Passive Source Quantum Superresolution Assisted by Physics-Informed Robust Deep Learning

Resolution is key to seeing fine details in imaging and sensing—whether it’s in biomedical imaging, observing distant stars, quantum measurements, or everyday optical systems. But all optical systems face a crucial barrier known as the diffraction limit, which sets a fundamental limit on how closely two points can be distinguished. Previous breakthroughs have achieved super-resolution by actively controlling or labeling the sample, which works well for certain biomedical systems. However, this approach isn’t possible for live biological samples that could be damaged by probes, for delicate quantum systems that can be disturbed by measurement, or for astronomical objects that we simply cannot manipulate. This project aims to overcome these challenges by developing new super-resolution methods that do not require controlling the source. The research team will combine advanced physical models of imaging with artificial intelligence (AI) to resolve details of passive, uncontrollable objects in real time. This research will promote the progress of imaging and sensing science by pushing the boundaries of what is possible in optical imaging, benefiting fields like medicine, astrophysics, and quantum sensing. In addition, by integrating artificial intelligence with optical engineering, the project will create unique educational opportunities in quantum and optical physics and AI for high school, undergraduate, and graduate students, helping to inspire and train the next generation of scientists and engineers. This project aims to achieve real-time quantum super-resolution imaging of practical passive point sources by developing a parameter-decoupled supper-resolution technique integrated with physics-informed robust deep learning. While prior super-resolution methods have overcome the Abbe-Rayleigh diffraction limit for controllable active sources, practical approaches for passive, incoherent, and unbalanced sources remain elusive due to real-world imperfections, multi-source complexities, and stability challenges. Building on the team’s previous theoretical advances and promising preliminary results (reaching 14 times better than the conventional resolution limit), the project proposes a three-step strategy: physics-informed machine-learning super-resolution, stability-driven real-time deep learning, and experimental validation. Scientifically, the primary objectives of this proposal are a high-precision, noise-tolerant technique that overcomes partial coherence, intensity imbalance, random phase and photon statistics limitations, with broader impacts across optical imaging in astrophysics, biomedical measurement, and quantum information science. Additionally, the objectives include comprehensive educational outreach and training efforts to promote STEM participation from high school through graduate education. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.