About

Some quant chem calculation tests using open source PSI4 package.
"proof of concept". Among that, to use a smiles to XYZ format instead of starting from XYZ as is the input format for PSI4.

heat of formation

Initial idea was to calculate energy of a compound to then calculate an approximation of reaction energy (delta) as a means to check if a reaction is endo- or exothermic (a chemical process safety background).

Accuracy on this level unfortunately not sufficient, maybe for simple mols like water, where I am close to the reported heat of formation. On the other hand already with e.g.
nitrobenzene + H2 -> aniline + water
the result is log units off .... so one can say that the package works, but some of the approaches (e.g. choosen parameters) are not always correct.

torsional energy profile

Another example (in the jupyter notebooks) are simplistic torsional energy calc/visualization, purpose to check the profile of a unhindered compound versus a atropisomeric compound.
The absolute values aren't correct? Also here, a q of choosen parameters....

Installation

had some issues with installation, e.g.:
tested, but didn't work:

conda create -n psi4_env python=3.7
conda activate psi4_env
conda install psi4 psi4
conda install -c conda-forge rdkit

a second attempt:

conda create --name psi4_env python=3.11
pip install psi4-step
pip install rdkit

did not work since not supported in 3.11 (for now)

finally, what worked:

conda create --name psi4_env python=3.10
conda install psi4 psi4
pip install rdkit

(jupyter also required for the notebooks)

temp files

regularly check you tmp files folder (or add an autodelete function), some of the files can get quite big (giga bytes of data).
in linux it's /tmp/psi*

references

e.g., but not exclusively: https://pubs.acs.org/doi/10.1021/jo1012898
code wise, using in one case a module by Steven Kearnes, license 3-clause BSD, check the conformers.py file for details.

license

MIT - bascially, if anyone finds this interesting they can do what they want with it (see license.md for details), after all, most of this is based on other open-source stuff.

Additions/Changes

added new logger functions to simplify/clarify future snippets. Demonstrated so far only in the test_xxxxxx.py files.
! as I slowly learned... there is a function psi4.set_output_file("output.dat", True) which can take care of all that out of the box.
Also a scratch function psi4.core.IOManager.shared_object().set_default_path("/scratch").
Added wfn.energy to the detla_H_formation_and_reaction_energy.py script. Shows that I don't really know the details of the Psi4 package outputs....