 |
|
|
$DFTB group (relevant for GBASIS=DFTB)
Density-functional tight-binding (DFTB) is turned on by
selecting GBASIS=DFTB in $BASIS. $DFTB controls optional
parameters for a DFTB calculation. DFTB is formulated in a
two-center approximation utilizing implicitly a minimal
pseudoatomic orbital basis set with corresponding,
pretabulated one- and two-center integrals. Because of
this, many properties (for instances, multipoles higher
than dipoles) and many options are ignored or not available
in the current implementations of DFTB. DFTB also uses an
independent SCF driver (SCF in DFTB is also called SCC, see
below), so most SCF options are not available for DFTB.
Only SCFTYP=RHF and UHF are implemented. SCFTYP=ROHF is
available, only when all SPNCST values are zero. DFTB does
not explicitly use symmetry (C1 throughout) since integrals
are never computed during the calculations. Slater-Koster
tables are only defined for spherical functions (5d) so
DFTB sets ISPHER=1. Most $GUESS options do not work for
DFTB (DFTB does not use initial orbitals in the usual
sense). Other than the default (METHOD=HUCKEL, which is
ignored), only METHOD=MOREAD works (note that SCC-DFTB can
use initial charges on atoms, derived from the orbitals).
RUNTYP=OPTIMIZE, HESSIAN and RAMAN are available for full
(non-FMO) DFTB, whereas RUNTYP=OPTIMIZE and OPTFMO are
available for FMO-DFTB. Excited state calculations for
full DFTB may be performed through the standard (linear-
response) time-dependent formalism (only closed shell). PCM
can be used for both ground and excited state calculations,
and energy and gradient can be evaluated.
In DFTB calculation, the atom type is determined by its
name, not its nuclear charge as elsewhere in GAMESS. The
nuclear charge (the second column in $DATA) is used only in
population analysis, but not in SCF. DFTB uses a notion of
"species", which means an atomic type. The species are
numbered according to the order in which atoms appear in
$DATA. For instances, in water there are two species, O and
H. An atomic type of each species needs MAXANG, which for
most but not all atoms is set automatically.
NDFTB order of the Taylor expansion of the total energy
around a reference density in the DFTB model.
= 1 NCC-DFTB, also called DFTB1.
NCC stands for non-charge-consistent, i.e., no
explicity charge-charge interaction term is
included in the energy calculation.
= 2 SCC-DFTB, also called DFTB2.
SCC means a self-charge-consistent approach,
and SCC implies that SCF iterations are carried
out that converge monopolar charges towards
self-consistency.
= 3 DFTB3, including 3rd order correction using
Hubbard derivatives (HUBDER).
In order to reproduce the published DFTB3
approach, it is necessary to also specify
DAMPXH=.TRUE. to add other terms.
Gaus, M. et al. J. Chem. Theory Comput. 2011,
7, 931-948 is referred to as Gaus2011 below.
Default: 2.
DAMPXH = a flag to include the damping function for X-H
atomic pair in DFTB3. See also DAMPEX, and eq 21
in Gaus2011.
The damping function is used when at least one
atom in a pair is "H". "HYDROGEN" and any other
name will turn off the damping.
Default: .FALSE.
DAMPEX = an exponent used in the damping function for X-H
atomic pairs. The default value is 4.2
(see Table 2 in Gaus2011 for more details).
SRSCC = a flag to perform shell-resolved SCC calculation.
If set to .FALSE., the code uses the Hubbard
value for an s orbital for p and d orbitals,
ignoring their Hubbard values defined in Slater-
Koster tables.
Using .TRUE. enables the use of proper Hubbard
values for p and d orbitals, implemented only
for DFTB1 and DFTB2.
Default: .FALSE.
ITYPMX Convergence method of SCC calculations.
= -1 Use standard GAMESS convergence methods.
SOSCF and DIIS are supported, but DEM is not.
= 0 Broyden's method.
Interpolation is applied for atomic
(or shell-resolved when SRSCC=.TRUE.)
charges, but not Hamiltonian matrix.
= 1 (reserved)
= 2 DIIS for charges.
Default: 0.
