This repository includes the python scripts to generate the surface/molecule descriptors and to run the developed iterative ML algorithms. Each folder has the main codes and an example.
- Prepare the molecule structures. Here we recommend to use the ASE to build the structure with the .traj format, other formats also work, such as .cif, .xyz.
- See the example to get the reordered Coulomb Matrix. ($python run_test.py ). The descriptors will be saved in a .csv file.
The extractDos.py file is in the path of /data/scripts_py.zip, this file is used to extract the DOS of active site.
- DFT calculated DOS was saved in the .pickle file.
- The site name and atom index on clean surface are required. ($python run_get_data.py) The descriptors are also saved in a .csv file.
The Iteration*.py files are in the folder of Iteration_I_AdsorptionE , all these scripts work for training process. We also show an example in this folder.
-
Max predicted error: a parameter to control the accuracy and number of DFT calculations.
-
$python run_test.py
The example is constructing the phase diagrams including ML phase diagrams and DFT phase diagrams. Please see the folder of Iteration_II_PhaseDiagram. Each folder have the main scripts and an example. The algorithm can construct the N dimensional phase diagrams (1<= N <=4).
- atomType_chemPot : a dictionary to specify the atomic type and the range of chemical potential.
- N_int : Resolution of phase diagram
- $python run_pd.py
The data folder includes the structure file (.cif) and python script files.
Paper: Scalable approach to high coverages on oxides via iterative training of a machine-learning algorithm (ChemCatChem)
More questions: [email protected]