NWChem

Computational chemistry package for quantum chemistry, QM/MM and molecular dynamics, driven by an NWChem input deck.

License: ECL-2.0. Source code.

What it simulates

NWChem is a computational chemistry package. It solves the electronic structure of molecules with Hartree-Fock (SCF), density functional theory and correlated methods, and it also covers QM/MM and molecular dynamics. You describe the job in an input deck: start, title, a geometry block, a basis block, and a task line that names the method and the operation.

Input is one text deck. Output is a long text file with the energy, orbital information and, for an optimisation, every geometry step. The run also leaves a database (.db), orbital vectors (.movecs) and scratch files in the working directory.

The key assumption is the Born-Oppenheimer approximation: nuclei sit at fixed points while the electrons are solved quantum mechanically in a finite Gaussian basis. Results depend on the method and basis you choose.

When to use it

Geometries, energies and charges of small molecules from first principles.
Comparing methods and basis sets on the same molecule.

Alternatives: DFTB+ answers the same kind of question much faster with a tight-binding model, at lower accuracy. GROMACS, OpenMM and LAMMPS run classical dynamics of large systems that NWChem would take too long to treat quantum mechanically.

Install it yourself

NWChem 7.3.1 from conda-forge. A dry run resolved nwchem 7.3.1 (an MPI build, 229 MB in total):

conda install -c conda-forge nwchem=7.3.1
nwchem water.nw

The conda activation script sets NWCHEM_BASIS_LIBRARY and NWCHEM_NWPW_LIBRARY. The recipe's image has no activation step, so it sets them directly.

Security: an NWChem deck can run operating system commands. A deck containing task shell "echo ran-a-shell-command" printed ran-a-shell-command in the recipe's container. Run only input decks you trust.

Run the tested example

This recipe builds NWChem in a container and runs the example below. It needs Docker and nothing else. It last ran on 2026-10-12 in 2 s.

Download nwchem-recipe.tar.gz (Dockerfile, example deck.nw, run.sh).

Unpack it and run the script:

tar -xzf nwchem-recipe.tar.gz
cd nwchem
sh run.sh

What run.sh does

docker build -t biosimhub/nwchem:7.3.1 .
mkdir -p out && chmod a+w out
docker run --rm \
  --user "$(id -u):$(id -g)" -e HOME=/tmp \
  -v "$PWD/in:/work/in:ro" -v "$PWD/out:/work/out" -w /work/out \
  biosimhub/nwchem:7.3.1 nwchem /work/in/deck.nw > out/nwchem_output.txt

Inputs are read from in/ and results land in out/. Change the values after each option to run your own case; the parameter table lists them.

Expected output

File	Size
`nwchem_output.txt`	75.1 KB
`water.b`	240 B
`water.b^-1`	240 B
`water.c`	96 B
`water.cfock`	427 B
`water.db`	831.2 KB
`water.drv.hess`	96 B
`water.movecs`	913 B
`water.p`	96 B
`water.zmat`	48 B

First lines of nwchem_output.txt

 argument  1 = /work/in/deck.nw
                                         
                                         


             Northwest Computational Chemistry Package (NWChem) 7.3.0
             --------------------------------------------------------


                    Environmental Molecular Sciences Laboratory
                       Pacific Northwest National Laboratory
                                Richland, WA 99352

Reading the output

The example deck: water at the starting geometry O (0, 0, 0.117), H (0, +-0.757, -0.469) angstroms, the sto-3g basis for every atom, and task scf optimize, which means a Hartree-Fock (SCF) geometry optimisation.

nwchem_output.txt is 76870 bytes. In it:

The optimisation table (lines starting with @) has four rows, steps 0 to 3. Energy goes from -74.96294665 to -74.96590119 hartree. Gmax, the largest gradient component, falls from 0.03812 to 0.00002. The log then prints Optimization converged.
The final geometry has O-H 0.98943 A and H-O-H 100.03 degrees. The symmetry group is C2v.
The Mulliken table gives oxygen 8.33 electrons and each hydrogen 0.83.
The job took 0.4 s of CPU time.

The other files:

water.db (851197 bytes) is the run database, named after start water. The restart directive reuses it.
water.movecs (913 bytes) holds the molecular orbital vectors.
water.b, water.b^-1, water.c, water.p, water.cfock, water.zmat and water.drv.hess are optimiser and Fock scratch files, 48 to 427 bytes.

The package banner in the output reads NWChem 7.3.0, although the conda package is 7.3.1.

Your own runs

dr() { docker run --rm --user "$(id -u):$(id -g)" -e HOME=/tmp -v "$PWD/in:/work/in:ro" -v "$PWD/out:/work/out" -w /work/out "$@"; }
dr biosimhub/nwchem:7.3.1 nwchem /work/in/deck.nw > out/nwchem_output.txt

A bigger basis: change sto-3g to 6-31g*. It converged in four steps (0 to 3) to -76.01074651 hartree, with O-H 0.94732 A and H-O-H 105.50 degrees. The wall time was 0.6 s.
B3LYP: use the 6-31g* basis, add a dft block with xc b3lyp, and set task dft optimize. It converged in four steps to -76.40895454 hartree, with O-H 0.96874 A and 103.65 degrees (0.7 s). The angle moved from 100.03 to 105.50 degrees between STO-3G and 6-31G* at the same SCF level, so the basis set changes the answer by more than 5 degrees here. Energies from different methods are not comparable.

Pitfalls

The recipe runs NWChem as one process (nproc = 1).
The recipe's image sets NWCHEM_BASIS_LIBRARY by hand because it has no conda activation. In your own environment, activate it so the variable points at the basis library.
start water names every output file, and restart water reuses that job's data. Use a fresh folder per job to keep outputs apart.
Energies are in hartree. Geometry coordinates are in angstroms because the deck says units angstrom.
Decks can run shell commands with task shell. Never run a deck from an unknown source.
Check for Optimization converged in the output before you trust a geometry.