lxf-gzu / dmcp

DMCP:DFT-based Machine learning method for Catalysis Program

DMCP is aimed to implement DFT-based and Machine-learning-accelerated (DFT-ML) scheme for intricate system catalysis. It is possible to predict the catalytic performance of intricate system catalysis such as double-atom catalysts (DACs) and to reveal the intrinsic descriptors which determine the underlying catalytic activity of them with appropriate algorithm and train data features.

Developer:

DMCP is developed within Prof. Yuzheng Guo's group in Wuhan University, in colloboration with Dr. Zhaofu Zhang in Cambridge University. Core developer: Xuhao Wan, Zhaofu Zhang, Wei Yu, and Yuzheng Guo Email: [email protected], [email protected], [email protected]

Major Features

1. Ten machine learning algorithms: GBR, KNR, SVR, GPR, FNN, RFR, ETR, KRR, LASSO, and ENR.
2. Multiple methods to improve model accuracy: dataset split, cross validation, repeated trails.
3. Visualization module for research.

Prerequisites

1. Generally, you need some data obtained from DFT calculations such as VASP, QE, and CP2K or available material database.
2. DMCP requires Python 3 with the packages specified in requirements.txt. This is taken care of by pip.

Citation

If you use DMCP in your research, please cite the following paper:

1. X. Wan, Z. Zhang*, W. Yu, Y. Guo*, A State-of-the-art Density-functional-theory-based and Machine-learning-accelerated Hybrid Method for Intricate System Catalysis. Submitted, 2021.

Reference

Our work applied DMCP are listed as following:

1. X. Wan, Z. Zhang*, H. Niu, Y. Yin, C. Kuai, J. Wang, C. Shao, Y. Guo*, Machine-Learning-Accelerated Catalytic Activity Predictions of Transition Metal Phthalocyanine Dual-Metal-Sites Catalysts for CO2 Reduction. Submitted, 2021.
2. H. Niu#, X. Wan#, X. Wang, C. Chen, J. Robertson, Z. Zhang*, Y. Guo*, Single-Atom Rhodium on Defective g-C3N4: A Promising Bifunctional Oxygen Electrocatalyst. ACS Sustainable Chem. Eng., 9(9), 3590-3599, 2021.