Autotuning of double dot devices in situ with machine learning.

Area of Science:

• Quantum Computing
• Machine Learning Applications
• Condensed Matter Physics

Background:

• Manual tuning of quantum dots (QDs) for qubit operation is slow and hinders scalability.
• Current methods rely on human expertise, which is not practical for large-scale quantum systems.

Purpose of the Study:

• To implement an in situ autotuning protocol for double-QD devices.
• To demonstrate the replacement of human heuristics with a machine learning algorithm for gate voltage tuning.

Main Methods:

• Utilized a machine learning (ML) algorithm trained on simulated data to classify the state of a double-QD device.
• Integrated the ML algorithm with an optimization routine for parameter space navigation.
• Implemented active feedback control for a functional double-dot device at millikelvin temperatures.

Main Results:

• Successfully demonstrated automated tuning of a double-QD device using an ML algorithm.
• Achieved active feedback control of a functional double-dot device at millikelvin temperatures.
• Analyzed success rates based on initial conditions and device performance.

Conclusions:

• The ML-based autotuning protocol effectively replaces manual tuning for quantum dot devices.
• The approach shows promise for scalable quantum computing by automating a critical calibration step.
• Further improvements in the ML network, fitness function, and optimizer can enhance success rates.