CSPML is a unique methodology for the crystal structure prediction (CSP) that relies on a machine learning algorithm (binary classification neural network model). CSPML predicts a stable structure for any given query composition, by automatically selecting from a crystal structure database a set of template crystals with nearly identical stable structures to which atomic substitution is to be applied. Pre-trained models are used to select the template crystals. The 33,153 stable compounds (all candidate crystals; obtained from the Materials Project database) and the pre-trained models are embedded in CSPML.

For more details, please see our paper: Crystal structure prediction with machine learning-based element substitution (Accepted 3 May 2022).

• pandas version = 1.3.3
• numpy version = 1.19.2 # tensorflow is compatible with numpy=<1.19.2 (01/14/2022).
• tensorflow version = 2.6.0
• pymatgen version = 2020.1.28
• xenonpy version = 0.4.2 (see this page for installation)
• torch version = 1.10.0 # peer dependency for xenonpy.
• matminer version = 0.6.2 (optional; for calculating the structure fingerprint with local structure order parameters)

1. First install the dependencies listed above.

2. Clone the CSPML github repository:

git clone https://github.com/Minoru938/CSPML.git

Note: Due to the size of this repository (about 500MB), this operation can take tens of minutes.

3. cd into CSPML directory.

4. Run jupyter notebook and open tutorial.ipynb to demonstrate CSPML.

• Python 3.8.8
• macOS Big Sur version 11.6

1. [Materials Project]: A. Jain, S. P. Ong, G. Hautier, W. Chen, W. D. Richards, S. Dacek, S. Cholia, D. Gunter, D. Skinner, G. Ceder, et al., Commentary: The materials project: A materials genome approach to accelerating materi- als innovation, APL materials 1 (1) (2013) 011002.

2. [XenonPy]: C. Liu, E. Fujita, Y. Katsura, Y. Inada, A. Ishikawa, R. Tamura, K. Kimura, R. Yoshida, Machine learning to predict quasicrystals from chemical compositions, Advanced Materials 33 (36) (2021) 2170284.

3. [Local structure order parameters]: N. E. Zimmermann, A. Jain, Local structure order parameters and site fingerprints for quantification of coordination environment and crystal structure similarity, RSC Advances 10 (10) (2020) 6063–6081.