DreamLufei/invDesMobility-alignn-bandgap-nonmetal

InvDesMobility ALIGNN Bandgap/Nonmetal Screening Model

This repository contains the ALIGNN model used in InvDesMobility for surrogate bandgap prediction and nonmetal screening before expensive first-principles mobility validation.

Paper and Repositories

• Paper: InvDesMobility: a reliability-gated first-principles feedback framework for closed-loop materials discovery
• arXiv: https://arxiv.org/abs/2606.16133
• Project page: https://dreamlufei.github.io/invDesMobility/
• GitHub: https://github.com/DreamLufei/invDesMobility
• Loop repository: https://github.com/DreamLufei/invdesmobility_loop
• Mobility workflow: https://github.com/DreamLufei/2d-mobility
• Zenodo: https://doi.org/10.5281/zenodo.20475023

Files

• last_model.pt: trained ALIGNN checkpoint used by the screening script.
• config.json: required ALIGNN model and graph-construction configuration.

config.json is not a disposable log file. It defines the ALIGNN architecture and inference graph parameters needed to reconstruct the model before loading the checkpoint.

Intended Use

Use this model to rank/filter generated 2D candidate structures by predicted bandgap/nonmetal behavior inside the InvDesMobility screening pipeline.

Training Data

The model was trained as a surrogate bandgap model for the project screening workflow. See the invDesMobility repository and the companion data archive for the associated dataset preparation scripts and records.

Training Parameters

The included config.json records the training and inference settings, including ALIGNNAtomWise architecture, cutoff, neighbor strategy, batch size, optimizer, learning rate, and split ratios.

Evaluation

Standalone training/validation/test prediction tables for this bandgap surrogate are not duplicated in this model repository. In InvDesMobility it is used as an intermediate nonmetal/bandgap screen, and the corresponding screening outputs are distributed in the companion DreamLufei/invDesMobility-data archive.

The model should be audited through those screening records together with the downstream deterministic DFT validation results, rather than treated as a final electronic-structure benchmark.

Limitations

This is a surrogate screening model. Final electronic structure and carrier mobility claims require downstream DFT validation. The model should not be used as a replacement for first-principles band-structure calculations.