Machine Learning-Assisted Nano-imaging and Spectroscopy of Phase Coexistence in a Wide-Bandgap Semiconductor

TL;DR

This paper introduces NISMA, a machine learning-assisted nano-spectroscopy technique that maps and analyzes phase coexistence in SrSnO3, revealing nanoscale structural and electronic inhomogeneities relevant for optoelectronic applications.

Contribution

It develops a novel nano-imaging and spectroscopy method with machine learning to analyze phase coexistence in complex oxides at the nanoscale.

Findings

Nanoscale phase coexistence in SrSnO3 was characterized.

Machine learning enabled robust quantitative analysis of optical responses.

The method can be applied to other complex oxide materials.

Abstract

Wide bandgap semiconductors with high room temperature mobilities are promising materials for high-power electronics. Stannate films provide wide bandgaps and optical transparency, although electron-phonon scattering can limit mobilities. In SrSnO3, epitaxial strain engineering stabilizes a high-mobility tetragonal phase at room temperature, resulting in a threefold increase in electron mobility among doped films. However, strain relaxation in thicker films leads to nanotextured coexistence of tetragonal and orthorhombic phases with unclear implications for optoelectronic performance. The observed nanoscale phase coexistence demands nano-spectroscopy to supply spatial resolution beyond conventional, diffraction-limited microscopy. With nano-infrared spectroscopy, we provide a comprehensive analysis of phase coexistence in SrSnO3 over a broad energy range, distinguishing inhomogeneous phonon and plasma responses arising from structural and electronic domains. We establish Nanoscale Imaging and Spectroscopy with Machine-learning Assistance (NISMA) to map nanotextured phases and quantify their distinct optical responses through a robust quantitative analysis, which can be applied to a broad array of complex oxide materials.