Probing oxygen vacancy distribution in oxide heterostructures by deep Learning-based spectral analysis of current noise

Abstract

Exploiting oxygen vacancies has emerged as a versatile tool to tune the electronic and optoelectronic properties of complex oxide heterostructures. For the precise manipulation of the oxygen vacancies, the capability of directly probing the defect distribution in nanoscale is essential, but still lacking. Here we estimate the spatial distribution of oxygen vacancies in LaAlO3/SrTiO3 (LAO/STO) heterostructures by deep learning-based spectral analysis of current noise. The Monte-Carlo simulation and the specifically-designed deep learning model allow us to evaluate the defect distribution from current noise signals, measured through two-dimensional electron gas at the LAO/STO interface. We show that the oxygen vacancies are uniformly distributed over ∼ 100 nm from the interface in the as-grown LAO/STO heterostructure, while they can be migrated and confined to the interface within ∼ 14 nm by a vertical electric field at room temperature. These results introduce a powerful strategy to quantitatively probe the spatial distribution of point defects in oxide heterostructures with nm-scale precision.