Page 62 - Read Online

Page 2 of 15 Hu et al. J Mater Inf 2023;3:1 I http://dx.doi.org/10.20517/jmi.2022.28

1. INTRODUCTION
[1]
Ever since the first discovery of an amorphous metal , also named metallic glass (MG) later, from Au-Si
system, developing new MGs with exceptional glass-forming ability (GFA), i.e., low critical cooling rate ,
has been one of the main goals in the field [2–5] . In turn, these new materials ensure the exploration of the
physics, chemistry, and mechanics of glasses in experiments [6–8] . In the past several decades, thousands of
new MGs with various GFAs have been synthesized successfully in labs all around the world. In addition, an
increasing amount of fascinating knowledge has been acquired. This greatly enriches the glass family.

Starting from the periodic table, the principal elements for MGs are transition metals, sometimes with metal-
loids as minor additions. Empirically, more components usually make a better glass former. There are usually
fourtofiveelementsinbulkMGs. Thismakestheglassformationproblemrathercomplextounderstand. First
[9]
of all, the parameter space is huge with enormous elemental features and alloy properties . These include but
are not limited to composition, atom size ratio, cohesive energy, pairwise and many-body interactions, and
their couplings. It is even impossible to sample the full space for a binary system by traditional methods. Sec-
ondly, with multiple components, there are many (metastable) phases involved during nucleation and growth
of the equilibrium crystalline product [10] . These metastable phases can have very complex crystal structures
and are hard to be captured by experimental observations. Thirdly, supercooled metallic liquids overall show a
disordered state, but there are abundant types of local structures formed [11] . They are favored by either energy
or entropy. Understanding the roles of these locally favored structures in glass transition and crystallization
of supercooled melts has been becoming a hot topic [12] . Due to these complexities, we are still far from well
understanding the crucial factors that govern MG formation.

In recent years, it is quite encouraging that advanced high throughput sputtering experimental technique has
shown its capability of synthesizing a library of ∼ 1, 000 compositions at the same time [4,5] . By sputtering from
multiple targets, a thin film of a system with continuous gradient composition is generated. These libraries are
likelyagoodstartingpointforminingtheGFAdatabybigdatamethods. However, becauseofthelargespecies
andcompositionalspace,theexperimentaldatasetscanberathersparse. Theintercorrelationbetweendifferent
libraries is obscure to understand the data. This will make either building the physical model or predicting
new materials challenging. The sophisticated design of the datasets (hence the experiments) is important for
material prediction.

To create a large dataset of GFA with continuous controlled parameters change, we have carried out very
large-scale molecular dynamics simulations to study the glass formation and crystallization process of binary
alloys in recent years [13–17] . On the one hand, by carefully analyzing the crystallization kinetics of supercooled
metallic liquids, the thermodynamic factor, interfacial energy, has been identified as the key to controlling the
crystallization rate and thus the GFA [15] . At the microscopic scale, the competing ordering effect is crucial
to determine the interfacial energy. In principle, the stronger the crystal-like preorder is frustrated by some
locally favored structures with incompatible symmetries (such as icosahedra), the higher the interfacial energy
will be [15,17] . Hence, the better GFA can be expected. The topological and especially the chemical properties of
these local structures are very crucial in determining the interfacial energy. Furthermore, by tuning the local
structures so as to decrease the wettability of the preorder at the liquid-crystal interface, the crystallization
speed can be manipulated over several orders of magnitude [17] . It is more interesting to find that the preorder
is very crucial not only in crystal nucleation but also in the crystal growth process. Accordingly, a critical
modification has been proposed to the classical nucleation theory. On the other hand, by characterizing the
(i.e., GFA) of binary alloys in a large parameter space, we have studied how different elemental features and
[13,14,16]
alloy properties affect . These include the atomic size mismatch, cohesive energy, mixing energy,
and ”atomic symmetry”. It is surprisingly found that local chemical ordering plays a deterministic role in .
In most of the previous studies, such as Cheng et al. [18] and Laws et al. [19] , the metallic glasses are usually
treated as hard-sphere-like systems where dense packing is critical. The local atomic packings, especially local