Skip to content

Latest commit

 

History

History
173 lines (128 loc) · 9.15 KB

README.md

File metadata and controls

173 lines (128 loc) · 9.15 KB
Raw

Material Screening

A python-based screening procedure for both Raspa2 and Zeo++ softwares

Installations

This code relies on 3 softwares that need a prior installation step

Python3.8.5

Download and install Python3.8.5 at https://www.python.org/ Other versions of python3 should be ok, but I have not tested them yet.

Raspa2

Follow the instructions of the Raspa2 Github repository to compile it: https://github.com/iRASPA/RASPA2 Remember the path to the directory containing the bin/ directory containing the binary file simulate

Zeo++

Download and compile Zeo++ at http://www.zeoplusplus.org/download.html Remember the path to the directory containing the compiled network binary file

Glost

Download and compile Glost at https://github.com/cea-hpc/glost It is used to run the glost list generated when using the -X glost of screen.py.

Initial set-up

This part explains how you will need to set-up your working environment

Bash environment

In the set_environment file, replace the example paths by the accurate ones.

RASPA environment

All molecules will be taken from ${RASPA_DIR}/share/raspa/molecules/TraPPE, which must contain at least the definition for helium (as well as all other molecules you need).

If the folder is missing, simply create it. The file for helium can be taken from https://github.com/numat/RASPA2/blob/master/molecules/TraPPE/helium.def.

Any used force field must be present in ${RASPA_DIR}/share/raspa/forcefields. In particular, the default force field UFF should be copied from ${MATSCREEN}/forcefield if not already present.

Python environment

I used a virtual environment for python3.8.5 but it is optional. To set-up a virtual environment and install the required packages in requirements.txt, follow those instructions:

./path_to_python3.8.5/bin/python3.* -m venv path_to_venv
source path_to_venv/bin/activate
pip install -r requirements.txt

If you want to go back to your global python environment type: deactivate

Database of materials and force fields

Force Field

For the forcefield, you can use the default ones in Raspa2 add your own force fields adapted to your applications: You can find in the forcefield directory an exmaple of UFF forcefield built on this paper. You can also build a Dreiding forcefield from this paper paper, usually it is mixed with UFF to complete the missing atoms.

Material Databases

  • CoRE MOF Database: The Computation-Ready, Experimental Metal--Organic Framework (MOF) database is an experimental based dabase described in this paper and can be downloaded via this link (about 14,000 structures)
  • QMOF Database: Quantum MOF (QMOF) database is a DFT-optimized MOFs database paper and can be downloaded via this link (about 14,000 structures)
  • MOFXDB: ToBaCCo and hMOF datasets can be downloaded via this link. The Topologically-Based Crystal Constructor generates automatically MOFs based on the metal, the linker and the topology. The algorithm is described in detail in this paper and the hMOF database is another hypothetical MOF database described in this paper

Zeolite, Porous Polymer Networks and Covalent Organic Framework can also be found as databases.

Running your first simulations

Using screen.py command line tool to generate adsorption or coadsorption simulation files and then run them.

Once the structures have been decided and put in ${RASPA_DIR}/share/raspa/structures/cif/, put their names in a one-column CSV file with "Structures" as its only header. Then, execute screen.py with the -t info option to extract the information necessary to run subsequent simulations. You should then launch ./data.sh to obtain an info.csv file which should replace the one in ${MATSCREEN}/data/.

Workflow example: simulating adsorption in cationic zeolite

Choose your target cationic zeolites and put their CIF files in ${RASPA_DIR}/share/raspa/structures/cif/ Store their name in a one-column CSV file with "Structures" as its only header, like in the previous paragraph. Give that file a name, for instance "foobar.csv".

Then, prepare electrostatic and VdW grids for each structure, to reduce the computation required subsequently: to do so, run

$MATSCREEN/screen.py -ppn X -n X -N 1 -s $MATSCREEN/data/foobar.csv -f FF -x "EXTRA" -m MOLECULES -t grid

where:

  • X is the number of processes to run in parallel.
  • FF is the name of the force field, which should be placed in ${RASPA_DIR}/share/raspa/forcefield.
  • EXTRA is additional command to add in each RASPA INPUT file. For instance, RemoveAtomNumberCodeFromLabel yes is useful if your CIF file numbers each atom. If you do not have any command to add, you can remove the -x "..." flag.
  • MOLECULES is the list of molecules that will be added in the framework. For instance, putting Na CO2 N2 is necessary to simulate coadsorption of CO2 and N2 on zeolites with sodium cation. If Na cations are already part of the framework, they should be removed from the list.

If the cations are not part of the framework, they should be equilibrated. This can be done with parallel tempering using the following command:

$MATSCREEN/screen.py -ppn X -n X -Ni INIT -N RUN -s $MATSCREEN/data/foobar.csv -f FF -x "EXTRA" -m CATION -t pt -T TEMPERATURES -M

where, in addition to the previous replacements:

  • INIT is the number of initialization steps.
  • RUN is the number of running steps (for which the movie will be recorded thanks to the -M option).
  • CATION is the name of the cation.
  • TEMPERATURES is the list of temperatures used for parallel tempering.

Different cation placements can be extracted from the movie, and should be transformed into restart files for later use. If only the last cation placement is required, you can remove the -M flag and directly take the restart file in the Restart folder.

Running ./data.sh (or ./data.jl if julia is available) outputs digests of RASPA's output giving the energy at each recorded step in the Movie, which is useful to check convergence. These are stored in DATAenergy... files.

Once cations have a starting position, the adsorption simulation itself can be run.

  • If the cations are fixed, then the most efficient strategy consists in making a new CIF file containing the cations
  • Otherwise, if cations are mobile, their starting position should be given as a restart file to be put in RestartInitial/System_0. Either one restart file should be made for each appropriate temperature and pressure, or a single restart file can be put whose name should be truncated to a trailing underscore, leaving the temperature and pressure blank. This file will then be copied and the name expanded for each combination of temperature and pressure.

The command to run to perform adsorption simulation is

$MATSCREEN/screen.py -ppn X -n X -Ni INIT -N RUN -s $MATSCREEN/data/foobar.csv -f FF -x "EXTRA" -m MOLECULES -c CODE -t ads -T TEMPERATURES -p PRESSURES

where

  • CODE is a list of 0 and 1, one for each molecule in MOLECULES: 1 indicate that this molecule is the cation, and 0 indicates that it is a gas whose adsorption is measured. If all molecules are gas (the cation is part of the framework), this -c ... flag can be removed.
  • PRESSURES is the list of pressures for which the simulation is carried.

Running ./data.sh (or ./data.jl if julia is available) outputs a digest of RASPA's output giving the loading at each recorded step in the Movie. These are stored in DATAloading... files.

In all cases, if a RASPA computation is interrupted, it can be restarted by moving the Restart subfolder into a new folder, renaming the subfolder into RestartInitial and then re-running the same command with an additional -R flag in this new folder. If there is no RestartInitial in the current folder and a -R command is issued, the Restart folder will be renamed into RestartInitial. Beware that previous simulation contained in the new folder will be overwritten by the new simulation so only restart a simulation in an occupied folder if you do not need these data.

Additionally, any command can generate files instead of directly running by specifying a -X option:

  • -X exe is the default option which runs the computation immediately.
  • -X glost generates a glost input file.
  • -X slurm generates a slurm input file.

Other

Example of usage in job_example.sh

Need help? Execute: screen.py --help.