We pioneered a recursive inverse-design algorithm that integrates multiscale simulations to overcome the “computational barrier” in photonic chip design, enabling automated end-to-end structural design. The hyperspectral photonic chips developed using this approach allow on-chip spectrometers to outperform costly commercial instruments across multiple key performance metrics, achieving on-chip spectroscopic measurements with sub-picometer resolution and a spectral coverage exceeding 800 nm. This work provides fundamental scientific support for the broader deployment and widespread adoption of spectroscopic sensing technologies in the future.

The co-first authors of the paper are Hao He and Zengji Tu, Ph.D. students at the School of Electronics, Peking University, and Yuanlei Wang, a Ph.D. student at the School of Physics, Peking University. Lin Chang of the School of Electronics and Qifan Yang of the School of Physics are the co-corresponding authors. Major collaborators also include Hongyan Zhao and Lei Zhang, postdoctoral researchers at the School of Electronics; Chuangxin Feng and Yujun Chen, graduate students at the School of Electronics; Ruoao Yang, a postdoctoral researcher at the School of Physics; Yongzhuo Zhou, an undergraduate student at the School of EECS; and Professor Jianjun Wu of the School of Electronics. The work was carried out with the National Key Laboratory of Photonic Transmission and Communications, School of Electronics, Peking University, as the primary affiliation.

Link: https://www.nature.com/articles/s41467-026-76328-3