Speaker
Description
Exceptional points (EPs) in non-Hermitian systems offer square-root‐enhanced response to perturbations, promising order‐of‐magnitude sensitivity gains in OPM-based MEG sensors. We formulate the two‐mode alkali‐vapor Hamiltonian with balanced gain and loss, and locate its second‐order EP by solving for the coalescence of eigenvalues and eigenvectors. To circumvent the ill‐conditioned eigenvalue derivatives and noise amplification inherent to EP calculations, we introduce StaR, an algorithmic transform of problem representation, that re‐expresses the non-Hermitian eigenproblem more suitably for complex-valued Neural Quantum State optimization. StaR‐NQS enables automatic differentiation through the EP degeneracy and Monte Carlo–regularized evaluation of the noise‐amplified frequency splitting. This methodology paves the way for compact, low‐power, EP‐enhanced OPM arrays in clinical BCIs.
