DIAMOND
Workflows
EN

Installation of aiida-diffusion-wf

Development mode

In the current stage, the package is not ready as a aiida plugin, otr pip package, and can only be installed by source :

git clone https://gricad-gitlab.univ-grenoble-alpes.fr/diamond/aiida/workflows/aiida-diffusion-wf.git
cd aiida-diffusion-wf

Create a python environment with python=3.11 (recommended), activate and install the dependencies in editable mode :

pip install -r requirements.txt
pip install -e .

Download the executable for LAMMPS

In principle, we could use any official executables for LAMMPS provided in the official channel. However, for the sake of reproductivity, we recommend the user to download a a containerized version of LAMMPS through the following command :

apptainer pull lammps-2Apr2025_serial.sif oras://gricad-registry.univ-grenoble-alpes.fr/diamond/apptainer/apptainer-singularity-projects/lammps_serial.sif:test

Setup Aiida

This workflow use the Aiida engine, which allows to submit calculations in remote HPC clusters. For simplicity of installation, we recommend to use the containerized version of aiida-core. This can be done following this tutorial in the DIAMOND website.

You have to follow all these steps in order to be able to run the workflows provided here :

  1. Setup a Aiida profile
  2. Setup a computer and configure it.
  3. Setup a code and configure it.

Hopefully, this is only done once.

Some examples of configuration files are given in this repository for computers (to manage scheduler in remote machine and transfer protocols) and codes (to manage the execution of a containerized version of LAMMPS in the remote machine).

For example, one can work with the aiida_raspa.sif image found in DIAMOND website.

Example : Run a single RAW calculation with pre-defined parameters

In this example, the material is pre-defined by providing the location of the CIF file containing 3D coordinates : examples/structures/0000[Ag][sra]3[ASR]1. The gas adsorbate is a monoatomic gas, the krypton (set by atom_symbol). The workflow automatically searchs for Lennard-Jones parameters in UFF forcefield in src/uff_non_bonded.py.

Tutorial
  1. Create a working directory in which the workflow script will be saved:
mkdir workdir_examples && cd workdir_examples
cp $PACKAGE_DIR/examples/workflows/run_raw_diffusion_Kr.py ./

Note that the environment variable $PACKAGE_DIR should be defined as the root path to the cloned package.

  1. Create an instance of the aiida-core image
apptainer instance start \
        --containall \
        --bind $PWD:$PWD \
        -B ~/.aiida:/.aiida \
        -B ~/.aiida/.ssh:$HOME/.ssh \
        -B ~/.aiida/postgres_run:/var/run/postgresql \
        -B ~/.aiida/rabbitmq/var/lib/rabbitmq:/var/lib/rabbitmq \
        -B ~/.aiida/rabbitmq/var/log:/var/log/rabbitmq \
        ~/apptainer-images/aiida_raspa.sif aiida

Note : You may need to adapt the paths of your aiida configuration (to link the aiida database). The aiida-core program is already contained in aiida_raspa.sif image.

  1. Run an instance to have access to aiida commands and predefined codes and computers
apptainer shell instance://aiida
  1. Activate the conda or python environment
source $PACKAGE_DIR/aiida-diffusion-wf/bin/activate

One need to adapt the path depending on where the python environment have been created. Note that this path must be accessible inside the aiida container, it can be useful to properly create the instance in the previous step in a appropriate location and using --bind $PWD:$PWD to acces to the python environment.

Note : in a future implementation, we will package the whole workflow in a single apptainer image (instead of using aiida_raspa, so that the libraries can be directly imported inside the container.

  1. Set the code name and the computer name in the workflow You will need to change manually these variables profile_name, code_name, project_namein the python script file :
nano-tiny run_raw_diffusion_Kr.py
  1. Run the workflow
python run_raw_diffusion_Kr.py

One need to change these lines at the end of the pthon script if we want to directly run the workflow (for testing in local computer) or if we want to submit the calculation ot a remote HPC compute :

# For local scheduler, the following line should be uncommented
#results, node = engine.run_get_node(builder)

# For remote scheduler, the following line should be uncommented
node = engine.submit(builder)

To see the current state of the calculation, one can use the following commands :

verdi process list # all running processes
verdi process list -a # all processes
Visualize the MD trajectory

As the simulations use a rigid framework, one may want to recover the total trajectory with all atoms (host + adsorbate).

  1. First dump the data repository
verdi node show <pk_process> # identify the pk of the retrieved node
verdi node  repo dump <pk_retrieved_node> repo_dump
cd repo_dump
  1. Concatenate the adsorbate trajectory with the rigid framework
python  $PACKAGE_DIR/src/aiida_diffusion_wf/merge_data_traj.py

A extended XYZ file containing the trajectory of all atoms and the cell information is generated and can be further opened by Ovito.

  1. Visualization with Ovito

diffusion ovito gif

A few steps of Krypton diffusion in 0000[Ag][sra]3[ASR]1 MOF