Exceptional points (EPs) in non-Hermitian systems offer a remarkably strong response to weak perturbations, but the nonorthogonal nature of the corresponding eigenvectors causes noise to diverge, hindering EPs practical application. Here, we report a twelve-fold enhancement of signal-to-noise ratio (SNR) in magnetic field sensing enabled by a third-order EP of coherent perfect absorption (CPA EP3) in a passive cavity magnonic system. This non-Hermitian magnonic platform comprises two identical yttrium iron garnet (YIG) spheres coherently coupled to a cavity mode, in which the CPA EP3 is realized by engineering the three-mode loss to form a pseudo-Hermitian absorption Hamiltonian. By independently tailoring the absorption EP apart from the resonance EP, the system circumvents the noise divergence caused by eigenbasis collapse. Notably, we harness the sensitivity of the minimum output intensity near CPA to perturbations, yielding a seventy-fold SNR improvement and a 400-fold increase in responsivity compared with non-CPA system. A comprehensive noise analysis over one hundred repeated measurements confirms the suppression of frequency noise near the CPA EP3. This demonstrates that our scheme not only avoids the noise divergence plaguing conventional higher-order EP sensors but also provides a general strategy to exploit both CPA and EP for SNR enhancement in passive non-Hermitian systems. Exceptional points (EPs) in non-Hermitian systems offer a strong response to perturbations, offering a promising platform for sensing. By using a third-order EP of coherent perfect absorption, here, authors demonstrate signal-to-noise ratio enhancement in a passive cavity magnonic system, providing a step forward towards future developments of EP sensing.
Yu et al. (2026) studied this question.