pynof 1.4

Package for performing Piris Natural Orbital Functionals (PNOF)

PyNOF

Citation

Formally read as Python-Natural-Orbital-Functionals, PyNOF is based on the original DoNOF software written in Fortran by Prof. Mario Piris, but it takes advantage of the Python capabilities such as optimizers, vectorization via numpy and gpu compatibility via cupy.

Installation

We recommend to perform the installation inside an Anaconda enviroment:

conda create -y -n pynof
conda activate pynof

PyNOF uses Psi4 for integrals, so it is necessary to install it first:

conda install -y psi4 -c conda-forge/label/libint_dev -c conda-forge

then, you can simply install PyNOF using pip

pip install pynof

[Optional] Integrals transformations can benefit from a GPU. If available, just install cupy

conda install -y -c conda-forge cupy

Example

A pynof input has the following parts:

• Import pynof
• A declaration of the molecule geometry
• A parameter object
• The calculation instruction

import pynof

mol = pynof.molecule("""
0 1
  O  0.0000   0.000   0.116
  H  0.0000   0.749  -0.453
  H  0.0000  -0.749  -0.453
""")

p = pynof.param(mol,"cc-pvdz")
p.autozeros()

p.ipnof=8

p.RI = True
#p.gpu = True

E,C,gamma,fmiug0 = pynof.compute_energy(mol,p,hfidr=True)

If everything worked, the job may be executed by

python -u test_energy.py

Note. The first run may be slightly slow due to jit precompilation.

For development

First, clone PyNOF from github and change to the project directory

git clone https://github.com/felipelewyee/PyNOF.git
cd PyNOF

In the PyNOF folder, execute the following code

conda create -y -n pynof
conda activate pynof
conda install -y psi4 -c conda-forge/label/libint_dev -c conda-forge
conda install -y -c conda-forge cupy # Optional
pip install build
python -m build && cd dist && pip install pynof*.whl && cd ..

Authors

The PyNOF code has been built by Juan Felipe Huan Lew Yee, Lizeth Franco Nolasco and Iván Alejandro Rivera under supervision of Jorge Martín del Campo Ramírez and Mario Piris.