PySlice: Routine Vibrational Electron Energy Loss Spectroscopy Prediction with Universal Interatomic Potentials

TL;DR

PySlice addresses the challenge of routine vibrational EELS prediction by unifying MD with universal machine learning interatomic potentials, GPU-accelerated multislice electron scattering, and the TACAW frequency-domain framework. The approach delivers automated, end-to-end predictions of momentum- and energy-resolved spectra, enabling phonon dispersions, spectral diffraction patterns, and spectrum images from atomic structures; a key equation is $I(\mathbf{k},\omega)=|\tilde{\psi}(\mathbf{k},\omega)|^2$ with $\tilde{\psi}(\mathbf{k},\omega)=\mathcal{F}_t[\psi(\mathbf{k},t)-\langle\psi(\mathbf{k},t)\rangle_t]$. The authors demonstrate the workflow on TMDCs (MoS$_2$, WS$_2$, MoSe$_2$, WSe$_2$) and show spatially-resolved phonon signals at interfaces and point defects, highlighting the system-wide applicability of the method. PySlice thus enables high-throughput spectroscopy predictions, training-data generation for ML models, and prediction-driven experimental design, all within a modular, reproducible Python framework that also supports conventional TEM/STEM simulations.

Abstract

Vibrational spectroscopy in the electron microscope can reveal phonon excitations with nanometer spatial resolution, yet routine prediction remains out of reach due to fragmented workflows requiring specialized expertise. Here we introduce PySlice, the first publicly available implementation of the Time Autocorrelation of Auxiliary Wavefunction (TACAW) method, providing an automated framework that produces momentum- and energy-resolved vibrational electron energy-loss spectra directly from atomic structures. By integrating universal machine learning interatomic potentials with TACAW, PySlice eliminates the bottleneck of per-system potential development. Users input atomic structures and obtain phonon dispersions, spectral diffraction patterns, and spectrum images through a unified workflow spanning molecular dynamics, GPU-accelerated electron scattering, and frequency-domain analysis. We outline the formulation behind the code, demonstrate its application to canonical systems in materials science, and discuss its use for advanced analysis and materials exploration. The modular Python architecture additionally supports conventional electron microscopy simulations, providing a general-purpose platform for imaging and diffraction calculations. PySlice makes vibrational spectroscopy prediction routine rather than specialized, enabling computational screening for experimental design, systematic exploration of phonon physics across materials families, and high-throughput generation of simulated data for training of future machine learning models.

Figures (4)

• Figure 1: PySlice workflow for automated vibrational EELS prediction. The pipeline integrates four stages: (1) input structures from CIF files, XYZ files, LAMMPS trajectories, or ASE Atoms objects; (2) optional molecular dynamics using universal machine learning interatomic potentials (ORB) with automated equilibration (NVT) and production (NVE) phases; (3) GPU-accelerated multislice simulation computing exit wavefunctions for each MD timestep; (4) TACAW analysis transforming time-domain wavefunctions to frequency-domain spectral intensities or conventional TEM/STEM such as frozen-phonon HAADF, TEM diffraction, or 4D-STEM. Data containers (Trajectory, WFData, TACAWData, HAADFData) manage information flow between stages.
• Figure 2: End-to-end automated TACAW spectra across the TMDC family. All simulations were generated from a single PySlice script that handled the complete workflow: crystal structure generation with ASE, ORB-driven molecular dynamics with automatic equilibration, GPU-accelerated multislice propagation through hundreds of timesteps, and TACAW frequency-domain analysis. No intermediate file manipulation, format conversion, or manual parameter adjustment was required between materials. The resulting phonon dispersions and spectral diffraction patterns for MoS$_2$, WS$_2$, WSe$_2$ and MoSe$_2$ capture material-specific vibrational signatures arising from differences in atomic masses and bonding.
• Figure 3: Spectrum Images of Localized Vibrations. (a-c) Bulk Si/Ge heterostructure interface showing optical phonon confinement: (a) Ge optical modes ($\sim$8-10 THz) confined to the Ge region, (b) Si optical modes ($\sim$14-16 THz) confined to the Si region, and (c) the normalized difference of the two energy regions. (d-f) Silicon substitution defect in graphene showing localized defect modes: (d) Silicon vibrations (10-15 THz) localized to the isolated atom, while (e) Carbon vibrations (40-45 THz) appear throughout the structure. The difference plot in (f) highlights the Silicon impurity. These results demonstrate PySlice's capability to predict spatially-resolved vibrational signals at interfaces and point defects.
• Figure 4: Extended simulation capabilities beyond vibrational spectroscopy. (a) Annular dark field (ADF) imaging validation against established multislice codes. (b) Large-angle convergent beam electron diffraction (LACBED) pattern demonstrating thickness-dependent dynamical diffraction effects. (c) Vortex beam probe with orbital angular momentum, illustrating PySlice's support for arbitrary wavefront profiles including chiral-sensitive illumination modes.