AI for quantum computing in silicon

Thesis

Spin qubits in silicon-based quantum devices are a candidate quantum computing architecture because of their high fidelities, long coherence times and pathway to scalability. However, their potential for scaling is tainted by device variability. Each device must be tuned to operation conditions. Automated artificial intelligence-based tuning methods are necessary as individual devices scale and the dimensions of the tuning parameter space increase. This thesis presents algorithms for the automatic tuning of silicon-based quantum device architectures. I demonstrate a machine learning-based algorithm that is capable of tuning a 4-gate Si FinFET, a 5-gate GeSi nanowire and a 7-gate Ge/SiGe heterostructure double quantum dot device without human intervention. I achieve double quantum dot tuning times of 30, 10, and 92 minutes, respectively. I construct a new classifier of quantum transport features using machine learning and obtain novel insights into the double quantum dot parameter space across the different device architectures. I demonstrate the first algorithm for the automatic tuning of an ion-implanted donor in silicon device up to the point of readout calibration within 10 minutes. Modules relying on computer vision perform signal processing of quantum transport measurements synonymous with donor spin in silicon devices and enable tuning and characterisation faster than human experts. Finally, using machine learning I infer true qubit states from imperfect measurements and cross-examine our method on simulated data. I estimate initialisation fidelities of 99.34% for a Si-MOS qubit at 1 kelvin, further validating silicon-based architectures as a platform for quantum computing. These results show that AI-enabled automation is integral to the wave which carries silicon-based quantum devices towards the shores of universal fault-tolerant quantum computation.

Authors: Severin, B

• DOI: 10.5287/ora-2zq6zqypz
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: silicon; phosphorus; donors; quantum computing; machine learning; germanium