Automated large-scale analysis of quantum dots in germanium bilayer heterostructures

As quantum dot (QD)-based qubits advance toward larger and more complex device architectures, rapid and automated device characterization and data analysis tools become critical. Bilayer Ge/SiGe heterostructures enable vertically coupled QDs across closely spaced quantum wells, introducing new degrees of freedom that enrich qubit operation but complicate analysis. Charge stability diagrams (CSDs) in such devices exhibit a diverse array of transitions, including interlayer tunneling and distinct capacitive signatures, which are not present in conventional planar systems. Manually interpreting these features is time-consuming and impractical.

We present an automated protocol for analyzing CSDs from bilayer devices, inspired by the MAViS protocol [1]. Our method integrates machine learning (ML), image processing, and object detection to identify and track charge transitions across large datasets. By combining ML and object tracking, we can distinguish between different types of transition with otherwise similar features. We explore various model architectures to enhance classification performance on bilayer devices of varying complexity. Finally, by analyzing the properties of many such CSDs, we can statistically estimate physical quantities, like lever arms and capacitive couplings. Our protocol allows for a rapid extraction of useful and nontrivial information about bilayer devices.



[1] Rao, et al. Phys. Rev. X (2025)