Data-Driven Insights into Oxygen Diffusion Barriers: A Closed-Loop Approach for Optimizing Superconducting Thin Films from First Principles

Surface oxides on superconducting thin films are known to introduce two-level systems (TLSs), which significantly contribute to losses in superconducting qubits. To mitigate this, a viable strategy is to introduce a metal capping layer to serve as an oxygen diffusion barrier. We present a machine learning (ML) approach that iteratively leverages experimental data to identify promising diffusion barrier candidates in a closed-loop manner. The model utilizes metal interstitial energies and oxide vacancy energies, either calculated from density functional theory (DFT) or inferred from the Materials Project using a novel neural network approach. These data, along with experimental observations and a quantification of their inherent uncertainties, are analyzed using logistic regression to predict metals that can prevent oxide formation at the metal/niobium interface. Further experiments validate our predictions, highlighting the model's ability to interpret fundamental materials processes and identify effective diffusion barriers. This work demonstrates the effectiveness of our ML technique in bridging theory and experiment, accelerating materials discovery, and providing valuable insights for the rational design of new materials for superconducting and other devices.