Physics-Informed Pre-training on Efficient Electron-Density Images for Organic Material Property Prediction

Zhixiang Cheng, Hongxin Xiang, Mingquan Liu, Tengfei Ma, Yingzhuo Tu, Wenjie Du, Bosheng Song, Yiping Liu, Xiangxiang Zeng
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:19032-19057, 2026.

Abstract

Precise property prediction of organic materials is pivotal for next-generation electronic and energy devices. In density functional theory (DFT), the electron density (ED) serves as the fundamental determinant of material properties. Yet, establishing it as an input modality for material property prediction has been impeded by two practical barriers: scarce large-scale ED data and the enormous computational complexity of ED representation. To bridge these gaps, we introduce VisionED, an efficient physics-informed model pre-trained on electron-density images. We curate a dataset of 2 million molecules and represent ED as multi-shot images that efficiently encode both geometric and electronic structure. VisionED is then pre-trained on 12 million multi-shot ED images via cross-scale, physics-informed pretext tasks. Empirical evaluations on photovoltaic and organic chromophore datasets show that VisionED outperforms state-of-the-art baselines by up to 27.0%, exhibiting superior robustness under distribution shifts and data scarcity. Notably, the model generalizes to unseen device-scale applications, successfully recovering experimental trends and mixing-ratio effects in ternary blends with an average accuracy of 92.77%. Moreover, relative to the previous ED point cloud, the ED image improves performance by 26.2% with 2.6$\times$ fewer memory and 4.6$\times$ lower time. The code and data are available at https://github.com/ZhixiangCheng/VisionED.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-cheng26m, title = {Physics-Informed Pre-training on Efficient Electron-Density Images for Organic Material Property Prediction}, author = {Cheng, Zhixiang and Xiang, Hongxin and Liu, Mingquan and Ma, Tengfei and Tu, Yingzhuo and Du, Wenjie and Song, Bosheng and Liu, Yiping and Zeng, Xiangxiang}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {19032--19057}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/cheng26m/cheng26m.pdf}, url = {https://proceedings.mlr.press/v306/cheng26m.html}, abstract = {Precise property prediction of organic materials is pivotal for next-generation electronic and energy devices. In density functional theory (DFT), the electron density (ED) serves as the fundamental determinant of material properties. Yet, establishing it as an input modality for material property prediction has been impeded by two practical barriers: scarce large-scale ED data and the enormous computational complexity of ED representation. To bridge these gaps, we introduce VisionED, an efficient physics-informed model pre-trained on electron-density images. We curate a dataset of 2 million molecules and represent ED as multi-shot images that efficiently encode both geometric and electronic structure. VisionED is then pre-trained on 12 million multi-shot ED images via cross-scale, physics-informed pretext tasks. Empirical evaluations on photovoltaic and organic chromophore datasets show that VisionED outperforms state-of-the-art baselines by up to 27.0%, exhibiting superior robustness under distribution shifts and data scarcity. Notably, the model generalizes to unseen device-scale applications, successfully recovering experimental trends and mixing-ratio effects in ternary blends with an average accuracy of 92.77%. Moreover, relative to the previous ED point cloud, the ED image improves performance by 26.2% with 2.6$\times$ fewer memory and 4.6$\times$ lower time. The code and data are available at https://github.com/ZhixiangCheng/VisionED.} }
Endnote
%0 Conference Paper %T Physics-Informed Pre-training on Efficient Electron-Density Images for Organic Material Property Prediction %A Zhixiang Cheng %A Hongxin Xiang %A Mingquan Liu %A Tengfei Ma %A Yingzhuo Tu %A Wenjie Du %A Bosheng Song %A Yiping Liu %A Xiangxiang Zeng %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-cheng26m %I PMLR %P 19032--19057 %U https://proceedings.mlr.press/v306/cheng26m.html %V 306 %X Precise property prediction of organic materials is pivotal for next-generation electronic and energy devices. In density functional theory (DFT), the electron density (ED) serves as the fundamental determinant of material properties. Yet, establishing it as an input modality for material property prediction has been impeded by two practical barriers: scarce large-scale ED data and the enormous computational complexity of ED representation. To bridge these gaps, we introduce VisionED, an efficient physics-informed model pre-trained on electron-density images. We curate a dataset of 2 million molecules and represent ED as multi-shot images that efficiently encode both geometric and electronic structure. VisionED is then pre-trained on 12 million multi-shot ED images via cross-scale, physics-informed pretext tasks. Empirical evaluations on photovoltaic and organic chromophore datasets show that VisionED outperforms state-of-the-art baselines by up to 27.0%, exhibiting superior robustness under distribution shifts and data scarcity. Notably, the model generalizes to unseen device-scale applications, successfully recovering experimental trends and mixing-ratio effects in ternary blends with an average accuracy of 92.77%. Moreover, relative to the previous ED point cloud, the ED image improves performance by 26.2% with 2.6$\times$ fewer memory and 4.6$\times$ lower time. The code and data are available at https://github.com/ZhixiangCheng/VisionED.
APA
Cheng, Z., Xiang, H., Liu, M., Ma, T., Tu, Y., Du, W., Song, B., Liu, Y. & Zeng, X.. (2026). Physics-Informed Pre-training on Efficient Electron-Density Images for Organic Material Property Prediction. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:19032-19057 Available from https://proceedings.mlr.press/v306/cheng26m.html.

Related Material