Leveraging Machine Learning Force Fields (MLFFs) to Simulate Large Atomistic Systems for Fidelity Improvement of Superconducting Qubits and Sensors

Abstract

Materials engineering using atomistic modeling is an essential tool for the development of qubits and quantum sensors. Traditional density-functional theory (DFT) does however not adequately capture the complete physics involved, including key aspects and dynamics of superconductivity, surface states, etc. There are also significant challenges regarding the system sizes that can be simulated, not least for thermal properties which are key in quantum-computing applications. The QuantumATK tool combines DFT, based on LCAO basis sets, with non-equilibrium Green's functions, to compute the characteristics of interfaces between superconductors and insulators, as well as the surface states of topological insulators. Additionally, the software leverages machine-learned force-fields to simulate thermal properties and to generate realistic amorphous geometries in large-scale systems. Finally, the description of superconducting qubits and sensors as two-level systems modeled with a double-well potential requires many-body physics, and this paper demonstrates how electron-electron interaction can be added to the single-particle energy levels from an atomistic tight-binding model to describe a realistic double-quantum dot system.