PyADF - A Scripting Framework for Multiscale Quantum Chemistry

The recommended way to install PyADF is to install it within a conda environment. You should also be able to manually add the src-folder to your $PYTHONPATH and the src/scripts-folder to your $PATH. In that case you will have to install the (optional and mandatory) requirements manually.

To install PyADF within a new conda environment with full functionality, follow these steps:

1. Create and activate a new conda environment:

conda create -n your_env_name
conda activate your_env_name

2. Install numpy and other dependencies via conda:

conda install -c conda-forge numpy scipy xcfun pyscf openbabel rdkit

3. Install PyADF from the top level folder (where the setup.py resides):

pip install .

4. Now you should be able to run the pyadf and test_pyadf scripts from the conda environment where you installed PyADF. There are some command line options which you can examine using the following commands:

pyadf --help
test_pyadf --help

There are two types of requirements for PyADF:

1. PyADF is written in Python and depends on some Python-packages and libraries.

2. PyADF is a scripting framework that can be used with a wide range of quantum chemistry software packages.

scipy and numpy are necessary to start the pyadf-script. xcfun and pyscf are needed for some calculations that PyADF can perform directly, especially embedding calculations. openbabel and rdkit are needed for extended functionality while handling molecules, but a molecule class that relies on neither exists as well. The src/pyadf/test/test_env.yml-file tells you which versions of the Python dependencies the provided version of PyADF was most recently tested against.

Similarly, the src/pyadf/test/test_pyadf.conf-file tells you which versions of the QC dependencies the provided version of PyADF was most recently tested against. The usage of such files is optional and depends on the Environment Modules package. A default configuration file can be provided in your home folder as ~/.pyadfconfig.

The file also includes some settings regarding the parallelization with an MPI im- plementation. Please correspond the manuals of the QC packages to set up your own MPI implementation. Often, the $LD_LIBRARY_PATH-variable has to be set for MPI to work.

Irrespective of whether you use the Environment Modules package and the jobrunnerconf functionality for this, you have to set some environment variables for each of the programs to function properly with PyADF. PyADF often uses these variables to find the scripts via explicit paths rather than relying on the $PATH-variable:

You should set the variables AMS needs (cf. the AMS documentation) like $AMSHOME, $AMSBIN, $AMSRESOURCES etc. PyADF directly depends on $AMSBIN to find the executables. Some tests rely on $AMSRESORCES.

You should set the variables DALTON needs (cf. the DALTON documentation) like $DALTONHOME, $DALTONBIN etc. PyADF directly depends on $DALTONBIN to find the executables.

You should set the variables DIRAC needs (cf. the DIRAC documentation) like $DIRACHOME, $DIRACBIN etc. PyADF directly depends on $DIRACBIN to find the executables.

For MOLCAS, the main folder has to be set as $MOLCAS. The bin folder within the main folder has to be added to the $PATH-varibale as the current implementation depends on the $PATH-variable.

NWChem depends on the $NWCHEMBIN-variable explicitly. Check the NWChem documentation whether variables like $NWCHEM_NWPW_LIBRARY, $NWCHEM and $NWCHEM_BASIS_LIBRARY need to be set as well as whether $NWCHEMBIN has to be added to the $PATH-variable.

The main folder of ORCA is usually added to the $PATH-variable. However, PyADF utilizes the $ORCA_PATH-variable. The MPI implementation is set via the $OPAL_PREFIX.

Has already been covered under Python Requirements and is currently not supported in the same way the other QC packages are.

Quantum ESPRESSO in PyADF depends on the $QEBINDIR-variable which has to be set on the bin directory in the main directory of Quantum ESPRESSO. Quantum ESPRESSO generally uses the variables $BIN_DIR, $PSEUDO_DIR and $TMP_DIR, with the first two depending on the path of the installed files.

PyADF currently utilizes the $PATH-variable to start the different scripts for pure TURBOMOLE calculations. So the $TURBOBIN-variable, which is used for SNF calculations, has to be added to the $PATH-variable. The $TURBODIR-variable is always needed for the installation to work.

The general workflow for PyADF relies on input scripts which are themselves written in Python and are run through pyadf:

pyadf --save example_script.pyadf

You can try running the full (some tests depend on OpenBabel and only run with OpenBabel installed) set of tests provided with PyADF by using:

test_pyadf --keep --timings --save --molclass=openbabel --tests=all --jobrunnerconf=YOURPATHTOPYADF/pyadf/test/test_pyadf.conf

Whether you can/need to use the jobrunnerconf-option depends on how you install and manage your QC software (see section QC dependencies).

A single test can be run by using:

test_pyadf --kep --save --singletest=Orca_SinglePoint_Dipole

see also:

pyadf --help
test_pyadf --help

The test-inputs which you can find in (src/pyadf/test/testinputs/SOME_TEST/SOME_TEST.pyadf) can give you further hints regarding the features of PyADF.