$FMO group (optional, activates FMO option)
The presence of this group activates the Fragment
Molecular Orbital option, which divides large molecules
(think proteins or clusters) into smaller regions for
faster computation. The small pieces are termed 'monomers'
no matter how many atoms they contain. Calculations within
monomers, then 'dimer' pairs, and optionally 'trimer' sets
act so as to approximate the wavefunction of the full
system. The quantum model may be SCF, DFT, DFTB, MP2, CC,
MCSCF, TDDFT, or CI.
Sample inputs, and auxiliary programs, and other
information may be found in the GAMESS source distribution
in the directory ~/gamess/tools/fmo.
NBODY = n-body FMO expansion:
0 only run initial monomer guess (maybe remotely
useful to create the restart file, or as an
alternative to EXETYP=CHECK).
1 run up to monomer SCF
2 run up to dimers (FMO2, the default)
3 run up to trimers (FMO3)
IEFMO = switch to turn on EFMO (effective fragment
potential based Fragment Molecular Orbital)
0 = use FMO (default)
1 = use EFMO
MODEFM array of five values controlling EFMO, each
allows a bit-wise combination of several options.
Default is MODEFM(1)=0,0,0,0,0
The first element is control over electrostatics
0 no screening of electrostatics
1 exponential screening of electrostatics by
fixed value to fit the classical potential to
the QM-potential, set SCREEN(1)=-1 in $FMO.
(experimental)
2 Add octupole energy into electrostatic energy
4 use Hui Li's density based multipole
expansion
8 ignore torque contributions to the gradient
16 generate electrostatics on bond midpoints too
The second element controls polarizabilities.
0 Tang-Toenis type screening
1 do not include any polarization at all
4 add percentage discrimination based on
distance to atoms
8 ignore torque contributions to the gradient
16 use full polarization tensors
32 move polarizability tensors to nearest atom
before induction
64 do not evaluate electrostatic field, induced
dipoles or gradient contributions from
neighbouring fragments. This assumes
fragments are made in a sequential fashion.
128 use Ruedenberg localization for localization
of orbitals
The third element affects dispersion
0 no dispersion interactions
1 include dispersion
The fourth element affects charge transfer
0 no charge transfer interactions
1 include charge transfer
The fifth element affects exchange repulsion
0 no exchange interactions
1 include exchange repulsion
MODFD = switch to freeze the electronic state of some
fragments. FMO/FD and FMO/FDD require
RUNTYP=OPTIMIZE and two layers in FMO.
0 = regular FMO
1 = FMO/FD (frozen domain)
3 = FMO/FDD (frozen domain and dimers)
NDUALB = switch to use dual basis approach with
auxilliary polarization, AP, (i.e., a different
basis set is used to estimate the polarization).
The two basis sets in FMO/AP are entered in
the multibasis fashion (not in the multilayer),
i.e., as H.1 and H.2 in $DATA, not as H-1 and H-2.
The dual basis set has some restrictions.
Gradient (but not Hessian) is available.
0 = usual FMO
1 = dual basis FMO/AP
I. The following parameters define layers.
NLAYER = the number of layers (default: 1)
MPLEVL = an array specifying n in MPn PT for each layer,
n=0 or 2. (default: all 0s).
Note that MCQDPT is not available and therefore
one may not choose this for MCSCF.
DFTTYP = an array specifying the DFT functional type for
each layer. (default: DFTTYP in $DFT).
See $DFT for possible functionals. All functionals
except dual hybrids may be used.
Only grid-based DFT is supported.
SCFTYP = an array specifying SCF type for each layer.
At present the only valid choices are RHF, ROHF,
UHF, and MCSCF.
(default: SCFTYP in $CONTRL for all layers).
CCTYP = an array specifying CC type for each layer, which
may be only the following choices from $CONTRL:
LCCD, CCD, CCSD, CCSD(T), CCSD(T), CCSD(TQ),
CR-CCL, or non-size extensive R-CC or CR-CC.
Since FMO's CC methods involve adding corrections
from pairs of monomers together, it is better to
choose a size extensive method.
TDTYP = an array specifying TDDFT type for each layer,
of the same kind as TDDFT in $CONTRL.
Default: TDDFT in $CONTRL for all layers.
CITYP = an array specifying CI type for each layer, see
CITYP in $CONTRL. At present, only CIS may be
used (FMO1-CIS energy only, i.e., nbody=1).
Default: CITYP from $CONTRL, for all layers.
