User's Guide for DeepPseudopot

Welcome! This site centralizes setup guides, reference details, and troubleshooting tips for running DeepPseudopot in production workflows.

DeepPseudopot is a machine-learned atomistic pseudopotential model that extends the semi-empirical pseudopotential method (SEPM) for simulating large and complex material systems.

It excels at capturing the electronic structure, photophysics, and charge-carrier dynamics in systems where ab initio methods such as GW or hybrid-functional DFT become computationally prohibitive — particularly in nanostructures, alloys, and polymorphic materials.

How to Cite

Please cite the following paper when referencing DeepPseudopot:

• Lin, K., Coley-O’Rourke, M.J. & Rabani, E. Deep-learning atomistic semi-empirical pseudopotential model for nanomaterials. npj Comput Mater 11, 381 (2025). https://doi.org/10.1038/s41524-025-01862-5

Table of Contents

• Environment & Installation
• Workflow Modes – Supported training and refinement workflow modes, plus a step-by-step execution flow.
• Input Data Description – Checklist and pointers for every configuration, lattice, and spectral input.
• Output Data Description – Catalogue of initialization, checkpoint, and final deliverables.
• Troubleshooting Guide – Quick answers to the most common failure modes.