Virtual gates enabled by digital surrogate of quantum dot devices

Advances in quantum technologies are often limited by slow device characterization, complex tuning, and scalability challenges. Spin qubits in electrostatically defined quantum dots present a promising platform but are not exempt from these limitations. Simulations improve our insight into the behaviour of such devices, and in many cases, fast feedback between measurements and simulations can provide a path towards developing optimal design and control strategies. Here, we introduce a modular graph-based simulator to serve as a digital surrogate of a semiconductor quantum dot device, where computationally expensive processes are accelerated with deep learning. We demonstrate the potential of our simulator by estimating the effect of crosstalk between gate electrodes and using this estimate to construct virtual gates in a quantum dot device. We compare our results with experiments using a double quantum dot defined in a Ge/SiGe heterostructure. We envision our approach to device simulation will facilitate progress in semiconductor-based quantum technologies including efficient design, characterization, and control of complex devices.