II. Parameters defining FMO fragments:
NFRAG = the number of FMO fragments (default: 1)
LAYER = an array defining the layer for each fragment.
Default: all fragments in layer 1, i.e.,
LAYER(1)=1,1,1,...,1
FRGNAM = an array of names for each fragment (each 1-8
character long) (default: FRG00001,FRG00002...).
INDAT = an array assigning atoms to fragments. Two styles
are supported (the choice is made based on
INDAT(1): if it is nonzero, choice (a) is taken,
otherwise INDAT(1) is ignored and choice (b) is
taken):
a) INDAT(i)=m assigns atom i is to fragment m.
INDAT(i) must be given for each atom.
b) the style is
a1 a2 ... ak 0
b1 b2 ... bm 0
...
Elements a1...ak are assigned to fragment 1,
then b1...bm are assigned to fragment 2,etc.
An element is one of the following:
I or I -J
where I means atom I, and a pair I,-J means
the range of atoms I-J. There must be no space
after the "-"!
Example:
indat(1)=1,1,1,2,2,1 is equivalent to
indat(1)=0, 1,-3,6,0, 4,5,0
Both assign atoms 1,2,3 and 6 to fragment 1,
and 4,5 to fragment 2.
ICHARG = an array of charges on the fragments
(default: all 0 charges)
MULT = an array of multiplicities for each fragment.
For MCSCF only the unique MCSCF fragment may be
something other than a singlet.
For ROHF and UHF multiple open-shell fragments
are allowed, which may have any multiplicity;
for dimers the high spin coupling will be used.
(default: all 1's)
SCFFRG = an array giving the SCF type for each fragment.
For MCSCF, only one fragment may be MCSCF and
the rest should be RHF. For ROHF, UHF and U-DFT
multiple open-shell fragments are allowed, but
for ground state runs only.
The values in SCFTYP overwrite SCFFRG, that is, if
you want to do a 2-layer calculation, the first
layer being RHF and the other MCSCF, then you
would use SCFTYP(1)=RHF,MCSCF and SCFFRG(N)=MCSCF,
where you should replace N by your MCSCF fragment
number. Then the first layer will be all RHF and
the other will have one MCSCF fragment. In special
cases, some SCFFRG values may be set to NONE, in
which case SCF is not performed. This is useful in
conjunction with ATCHRG.
(default: SCFTYP in $CONTRL).
MOLFRG = an array listing fragments for selective FMO,
where not all dimers (and/or trimers) are
computed. Setting MOLFRG imposes various
restrictions, such as RUNTYP=ENERGY only.
See MODMOL. For subsystem analysis (MODMOL=8),
MOLFRG(i) defines which subsystem fragment i
belongs to. Default: all 0.
MODMUL Use the multipole expansion to compute
electrostatic interactions exactly, bit additive.
1 Compute individual contributions for each ES
dimer.
2 Compute the sums for all ES dimer contributions
and add them to the energy and gradient.
8 Compute one-electron ESP gradients (implemented
for RESPPC<=0 only).
Only one of bits 1 or 2 may be turned on.
Default: 0
IACTFG = array specifying fragments in the active domain in
FMO/FD(D). Ranges can be specified as in INDAT, so
INDAT(1)=1,2,-5,8 means fragments 1,2,3,4,5,8.
All IACTFG fragments should be in the 2nd layer,
and the interfragment distance between fragments
in IACTFG and the 1st layer's fragments should not
be zero (i.e., no detached bonds between them).
Default: all zeroes.
NOPFRG = printing and other additive options, specified for
each fragment,
1 set the equivalent of $CONTRL NPRINT=7 (printing
option). Useful if you want to print orbitals
only for a few selected monomers.
2 set MVOQ to +6 to obtain better virtual orbitals
(ENERGY runs only, useful mostly to prepare good
initial orbitals for MCSCF).
4 generate cube file for the specified fragment,
the grid being chosen automatically.
(default: all 0s)
64 use frozen atomic charges (defined in ATCHRG)
instead of the variational ones to compute
converged fragment densities, to describe the
electrostastic field from a fragment acting upon
other fragments.
128 apply options 1 and 4 above only at the final SCF
iteration (correlation or GRADIENT only).
NACUT = automatically divides a molecule into fragments by
assigning NACUT atoms to each fragment (useful for
something like water clusters). This sets FRGNAM
and INDAT, so they need not be given. If 0, the
automatic option is disabled. (default: 0)
IEXCIT = options for FMO based TDHF, TDDFT, or CI
calculations:
IEXCIT(1): ordinal number for the excited state fragment.
