Scientific Research

In the research on artificial intelligence empowering the discipline of chemistry, dimensionality reduction encoding of complex substances is extremely critical. Encoding involves digitizing complex structures into parameterized descriptors. Spectroscopy, capable of compressing structural, property, and functional information based on physical rules, serves as a natural descriptor of the material world. The research group has long been committed to the study of spectral descriptors, addressing challenges in complex chemical systems such as the curse of dimensionality, the disconnect between theory and practice, sparse and discrete data, and difficulties in screening and optimization. Leveraging computable and measurable spectral descriptors, the group employs machine learning to establish clear models of spectral-structure-activity relationships, revealing the evolutionary dynamics of chemical reactions and enabling high-throughput screening and intelligent generation of chemicals. Several theoretical works have been experimentally validated, demonstrating the effective role of spectral intelligence in driving chemical innovation and offering new approaches to increasingly complex chemical problems.

To date, the group has published over 50 SCI-indexed papers in top-tier international academic journals, including 35 papers as corresponding and first authors (including co-authors), in journals such as Nat. Catal., J. Am. Chem. Soc., Angew. Chem. Int. Ed., and Proc. Natl. Acad. Sci. U.S.A.. These works have garnered significant attention from peers in related fields and have been cited and discussed in authoritative review articles in journals like Chem. Soc. Rev., Chem. Rev., and Nat. Rev. Mater.. Additionally, the proposed spectral descriptors have assisted experimental collaborators in achieving efficient design and high-throughput screening for various catalytic systems, guiding research practices on the machine chemist platform. The group has led projects supported by the National Natural Science Foundation of China, including the Youth Program and General Program, and has contributed as a key member to the National Major Scientific Instrument Development Project.

Recent representative achievements are as follows:

[1] Infrared spectroscopybased zero-shot learning for identifying reaction intermediates in unseen systems, PNAS, 2025, 122(32): e2506834122

[2] Quantitative Insight into the Electric Field Effect on CO2 Electrocatalysis via Machine Learning Spectroscopy, JACS, 2024, 146(50): 34551-34559

[3] Catalytic Structure Design by AI Generating with Spectroscopic Descriptors, JACS, 2023, 145(49): 26817-26823

[4] Machine Learning of Spectra-Property Relationship for Imperfect and Small Chemistry Data, PNAS, 2023, 120(20): e2220789120

[5] Identifying a highly efficient molecular photocatalytic CO2 reduction system via descriptor-based high-throughput screening, Nature Catalysis, 2025, 8(2): 126-136

[6] Identifying Chemical Reaction Processes by Machine Learned Spectroscopy, CCS Chemistry, 2025, 7(8): 2315-2324

[7] Cross-database property prediction and multifunctional molecule screening via spectral descriptor, Chinese Chemical Letters

[8] Fusion of Multiple Spectra for Investigating Chemical Bonding Properties via Machine Learning, The Journal of Physical Chemistry Letters, 2023, 14(33): 7461-7468

[9] Decoupling Analysis of O2 Adsorption on Metal–N–C Single-Atom Catalysts via Data-Driven Descriptors, The Journal of Physical Chemistry Letters, 2023, 14(20): 4760-4765

[10] Interlayer Charge Transfer Regulates Single-Atom Catalytic Activity on Electride/Graphene 2D Heterojunctions, JACS, 2023, 145(8): 4783-4783

[11] Accelerated Screening of Alternative DNA Base‐Organic Molecule‐Base Architectures via Integrated Theory and Experiment, Angew, 2024, 63(35)

[12] Spectra-based clustering of high-entropy alloy catalysts: improved insight over use of atomic structure, Chemical Science, 2025, 16(11): 4646-4653

[13] Cross-Modal Prediction of Spectral and Structural Descriptors via a Pretrained Model Enhanced with Chemical Insights, The Journal of Physical Chemistry Letters, 2024, 15(34): 8766-8772

[14] Deep Learning Accelerated Determination of Hydride Locations in Metal Nanoclusters, Angew, 2021, 60(22): 12289-12292

[15] Prediction by Convolutional Neural Networks of CO2/N2 Selectivity in Porous Carbons from N2 Adsorption Isotherm at 77 K, Angew, 2020, 59(44): 19645-19648