Methods for automating quantum dot gate-voltage tuning from a cold start

The future scaling up of quantum dot qubits into large arrays will depend on efficient methods to select gate voltages that put each dot into the correct operating regime. These gate voltages, when chosen correctly, confine electrons within not only the quantum dots but also the reservoirs and nearby auxiliary devices, such as single-electron transistor charge sensors.  In this talk, we present a step-by-step procedure for choosing – tuning up – gate voltages starting from a blank slate.  Some of the steps, such as checking for gate-to-gate leakage, can be accomplished with automated measurements and simple conditional logic.  The important functionality of other gates, such as those controlling reservoirs, can be parametrized by automated fitting to phenomenological functional forms. In addition, there are important aspects of tune up that require two or more gate voltages to be correct before success can be assessed. In these cases, machine learning (ML) methods are especially useful. We present results using an explainable boosting machine (EBM) to automate the analysis of such data, developed from measurements of over 25 quantum dot samples. Finally, we discuss a class-based approach, built on top of QCoDeS [1], designed to aid the organization and implementation of the steps needed to tune quantum dot devices from a cold start.

Worked performed with M. A. Wolfe, Piotr Marciniec, Patrick J. Walsh, E.S. Joseph, Brighton X. Coe, J. A. Reily, Alysa Rogers, Jared Benson, Danielle Middlebrooks, Daniel Schug, G. J. Bernhardt, Justyna P. Zwolak, M. A. Eriksson.

[1] https://github.com/QCoDeS/Qcodes