SE156: Machine learning-assisted discovery and design of enhanced interfacial superconductivity

Monolayer FeSe on SrTiO3 (FeSe/STO) exhibits a superconducting transition temperature far exceeding that of bulk FeSe, making it a model system for studying interfacial superconductivity. Recent studies suggest that the interfacial structure at the FeSe/STO interface plays a critical role in determining electron-phonon coupling and superconducting properties. However, the enormous configurational space associated with interface reconstructions makes systematic first-principles exploration computationally prohibitive.In this work, we develop a machine learning-assisted framework to accelerate the discover and design FeSe/STO interfaces. A machine learning surrogate model trained on density functional theory (DFT) calculations is used to efficiently identify energetically favorable interface structures. These predicted structures are subsequently investigated using first-principles calculations to understand how different interface configurations influence electron-phonon coupling. By combining data-driven structural exploration with physics-based analysis, our goal is not only to identify promising interface structures but also to uncover the structural features that govern interfacial superconductivity. This work establishes an efficient computational strategy for the discovery and design of superconducting heterostructures and provides insights for engineering interfaces with potentially enhanced superconducting transition temperatures.