Kramers unrestricted Coupled-Cluster methods

See [Visscher1996] for the initial CCSD(T) implementation. The Fock-space CC implementation is described in [Visscher2001].

**RELCC

Specification of reference determinant, type of calculation, and general settings.

.FOCKSP

Activate the Fock space module. This option should be used for multireference calculations. See further *CCFSPC. Because the first sector of Fock space gives the same result as a regular CCSD calculation, the latter calculation is switched off. If you do wish to perform also a regular CC calculation (e.g. to get the CCSD(T) energy) you need to activate this explicitly via .ENERGY (see below).

.ENERGY

Activate the energy calculation. This is the preferred option for calculations on closed shell or simple open shell systems and need not be specified explictly in such cases. For Fock space calculations the keyword switches on a separate single reference calculation done prior to the FS calculation.

Default:

Perform energy calculation.

.GRADIENT

Calculate the effective 1-particle density matrix. This option can be used to calculate molecular properties. For closed-shell MP2 see [vanStralen2005]. For closed-shell CC see [Shee2016].

Default:

No gradient calculation.

.NELEC

Number of correlated electrons. This variable determines the reference determinant to be used in the exponential expansion of the wave function. Since the default values correspond to the information passed on by the MOLTRA code on basis of the Hartree-Fock occupations and chosen range of active orbitals in MOLTRA, for CLOSED SHELL systems there is usually no need to specify this variable manually. If you do chose to specify this manually, you should make sure to count electrons carefully as the numbers relate to the number of correlated electrons rather than the total number. This number may therefore change if you change tresholds in the integral transformation.

For OPEN SHELL systems it is usually better to employ the multireference Fock space approach, except for simple cases such as a high-spin open shell state. In this case the single reference CCSD(T) ansatz works rater well. For such cases it is easier to use the .NELEC_OPEN keyword that allows to just specify the distribution of open shell electrons over the irreps, but for backwards compatibility we retain the older NELEC option as well.

Arguments:

Integer (NELEC(I),I=1,NFSYM*2).

Default:

Number of correlated electrons in closed shells in these irreps (written by **MOLTRA).

.NELEC_OPEN

Distribution of correlated open shell electrons over the irreps. This determines the reference determinant for open shell single reference calculations.

This input should always be given for average-of-configuration SCF calculations that are followed by a CC calculation. The total number of electrons that is given should correspond to the number of open shell electrons.

Arguments:

Integer (NELEC_OPEN(I),I=1,NFSYM*2).

Default:

Zero. (note that this default leads to wrong results because a closed shell ion will be calculated if no input is given).

.NEL_F1

Number of electrons in the gerade irreps of the Abelian symmetry group. This works like the NELEC keyword, but uses the supersymmetry possible for linear groups in which irreps are ordered as 1/2, -1/2, 3/2, -3/2, 5/2, ... (with the number being the m_j value).

.NEL_F2

Number of electrons in the ungerade irreps of the Abelian symmetry group.

.PRINT

Print level.

Default:

.PRINT
 0

.TIMING

Print detailed timing information.

Default:

Only limited timing information is printed.

.DEBUG

Print debug information.

Default:

Debug information is not printed.

*CCENER

Covers options related to energy.

.NOMP2

Deactivate MP2 calculation.

.NOSD

Deactivate CCSD calculation.

.NOSDT

Deactivate the calculation of perturbative triples. This is potentially useful when running into memory problems for very big calculations and will also save some CPU time.

.MAXIT

Set maximum number of iterations allowed to solve the CC equations.

.MAXDIM

Set maximum number of amplitude vectors used in the DIIS extrapolation.

.NTOL

Specify requested convergence (10^-NTOL) in the amplitudes.

.NOSING

Eliminate T1 amplitudes in the calculation (only interesting for test purposes, this gives no computational speed-up).

.NODOUB

Eliminate T2 amplitudes in the calculation (only interesting for test purposes, this gives no computational speed-up). Deactivate contribution from doubles; corresponds to a CCS calculation.

*CCFOPR

Calculate first-order properties (expectation values) for the MP2 and CCSD wave function.

.MP2G

Use MP2 wave function.

.CCSDG

Use CCSD wave function.

.NATORB

Calculate natural orbitals (currently only for MP2 density matrix)

.RELAXED

Use orbital-relaxed density matrix (currently only for MP2). The current default for MP2 and CC wave functions is to use the unrelaxed density matrix. This is computationally less expensive, but also less accurate.

*CCFSPC

Perform a Fock space MRCC calculation in which a model space is correlated and then diagonalized to give CC energies for a set of states.

.DOIH

Use the Intermediate Hamiltonian formalism in which an auxiliary space is used to prevent the “intruder state” problem. Default: IH formalism not used.

.DOEA

Calculate electron affinities (add one electron to the reference state, allowing occupation of the active virtual orbitals)

.DOIE

Calculate ionization energies (remove one electron from the reference state, allowing depletion of the active occupied orbitals)

.DOEA2

Calculate second electron affinities (add two electrons to the reference state, allowing occupation of the active virtual orbitals)

.DOIE2

Calculate second ionization energies (remove two electrons from the reference state, allowing depletion of the active occupied orbitals)

.DOEXC

Calculate excitation energies (allow excitation from the set of active occupied orbitals to the set of active virtual orbitals)

.NACTH

Specification of the set of active hole orbitals (from which ionization/excitation takes place)

.NACTP

Specification of the set of active particle orbitals (to which electron attachment/excitation takes place)

.MAXIT

Maximum number of iterations allowed to solve the FSCC equations

.MAXDIM

Set maximum number of amplitude vectors used in the DIIS extrapolation.

