Predicting Fracture Propensity in Amorphous Alumina from Its Static Structure Using Machine Learning

Exceptional Microscale Plasticity in Amorphous Aluminum Oxide at Room Temperature

Oxide glasses are an elementary group of materials in modern society, but brittleness limits their wider usability at room temperature. As an exception to the rule, amorphous aluminum oxide (a-Al2 O3 ) is a rare diatomic glassy material exhibiting significant nanoscale plasticity at room temperature. Here, it is shown experimentally that the room temperature plasticity of a-Al2 O3 extends to the microscale and high strain rates using in situ micropillar compression. All tested a-Al2 O3 micropillars deform without fracture at up to 50% strain via a combined mechanism of viscous creep and shear band slip propagation. Large-scale molecular dynamics simulations align with the main experimental observations and verify the plasticity mechanism at the atomic scale. The experimental strain rates reach magnitudes typical for impact loading scenarios, such as hammer forging, with strain rates up to the order of 1 000 s-1 , and the total a-Al2 O3 sample volume exhibiting significant low-temperature plasticity without fracture is expanded by 5 orders of magnitude from previous observations. The discovery is consistent with the theoretical prediction that the plasticity observed in a-Al2 O3 can extend to macroscopic bulk scale and suggests that amorphous oxides show significant potential to be used as light, high-strength, and damage-tolerant engineering materials.

Predicting the failure of two-dimensional silica glasses

Being able to predict the failure of materials based on structural information is a fundamental issue with enormous practical and industrial relevance for the monitoring of devices and components. Thanks to recent advances in deep learning, accurate failure predictions are becoming possible even for strongly disordered solids, but the sheer number of parameters used in the process renders a physical interpretation of the results impossible. Here we address this issue and use machine learning methods to predict the failure of simulated two dimensional silica glasses from their initial undeformed structure. We then exploit Gradient-weighted Class Activation Mapping (Grad-CAM) to build attention maps associated with the predictions, and we demonstrate that these maps are amenable to physical interpretation in terms of topological defects and local potential energies. We show that our predictions can be transferred to samples with different shape or size than those used in training, as well as to experimental images. Our strategy illustrates how artificial neural networks trained with numerical simulation results can provide interpretable predictions of the behavior of experimentally measured structures.

The sheer number of parameters in deep learning makes the physical interpretation of failure predictions in glasses challenging. Here the authors use Grad-CAM to reveal the role of topological defects and local potential energies in failure predictions.

Advancing the Mechanical Performance of Glasses: Perspectives and Challenges

Glasses are materials that lack a crystalline microstructure and long-range atomic order. Instead, they feature heterogeneity and disorder on superstructural scales, which have profound consequences for their elastic response, material strength, fracture toughness, and the characteristics of dynamic fracture. These structure-property relations present a rich field of study in fundamental glass physics and are also becoming increasingly important in the design of modern materials with improved mechanical performance. A first step in this direction involves glass-like materials that retain optical transparency and the haptics of classical glass products, while overcoming the limitations of brittleness. Among these, novel types of oxide glasses, hybrid glasses, phase-separated glasses, and bioinspired glass-polymer composites hold significant promise. Such materials are designed from the bottom-up, building on structure-property relations, modeling of stresses and strains at relevant length scales, and machine learning predictions. Their fabrication requires a more scientifically driven approach to materials design and processing, building on the physics of structural disorder and its consequences for structural rearrangements, defect initiation, and dynamic fracture in response to mechanical load. In this article, a perspective is provided on this highly interdisciplinary field of research in terms of its most recent challenges and opportunities.