Materials Discovery Cloud: Decoding Defect Mediated Mechanisms Governing Stability and Function

A physics-informed AI framework that fuses multimodal experiments and multiscale simulations to predict how defects form, evolve and impact electrical and thermal performance in next-generation microelectronics.

Microelectronics performance and reliability are increasingly limited by defects in thin films and interfaces — vacancies, dislocations, voids and chemical disorder — that evolve during deposition and under real operating conditions. As devices move toward heterogeneous 3D integration, the number and complexity of interfaces grow, making it difficult for single-modality characterization or physics-only simulation to predict how defect networks drive leakage, switching instability, adhesion loss and ultimately device lifetime. This project aims to create the Materials Discovery Cloud: a physics-informed AI framework that learns the governing relationships between composition, geometry, bias and temperature and the resulting defect landscapes and functional fields such as electrostatic potential, current density and temperature.

The team will integrate AI-ready datasets from DOE user facilities and leadership computing, including advanced electron microscopy (e.g., STEM/EELS, 4D-STEM), coherent X-ray methods (e.g., X-ray photon correlation spectroscopy, Bragg coherent diffraction imaging, ptychographic laminography), spectroscopy, and device-level electrical and thermal measurements, complemented by multiscale simulations (DFT/MD/KMC/phase-field/FEM). The AI approach combines multimodal data fusion into a shared latent representation, physics-informed neural networks to enforce governing transport equations, active learning to guide new measurements and simulations, and causal discovery to identify mechanistic pathways linking defect evolution to electrical/thermal/mechanical degradation. The outcome will be predictive models and workflows that accelerate the design of robust, energy-efficient microelectronics by anticipating defect-driven failure before it occurs.

The Materials Discovery Cloud is part of DOE’s Genesis Mission, a bold national initiative to double America’s research and development productivity within a decade while strengthening U.S. technological leadership and global competitiveness.