From spectra to performance: Multimodal machine learning for organic optoelectronic materials discovery
The discovery and rational optimization of optoelectronic materials increasingly rely on heterogeneous characterization data encoding structure, morphology, electronic states, and device properties. Conventional spectral interpretation remains expert-dependent and often treats each modality in isolation, limiting information extraction. Machine learning has improved information extraction from individual spectroscopic modalities, yet single modality approaches often suffer from incomplete structural and physical representations. Multimodal machine learning enables integration of diverse spectroscopic signals into unified representations. In this review, we summarize recent advances in multimodal spectral learning and heterogeneous data fusion for spectra-structure-property relationships in optoelectronic materials. We discuss how the synergistic combination of infrared, Raman, nuclear magnetic resonance, and mass spectrometry improves functional-group recognition, substructure inference, and spectrum-based molecular structure elucidation and raw-spectrum learning for molecular elucidation. We further elaborate on how the fusion of optical spectra, morphology, processing parameters, and molecular representations facilitates quantitative property prediction and materials discovery. Key challenges include simulation-experiment domain gaps, limited multimodal datasets, missing modalities, and insufficient model interpretability. Overall, multimodal spectral learning is emerging as a promising approach for transforming characterization from a descriptive tool into a predictive framework for rational materials design.
Authors
- Xuefeng Guo (ORCID: https://orcid.org/0000-0001-5723-8528)
- Linghai Xie (ORCID: https://orcid.org/0000-0001-6294-5833)
- Mengna Yu
- Xinmiao Xie (ORCID: https://orcid.org/0009-0004-3398-3901)
- Yong Yan (ORCID: https://orcid.org/0000-0002-5784-0014)
- Hui Zhu
- Xun Fu (ORCID: https://orcid.org/0009-0005-2941-0788)
- Jie Hao
Institutions
- Peking University (CN)
- Nanjing University of Posts and Telecommunications (CN)
- Beijing National Laboratory for Molecular Sciences (CN)
- South China University of Technology (CN)
Publication Details
- Journal
- The Journal of Chemical Physics
- Published
- 2026-11-16
- DOI
- https://doi.org/10.1063/5.0356266
- Primary Topic
- Machine Learning in Materials Science
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Natural Science Foundation of China
- Basic Research Program of Jiangsu Province
- National Key Research and Development Program of China