in MD simulations. With increasing Zr concentrations, W decreases gradually. The W of the first four components (x = 50, 60, 70 and 80) are at a relatively high level, especially for Zr 50 Cu 50 . Zr 50 Cu 50 has the strongest GFA and the largest W value, and exhibits perfect amorphous structure in our simulations. components (x = 50, 60, 70 and 80) are at a relatively high level, especially for Zr50Cu50. Zr50Cu50 has the strongest GFA and the largest W value, and exhibits perfect amorphous structure in our simulations. As shown in Figure 10, the increase of FFLS with decreasing the mean kinetic energy of incoming atoms indicates that FFLS shows energy dependence expect for Zr90Cu10. The largest W value occurs at the lowest energy shows that reducing the mean kinetic energy of incoming atoms will increase the five-fold local symmetry of the amorphous alloy. The GFA of Zr90Cu10 is the smallest, and the decrease of W is the most obvious. The difference between Zr90Cu10 and ZrxCu100−x (x = 50, 60, 70 and 80) is that the maximum W of Zr90Cu10 appears in high energy conditions. This may be because of the well crystal characteristic exhibited in Zr90Cu10, and the disorder layer of 1.5 nm appearing on the surface of the film at high energy. At the same time, it shows that the FFLS is not suitable for the description of the long-range order structure.
The component dependence of the formation enthalpy of Zr-Cu amorphous at 298 K is shown in the dash line ( Figure 10) [28]. It can be seen that the amorphous formation enthalpy of the Zr-Cu alloy is negative, and the enthalpy of the amorphous formation increases with the increase of Zr concentration when the Zr concentration is greater than 50%. As the value of x increases, its GFA decreases, which is consistent with our simulation result.
Nearly eutectic Zr70Cu30 films deposited at three energy conditions have amorphous structures shown in our simulations. Combined with its FFLS and Voronoi polyhedral index, it is easy to find that the deposited film has better five-fold local symmetry and the highest content of icosahedronlike clusters under low energy, most of which are Cu-centered. At the same time, using the average degree of FFLS, it is understandable that under low energy the deposited atoms formed film firstly at 5 ns (Figure 4), because it has largest degree of FFLS. On the other hand, as shown in Figure 9, the GFA of five alloys decrease with increasing x, which is coincident with the variation of the five-fold local symmetry that decrease with the x. Our results demonstrate that there is a positive correlation between GFA and the degree of FFLS of Zr-Cu systems.

Conclusions
We have investigated the correlation between the glass forming ability and the degree of the FFLS in the deposition process of ZrxCu100−x (x = 50, 70, 90) alloy thin films via molecular dynamics simulations. The structural state of alloy films was analyzed by RDF, XRD, Voronoi polyhedron index As shown in Figure 10, the increase of FFLS with decreasing the mean kinetic energy of incoming atoms indicates that FFLS shows energy dependence expect for Zr 90 Cu 10 . The largest W value occurs at the lowest energy shows that reducing the mean kinetic energy of incoming atoms will increase the five-fold local symmetry of the amorphous alloy. The GFA of Zr 90 Cu 10 is the smallest, and the decrease of W is the most obvious. The difference between Zr 90 Cu 10 and Zr x Cu 100−x (x = 50, 60, 70 and 80) is that the maximum W of Zr 90 Cu 10 appears in high energy conditions. This may be because of the well crystal characteristic exhibited in Zr 90 Cu 10 , and the disorder layer of 1.5 nm appearing on the surface of the film at high energy. At the same time, it shows that the FFLS is not suitable for the description of the long-range order structure.
The component dependence of the formation enthalpy of Zr-Cu amorphous at 298 K is shown in the dash line ( Figure 10) [28]. It can be seen that the amorphous formation enthalpy of the Zr-Cu alloy is negative, and the enthalpy of the amorphous formation increases with the increase of Zr concentration when the Zr concentration is greater than 50%. As the value of x increases, its GFA decreases, which is consistent with our simulation result.
Nearly eutectic Zr 70 Cu 30 films deposited at three energy conditions have amorphous structures shown in our simulations. Combined with its FFLS and Voronoi polyhedral index, it is easy to find that the deposited film has better five-fold local symmetry and the highest content of icosahedron-like clusters under low energy, most of which are Cu-centered. At the same time, using the average degree of FFLS, it is understandable that under low energy the deposited atoms formed film firstly at 5 ns (Figure 4), because it has largest degree of FFLS. On the other hand, as shown in Figure 9, the GFA of five alloys decrease with increasing x, which is coincident with the variation of the five-fold local symmetry that decrease with the x. Our results demonstrate that there is a positive correlation between GFA and the degree of FFLS of Zr-Cu systems.

Conclusions
We have investigated the correlation between the glass forming ability and the degree of the FFLS in the deposition process of Zr x Cu 100−x (x = 50, 70, 90) alloy thin films via molecular dynamics simulations. The structural state of alloy films was analyzed by RDF, XRD, Voronoi polyhedron index and FFLS. We find that Zr 90 Cu 10 films all formed crystals under different energy deposition conditions. There are polycrystalline structure with two crystal orientations under low energy, single crystal under middle energy, and a crystalline-amorphous composite structure under high energy. The formation of crystalline-amorphous structure in Zr 90 Cu 10 indicates that alloys located at the edge of area B can be prepared into a crystalline-amorphous composite by magnetron sputtering under certain conditions.
The Voronoi polyhedron analysis shows that the greater the number of icosahedral clusters and their deformed structures, the greater the GFA. The Voronoi polyhedron has an important influence on the GFA of Zr-Cu system, and also reveals the relationship between composition and GFA. There is a positive correlation between the degrees of FFLS and GFA. Our results imply that the degrees of FFLS could serve as a measure of the GFA, providing new approach to characterize the GFA for the design of amorphous composite materials. More importantly, our results manifest that sputtering conditions greatly affect the local atomic structures and consequential properties. This result could be extended to other glassy materials as the universality of FFLS for short-range order structure in amorphous. Our study might be useful in optimizing sputtering conditions in real experiments, as well as promising implications in material design of advanced glassy materials.

Conflicts of Interest:
The authors declare no conflict of interest.