Machine Learning-Assisted Measurement Device-Independent Quantum Key Distribution on Reference Frame Calibration.

Area of Science:

• Quantum Information Science
• Applied Machine Learning
• Optical Communication Systems

Background:

• Measurement device-independent quantum key distribution (MDI-QKD) systems require robust feedback controls for stability against environmental disturbances and component imperfections.
• Traditional phase reference frame calibration methods in MDI-QKD systems are complex and reduce transmission efficiency.

Purpose of the Study:

• To develop and demonstrate a novel machine learning-assisted MDI-QKD system for efficient real-time phase calibration.
• To enhance the stability and key transmission efficiency of MDI-QKD systems.

Main Methods:

• Implementation of a machine learning model, specifically a long short-term memory (LSTM) network, for predicting phase drift in MDI-QKD.
• Active real-time phase compensation based on LSTM predictions.
• Experimental validation over 100 km and 250 km of standard single-mode fiber.

Main Results:

• The LSTM network successfully predicted phase drift in real-time.
• Active phase compensation significantly improved system stability and interference visibility.
• Dramatically increased key transmission efficiency in the MDI-QKD system.

Conclusions:

• Machine learning, particularly LSTM networks, offers an effective solution for real-time reference frame calibration in MDI-QKD systems.
• The developed approach enhances MDI-QKD system performance and efficiency over long distances.
• This work paves the way for more practical and efficient quantum communication networks.