OGCNN This is the repository for our work on property prediction for crystals. In this work we have used ideas from the Orbital Field matrix and Crystal Graph Convolutional Neural Network to predict material properties with a higher accuracy. Paper link:https://journals.aps.org/prmaterials/abstract/10.1103/PhysRevMaterials.4.093801

The two important papers referenced for this work are:

1. Crystal Graph Convolutional Neural Networks for an Accurate and Interpretable Prediction of Material Properties (https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.120.145301)
2. Machine learning reveals orbital interaction in crystalline materials, Science and Technology of Advanced Materials Volume 18, 2017 - Issue 1.(https://www.tandfonline.com/doi/full/10.1080/14686996.2017.1378060) We used the ideas from these papers and did some of our modifications to develop the OGCNN which gives a higher performance than the seminal work of CGCNN

To run the OGCNN code the following packages are required

• PyTorch
• scikit-learn
• pymatgen. It is preferable to install this package via pip
• ase It is advised to create a new conda environment and then install these packages. To create a new environment please refer to the conda documentation on managing environments (https://docs.conda.io/projects/conda/en/latest/user-guide/tasks/manage-environments.html)

To input crystal structures to OGCNN, you will need to define a customized dataset. Note that this is required for both training and predicting. The dataset that we use for this work are in the cif format.

• CIF files recording the structure of the crystals that you are interested in
• The values of the target properties for each crystal in the dataset.

You can create a customized dataset by creating a directory root_dir with the following files: You can create a customized pre-training dataset by creating a directory root_dir with the following files:

1. id_prop.csv: a CSV file with two columns. The first column recodes a unique ID for each crystal, and the second column recodes the value of target property. If you want to predict material properties with predict.py, you can put any number in the second column. (The second column is still needed.)

2. atom_init.json: a JSON file that stores the initialization vector for each element. An example of atom_init.json is data/sample-regression/atom_init.json, which should be good for most applications. The atom_init.json file has some of the basic atomic features encoded. Please refer the supplementary information of the paper to find out more about the basic atomic features.

3. atom_init.json: a JSON file that stores the initialization vector for each element. An example of atom_init.json is data/sample-regression/atom_init.json, which should be good for most applications. The atom_init.json file has some of the basic atomic features encoded. Please refer the supplementary information of the paper to find out more about the basic atomic features.

4. ID.cif: a CIF file that recodes the crystal structure, where ID is the unique ID for the crystal.

5. ID.cif: a CIF file that recodes the crystal structure, where ID is the unique ID for the crystal.

The structure of the root_dir should be:

root_dir
├── id_prop.csv
├── atom_init.json
├── id0.cif
├── id1.cif
├── ...

There are two examples of customized datasets in the repository: data/sample-regression for regression and data/sample-classification for classification.

For advanced PyTorch users The above method of creating a customized dataset uses the CIFData class in ogcnn.data. If you want a more flexible way to input crystal structures and more feture descriptors to the model, PyTorch has a great Tutorial for writing your own dataset class.

Before training a new CGCNN model, you will need to:

• Define a customized dataset at root_dir to store the structure-property relations of interest.

Then, in the directory that you choose to have main.py, you can train a OGCNN model for your customized dataset by:

python main.py root_dir

You can set the number of training, validation, and test data with labels --train-size, --val-size, and --test-size. Alternatively, you may use the flags --train-ratio, --val-ratio, --test-ratio instead. Note that the ratio flags cannot be used with the size flags simultaneously. For instance, data/sample-regression has 10 data points in total. You can train a model by:

python main.py --train-size 8 --val-size 1 --test-size 1 data/sample-regression

or alternatively

python main.py --train-ratio 0.8 --val-ratio 0.1 --test-ratio 0.1 data/sample-regression

You can also train a classification model with label --task classification. For instance, you can use data/sample-classification by:

Although in the OGCNN work, we have not done any classification tasks. OGCNN similar to CGCNN has a switch to do the classification tasks which can run by using the following commands.

python main.py --task classification --train-size 5 --val-size 2 --test-size 3 data/sample-classification

After training, you will get three files in the same directory as the main.py file.

• model_best.pth.tar: stores the CGCNN model with the best validation accuracy.
• checkpoint.pth.tar: stores the CGCNN model at the last epoch.
• test_results.csv: stores the ID, target value, and predicted value for each crystal in test set.

Before predicting the material properties, you will need to:

• Define a customized dataset at root_dir for all the crystal structures that you want to predict.
• Obtain a pre-trained OGCNN model named pre-trained.pth.tar.

Then, in directory where you have your predict.py script, you can predict the properties of the crystals in root_dir:

python predict.py pre-trained.pth.tar root_dir

For instance, you can predict the formation energies of the crystals in data/sample-regression:

python predict.py pre-trained/formation-energy-per-atom.pth.tar data/sample-regression

After predicting, you will get one file in ogcnn directory:

• test_results.csv: stores the ID, target value, and predicted value for each crystal in test set. Here the target value is just any number that you set while defining the dataset in id_prop.csv, which is not important.

To reproduce our paper, you can download the corresponding datasets following the instruction.

This work was primarily done by Rishikesh Magar,Mohammadreza Karamad and Yuting Shi and was advised by Prof. Amir Barati Farimani, CMU