Reinforcement-Learned Electric-Field Sensing with Asymmetrically Blockaded Rydberg Arrays

This paper addresses the challenge of enhancing the sensitivity and dimensionality of electric-field sensing using Rydberg atom arrays. Traditional Rydberg electrometers often rely on fixed pulse sequences and are primarily limited to scalar field measurements, which restricts their utility in complex, multi-dimensional environments. The authors aim to overcome these limitations by integrating microwave-dressed asymmetric blockade, Förster-enhanced interactions, and reinforcement learning (RL) to optimize quantum sensing protocols.

The researchers utilize a neutral atom array trapped in optical tweezers, where a central control atom interacts with surrounding target atoms. They employ microwave dressing to induce an asymmetric blockade, which suppresses interactions between target atoms while maintaining a field-tunable coupling between the control and target atoms near a Förster resonance. To maximize sensitivity, they employ an RL agent to discover optimal composite pulse sequences that steer the many-body state evolution. For vector electrometry, they propose a spherical six-atom configuration, using a weak bias field to resolve angular ambiguities inherent in dipole-dipole interactions.

The study reports several key results. First, in planar arrays, the system exhibits near-quadratic scaling of classical Fisher information with the number of atoms, approaching the Heisenberg limit. Second, RL-optimized pulse sequences significantly enhance quantum Fisher information, providing up to an order of magnitude improvement over standard single-pulse protocols. Third, the spherical six-atom array successfully reconstructs the orientation of vector electric fields by mapping axial population signals to field directions. Finally, numerical simulations confirm that the protocol remains robust against common experimental imperfections, such as Rabi frequency deviations, positional errors, and projection noise.