DFTD3¶
Introduction¶
The DFTD3 calculator class wraps the ‘dftd3’ command line utility by the research group of Stefan Grimme. This can be used to calculate classical vdW dispersion corrections to a large number of common DFT functionals. This calculator can be used in conjunction with other DFT calculators such as GPAW to allow seamless calculation of dispersioncorrected DFT energies, forces, and stresses.
This is a list of all supported keywords and settings:
Keyword 
Default value 
Description 



Use parameters optimized for the selected XC functional. 


Alternative to 


Enable or disable calculation of gradients (forces, stress tensor). 


Enable threebody ATM correction. 

40 Bohr 
Cutoff radius for coordination number and threebody calculations. 

95 Bohr 
Cutoff radius for twobody dispersion calculations. 


Enable older DFTD2 dispersion correction method. 


Damping method. Valid options are 


Custom parameters optimized for triplezeta basis sets. 

Custom damping parameter used in all damping methods. 


Custom damping parameter used in 


Custom damping parameter used in all damping methods. 


Custom damping parameter used in 


Custom damping parameter used in all damping methods. 


Custom damping parameter used in 


Custom damping parameter used in 


Custom damping parameter used in 
Examples¶
DFTD3 can be used by itself to calculate only the vdW correction to a system’s energy, forces, and stress. Note that you should not use these properties alone to perform dyanmics, as DFTD3 is not a full classical potential.
from ase.calculators.dftd3 import DFTD3
from ase.build import bulk
diamond = bulk('C')
d3 = DFTD3()
diamond.calc = d3
diamond.get_potential_energy()
If used in conjunction with a DFT calculator, DFTD3 returns dispersioncorrected energies, forces, and stresses which can be used to perform dynamics.
import numpy as np
from gpaw import GPAW, PW
from ase.calculators.dftd3 import DFTD3
from ase.build import bulk
from ase.constraints import UnitCellFilter
from ase.optimize import LBFGS
np.random.seed(0)
diamond = bulk('C')
diamond.rattle(stdev=0.1, seed=0)
diamond.cell += np.random.normal(scale=0.1, size=(3,3))
dft = GPAW(xc='PBE', kpts=(8,8,8), mode=PW(400))
d3 = DFTD3(dft=dft)
diamond.calc = d3
ucf = UnitCellFilter(diamond)
opt = LBFGS(ucf, logfile='diamond_opt.log', trajectory='diamond_opt.traj')
opt.run(fmax=0.05)
Additional information¶
This calculator works by writing either an xyz
file (for nonperiodic
systems) or a POSCAR
file (for periodic systems), calling the
dftd3
executable, and parsing the output files created. It has been
written such that its interface should match that of the dftd3
utility
itself as closely as possible, while minimizing the possibility of setting
redundant and contradictory options. For example, you can only select one
damping method, and the interface will sanitycheck any provided custom
damping parameters.
Without any arguments, the DFTD3 will default to calculating the PBED3
dispersion correction with 'zero'
damping. If a DFT calculator is
attached, DFTD3 will attempt to glean the XC functional from the DFT
calculator. This will occasionally fail, as dftd3
is very particular
about how the names of XC functionals are to be formatted, so in general
you should supply the XC functional to both the DFT calculator and the DFTD3
calculator.
Caveats¶
The dftd3
does not handle systems with only 1D or 2Dperiodic boundary
conditions. If your system has 1D or 2D PBC, DFTD3 will calculate the
dispersion correction as though it was fully 3D periodic.
If your system is very large, the dispersion calculation can take quite long,
especially if you are including threebody corrections (abc=True
). For
highly parallel calculations, this may result in the dispersion correction
taking longer than the DFT calculation! This is because the dftd3
utility
is not parallelized and will always run on a single core. Be sure to
benchmark this calculator interface on your system before deploying large,
heavily parallel calculations with it!