Measurement-Based Feedback Quantum Control with Deep Reinforcement Learning for a Double-Well Nonlinear Potential

Area of Science:

• Quantum physics
• Artificial intelligence
• Control theory

Background:

• Closed-loop quantum control utilizes measurements to steer quantum systems.
• Optimal control methods are established for quadratic Hamiltonians but fail for nonlinear systems.
• Nonlinear quantum systems, like the double well, present significant control challenges.

Purpose of the Study:

• To apply deep reinforcement learning (DRL) for controlling nonlinear quantum systems.
• To drive a nonlinear quantum system towards its ground state with high fidelity.
• To investigate DRL's ability to learn effective control strategies for complex quantum dynamics.

Main Methods:

• Utilized a deep reinforcement learning agent to learn quantum control strategies.
• Implemented a control strategy motivated by continuous, weak quantum measurements.
• Trained the neural agent using feedback from system measurements.

Main Results:

• The DRL agent successfully learned to control the quantum evolution of a nonlinear double-well system.
• Achieved high-fidelity driving of the system towards its ground state.
• Discovered counterintuitive cooling strategies leading to a near-pure "cat" state.

Conclusions:

• Deep reinforcement learning offers a powerful approach for controlling complex, nonlinear quantum systems.
• DRL can overcome limitations of traditional control techniques in quantum dynamics.
• The learned DRL strategy effectively prepares a high-fidelity ground state analogue.