 |
|
|
$CCINP group (optional, relevant for any CCTYP)
This group controls a coupled-cluster calculation
specified by CCTYP in $CONTRL. The reference orbitals may
be RHF or high spin ROHF. If this group is not given, as
is often the case, all valence electrons are correlated.
Several ground state CCTYP choices obey at least a few of
the keywords from $EOMINP, so please see that group too.
Excited state runs such as CCTYP=EOM-CCSD or CR-EOML read
$CCINP to define orbital ranges for the ground state CCSD
prior to generating excitations under $EOMINP's control.
A number of CCTYP choices have been superceded by more
advanced equations. For example, R-CC and CR-CC were
developed prior to their CR-CCL replacement, while CR-EOML
supercedes CR-EOM. CR-CCL provides a good approximation to
the fully iterated CCSDT method, and so is superior to the
familiar CCSD(T). A reasonable menu is:
RHF ROHF (high spin)
--- ----------------
ground states [properties]:
CCSD [CCPRP] CCSD [n/a]
CCSD(T) n/a
CR-CCL CR-CCL
excited states [properties]:
EOM-CCSD [CCPRPE] EOM-CCSD
MULT=1 MULT=1,3,5 or
spin-contaminated
CR-EOML n/a
ionization processes:
EA-EOM3a n/a
IP-EOM3a n/a
MULT=2,4
CR-CCL =left-CCSD + CR-CC(2,3) perturbative triples.
CR-EOML=left-EOM-CCSD + CR-CC(2,3) perturbative triples.
Parallel computation is possible for RHF references
only, and only for CCTYP=CCSD or CCSD(T). Memory use in
parallel runs is exotic: use EXETYP=CHECK with PARALL in
$SYSTEM set on prints the per node memory requirements.
See the "Further Information" section of this manual
for more details about coupled-cluster runs.
**** The first four pertain to both RHF and ROHF ****
NCORE = gives the number of frozen core orbitals to be
omitted from the CC calculation. The default
is the number of chemical core orbitals.
NFZV = the number of frozen virtual orbitals to be
omitted from the calculation. (default is 0)
MAXCC = defines the maximum number of CCSD (or LCCD, CCD)
iterations. This parameter also applies to ROHF's
left CC vector solver, but not RHF's left vector.
See MAXCCL for RHF. (default=30)
ICONV = defines the convergence criterion for the cluster
amplitudes, as 10**(-ICONV). The ROHF reference
also uses this for its left eigenstate solver, but
see CVGEOM in $EOMINP for RHF references.
(default is 7, but it tightens to 8 for FMO-CC.)
**** the next group pertains to RHF reference only ****
CCPRP = a flag to select computation of the CCSD level
ground state density matrix (see also CCPRPE in
$EOMINP for EOM-CCSD level excited states). The
computation takes significant extra time, to
obtain left eigenstates, so the default is .FALSE.
except for CCTYP=CR-CCL or CR-EOML, where the work
required for properties must be done anyway.
This keyword is only available in serial runs.
Notes: CCSD is the only level at which properties can be
obtained. Therefore this option can only be chosen for
CCTYP=CCSD, CR-CCL, EOM-CCSD, CR-EOM, or CR-EOML. A CCSD
run requesting CCPRP=.TRUE. will internally change itself
to EOM-CCSD to run the left CCSD, but since NSTATE of
$EOMINP will still be zero, this remains a ground state
calculation. Note that the convergence criterion for left
eigenstates is CVGEOM in $EOMINP, which is set to obtain
excitation energies, and may need tightening.
There is little reason to select any of these:
MAXCCL = iteration limit on the left eigenstate needed by
CCSD properties, or CR-CCL energies.
This is just a synonym for MAXEOM in $EOMINP.
If you want to alter the left state's convergence
tolerance, use CVGEOM in $EOMINP. The right CCSD
state's convergence is set by MAXCC and ICONV.
NWORD = a limit on memory to be used in the CC steps.
The default is 0, meaning all memory available
will be used.
IREST = defines the restart option. If the value of IREST
is greater or equal 3, program will restart from
the earlier CC run. This requires saving the disk
file CCREST from the previous CC run. Values of
IREST between 0 and 3 should not be used. In
general, the value of IREST is used by the program
to set the iteration counter in the restarted run.
The default is 0, meaning no restart is attempted.
MXDIIS = defines the number of cluster amplitude vectors
from previous iterations to be included in the
DIIS extrapolation during the CCSD (or LCCD, CCD)
iterative process. The default value of MXDIIS is
5 for all but small problems. The DIIS solver can
be disengaged by entering MXDIIS = 0. It is not
necessary to change the default value of MXDIIS,
unless the CC equations do not converge in spite
of increasing the value of MAXCC.
AMPTSH = defines a threshold for eliminating small cluster
amplitudes from the CC calculations. Amplitudes
with absolute values smaller than AMPTSH are set
to zero. The default is to retain all small
amplitudes, meaning fully accurate CC iterations.
Default = 0.0.
**** the next group pertains to ROHF reference only ****
There is little reason to select any of these.
MULT = spin multiplicity to use in the reference
determinant during the CC computation.
The value of MULT given in the $CONTRL input
determines the spin state for the ROHF orbital
optimization, and is the default for the CC.
It would be quite unusual to use a different
spin in the SCF than in the CC. The MULT keyword
in $EOMINP is of greater physical interest.
IOPMET = method for the CR-CC(2,3) triples correction.
= 0 means try 1 and then try 2 (default)
= 1, the high memory option
This option uses the most memory, but the least
disk storage and the least CPU time.
= 2, the high disk option
This option uses least memory, by storing a large
disk file. Time is slightly more than IOPMET=1,
but the disk file is (NO**3 * NU**3)/6 words,
where NO = correlated orbitals, and NU= virtuals.
= 3, the high I/O option
This option requires slightly more memory than 2,
and slightly more disk than 1, but does much I/O.
It is also the slowest of the three choices.
Check runs will print memory needed by all three options.
KREST = 0 fresh start of the CCSD equations (default)
= 1 restart from AMPROCC file of a previous run
KMICRO = n performs DIIS extrapolation of the open shell
CCSD, every n iterations (default is 6)
Enter 0 to avoid using the DIIS converger.
LREST = 0 fresh start of the left CCSD equations (default)
= 1 restart from AMPROCC file of a previous run
LMICRO = n performs DIIS extrapolation of the open shell
left equations, every n iterations (default is 5)
Enter 0 to avoid using the DIIS converger.
KMICRO and LMICRO are ignored for trivial
problem sizes.
==========================================================
==========================================================
generated on 7/7/2017