diffcrysgen 0.1.1

"DiffCrysGen is a score-based diffusion model for accelerated design of diverse inorganic crystalline materials."

DiffCrysGen

DiffCrysGen is a score-based diffusion model. It treats the entire materials representation with a single, unified diffusion process, allowing a single denosing neural network to predict a holistic score for the entire noisy crystal data. This unified treatment significantly simplifies the architecture and improves the computational efficiency.

Generative diffusion framework in DiffCrysGen

---

Architecture of the denoising neural network

---

Installation

Prerequisites

The package requires specific environments and dependencies. Using a virtual environment is highly recommended. Environment Setup using Conda

conda create -n diffcrysgen python=3.11
conda activate diffcrysgen

Install from PyPI

pip install diffcrysgen

Install from Source Code

git clone https://github.com/SouravMal/DiffCrysGen.git
cd DiffCrysGen
pip install -e .

Quick Start

For a simple walkthrough of generating materials and analyzing them, see the DiffCrysGen Demo Notebook.

License

This project is licensed under the MIT License.

See the LICENSE file for details.

Developed by: Sourav Mal at Harish-Chandra Research Institute (HRI) (https://www.hri.res.in/), Prayagraj, India.

Citation

Please consider citing our work if you find it helpful:

@misc{mal2025generativediffusionmodeldiffcrysgen,
      title={Generative Diffusion Model DiffCrysGen Discovers Rare Earth-Free Magnetic Materials}, 
      author={Sourav Mal and Nehad Ahmed and Subhankar Mishra and Prasenjit Sen},
      year={2025},
      eprint={2510.12329},
      archivePrefix={arXiv},
      primaryClass={cond-mat.mtrl-sci},
      url={https://arxiv.org/abs/2510.12329}, 
}

Contact

If you have any questions, feel free to reach us at: Sourav Mal [e-mail]