Skip to content


GPAW is an efficient program package for electronic structure calculations. It is based on the density-functional theory (DFT) implemented within the projector augmented wave (PAW), and it can utilize various basis sets (uniform real-space grids, plane waves, localized atomic orbital basis).

Some features of the software:

  • total energy calculations
  • structural optimizations
  • different boundary conditions (finite, wire, slab, bulk)
  • efficient parallelization
  • excited state properties within time-dependent density-functional theory


  • Puhti: 20.1.0, 20.10.0, 21.1.0
  • Mahti: 20.10.0, 21.1.0, 21.6.0, 22.1.0
  • Check all available versions (and default version) with module avail gpaw
  • Modules ending with -omp have the optional OpenMP parallelization enabled, see GPAW documentation about parallel runs for more details.

PAW Setups

All installations use version 0.9.20000 of GPAW's PAW Setups.


GPAW is free software available under GPL, version 3+


Since the default version, that is available with module load gpaw, is subject to change when new versions are installed, we recommend to always load a specific GPAW version:

module load gpaw/version


In CSC environment GPAW calculations are run with the gpaw-python command.

Enabling ELPA

In Mahti, GPAW can use the ELPA library to speed up the diagonalization step. Especially with LCAO calculations ELPA can improve the performance. In order to use ELPA, the 'use_elpa' : True` setting needs to be included in the parallelization options in GPAW input, see GPAW documentation for more information.

Batch script examples

Example parallel batch script for Puhti

#!/bin/bash -l
#SBATCH --time=00:30:00
#SBATCH --partition=large
#SBATCH --ntasks-per-node=40
#SBATCH --nodes=2
#SBATCH --mem-per-cpu=2GB
#SBATCH --account=<project>
##SBATCH --mail-type=END #uncomment to get mail

# this script runs a 80 core (2 full nodes) gpaw job, requesting
# 30 minutes time and 2 GB of memory for each core

module load gpaw/21.1.0

srun gpaw-python

Example batch script for Mahti with hybrid MPI/OpenMP parallelization

#!/bin/bash -l
#SBATCH --time=00:30:00
#SBATCH --partition=medium
#SBATCH --nodes=10
#SBATCH --ntasks-per-node=32
#SBATCH --cpus-per-task=4
#SBATCH --account=<project>
##SBATCH --mail-type=END #uncomment to get mail

# this script runs a 1280 core (10 full nodes) gpaw job, using hybrid
# MPI/OpenMP parallelization with 4 OpenMP threads per node,
# requesting 30 minutes time.
# Please experiment with optimum MPI task / OpenMP thread ratio with
# your particular input

# Note: only the modules with "-omp" ending supports OpenMP
# (default version in Mahti is OpenMP enabled)
module load gpaw/21.1.0-omp


srun gpaw-python

Example batch script for Mahti with pure MPI parallelization

#!/bin/bash -l
#SBATCH --time=00:30:00
#SBATCH --partition=medium
#SBATCH --nodes=10
#SBATCH --ntasks-per-node=128
#SBATCH --account=<project>
##SBATCH --mail-type=END #uncomment to get mail

# this script runs a 1280 core (10 full nodes) gpaw job, using pure
# MPI parallelization requesting 30 minutes time.

module load gpaw/21.1.0


srun gpaw-python

More information

Last edited Tue May 3 2022