Additional features

The dftd4.parameters module becomes available if a TOML parser is available, either tomlkit or toml can be used here. The returned dict can be used to supply parameters to the constructor of the DampingParam object, only the s6, s8, s9, a1, a2 and alp entries will be used, the remaining entries are meta data describing the damping parameters.

>>> from dftd4.parameters import get_damping_param
>>> get_damping_param("b97m")
{'s6': 1.0, 's9': 1.0, 'alp': 16.0, 's8': 0.6633, 'a1': 0.4288, 'a2': 3.9935}
>>> get_damping_param("r2scan", keep_meta=True)
{'s6': 1.0, 's9': 1.0, 'alp': 16.0, 'damping': 'bj', 'mbd': 'approx-atm', 's8': 0.6018749, 'a1': 0.51559235, 'a2': 5.77342911, 'doi': '10.1063/5.0041008'}

QCSchema Integration

This Python API natively understands QCSchema and the QCArchive infrastructure. If the QCElemental package is installed the dftd4.qcschema module becomes importable and provides the run_qcschema function. If the QCElemental package is >=0.50.0, dftd4.qcschema supports QCSchema v1 and v2, returning whichever version was submitted. Note that only QCSchema v2 works with Python 3.14+ due to Pydantic restrictions.

>>> from dftd4.qcschema import run_qcschema
>>> import qcelemental as qcel
>>> atomic_input = qcel.models.AtomicInput(
...     molecule = qcel.models.Molecule(
...         symbols = ["O", "H", "H"],
...         geometry = [
...             0.00000000000000,  0.00000000000000, -0.73578586109551,
...             1.44183152868459,  0.00000000000000,  0.36789293054775,
...            -1.44183152868459,  0.00000000000000,  0.36789293054775
...         ],
...     ),
...     driver = "energy",
...     model = {
...         "method": "TPSS-D4",
...     },
...     keywords = {},
... )
...
>>> atomic_result = run_qcschema(atomic_input)
>>> atomic_result.return_result
-0.0002667885779142513

ASE Integration

To integrate with ASE this interface implements an ASE Calculator. The DFTD4 calculator becomes importable if an ASE installation is available.

>>> from ase.build import molecule
>>> from dftd4.ase import DFTD4
>>> atoms = molecule('H2O')
>>> atoms.calc = DFTD4(method="TPSS")
>>> atoms.get_potential_energy()
-0.007310393443152083
>>> atoms.calc.set(method="PBE")
{'method': 'PBE'}
>>> atoms.get_potential_energy()
-0.005358475432239303
>>> atoms.get_forces()
array([[-0.        , -0.        ,  0.00296845],
       [-0.        ,  0.00119152, -0.00148423],
       [-0.        , -0.00119152, -0.00148423]])

To use the DFTD4 calculator as dispersion correction the calculator can be combined using the SumCalculator from the ase.calculators.mixing module.

>>> from ase.build import molecule
>>> from ase.calculators.mixing import SumCalculator
>>> from ase.calculators.nwchem import NWChem
>>> from dftd4.ase import DFTD4
>>> atoms = molecule('H2O')
>>> atoms.calc = SumCalculator([DFTD4(method="PBE"), NWChem(xc="PBE")])

For convenience DFTD4 allows to combine itself with another calculator by using the add_calculator method which returns a SumCalculator:

>>> from ase.build import molecule
>>> from ase.calculators.emt import EMT
>>> from dftd4.ase import DFTD4
>>> atoms = molecule("C60")
>>> atoms.calc = DFTD4(method="pbe").add_calculator(EMT())
>>> atoms.get_potential_energy()
6.348142387048062
>>> [calc.get_potential_energy() for calc in atoms.calc.calcs]
[-6.015477436263984, 12.363619823312046]

The individual contributions are available by iterating over the list of calculators in calc.calcs. Note that DFTD4 will always place itself as first calculator in the list.

PySCF support

Integration with PySCF is possible by using the dftd4.pyscf module. The module provides a DFTD4Dispersion class to construct a PySCF compatible calculator for evaluating the dispersion energy and gradients.