.NTOL

Specify requested convergence (10^-NTOL) in the amplitudes.

.GESTAT

Specify the state number in the last active sector to pick the energy from (remember to account for degeneracies) for a state-specific FSCC geometry optimization based on a numerical gradient.

*CCIH

Options for intermediate hamiltonian in FSCC.

.EHMIN

Minimum orbital energy of occupied orbitals forming the auxiliary (Pi) space. Orbitals with energies lower than this energy are taken in the secundary (Q) space and do not contribute to the model space.

low limit of orbital energies of active occupied orbitals, which constitute the secondary Pi space. Could be used in (1,0), (2,0) and (1,1) sectors. Arguments: real.

.EHMAX

Maximum orbital energy of occupied orbitals forming the auxiliary (Pi) space. Orbitals with energies higher than this energy are taken in the primary (Pm) space.

This is upper limit of one-electronic energies of active occupied orbitals, which constitute the secondary Pi space. Could be used in (1,0), (2,0) and (1,1) sectors. Arguments: real.

.EPMIN

Minimum orbital energy of virtual orbitals forming the auxiliary (Pi) space. Orbitals with energies lower than this energy are taken in the primary (Pm) space.

This is the low limit of orbital energies of active virtual orbitals, which constitute the secondary Pi space. Could be used in (0,1), (0,2) and (1,1) sectors. Arguments: real.

.EPMAX

Maximum orbital energy of virtual orbitals forming the auxiliary (Pi) space. Orbitals with energies higher than this energy are taken in the secundary (Q) space and do not contribute to the model space.

This is the upper limit of one-electronic energies of active virtual orbitals, which constitute the secondary Pi space. Could be used in (0,1), (0,2) and (1,1) sectors. Arguments: real.

Other Intermediate Hamiltonian (IH) input parameters

For experts only.

Following keyowrds belong to the CCIH namelist section.

.IHSCHEME

Choose particular IH scheme. Arguments: Integer IHSCHEME = 1, or 2.

The IHSCHEME =1 corresponds to the extrapolated IH (XIH) approach, described in the paper [Eliav2005].

Main idea: proper modification of the energetic denominators, containing problematic Pi -> Q transition. The original denominator 1/(E_Pi - E_Q) , used during CC iterations, is substituted by the following expression (1)

\[\frac{(1-[\frac{AIH*SHIFT}{(E_{Pi} - E_{Q} + SHIFT)}]^{NIH})}{\frac{(1-AIH*SHIFT}{(E_{Pi} - E_{Q} + SHIFT))}},\]

here AIH, SHIFT,NIH are parameters, specially chosen for overcoming of the intruder states problem. These parameters could be used in the procedure of the extrapolation of the “exact” effective Hamiltonian solutions from corresponding IH CC energies and wave functions.

IHSCHEME =2 corresponds to the simplified IH-2 approach, described in the paper [Landau2004].

Here the problematic denominators \(1/(E_{Pi} - E_{Q})\) are substituted simply by the factor 0.

Default: IHSCHEME = 2

Next key options are used only in case of XIH (IHSCHEME = 1).

.SHIFTH11

Energy shift for the one-electronic excitations in (1,0) sector. Arguments: real.

.SHIFTH12

Energy shift for the two-electronic excitations in (1,0) sector. Arguments: real.

.SHIFTH2

Energy shift for the two-electronic excitations in (2,0) sector. Arguments: real.

.SHIFTP11

Energy shift for the one-electronic excitations in (0,1) sector. Arguments: real.

.SHIFTP12

Energy shift for the two-electronic excitations in (0,1) sector. Arguments: real.

.SHIFTP2

Energy shift for the two-electronic excitations in (0,2) sector. Arguments: real. Usually we choose the approximate difference between the highest orbital energy belonging to Pi and the lowest orbital energy belonging to the Pm space. Works only with the old style of RELCC input.

.SHIFTHP

Energy shift for the two-electronic excitations in (1,1) sector. Arguments: real

.AIH

Compensation factor, used in expression (1). Arguments: real positive, not greater then 1.0.

.NIH

Compensation power, used in expression (1). Arguments: integer.

In the case of the limit: AIH=1.0 and NIH -> “infinity” ( NIH>100, in practice) we have so called “full compensation” method, corresponding to the extrapolation of the effective Hamiltonian from the intermediate one.

*CCSORT

Specialist options related to the sorting of two-electron integrals and the calculation of the reference Fock matrix.

.NORECMP

Do not recompute the Fock matrix, but assume a diagonal matrix with the orbital energies taken from the SCF program on the dioagonal. This is usually not recommended as the latter correspond to a restricted open shell expression and RELCCSD uses an unrestricted formalism. For closed shell systems the two expressions are identical and this option merely suppresses a build-in check on the accuracy of transformed integrals.

.USEOE

Ignore recomputed Fock matrix and use orbital energies supplied by the SCF program. This option is sometimes useful for degenerate open shell cases in which case the perturbation theory for the unrestricted formalism is not invariant for rotations among degenerate orbitals. It should only change the outcome of the [T], (T) and -T energy corrections.