This repository allows the usage of 3T structure optimization / energy minimization algorithm on solid surface/liquid (multi molecule) interface. Depending on your system setup, this can effectively be used for bulk liquid system as well. The 3T structure transformation is performed within PyTorch, while the structure energy evaluation is done using an external software outside of PyTorch such as VASP, etc. In fact we also use PyTorch as an 'external calculator' to calculate classical force field energy in this repository.
Start by setting up your conda environment and Python dependencies:
conda create --name 3T python=3.8
conda activate 3T
conda install --file requirements.txt -c pytorch -c conda-forge -c rdkit
Install the conda dependencies separately if necessary.
In addition to the python dependencies in requirements.txt, you should also ensure that Gromacs is properly installed. We suggest installing gromacs from source. For example, follow these download and installation instructions:
https://manual.gromacs.org/documentation/2021.3/download.html
https://manual.gromacs.org/2021.3/install-guide/index.html
Or on CentOS, simply do:
yum -y install gromacs
We only use Gromacs for 3T-FF force field assignment and structure data creation, not for MD. If the molecule force field parameters are already cached in the cache folder, you do not need Gromacs.
Ensure that the command python refers to the python library of your conda environment. This is not always the case. For example, this may not be true in centOS image in Tencent Cloud with VASP installed. In my case, I need to do:
conda activate 3T
alias python='/opt/intel/oneapi/intelpython/latest/envs/3T/bin/python3.8'
Finally, after installing these python libraries and Gromacs, you should install the Gromacs-LAMMPS file format converter. The InterMol library is taken from the InterMol Github page https://github.com/shirtsgroup/InterMol, but has been modified to fix some bugs related to Gromacs-LAMMPS file conversion (so using the original Github's code with 3T won't work).
If you are interested, the modification is primarily done in the utils/Convert_Gromacs_LAMMPS/InterMol/intermol/gromacs/grofile_parser.py file.
Note that InterMol is only used for preparing new 3T molecule data GL_data object. Because this 3T codebase uses molecule force field caching, you do not need InterMol if you are working with just old molecules that you have already cached in your database. See utils/process_molecule.py function convert_molecule.
cd utils/Convert_Gromacs_LAMMPS/InterMol
python setup.py build
python setup.py install
cd ../../..
Assuming that you are using this codebase for its VASP-related functionalities, please ensure that your VASP installation is valid, and that the correct command is utilized in the file utils/calculator_3T_VASP.py function run_VASP. For Tencent TEFS VASP 6.2.1 system, the command we use to make external system call to the VASP software is:
os.system('nohup mpirun -n '+n_gpu+' --allow-run-as-root ~/software/vasp.6.2.1/bin/vasp_std')
Please change this Python system call into something specific to your VASP computing environment.
Please ensure that the following commands are valid inside your 3T conda environment: gmx, wget, unzip, and packmol.
At this point, you are ready to run a test example.
For paper example on bulk electrolyte reduction, run the command:
python randomize_3T_bulk_electrolyte_reduction.py
For paper example on bulk electrolyte oxidation, run the command:
python randomize_3T_bulk_electrolyte_oxidation.py
For trajectory analysis and result visualization of running results, please refer to the Examples folder.
On a separate terminal, you can track the progress of the cycles by checking the content of the log file default.log, which is updated in real-time for both the classical force field and VASP components of the structure minimization.
Questions about this repository may be addressed to Jonathan Mailoa ( jpmailoa [AT] alum [DOT] mit [DOT] edu ).