>>> from pyscf import gto
>>> import dftd4.pyscf as disp
>>> mol = gto.M(
...     atom="""
...          C   -0.755422531  -0.796459123  -1.023590391
...          C    0.634274834  -0.880017014  -1.075233285
...          C    1.406955202   0.199695367  -0.653144334
...          C    0.798863737   1.361204515  -0.180597909
...          C   -0.593166787   1.434312023  -0.133597923
...          C   -1.376239198   0.359205222  -0.553258516
...          I   -1.514344238   3.173268101   0.573601106
...          H    1.110906949  -1.778801728  -1.440619836
...          H    1.399172302   2.197767355   0.147412751
...          H    2.486417780   0.142466525  -0.689380574
...          H   -2.454252250   0.422581120  -0.512807958
...          H   -1.362353593  -1.630564523  -1.348743149
...          S   -3.112683203   6.289227834   1.226984439
...          H   -4.328789697   5.797771251   0.973373089
...          C   -2.689135032   6.703163830  -0.489062886
...          H   -1.684433029   7.115457372  -0.460265708
...          H   -2.683867206   5.816530502  -1.115183775
...          H   -3.365330613   7.451201412  -0.890098894
...          """
... )
>>> d4 = disp.DFTD4Dispersion(mol, xc="r2SCAN")
>>> d4.kernel()[0]
array(-0.0050011)

To make use of the dispersion correction together with other calculators, the energy method allows to apply a dispersion correction to an existing calculator.

>>> from pyscf import gto, scf
>>> import dftd4.pyscf as disp
>>> mol = gto.M(
...     atom="""
...          O  -1.65542061  -0.12330038   0.00000000
...          O   1.24621244   0.10268870   0.00000000
...          H  -0.70409026   0.03193167   0.00000000
...          H  -2.03867273   0.75372294   0.00000000
...          H   1.57598558  -0.38252146  -0.75856129
...          H   1.57598558  -0.38252146   0.75856129
...          """
... )
>>> mf = disp.energy(scf.RHF(mol)).run()
converged SCF energy = -149.939098424774
>>> grad = mf.nuc_grad_method().kernel()
--------------- DFTD4 gradients ---------------
         x                y                z
0 O     0.0172438133     0.0508406920     0.0000000000
1 O     0.0380018285    -0.0460223790    -0.0000000000
2 H    -0.0305058266    -0.0126478132    -0.0000000000
3 H     0.0069233858    -0.0382898692    -0.0000000000
4 H    -0.0158316004     0.0230596847     0.0218908543
5 H    -0.0158316004     0.0230596847    -0.0218908543
----------------------------------------------

Installing

This project is packaged for the conda package manager and available on the conda-forge channel. To install the conda package manager we recommend the miniforge installer. If the conda-forge channel is not yet enabled, add it to your channels with

conda config --add channels conda-forge

Once the conda-forge channel has been enabled, this project can be installed with:

conda install dftd4-python

Now you are ready to use dftd4, check if you can import it with

>>> import dftd4
>>> from dftd4.libdftd4 import get_api_version
>>> get_api_version()
'4.3.0'

Building the extension module

The released Python source distribution bundles the native dftd4 sources. When building directly from the repository, an installed dftd4 must be available through CMake or pkg-config, depending on the selected build system. Try to find the installation first with

pkg-config --modversion dftd4

Adjust the PKG_CONFIG_PATH environment variable to include the correct directories to find the installation if necessary.

export CC=gcc

pip install .

pip install '.[parameters,ase,qcschema]'

pip install "https://github.com/dftd4/dftd4/archive/refs/heads/main#egg=dftd4-python&subdirectory=python"

cmake -S . -B _build -G Ninja -DPython3_EXECUTABLE=$(which python3)
cmake --build _build
cmake --install _build --prefix=/path/to/install --component python

meson setup _build -Dpython_version=$(which python3)

meson compile -C _build

meson configure _build --prefix=/path/to/install
meson install -C _build