Sovereign 8s.is mirror of standardgalactic/Struct2Flat
Find a file
Repository files (latest commit first)
Filename Latest commit message Latest commit date
2026-02-13 15:34:58 +00:00
data Add files via upload 2025-06-27 22:33:17 +01:00
model Add files via upload 2026-02-13 15:34:58 +00:00
results Add files via upload 2025-06-29 13:37:36 +01:00
scripts Add files via upload 2025-05-16 15:03:08 +01:00
LICENSE Initial commit 2025-05-14 20:04:15 +01:00
README.md Update README.md 2025-05-17 10:48:10 +01:00
requirements.txt Add files via upload 2025-05-15 19:46:02 +01:00

Struct2Flat

This repository contains the code, datasets, and candidate results for the structure-informed discovery of flat bands in two-dimensional (2D) materials. Our framework integrates a physically motivated flatness score, combination of band dispersion and density of states, with a multi-modal deep learning model trained on atomic structures and compositional descriptors. Without requiring electronic band structure inputs, the model enables scalable and interpretable prediction of flat-band candidates across large 2D material databases.

Features

Physically motivated flatness score combining band dispersion and DOS sharpness Multi-modal model using graph-based encoders and text-based descriptors Fully structure-only prediction — no DFT band structure required at inference time Applied to predict 10,000+ materials from C2DB Includes DFT validation and topological analysis for selected candidates

Repository Structure

Struct2Flat/

data/ # Processed and labeled datasets

model/ # Deep learning architecture

scripts/ # Flatness score computation, kagome-like and atom projection screening

results/ # Candidate summaries and plots

requirements.txt # Environment of the deep learning model

README.md

Data

This repository includes only the processed datasets with algorithm-labeled flatness scores used for model training and inference.

The original raw data can be accessed from the official databases:

2DMatPedia: http://www.2dmatpedia.org

C2DB: https://cmr.fysik.dtu.dk/c2db/c2db.html

Getting Started

  1. Clone the repository

  2. git clone https://github.com/your-username/Struct2Flat.git

cd Struct2Flat

  1. Set up the environment

pip install -r requirements.txt

  1. Train the model

All parameters and file paths can be configured in configs.py.

cd model

python train.py

  1. Predict with the model

python prediction.py

We provided a trained model in results folder.

  1. Run inference for c2db

python predict_c2db.py

Results

Over 10,000 2D materials screened

Selected candidates form a new flat-band database

Several top candidates validated with DFT and topological quantum chemistry

Results and figures from the paper are available under /results/.

Citation