Finding defects in glasses through machine learning

Abstract

Structural defects control the kinetic, thermodynamic and mechanical properties of glasses. For instance, rare quantum tunneling two-level systems (TLS) govern the physics of glasses at very low temperature. Due to their extremely low density, it is very hard to directly identify them in computer simulations. We introduce a machine learning approach to efficiently explore the potential energy landscape of glass models and identify desired classes of defects. We focus in particular on TLS and we design an algorithm that is able to rapidly predict the quantum splitting between any two amorphous configurations produced by classical simulations. This in turn allows us to shift the computational effort towards the collection and identification of a larger number of TLS, rather than the useless characterization of non-tunneling defects which are much more abundant. Finally, we interpret our machine learning model to understand how TLS are identified and characterized, thus giving direct physical insight into their microscopic nature.

Applications of amorphous inorganics as novel functional materials

“…3b) commonly raised by disorder structures. 66 There are abundant defects that occur in the skeleton of the disorderliness structure, originating during its amorphization. Owing to their vast quantities and activities, these defects ( e.g.…”

“…In amorphous materials, the quantity of oxygen vacancies is far more than that in crystals, performing distinctly enhanced functions, especially in catalysis. 65–70 Additionally, the design of amorphous electrodes with a combination of oxygen vacancy and disordered structure was capable of enhanced ion storage. In alkali metal ion batteries, storages of Na + and K + in electrodes were restricted by lattice owing to their relatively large atomic radius.…”

Vibrations and Heat Transfer in Glasses: The Role Played by Disorder

“…It is the case of double well potential effects, and two-level systems that seems to play an important role in the low-frequency / low temperature cases. They already have been reviewed in details in [37], and there are now new attemps to identify them in real systems at the microscopic level from numerical simulations [89][90][91][92] as well as from experimental devices [88].…”

Classifying the age of a glass based on structural properties: A machine learning approach

“…Based on this approach, several recent works [30][31][32][33][34][35][36][37][38][39][40][41] extended our conceptual understanding of glassy liquids by convincingly demonstrating that machine learning is able to accurately connect structural properties with the corresponding dynamics. In particular, standard machine learning tools like support vector machines have been able to compute the relaxation time through softness [42] and collective effects like fragility [36] and low-temperature defects [43]. More sophisticated models like graph neural networks [44] give accurate predictions of dynamic propensity, but similar results can be achieved by simpler models with accurate structural indicators [45].…”