ETEMP = electronic temperature in Kelvin. Non-zero values
of ETEMP help SCF convergence of nearly-degenerate
systems by smearing occupation numbers around the
Fermi-level. Only the Fermi-Dirac distribution
function is available as a smearing function. The
default value is 0 Kelvin, meaning the smearing
function is not used.
ETEMP is implemented only for SCFTYP=RHF and when
FMO is not used.
DISP dispersion model for DFTB.
= NONE no Dispersion correction.
= UFF UFF-type dispersion correction.
Parameters for atomic numbers up to 54 are
available internally or can be supplied in
DISPPR for any atom.
Built-in parameters are taken from Rappe
et al. J. Am. Chem. Soc. 1992, 114, 10024.
= SK The Slater-Kirkwood type dispersion
correction omitting the change polarizability
depending on the number of bonds.
No default values of DISPPR are available.
Some are listed in the manual of the DFTB+
program.
DISPPR an array of parameters used for dispersion
correction, listed in sets for each species.
For DISP=UFF, DISPPR(1) and DISPPR(2) define the
non-bonded distance (Angs.) and energy (kcal/mol)
for the first species, respectively, and so on.
For DISP=SK, a set for a species has 3 parameters,
the polarizability (Angstrom^3), cutoff length
(Angstrom), and atomic charge.
Default: see DISP.
HUBDER an array of Hubbard derivatives for each species
(1 per species) used only for DFTB3 calculations.
Default values are set only for C, H, N, O, and P
using the final row of Table 2 in Gaus2011 (see
the paper for other choices).
MAXANG array of maximum angular momentum of each species,
which determines the number of basis functions.
DFTB uses only valence orbitals and electrons!
Most elements have proper default values, but for
some atomic types (i.e., species) you need to
manually define the values.
QREF array of the number of reference electrons of each
species. QREF is usually automatically taken from
Slater-Koster parameters, so this option is seldom
used.
SPNCST an array of spin constants used in unrestricted
(UHF) DFTB calculation. Provide 6 spin constants,
W_{ss}, W_{sp}, W_{pp}, W_{sd}, W_{pd}, & W_{dd},
for each species in a continuous array. Constants
for some elements can be found in the manual of
the DFTB+ program.
MODHSS controls the behavior of the computation of
analytic Hessian (bit additive).
1 Do not write integrals on disk.
2 Use a faster algorithm for solving CP-DFTB
requiring a lot of memory.
4 Parallelize integral transformation using GDDI.
8 Hessian contributions are calculated one by one.
By default, all of these flags are set to true,
unless there is not enough memory or for some
other reason.
* * *
The following options are FMO-DFTB specific (Nishimoto, Y.
et al. J. Chem. Theory Comput. 2014, 10, 4801-4812.).
FMO-DFTB has many limitations and some FMO options are not
supported (for instance, AFO, multilayer FMO etc). Only
single layer, restricted closed-shell FMO2-DFTB1, 2, and 3
are implemented at present. SRSCC, ETEMP etc are not
available. The analytic gradient is available for FMO-DFTB,
requiring solving SCZV as in other FMO methods.
MODESD = controls the behavior of ES-DIM (electrostatic
dimer) approximation (bit additive).
1 Calculate interfragment repulsive energy for ES
dimers (almost never used).
2 Add up all ES-DIM energies. This means that
individual ES dimer energies are not calculated,
but only their total lump sum, computed with the
dynamic load balancing.
4 Lump ES-DIM routine with static load balancing.
The bits of 2 or 4 are mutually exclusive.
Default: 0 (i.e., individual ES dimer energies).
MODGAM = controls the calculation of gamma values
(interatomic 1/R-like function) in FMO-DFTB2 and
FMO-DFTB3 (bit additive).
0 Calculate gamma values on the fly. (default)
1 Calculate once and prestore gamma values in
triangular matrix.
2 Calculate once and prestore gamma values in
square matrix.
4 With the bits of 1 or 2, the calculation of
gamma values is parallelized with GDDI.
The bits of 1 or 2 are mutually exclusive. These
options are faster but takes more memory.
Default: 0
==========================================================
==========================================================
generated on 7/7/2017