AI-Enabled Quantum Materials Research
13th November 2025
Timing : 1 pm ET
Please use this zoom link for joining the webinar
For a list of all talks at the NanoBio seminar Series Fall'25, see here
For high data volumes (e.g., density-functional-theory studies in weakly correlated systems), ML can streamline prediction of lower-dimensional properties. We introduce a convolutional neural network classifying band topology from X-ray absorption signals [1], and demonstrate an autoencoder protocol that boosts resolution in polarized neutron reflectometry for the magnetic proximity effect [2]. When data are scarce due to high computational costs, incorporating symmetry helps reduce dataset size. For instance, the O(3) Euclidean neural network predicts phonon density-of-states [3], dielectric functions, and quantum weight [4], with differential equations as constraints [5]. For high-dimensional outputs like phonon dispersion relations, we employ graph neural networks enhanced by virtual nodes [6], achieving higher efficiency without sacrificing accuracy. To address unreliable ground truth, we apply ML to distinguish Majorana zero modes in scanning tunneling spectroscopy [7]. For quantum spin liquids, where experiments and calculations are both challenging, we generate geometrically frustrated materials via SCIGEN as a starter, which produces 10 million Archimedean-lattice structures, over half passing DFT stability checks and 2 synthesized in lab [8].
Despite these advances, ML for quantum materials remains in its infancy. We must confront defect problems [9], out-of-distribution problems [10], accuracy constraints, and the pursuit of genuinely new phenomena, especially in complex quantum systems and phase diagram studies.
[1] NA, ML, “Machine learning spectral indicators of topology,” Advanced Materials 34, 202204113 (2022).
[2] NA, ML, "Elucidating proximity magnetism through polarized neutron reflectometry and machine learning," Applied Physics Review 9, 011421 (2022).
[3] ZC, ML, “Direct prediction of phonon density of states with Euclidean neural networks,” Advanced Science 8, 2004214 (2021).
[4] NH, ML, “Universal Ensemble-Embedding Graph Neural Network for Direct Prediction of Optical Spectra from Crystal Structure,” Advanced Materials 36, 2409175 (2024).
[5] ZC, ML, "Panoramic mapping of phonon transport from ultrafast electron diffraction and machine learning," Advanced Materials 35, 2206997 (2023).
[6] RO, ML, "Virtual Node Graph Neural Network for Full Phonon Prediction," Nature Computational Science 4, 522 (2024).
[7] MC, ML, “Machine Learning Detection of Majorana Zero Modes from Zero Bias Peak Measurements,” Matter 7, 2507 (2024).
[8] RO, ML, "Structural Constraint Integration in Generative Model for Discovery of Quantum Material Candidates," DOI:10.1038/s41563-025-02355-y, Nature Materials (2025).
[9] CF, ML, “AI-driven defect engineering in advanced thermoelectric materials,” Advanced Materials 37, 2505642 (2025).
[10] MC, ML, arXiv:2502.02905, Nature Materials, in production (2025).
Dr. Mingda Li
Associate Professor
Mingda Li is an Associate Professor in the Department of Nuclear Science and Engineering at MIT. He earned his BS in Engineering Physics from Tsinghua University in 2009 and his PhD from MIT in 2015. Following his doctoral studies, he conducted postdoctoral research at MIT MechE. Li's research focuses on Quantum, Nuclear and AI. In 2025 he has been named a Fellow of American Physical Society (APS).
Associate Professor
Mingda Li is an Associate Professor in the Department of Nuclear Science and Engineering at MIT. He earned his BS in Engineering Physics from Tsinghua University in 2009 and his PhD from MIT in 2015. Following his doctoral studies, he conducted postdoctoral research at MIT MechE. Li's research focuses on Quantum, Nuclear and AI. In 2025 he has been named a Fellow of American Physical Society (APS).