Abstract:Gate-defined quantum dots (QDs) have appealing attributes as a quantum computing platform, however, near-term devices possess a range of possible imperfections that need to be accounted for during the tuning and operation of QD devices. One such problem is the capacitive cross-talk between the metallic gates that define and control QD qubits. A way to compensate for the capacitive cross-talk and enable targeted control of specific QDs independent of coupling is by the use of virtual gates. Here, we demonstrate a reliable automated capacitive coupling identification method that combines machine learning with traditional fitting to take advantage of the desirable properties of each. We also show how the cross-capacitance measurement may be used for the identification of spurious QDs sometimes formed during tuning experimental devices. Our systems can autonomously flag devices with spurious dots near the operating regime which is crucial information for reliable tuning to a regime suitable for qubit operations.
Abstract:Quantum computers based on gate-defined quantum dots (QDs) are expected to scale. However, as the number of qubits increases, the burden of manually calibrating these systems becomes unreasonable and autonomous tuning must be used. There have been a range of recent demonstrations of automated tuning of various QD parameters such as coarse gate ranges, global state topology (e.g. single QD, double QD), charge, and tunnel coupling with a variety of methods. Here, we demonstrate an intuitive, reliable, and data-efficient set of tools for automated global state and charge tuning in a framework deemed physics-informed tuning (PIT). The first module of PIT is an action-based algorithm that combines a machine learning (ML) classifier with physics knowledge to navigate to a target global state. The second module uses a series of one-dimensional measurements to tune to a target charge state by first emptying the QDs of charge, followed by calibrating capacitive couplings, and navigating to the target charge state. The success rate for the action-based tuning consistently surpasses $95~\%$ on both simulated and experimental data suitable for off-line testing. The success rate for charge setting is comparable when testing with simulated data, at $95.5(5.4)~\%$, and only slightly worse for off-line experimental tests, with an average of $89.7(17.4)~\%$ (median $97.5~\%$). It's noteworthy that the high performance is demonstrated both on data from samples fabricated in an academic cleanroom as well as on an industrial 300 mm process line, further underlining the device-agnosticity of PIT. Together, these tests on a range of simulated and experimental devices demonstrate the effectiveness and robustness of PIT.
Abstract:The current autotuning approaches for quantum dot (QD) devices, while showing some success, lack an assessment of data reliability. This leads to unexpected failures when noisy data is processed by an autonomous system. In this work, we propose a framework for robust autotuning of QD devices that combines a machine learning (ML) state classifier with a data quality control module. The data quality control module acts as a ``gatekeeper'' system, ensuring that only reliable data is processed by the state classifier. Lower data quality results in either device recalibration or termination. To train both ML systems, we enhance the QD simulation by incorporating synthetic noise typical of QD experiments. We confirm that the inclusion of synthetic noise in the training of the state classifier significantly improves the performance, resulting in an accuracy of 95.1(7) % when tested on experimental data. We then validate the functionality of the data quality control module by showing the state classifier performance deteriorates with decreasing data quality, as expected. Our results establish a robust and flexible ML framework for autonomous tuning of noisy QD devices.