CataLiUst

Welcome to the CataLiUst initiative!

We develop DFT datasets and tools, and train machine learning models for use in computational catalysis and surface chemistry research.

---

Datasets

Here you will find the datasets we have released so far. Each entry includes a short description, the type of data, the level of theory, a link to Zenodo where the dataset is hosted, the repository containing code for using the dataset, and the publication introducing the dataset.

If you use a dataset, please cite the corresponding publication and acknowledge CataLiUst.

Name	Description	Data type	Level of theory	Zenodo DOI	Repository	Publication
CataLiUst Ti2C-MXene	Molecules (CO2, HCOOH, H2O, O2, H2) on Ti2C MXenes	Formation energies and forces	DFT (rev-vdW-DF2 1)	doi.org/10.5281/zenodo.17652071	equimodel-mxene	P. Melnyk, et al., Equivariant Modelling for Catalysis on 2D MXenes, EurIPS 2025 Workshop on SIMBIOCHEM

---

Machine Learning Models

Here you will find the machine learning models we have trained so far. Each entry may include multiple models that differ, for example, in whether they were pretrained or trained from scratch, and in the number of parameters (i.e., model size).

For each model set, we list the dataset used for training, the model type, the underlying architecture, the Zenodo DOI where the trained model files are hosted, the repository containing code for using the models, and the publication introducing the work.

If you use a model, please cite the corresponding publication and acknowledge CataLiUst.

Trained on dataset	Model type	Architecture	Zenodo DOI	Repository	Publication
CataLiUst Ti2C-MXene	MLIP	EquiformerV22	doi.org/10.5281/zenodo.17654434	equimodel-mxene	P. Melnyk, et al., Equivariant Modelling for Catalysis on 2D MXenes, EurIPS 2025 Workshop on SIMBIOCHEM

---

Citation

Please cite the associated publications when using data or models.

Footnotes

1. I. Hamada. van der Waals density functional made accurate. Phys. Rev. B 2014, 89, 121103(R). ↩

2. Y.-L. Lioa, B. Wood, A. Das, and T. Smidt. EquiformerV2: Improved equivariant transformer for scaling to higher-degree representations. ICLR 2024 ↩