Away from voltages: Generating and using abstractions to operate arrays of quantum dots

As quantum dot qubits mature as a technology, operation and characterization of larger and larger devices must become robust and routine. This talk describes automated tuning and characterization methods that focus on extracting and storing relevant and physically-meaningful information contained in measurements of dot devices. These methods, which use a combination of machine learning techniques and simple physical models, are shown to work with a variety of data types of varying quality acquired from Si/SiGe SLEDGE arrays of quantum dots. The resulting information forms a high-level 'device API' that allows interaction with, e.g., quantities of charge and tunnel coupling rather than applied gate voltages.