Spectroscopic Descriptors for Complex Chemical Systems

• My research focuses on intelligent chemistry, with key representative achievements including: (1) Proposing dimensionally unified and theory-practice aligned spectroscopic descriptors, establishing a spectral-structure-activity intelligent model with transferable prediction capabilities, and achieving prediction and screening for multiple types of catalytic systems. (2) Developing a correlation model between spectroscopic descriptors and dynamic reaction processes, significantly enhancing its generalization ability through zero-shot learning, and enabling the identification of chemical reaction intermediates in experimental scenarios. (3) Revealing the capability of spectroscopic descriptors for cross-modal prediction and cross-dataset modeling, and leveraging their inherent physical constraints to achieve functional label completion and catalytic structure generation. In future work, the applicant aims to develop novel generative artificial intelligence strategies based on multimodal spectroscopic descriptors for the rational design of high-value, multifunctional chemicals, enabling on-demand generation of molecules and materials that simultaneously satisfy multiple predefined properties. Additionally, the applicant plans to map extensive chemical reaction networks into a continuously inferable spectroscopic space, providing innovative solutions for synthetic route planning and functional evolution assessment of multifunctional chemicals designed by generative models.