Graph-Based Calibration for the control of silicon Spin Qubits
ORAL
Abstract
Although silicon spin-qubits have demonstrated the potential for fault-tolerant quantum computing, they need fast and reliable calibrations to tune and maintain optimal performance.
We perform automated calibrations using a directed acyclic graph [1] to tune a spin qubit device, fabricated by an industrial pilot line, to achieve Pauli Spin Blockade (PSB) from a cold start. Each node calibrates a specific experiment and feeds the extracted parameters to the subsequent node. The first nodes optimize the readout and identify sensor voltages for maximum signal. Subsequent steps map the voltages to load n-electrons into an isolated dot, apply interpolation to auto-adjust the sensor during qubit scans, load user-defined number of electrons, and identify regions of PSB stochasticity. The process ends when all nodes are within specifications. We employ a diverse set of data analysis methods including mathematical optimization, image and signal processing, and unsupervised learning methods. Our physics-driven approach minimizes measurement duration and computational overhead, needed for scaling up this platform. This pipeline enables device calibration in under 20 minutes.
References
[1]: J. Kelly et al., Physical qubit calibration on a directed acyclic graph, arXiv:1803.03226 (2018)
We perform automated calibrations using a directed acyclic graph [1] to tune a spin qubit device, fabricated by an industrial pilot line, to achieve Pauli Spin Blockade (PSB) from a cold start. Each node calibrates a specific experiment and feeds the extracted parameters to the subsequent node. The first nodes optimize the readout and identify sensor voltages for maximum signal. Subsequent steps map the voltages to load n-electrons into an isolated dot, apply interpolation to auto-adjust the sensor during qubit scans, load user-defined number of electrons, and identify regions of PSB stochasticity. The process ends when all nodes are within specifications. We employ a diverse set of data analysis methods including mathematical optimization, image and signal processing, and unsupervised learning methods. Our physics-driven approach minimizes measurement duration and computational overhead, needed for scaling up this platform. This pipeline enables device calibration in under 20 minutes.
References
[1]: J. Kelly et al., Physical qubit calibration on a directed acyclic graph, arXiv:1803.03226 (2018)
–
Presenters
-
Daniel Solis
- Quobly