Page 64 - Read Online

Page 4 of 15 Hu et al. J Mater Inf 2023;3:1 I http://dx.doi.org/10.20517/jmi.2022.28

0.0

)
A 0.2
A
+
B
B
0.4
(
/
)
A
A
0.6
B
B
(
0.8

1 2 3 4 5
2 AB /( BB + AA )

Figure 1. Sampling library of the energetic parameters for the pairwise Lennard-Jones binary systems.

fraction at each . 10 independent simulations are generally performed for better statistics. Note that crystal-
lizing a good glass former can take a very long time and even out the capability of the computational power.
The basic units for energy, length, and mass are AA, AA, and , respectively. The pressure, temperature, and
√
time scale are reported in reduced units of / , / , and 2 / , where B is the Boltzmann
3

√
constant. The derived units for c, , and Δ imix are 3 / 2 , / , and AA / B, respectively. More
2
3
AA B AA
details about the technical details are available in our previous works [13,14,16] .
In addition to the above grid search of binary systems, we also performed extensive simulations to simulate
[9]
many binary systems inspired by experiments . The detailed information of these binary systems is provided
in Table 1. Based on the experimental values of the elemental features, including particle size, cohesive energy
and mass, we map them to the reduced units and ran the simulations. During the mapping, we also keep
√
≥ . To capture the inter-species interactions, we follow the classical London’s rule that =
[13]
and set = ( + )/2 . In this way, we can include more realistic models and explore a larger
parameter space. To better display the data, we plot the data in Figure 2 in two dimensions with respect to
( − )/( + ) and / . Obviously, the parameter ranges are quite broad for binary MGs and
those with different GFA can overlap in the two-dimensional space. This indicates that the GFA problem is not
single-parameter deterministic. The energetic parameters and geometrical one may couple in a higher order.
We should emphasize that this set of simulations is not aiming to compare directly to experiments to model
each specific system. Instead, we hope to explore a larger space with some sort of connection to experiments.
Furthermore, an alloy with an element having both larger cohesive energy and particle size than the other has
a higher probability of becoming glass. From Table 1, there are 40 samples out of the total (62) that fall into
this group. If we include the mass comparison, this number decreases to 31 out of 62. This demonstrates
the neutral effect of particle masses. Meanwhile, there is also a higher chance for glass formation when both
elements are metals (39/62). These insights are helpful for future experimental glass design, but are still subject
to the small number of binary MGs being developed.