DFTB+ interface

DFTB+ is a density functional tight-binding (DFTB) program, intended for running semi-empirical DFT calculations. Since DFTB can run at least 2-orders of magnitude faster than a DFT calculation it is an attractive option for DFTB+ can be used for both regular DFTB Hamiltonians as well as extended tightbinding Hamiltonians (xTB). See also xTB interface for xTB calculations.

The ASH interface to DFTB+ is fairly new and not very extensive. It supports energies and gradients and can thus be used within ASH for optimization, frequencies, dynamics etc. It also supports pointcharge embedding and can thus be used for electrostatic embedding QM/MM.

Periodic boundary conditions were recently made available in the interface.

DFTBTheory class:

class DFTBTheory():
    def __init__(self, dftbdir=None, hamiltonian="XTB", xtb_method="GFN2-xTB", printlevel=2, label="DFTB",
                numcores=1, slaterkoster_dict=None, maxmom_dict=None, Gauss_blur_width=0.0,
                SCC=True, ThirdOrderFull=False, ThirdOrder=False, MaxSCCIterations=300,
                periodic=False, periodic_cell_vectors=None, periodic_cell_dimensions=None, kpoint_values=[1,1,1]):

Keyword	Type	Default value	Details
`dftbdir`	string	None	Directory where DFTB+ binaries are.
`printlevel`	integer	2	Printlevel
`numcores`	integer	1	The number of CPU cores used.
`hamiltonian`	string	'XTB'	Type of Hamiltonian to use. Options: 'DFTB', 'XTB'
`xtb_method`	string	1	For XTB hamiltonian, what xTB method to use. Options: 'GFN1-xTB', 'GFN2-xTB', 'IPEA1-xTB'
`slaterkoster_dict`	dict	None	Dictionary pointing to Slaterkoster parameter files for each pair of elements.
`SCC`	Boolean	True	Whether self-consistence charge option should be enabled or not.
`MaxSCCIterations`	integer	300	Max number of SCC iterations.
`ThirdOrderFull`	Boolean	False	Whether full third-order term should be included or not.
`ThirdOrder`	Boolean	False	Whether on-site third-order term should be included or not. Typically not recommended.
`maxmom_dict`	dict	None	Dictionary pointing to max mom for each element.
`Gauss_blur_width`	float	0.0	Gaussian blur width for pointcharge embedding
`periodic`	Boolean	False	Enable periodic boundary conditions.
`periodic_cell_vectors`	2d array	None	Cell vectors in Å as a Numpy 2D array.
`periodic_cell_dimensions`	list	None	Cell dimensions as a list of [a,b,c,alpha,beta,gamma] parameters in Å and °.
`kpoint_values`	list	[1,1,1]	Specify k-point grid by a list. [1,1,1] indicates gamma point in all three X,Y,Z directions.

DFTB+ installation

See DFTB+ download and installation page: https://dftbplus.org/download/index.html

In addition to the program, DFTB parameters (Slater-Koster files) need to be downloaded for the specific parameterization. Popular ones include e.g. '3ob', 'ob2', 'mio' See Github repository https://github.com/dftbparams

How to use

The ASH interface to DFTB+ is basic and currently allows basic usage of the implemented DFTB and xTB methods. ASH takes care of creating the DFTB+ inputfile.

DFTB and xTB methods inside DFTB+ are first differentiated by selecting hamiltonian and then by different options.

To use xTB methods within the DFTB program you select hamiltonian to be 'XTB' and then select the appropriate xtb_method string.

# GFN1-xTB
DFTBTheory(hamiltonian="XTB", xtb_method="GFN1-xTB")

# GFN2-xTB
DFTBTheory(hamiltonian="XTB", xtb_method="GFN2-xTB")

To use DFTB methods you select hamiltonian to be 'DFTB' and you can then enable/disable SCC and ThirdOrderFull options to get the desired DFTB Hamiltonian

# Original DFTB, a.k.a. DFTB1
DFTBTheory(hamiltonian="DFTB", SCC=False)

# DFTB2 a.k.a. SCC-DFTB.
DFTBTheory(hamiltonian="DFTB", SCC=True)

# DFTB3 a.k.a. SCC-DFTB with full third-order term
DFTBTheory(hamiltonian="DFTB", SCC=True, ThirdOrderFull=True)

However, to use DFTB methods one must additionally specify the specific DFTB parameter set by defining a dictionary for all element-pairs of the system and point to the individual Slater-Koster files. The parameter sets must be downloaded separately. Additionally third-order DFTB (DFTB3) requires the specification of atomic Hubbard derivatives, and additionally, depending on the parameter set, a hydrogen bond correction may also have to be specified.

