Deep learning for online electron spin relaxation measurement and its application in spin-exchange relaxation-free co-magnetometers

This paper presents a novel online method for measuring electron spin relaxation in a spin-exchange relaxation-free comagnetometer (SERFCM). The stability of electron spin relaxation is crucial for ensuring the accuracy of SERFCM measurements. First, we introduce a pumping technique based on elliptically polarized light, which not only enables hyperpolarization of the atomic spin ensemble but also facilitates high-precision online monitoring of electron spin polarization. Second, to account for the potential influence of stochastic disturbances in the complex operating environment of the SERFCM, we model the pumping process of the atomic spin ensemble as a stochastic process. It is demonstrated that this process can be effectively described using a Markov chain framework. Based on this, we develop a deep learning-based intelligent recognition algorithm combined with fast sampling to enable real-time acquisition of electron spin relaxation data. Finally, the proposed method is applied to the SERFCM, and experimental results demonstrate that it overcomes the limitations of traditional relaxation measurements, which require excitation signals, successfully enabling online monitoring of electron spin relaxation under normal operating conditions.