DMREF: Collaborative Research: AI-Driven Design of Tough 2D Metal Halide Perovskites for Next-Generation Semiconductor Applications

Metal halide perovskites are burgeoning low-cost, high-performance semiconductor materials for various energy and electronic applications, but they are brittle and tend to crack, causing reliability issues that hinder the practical deployment of devices. 2D layered metal halide perovskites manifest permanent deformation, which can be harnessed to greatly improve the materials’ resistance to fracture without significantly sacrificing their semiconductor properties. However, the design of tough 2D metal halide perovskites faces major challenges arising from limited experimental and computational data on fracture behavior and nearly unlimited compositional choices. By integrating machine learning, multiscale modeling and simulations, as well as high-throughput material synthesis and mechanical experiments, this Designing Materials to Revolutionize and Engineer our Future (DMREF) project will establish an AI-driven design framework to effectively discover new 2D metal halide perovskites with enhanced fracture resistance, nearly two orders better than the best reported values to date. The project will resolve the bottleneck in mechanical reliability, removing a major roadblock to the long-term durability and commercial viability of semiconductor devices based on these materials. The project also includes training of students in interdisciplinary research comprising machine learning, semiconductor materials, and material design and synthesis, development of new curriculum modules for undergraduate and graduate education, recruitment and mentoring of undergraduate students through research opportunities, and on-campus K-12 and community outreach to promote STEM education and careers. The current search for tough 2D metal halide perovskites mainly follows an inefficient trial-and-error paradigm. Limited high-quality fracture data also impedes the development and training of fracture and AI models, respectively, imposing enormous challenges in navigating through the vast compositional space. The project aims to discover new 2D metal halide perovskites with high fracture energy by leveraging an AI-driven design framework that integrates multi-fidelity and multi-source active learning, multiscale fracture simulations, as well as high-throughput material synthesis and nanomechanical tests. The framework first leverages low-fidelity, high-throughput atomic force microscopy fracture data to train an ensemble of machine learning models that can find new metal halide perovskite candidates. The multi-source active learning strategy employing median-fidelity multiscale simulations and high-fidelity experimental fracture energy measurements will then form a continuous feedback loop among experiments, simulations, and computational design, and enable continuous update of the machine learning models. The fracture energy of top candidates will be experimentally validated and used to enhance machine learning models for greater efficiency and output. It is expected that this project will produce rules and tools for accelerated discovery of hybrid organic-inorganic 2D materials. 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.