Towards autonomous time-calibration of large quantum-dot devices: detection, real-time feedback, and noise spectroscopy

The scalability of spin-qubit registers is currently hindered by device inhomogeneity and charge noise, which give rise to slow, time-dependent drifts and necessitate frequent, time-consuming manual calibration. To address these challenges, we introduced an automated framework for real-time monitoring, characterization, and correction of drift in large quantum-dot systems. Built on the Modular Autonomous Virtualization System (MAViS) [1], our framework leverages time traces of intra- and inter-dot charge transition lines extracted from double-quantum-dot charge-stability diagrams. By combining an ensemble of convolutional neural networks, dynamic windowed Hough transforms, and physics-informed heuristics, we reliably identify the charge transitions defining the relevant charge state and track their time evolution, enabling both on-the-fly electrostatic drift compensation and in-situ noise spectroscopy.

From the recorded time traces, we extract the power spectral density to characterize the local noise affecting each quantum dot. This allows us to autonomously identify dominant noise sources and signatures of two-level fluctuators, independently of charge-sensor proximity. We benchmark our approach against simulations and determine minimum noise thresholds compatible with high-fidelity device performance. In addition, we demonstrate high-accuracy tracking of stability maps and apply real-time feedback to correct random drifts.

We deploy our framework on a 10-dot hole-based device in planar germanium, performing continuous autonomous monitoring for over 48 hours. To benchmark sensitivity, we inject artificial voltage pulses and demonstrate reliable detection of electrostatic shifts as small as 0.1 mV. We autonomously identify dominant noise sources for each gate and quantify their correlations, revealing a noise-correlation length of approximately 200 nm. Our methods establish a powerful toolkit for autonomous stabilization and scalable operation of quantum-dot arrays–an essential capability on the path to fault-tolerant quantum computing.