We create machine-learning-based open-source packages for neutron, X-ray, and electron scattering and spectroscopy community.
Diffraction
- Thermal transport analyzer from time-resolved diffraction
- Magnetism classifier from crystal structures
Reflectometry
Spectroscopy
- Topology classifier from X-ray absorption (XAS)
- Majorana zero mode predictor from scanning tunneling spectroscopy
Inelastic Scattering
- Phonon density-of-states predictor from crystal structures
- Phonon dispersion relation predictor from crystal structures
Thermal transport analyzer from time-resolved diffraction
Input: Time-dependent diffraction intensities, heat capacities, phonon group velocities of each layer in a multi-layered sample.
Output: phonon-frequecny-dependent phonon relaxation time and interfacial phonon transmittance.
Machine learning: Adjoint-state neural diffential equation.
Reference: “Panoramic mapping of phonon transport from ultrafast electron diffraction and machine learning,” Advanced Materials 35, 2206997 (2023)
Magnetism classifier from crystal structures
Input: Crystal structure data (e.g. cif file)
Output: The class of magnetism of the crystal (ferromagnetism, antiferromagnetism, or non-magnetism)
Machine learning: Equivariant graph nerual network.
Reference: “Machine Learning Magnetism Classifiers from Atomic Coordinates,” iScience: Cell Press 25, 105192 (2022).
Polarized neutron reflectometry data fitting
Input: Reflectometry data (either unpolarized or polarized) of a multilayered sample
Output: The sturcture information of the sample (thickness, roughness, density, and magentism of each layer)
Machine learning: Variational autoencoder.
Reference: “Elucidating proximity magnetism through polarized neutron reflectometry and machine learning”, App. Phys. Rev. 9, 011421 (2022).
Topology classifier from X-ray absorption
Input: X-ray absorption data
Output: The class of band topology (topologically trivial, topological insulator, or topological semimetals)
Machine learning: Convolutional neural networks.
Reference: “Machine learning spectral indicators of topology”, Advanced Materials 34, 2204113 (2022).
Majorana zero mode predictor from scanning tunneling spectroscopy
Input: Zero-bias-peak from scanning tunneling spectroscopy (STS), either 1D or 2D with Zeeman field
Output: The prediction if the peak indicates it is from Majorana zero modes, Andreev bound states, or other topological trivial states.
Machine learning: Topological data analysis and gredient boosting.
Reference: “Machine Learning Detection of Majorana Zero Modes from Zero Bias Peak Measurements,” arXiv:2310.18439 (2023).
Phonon density-of-states predictor from crystal structures
Input: Crystal structure data (e.g. cif file)
Output: The curve of phonon density-of-states.
Machine learning: Equivariant graph neural networks.
Reference: “Direct prediction of phonon density of states with Euclidean neural networks”, Adv. Sci. 8, 2004214 (2021).
Phonon dispersion relation predictor from crystal structures
Input: Crystal structure data (e.g. cif file) and pathways
Output: The full 3N phonon bands in the Brioullin zone.
Machine learning: Virtual-node graph neural networks.
Reference: “Virtual Node Graph Neural Network for Full Phonon Prediction,” arXiv:2301.02197 (2023).
