Collaborative Research: DMREF: Machine-learning accelerated design of tough, hierarchically heterogeneous ceramic composites

Suny At Stony Brook, Stony Brook NY

Investigators

Abstract

The intrinsic high strength, light weight, and heat, corrosion, and irradiation resistance of ceramics positions them as nearly ideal structural materials. However, their inability to resist the growth of cracks causes any tiny flaw to grow into a catastrophically large crack; this renders ceramics brittle and impractical as structural materials for automotive, energy, aerospace, and defense applications, to name a few examples. The overarching goal of this Designing Materials to Revolutionize and Engineer our Future (DMREF) project is to transform the fracture resistance of ceramics by introducing heterogeneous metallic features across multiple length scales into the ceramic material. The fundamental challenges to be overcome in the project are: (1) how can heterogeneous ceramics be computationally designed when the space of possible designs is massive? and (2) how can such engineered ceramics be manufactured? These challenges will be overcome by leveraging recent advances in machine learning for material modeling in conjunction with advanced low-temperature ceramic processing techniques. The revolutionary new class of materials designed through the project can be directly implemented in commercial applications, such as satellite structures, low-wear medical devices, armor, and hypersonic vehicles. Insights gained on the design, processing, and fracture of heterogeneous ceramics will drive future innovations enabling next-generation structural materials. Despite many decades of development of advanced ceramics, fracture toughness values have remained consistently below about 15 MPa-m0.5, a factor of three less than typical structural metals. This DMREF project aims to transform the toughness of ceramics through the introduction of hierarchically heterogeneous metallic interphases that will drive toughening via crack multiplication and deflection ahead of an advancing crack tip. The structure of the interphases and their distribution in the microstructure will be designed using molecular dynamics and finite element simulations of fracture. To accelerate computational materials design, finite-element-based physics-informed neural networks will be employed after training on cohesive surface and phase field finite element fracture models. The computationally designed ceramic microstructures will be produced using non-conformal coating of ceramic powders in conjunction with low-temperature sintering enabled by demonstrated doping technologies to avoid melting of the metallic interphases. R-curves will be used to quantify fracture response via macro and micro-scale mechanical testing. After establishing all necessary modeling, processing, and mechanical testing capabilities, a closed-loop design cycle will be executed utilizing the machine-learning accelerated model for microstructure design. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.