There is no default for IEXCIT(1), you should
always set it.
IEXCIT(2): chooses the many-body level excitation n, e.g.
for FMOn-TDDFT.
n=1 means only the fragment given in IEXCIT(1)
will be excited.
n=2 adds dimer corrections (from fragment
pairs involving IEXCIT(1)).
IEXCIT(2) must not exceed NBODY. Default: 1.
IEXCIT(3): (relevant for FMO2-TDDFT only)
= 0 economic mode: only TDDFT dimer calculations
are performed (skipping all other dimers).
= 1 all dimer calculations are performed to
obtain not just the excitation but also the
total excited state energy.
Default: 0.
IEXCIT(4): excited state matching method in FMO2-TDDFT used
to determine which excitations in dimers
correspond to those in the TDDFT fragment given
in IEXCIT(1). Default=2.
= 0 trivial or identity matching (assume the
same order of the excited states in monomers
and dimers.
= 1 match the dominant orbital pair (aka DRF)
coefficient.
= 2 match the whole excitation vector.
Methods 1 and 2 try to match monomer dimer orbitals
first, and then use DRF coefficients. In difficult
cases (i.e., if the orbitals in a dimer are very
delocalised), methods 1 and 2 may not be able to find
the right transition, so some visual checking is
recommended.
ATCHRG = array of atomic charges, to be used with NOPFRG,
set for some fragments to 64 (i.e., to freeze some
of fragment electrostatic potentials during SCC).
Nota bene: the order of atoms in ATCHRG is not the same as
in FMOXYZ. In ATCHRG, you should specify atomic charges for
all atoms in fragment 1, then for fragment 2 etc, as a
single array. For covalently connected fragments there are
formally divided atoms (some redundant), and ATCHRG should
then list charges for them as well, all in the exact order
of atoms in which fragments are defined in FMO. The number
of entries in ATCHRG is NATFMO+NBDFG, where NATFMO is the
number of atoms in $FMOXYZ and NBDFG is the number of bonds
defined in $FMOBND.
NATCHA = option applicable to molecular clusters made
exclusively of the same molecules. Only NATCHA
atoms are then specified in ATCHRG, and the rest
are copied from the first set.
RAFO = array of three thresholds defining model systems
in FMO/AFO. All of them are multiplicative factors
applied to distances. Two atoms are considered
covalently bonded if they are separated by the
predefined distance determined by their van der
Waals radii. Larger RAFO values make further
separated atoms to be considered as bonded.
All atoms within RAFO(1) distance from BDA or BAA are
included into the model system in AFO ($FMOBND lists BDAs
and BAAs in this order as -BDA BAA). Atoms within RAFO(2)
from the set defined by RAFO(1) are replaced by hydrogens.
AO coefficients expanding localized orbitals to be frozen
are saved for use in FMO for atoms within RAFO(3) from BDA
or BAA. A nonzero RAFO(1) turns on FMO/AFO, else FMO/HOP is
used. Default: 0,0,0.
MODMOL = additive options for dimers and trimers in the
selective FMO based on MOLFRG.
1 limits correlated calculations to a) dimers/
trimers including one fragments in MOLFRG, and
b) monomers appearing in such dimers/trimers.
In other words, this is a cross option, to study
interactions between MOLFRG and the rest.
2 modifies the choice of dimers/trimers to those
in which all fragments are listed in MOLFRG
(i.e., option 2 requires also 1, resulting in 3).
In other words, this is an intra option, to study
interactions within MOLFRG.
4 do not store NFRAG**3 arrays in FMO3, to be used
with MODMOL=2, to reduce memory in very special
cases. No property summary will be provided, just
whatever is printed in SCF for each trimer.
8 do subsystem analysis. See MOLFRG.
Default: 0 (do not use MOLFRG)
NFRND = additive options controlling interface and
compatibility of GAMESS' FMO with other programs.
2 output basis set for each n-mer. Such an FMO
output can be split with tools/fmo/misc/frgout,
and thus obtained fragment output files can be
read into various GUIs (e.g., MacMolPlt), for
example to plot MOs of individual n-mers (but not
of the whole system), e.g., to help understand an
excited state calculation.
4 punch normal modes in RUNTYP=FMOHESS for
GUIs (e.g., MacMolPlt) to visualize vibrations.
This also prints a frequency table in the output.
8 write out coordinates
III. Parameters defining FMO approximations
MODESP = options for ESP calculations.
