SCF Guess Project

Contents

• Prerequisites
• Installing
• Usage
• API
• Acknowledgements

Prerequisites

To use this repository you will need to have the following installed:

• git
• Python 3

Both of these dependencies are very common dependencies and are likely available in your operating system's package repository (e.g., Aptitude on Ubuntu)

Installing

To obtain and install this repo in a Linux-like environment run the following commands in a terminal:

# Obtain the source code and enter the directory
git clone https://github.com/rmrresearch/SCF-Guess-Project.git scf_guess
cd scf_guess

# Create a virtual Python environment to keep our dependencies isolated from the
# system (also ensures reproducibility)
python3 -m venv venv
source venv/bin/activate

# Install this repository's dependencies
pip install -r requirements.txt

The above assumes that git and python3 are installed in a location included in your path. If they are not you will need to specify the full path to the git and python3 executables.

Usage

Once installed this repo is designed to be usable as a Python module by another Python script or Jupyter notebook. For example, put the following inside a script h2_aug_cc_pvdz.py

from wrap_pyscf import *

# Define the molecule in Cartesian coordinates
mol = [("H", 0.0, 0.0, 0.0),
       ("H", 0.0, 0.0, 0.89)]

# The name of the atomic basis set
bs  = "aug-cc-pvdz"

# Compute the density of the molecule at the SCF/aug-cc-pVDZ level of theory
rho = density(mol, bs)

# Test that the density is converged (return the number of iterations needed to)
# converge the SCF using rho as a guess)
nitr = run_scf(mol, bs, rho)

print(nitr) # This should print out 1 (1 cycle is needed to test convergence)

This script is then run as:

python3 h2_aug_cc_pvdz.py

and the output should be something along the lines of (all but the last line are output from PySCF's SCF run):

******** <class 'pyscf.scf.hf.RHF'> ********
method = RHF
initial guess = minao
damping factor = 0
level_shift factor = 0
DIIS = <class 'pyscf.scf.diis.CDIIS'>
diis_start_cycle = 1
diis_space = 8
SCF conv_tol = 1e-09
SCF conv_tol_grad = None
SCF max_cycles = 50
direct_scf = True
direct_scf_tol = 1e-13
chkfile to save SCF result = /home/ryan/CLionProjects/wrap_pyscf/tmpmc9v4foe
max_memory 4000 MB (current use 115 MB)
Set gradient conv threshold to 3.16228e-05
init E= -1.11798956774344
  HOMO = -0.551284046862827  LUMO = 0.0586573242914685
cycle= 1 E= -1.1179895677435  delta_E= -6.31e-14  |g|= 7.54e-08  |ddm|= 3.68e-07
  HOMO = -0.551284016599658  LUMO = 0.0586573291817753
Extra cycle  E= -1.1179895677435  delta_E= -4.44e-15  |g|= 2.03e-08  |ddm|= 1.02e-07
converged SCF energy = -1.1179895677435
1

API

The wrap_pyscf module defines two functions density and run_scf whose APIs are detailed here.

density

density is used to compute the atomic density of the provided molecular system using the provided atomic basis set.

the_density density(molecule, basis_set, verbose=0)

where:

• the_density is a numpy.ndarray instance containing the converged SCF density matrix in the AO basis set.
• molecule is a [(str, float, float, float)] (i.e. a list of four element tuples) such that the i-th tuple is the atomic symbol and x, y, and z Cartesian coordinates (in Angstroms) for the i-th atom
• basis_set is the name of the atomic basis set to use
• verbose is an optional integer controlling how much printing is output by PySCF. Default is 0, i.e., no printing.

run_scf

run_scf is used to compute the SCF energy of the provided molecular system using the provided atomic basis set and optionally the provided atomic density.

nitr run_scf(molecule, basis_set, the_density = None, verbose = 4)

where:

• nitr is the number of iterations that the SCF took to converge
• molecule is a [(str, float, float, float)] (i.e. a list of four element tuples) such that the i-th tuple is the atomic symbol and x, y, and z Cartesian coordinates (in Angstroms) for the i-th atom
• basis_set is the name of the atomic basis set to use
• the_density is a numpy.ndarray instance containing the density to use as a guess for the SCF density matrix. Passing a value of None, the default, indicates that the program should use the default guess.
• verbose is an optional integer controlling how much printing is output by PySCF. Default is 4, which will print out basic SCF information.

Acknowledgements

• Jonathan Waldrop - Created the simplified API for calling PySCF