NQCP Life in Quantum - Ass. Professor Vincent Philippe Michal, NQCP Niels Bohr Institute

A deep-learning enabled digital surrogate of semiconductor based quantum devices

General control of quantum devices is a central question of modern quantum science research and underlies foreseen applications that include quantum sensing, networking, computation and simulation of quantum many-body physics. However this objective is often obstructed by inefficient characterization of the devices, difficulty in tuning, and challenges in generating appropriate designs. Such limitations are important for carriers in electrostatically-defined quantum dots, particularly susceptible to capacitive crosstalk effects and variations in fabrication. Simulations can improve our insight on these issues, and in many cases fast feedback between measurements and simulations would help resolve them. 

We introduce a modular graph-based simulator to serve as a digital surrogate for tuning and control of charges in semiconductor quantum dots. Nodes of the graph each predict a property of the device, and slow computations may be accelerated using machine learning models trained using the original node. We model an experimental device and compare charge sensing simulation outputs with measurement data in practical timescales. To demonstrate the utility of our simulator we address the issue of capacitive crosstalk through comparison of different virtual gate voltage computations. Our comparison shows that crosstalk is better addressed holistically for independent and consistent control of all device properties. Furthermore, we demonstrate that these simulated virtual gates can remove unwanted crosstalk in the physical experimental device. We envision our approach to device simulation will facilitate progress in semiconductor-based quantum technologies including efficient design, characterization, and full control of experimental devices.

  1. A. Lidiak, J. Swain, D. L. Craig, J. Hickie, Y. Yang, G. Katsaros, D. T. Lennon, VPM, E. M. Gauger, and N. Ares (to appear)