0 the original distance definition (uniform),
1 an improved distance definition (many-body
consistent, applied to unconnected n-mers),
2 an improved distance definition (many-body
consistent, applied to all n-mers).
(default: 0 (FMO2) or 1 (FMO3))
MODGRD = 0 subtract the external potential from the
Lagrangian (default).
1 do not do that.
2 add ESP derivatives (MODESP should be 0)
8 add Mulliken charge derivatives to MODGRD=2
16 do not add HOP derivatives (required for AFO)
32 add CPHF-related terms (known as SCZV) needed
for the fully analytic gradient, which may
be combined with EFP or PCM<1>.
This option requires MODESP=0 and for MP2
also RESPPC=0.
Note that 2+8 terms should be added, too,
so MODGRD=42 (=2+8+32) gives the fully analytic
gradient.
There are three main usages (some further limitations
are not listed below, e.g., for combinations with PCM
or EFP):
MODGRD=0 gives the least accurate gradient,
available for almost any FMO method
(except CIS and when ab initio gradient
in GAMESS is not available, e.g., CC).
MODGRD=10 is medium accurate, unavailable for CI,
CC, ROHF and MCSCF.
MODGRD=42 is analytic, only for RHF, UHF, ROHF,
RDFT, UDFT, and RMP2. RMP2 requires
RESPCC=0.
Note that RUNTYP=FMOHESS should use MODGRD=2, and
such runs cannot calculate analytic gradient.
Default: 10 (=2+8, for FMO2) or 0 (for FMO3).
RESPAP = cutoff for Mulliken atomic population approx,
namely, usage of only diagonal terms in ESPs.
It is applied if the distance between two monomers
is less than RESPAP, the distance is relative to
van der Waals radii; e.g. two atoms A and B
separated by R are defined to have the distance
equal to R/(RA+RB), where RA and RB are van
der Waals radii of A and B). RESPAP has no units,
as may be deduced from the formula.
RESPAP=0.0 disables this approximation.
(default: 0.0)
RESPPC = cutoff for Mulliken atomic point charge
approximation, namely replacing 2e integral
contributions in ESPs by effective 1e terms).
See RESPAP. (default: 2.0 (FMO2) or 2.5 (FMO3))
RESDIM = cutoff for approximating the SCF energy by
electrostatic interaction (1e terms), see RESPAP.
This parameter must be nonzero for ab initio
electron correlation methods. RESDIM=0 disables
this approximation. (default: 2.0 (FMO2) or
RITRIM(1)+RITRIM(3) for FMO3 energy, 0 for FMO3
gradient)
RCORSD = cutoff that is compared to the distance between
two monomers and all dynamic electron correlation
during the dimer run is turned off if the
distance is larger than this cutoff. RCORSD must
be less than or equal to RESDIM and it affects
only MP2, CC, CI, and TDDFT.
(default: 2.0 (FMO2), RITRIM(1)+RITRIM(4) for
FMO3 energy, 0 for FMO3 gradient)
RITRIM = an array of 4 thresholds determining neglect of
3-body terms (FMO3 only). The first three are for
uncorrelated trimers and the exact definition can
be found in the source code. The fourth one
neglects correlated trimers with the separation
larger than the threshold value. RITRIM(4) should
not exceed RITRIM(3).
(default: 1.25,-1.0,2.0,2.0, which corresponds to
the medium accuracy with medium basis sets, see
REFS.DOC).
SCREEN = an array of two elements, alpha and beta, giving
the exponent and the multiplicative factor
defining the damping function
1-beta*exp(-alpha*R**2).
This damping function is used to screen the
potential due to point charges of bond detached
atoms and it can only be applied for RESPPC=-1,
i.e., when ESP is approximated by point charges.
Default: 0,0 (no screening). Other sensible
values are 1,1.
ORSHFT = orbital shift, the universal constant that
multiplies all projection operators. The value of
1e+8 was sometimes erroneously quoted instead of
the actual value of 1e+6 in some FMO publications.
(default: 1e+6).
MAXKND = the maximum number of hybrid orbital sets (one set
is given for each basis set located at the atoms
where bonds are detached). See also $FMOHYB.
(default: 10)
MAXCAO = the maximum number of hybrid orbitals in an LMO
set. (default: 5)
MAXBND = the maximum number of detached bonds.
(default: NFG*2+1)
==========================================================
==========================================================
467 lines are written.
Edited by Shiro KOSEKI on Mon Feb 13 10:50:16 2017.