Examples

In the examples below, the DFTB parameter sets have already been downloaded from: https://github.com/dftbparams

DFTB2 with mio paremeter set

from ash import *

#H2O fragment
frag = Fragment(databasefile="h2o.xyz")


# DFTB2-mio
skdir="/Users/rb269145/ash-tests/dftb_interface/dftb_for_website/DFTB2-mio/mio-main/skfiles"
sldict_mio= {'O-O':f'{skdir}/O-O.skf', 'H-O':f'{skdir}/H-O.skf',
        'O-H':f'{skdir}/O-H.skf', 'H-H':f'{skdir}/H-H.skf'}

# Defining DFTB2-mio Hamiltonian
theory = DFTBTheory(hamiltonian="DFTB", SCC=True, slaterkoster_dict=sldict_mio)

Singlepoint(theory=theory, fragment=frag)

DFTB3 with 3ob paremeter set

For DFTB3 calculations we have to enable the third-order term by ThirdOrderFull keyword. We also have to use a compatible parameter set and here we use the 3ob set . Additionally, because of the third-order term we have to provide Hubbard derivatives for each element. This information should be available with the parameter set (here in the README file of 3ob). We pass the Hubbard derivatives as a dictionary with the hubbard_derivs_dict keyword. Finally, we should also enable damping of the hydrogen interaction and here we set hcorrection_zeta to be 4.0 as recommended (see README of 3ob).

from ash import *

#H2O fragment
frag = Fragment(databasefile="h2o.xyz")


# DFTB3-3ob
skdir="/Users/rb269145/ash-tests/dftb_interface/dftb_for_website/DFTB3-3ob/3ob-main/skfiles"
sldict_3ob = {'O-O':f'{skdir}/O-O.skf', 'H-O':f'{skdir}/H-O.skf',
        'O-H':f'{skdir}/O-H.skf', 'H-H':f'{skdir}/H-H.skf'}

hubbard_derivs_dict={'O':-0.1575, 'H':-0.1857}
zeta=4.0 #damping parameter for H

# Defining DFTB3-3ob Hamiltonian
theory = DFTBTheory(hamiltonian="DFTB", SCC=True, ThirdOrderFull=True,  slaterkoster_dict=sldict_3ob,
   hubbard_derivs_dict=hubbard_derivs_dict, hcorrection_zeta=zeta)

Singlepoint(theory=theory, fragment=frag)

Periodic DFTB3 with 3ob paremeter set with PBCs

Here showing how a periodic DFTB3 geometry optimization (optimization of both atom and cell parameters) can be carried out.

from ash import *

numcores=1
frag = Fragment(xyzfile="ammonia.xyz", charge=0, mult=1)

#periodic_cell_dimensions=[5.01336, 5.01336, 5.01336, 90.0, 90.0,90.0]
cell_vectors = np.array([[5.01336,0.0,0.0],[0.0,5.01336,0.0],[0.0,0.0>

# DFTB3-3ob
skdir="/Users/rb269145/ash-tests/dftb_interface/dftb_for_website/DFTB>
sldict_3ob = {'N-N':f'{skdir}/N-N.skf', 'H-N':f'{skdir}/H-N.skf',
        'N-H':f'{skdir}/N-H.skf', 'H-H':f'{skdir}/H-H.skf'}

hubbard_derivs_dict={'N':-0.1575, 'H':-0.1857}
zeta=4.0 #damping parameter for H

# Defining DFTB3-3ob Hamiltonian
theory = DFTBTheory(hamiltonian="DFTB", SCC=True, ThirdOrderFull=True>
  hubbard_derivs_dict=hubbard_derivs_dict, hcorrection_zeta=zeta,
    periodic=True, periodic_cell_vectors=cell_vectors)

Optimizer(theory=theory, fragment=frag, coordsystem="hdlc")

Periodic GFN2-xTB with PBCs

Here showing how a periodic GFN2-xTB geometry optimization (optimization of both atom and cell parameters) can be carried out.

from ash import *

numcores=1
frag = Fragment(xyzfile="ammonia.xyz", charge=0, mult=1)

#periodic_cell_dimensions=[5.01336, 5.01336, 5.01336, 90.0, 90.0,90.0]
cell_vectors = np.array([[5.01336,0.0,0.0],[0.0,5.01336,0.0],[0.0,0.0,5.01336]])
theory = DFTBTheory(hamiltonian="XTB", xtb_method="GFN2-xTB",
    periodic=True, periodic_cell_vectors=cell_vectors)

Optimizer(theory=theory, fragment=frag, coordsystem="hdlc")