Skip to content

Latest commit

 

History

History
112 lines (89 loc) · 4.03 KB

Gaussian-interface.rst

File metadata and controls

112 lines (89 loc) · 4.03 KB
Raw

Gaussian interface

Gaussian is one of the oldest and widely used quantum chemistry programs.

ASH features a basic interface to Gaussian, allowing you to use it for DFT and WFT based methods within the program. Energies and gradients are available in the interface so GaussianTheory in ASH can be used for single-point energies, geometry optimizations, numerical frequencies, surface scans, NEB and molecular dynamics. Warnign: Pointcharge-support is available, however, as the pointcharge gradient is not available, electrostatic embedding QM/MM is limited to single-point energy calculations, for now.

GaussianTheory class:

class GaussianTheory:
    def __init__(self, gaussiandir=None, gauss_executable='g16', filename='gaussian', 
                file_extension='.com', printlevel=2, label="Gaussian",
                gaussianinput=None, memory='800MB', numcores=1):
Keyword Type Default value Details
gaussiandir string None Directory where Gaussian binaries are.
gauss_executable string 'g16' Name of the Gaussian executable to be used (e.g. 'g16', 'g09', 'g03').
filename string 'gaussian' Filename used for Gaussian input/output files.
file_extension string '.com' File extension used for Gaussian input/output files created by ASH, e.g. '.com' or '.gjf'.
gaussianinput string None Multi-line string containing the Gaussian input-lines.
memory string '800MB' A string specifying the memory to be given to Gaussian.
printlevel integer 2 Printlevel
numcores integer 1 The number of CPU cores used for OpenMP parallelization of Gaussian.
label string 'Gaussian' Optional label for GaussianTheory object.

Gaussian installation

Gaussian needs to be installed separately. As we do not have access to the Gaussian source code, we cannot offer any installation instructions.

Once Gaussian has been installed you need to either : i) use the gaussiandir keyword option to GaussianTheory to provide the location of the directory where the Gaussian binary resides or ii) Make sure the g16/g09/g03 binary is in the PATH variable of your shell environment. ASH will try to find an executable the gauss_executable (e.g. 'g16') automatically when creating the GaussianTheory object.

Parallelization

Gaussian can be parallelized using the built-in OpenMP. This parallelization will only work for single node parallelization. The parallelization is controlled via the numcores keyword.

Note that you should not add any parallelization options to the Gaussian input string, as ASH will handle this.

Examples

DFT-SCF example:

from ash import *

#Variable defining number of cores
numcores=4

#Fragment
frag = Fragment(databasefile="h2o.xyz", charge=0, mult=1)

#Multi-line inputstring to define the main Gaussian options
#No geometry-input, job-keywords (opt, freq etc.) or parallelization options should be provided here (handled by ASH)
input_string=""""
# b3lyp cc-pVDZ scf=tight int=ultrafine
""""

gaussian_object = GaussianTheory(gauss_executable='g09', gaussianinput=input_string, numcores=numcores)

Singlepoint(theory=gaussian_object, fragment=frag)