Adaptive Self-Optimization of Polarization Efficiency for Low-Power NV Center Thermometers Based on PAS-Q

Abstract

Nitrogen-vacancy (NV) centers in diamond have attracted considerable attention for micro- and nanoscale thermometry owing to their high precision and spatial resolution. However, achieving sufficient electron spin polarization and signal-to-noise ratio (SNR) typically requires high-power or long-duration laser excitation, which introduces heating and degrades measurement accuracy and stability. To address this trade-off, a reinforcement learning (RL)-based approach is proposed to optimize spin polarization under low-heating constraints. First, a physical model integrating five-level optical pumping rates and NV spin polarization dynamics is established. A temperature-rise constraint model is then developed based on pulsed optically detected magnetic resonance (P-ODMR) measurements under varying laser powers and pulse widths in an external magnetic field. Subsequently, a Progressive Action-Space Q-learning (PAS-Q) algorithm is employed to explore the parameter space and identify optimal laser power and pulse-width combinations. The agent is trained through numerical simulations and the resulting strategy is implemented in a hardware system. Experimental results demonstrate a 44% reduction in laser-induced heating, a 33% increase in spin polarization rate based on T1 relaxation analysis, and a 71% reduction in optimization time compared with traditional methods. This approach enables efficient spin polarization with reduced laser heating and improved optimization efficiency, and may be extended to other heat-sensitive quantum spin sensors, including silicon-vacancy centers in diamond and divacancy defects in silicon carbide.