Moved to MolecularMaterials/nfp! Please follow the link to the new page for the latest updates.

MPNN-Mo2C

Python codes and graph data as described in:
"Accelerating the Evaluation of Crucial Descriptors for Catalyst Screening via Message Passing Neural Network"
Hieu A. Doan, Chenyang Li, Logan Ward, Mingxia Zhou, Larry A. Curtiss, and Rajeev S. Assary.

Dependencies

• Python (3.6.6)
• Numpy (1.19.4)
• Tensorflow (2.5.0)
• Pandas (1.1.5)
• Scikit-learn (0.23.1)
• Seaborn (0.11.1)
• Matplotlib (3.3.1)

Instructions

Preprocess .graphml files and create an input dataframe

• Unzip notebooks/Oads_Mo2C_graphml.tar.gz that contains all .graphml files of adsorption geometries:
  tar -xvf Oads_Mo2C_graphml.tar.gz
• Run notebooks/Oads_Mo2C_catalysts_PreprocessGraphStructure.ipynb to convert graph data into a dataframe input for MPNN model

Train the model and make predictions on the test set

• Run notebooks/Oads_Mo2C_catalysts.ipynb

Additional instructions for analyzing Atomic Simulation Environment (ASE) database and generating graphs in .graphml format

If you want to experiment with building your own adsorption graphs from the VASP outputs used in the paper, please follow the steps below:

1. Download the data in ASE database format (.db) from the Materials Data Facility here
2. Run notebooks/DescriptorGen-networkxGraph-pristine-Mo2C.ipynb and notebooks/DescriptorGen-networkxGraph-doped-Mo2C.ipynb with the downloaded "Oads_Mo2C_pristine_MDF.db" and "Oads_Mo2C_doped_MDF.db", respectively