A transferable machine-learning framework linking interstice distribution and plastic heterogeneity in metallic glasses

Abstract

When metallic glasses (MGs) are subjected to mechanical loads, the plastic response of atoms is non-uniform. However, the extent and manner in which atomic environment signatures present in the undeformed structure determine this plastic heterogeneity remain elusive. Here, we demonstrate that novel site environment features that characterize interstice distributions around atoms combined with machine learning (ML) can reliably identify plastic sites in several Cu-Zr compositions. Using only quenched structural information as input, the ML-based plastic probability estimates (“quench-in softness” metric) can identify plastic sites that could activate at high strains, losing predictive power only upon the formation of shear bands. Moreover, we reveal that a quench-in softness model trained on a single composition and quench rate substantially improves upon previous models in generalizing to different compositions and completely different MG systems (Ni62Nb38, Al90Sm10 and Fe80P20). Our work presents a general, data-centric framework that could potentially be used to address the structural origin of any site-specific property in MGs.

Predicting the propensity for thermally activated β events in metallic glasses via interpretable machine learning

“…We make use of a set of interstice distribution descriptors to represent the local atomic environment 19 . The basic fingerprinting procedure is to extract groups of bonds, facets and tetrahedra from the coordination polyhedron of an atom, and then featurize the distribution of interstitial spaces present in these bond, facet, and tetrahedron groups.…”

“…Over the past several decades, many efforts have been devoted to addressing this critical question. Recently, the emerging machine learning (ML) technique, based on wellcrafted representations of the atomic environment, has been proven to be promising for establishing atomic-level structureproperty relationships in liquids and glasses [17][18][19][20][21][22][23][24] . For example, Schoenholz et al 17 studied L-J model liquids and utilized ML to derive a structural parameter called "softness", which was found to correlate well with the particle's propensity for hopping, reflecting its susceptibility to β relaxation of liquids 10 .…”

“…We, therefore, have to resort to local perturbation methods, to activate the local group of atoms into excited states by stress or thermal stimulus, as a probe into the susceptibility to elementary excitations. Several recent ML studies have focused on quantitatively gauging how the local environment influences the propensity for stress-activated β processes (i.e., shear transformations) in MGs [18][19][20] . For example, pioneering works of Cubuk et al 18 performed ML on disordered materials such as L-J glasses and granular systems and showed that radial and bond-angle distribution information can be used to identify atoms with a high propensity to shear transformation.…”

“…For example, pioneering works of Cubuk et al 18 performed ML on disordered materials such as L-J glasses and granular systems and showed that radial and bond-angle distribution information can be used to identify atoms with a high propensity to shear transformation. Wang et al 19 developed interstice distribution as a new local structural representation for MGs, which is proven to be robust in predicting plastic sites of several MGs and has advantages in generalizing between compositions even chemical systems. However, the accuracy achieved in these attempts is not yet sufficiently high, and the reported scoring metric, e.g., recall or area under the receiver operating characteristic curve (AUC-ROC), is typically below 80%.…”

Structural insights into metal-metalloid glasses from mass spectrometry

“…In view of their distinct physical and chemical properties, metallic glasses have gained popularity in several scientific disciplines. Advanced synthesis and processing methods have been employed to modify the local structure of these glassy systems by the controlled introduction of defects, such as interfaces and precipitates [4][5][6][7][8][9][10][11][12][13][14] . The absence of long range ordering in metallic glasses complicates the understanding of their local structure 3 .…”