Gate-defined quantum dots (QD) have many appealing attributes as a quantum computing platform. However, their tune-up requires precise and often manual calibration of many parameters. Recent efforts, largely leveraging image processing and machine learning tools [1], have demonstrated approaches for automated QD device calibration. Unfortunately, the imperfections present in current devices cause unexpected failure of tuning algorithms, especially when models used to assess the device state are prepared using idealized noiseless data. To improve robustness of autotuning systems and avoid failure when tuning up nonideal devices, we have proposed a tuning framework that integrates quality assessment into the tuning system [2]. Beyond enabling development of the quality control module, the synthetic noise we introduced into the simulated data [3] also significantly improved autotuning robustness [2]. We have now expanded the framework developed in Ref. [3] to go beyond rejecting data with insufficient quality and include identification of common defects. In addition to providing high-level information about device issues, this will also allow for targeted autonomous device recalibration. Specifically, we are developing tools to autonomously flag devices with unintended quantum dots near the operating regime and to identify high levels of telegraph noise. Overall, our autonomous systems will enable both high throughput screening of quantum dot devices as well as more reliable tuning to a regime suitable for qubit operations.

[1] J.P. Zwolak, et al., Autotuning of Double-Dot Devices In Situ with Machine Learning, Phys. Rev. Applied 13, 034075, (2020).

[2] J. Ziegler, et al., Towards Robust Autotuning of Noisy Quantum Dot Devices, arxiv:2108.00043, (2021).

[3] J.P. Zwolak, S.S. Kalantre, X. Wu, S. Ragole, J.M. Taylor. QFlow lite dataset: A machine-learning approach to the charge states in quantum dot experiments. PLoS ONE, 13 (10): 1–17, (2018).