Noise-robust discrimination of incoherent point sources with spatial-mode demultiplexing

We theoretically predict and experimentally demonstrate that a reduced spatial-mode demultiplexing (SPADE) measurement using only the two lowest-order Hermite-Gaussian modes exhibits remarkable robustness against background noise in discriminating between a single source and two incoherent point sources. We establish a theoretical framework incorporating uniform background noise and derive an analytical Chernoff exponent expression, showing that SPADE consistently outperforms direct imaging (DI) across the sub-Rayleigh separations. Experimental results confirm that SPADE-based hypothesis testing approaches the quantum limit even when the background-to-signal photon ratio per pixel is 0.11. This advantage stems from SPADE's ability to concentrate source information into minimal detection modes, reducing the cumulative background noise impact. Our findings provide a practical detection scheme for applications where background noise is inevitable, such as astronomical observations and quantum sensing.