How to run a calculation

Options/Description

on: The "alternative" Davidson scheme - see Kitsaras, M.-P. (2023) https://doi.org/10.25358/openscience-9599 - will be used for the EOM-CCn calculation (default).

 off: Standart Davidson will be used for the EOM-CCn calculation. Wrong option is equivalent to on. Do not use with davidson-method=multi

on: Depending on the topic (i.e., equation-of-motion or coupled-cluster, the eom amplitudes or the cc amplitudes are enforced to be antisymmetric. 

off: the amplitudes are not explicitly antisymmetrized and may contain numerical noise (default). 

on: Stops the generalized Davidson (multi, see "davidson-method"), when all of the roots are locked and half of them are converged to converge the rest of the roots with the single Davidson algorithm. 

off: the "multi Davidson scheme" stops when maximum number of iterations has been reached (default)

yes: depending on topic, the abcd intermediate for EOM-CCSD or CCSD is built (default). 

no/false/0: abcd intermediate is not built.

Give a number. The reference state CC3 and CCSDT calculations make use of the Matthews scheme for intermediate CCSD subiterations which accelerates convergence. Specify number of sub-iterations. Maximum 5. Default 0.

on: Assumes that Cholesky vectors (CVs) rather than two-electron integrals were written out by the SCF program and reads CVs. 

off: Assumes no Cholesky decomposition (default)

Depending on the topic, specifies either the convergence criterion for the reference-state or the excited-state calculations. Default 1e-7

on: The program creates only the interface data MOL_ORB for the COrbit19 visualisation program and then stops. 

off: The program runs normally (default)

1: diabatic couplings following Pieniazek, Arnstein, Bradforth, Krylov, Sherrill, J. Chem. Phys. 2007, 127, 164110 are calculated

0/false: diabatic couplings are not calculated (default)

single: The Davidson method will converge one root at a time (default).

 multi: The generalized Davidson method (see SIAM, J. Sci. Comput., 15, 62, 1994) will be used. It attempts to converge multiple roots in one run. Best used in combination with auto-multi-break=on. Avoid using with alternative-davidson=on.

Specifies the spin change for the spin/particle non-conserving variants of EOM 

-1: -1 for SF-EOM and -1/2 for EA/IP-EOM. 

 0: for EE-EOM. 

+1: +1 for SF-EOM and +1/2 for EA/IP-EOM

Defines the maximum DIIS subspace for the DIIS extrapolation of the reference state energy and properties or the properties of the excited states. Default 7

Specifies the electric field strength in the x, y, and z direction for the calculation in the presence of electric field in a.u. Give three numbers separated by a comma. x,y,z. E.g. 5e-5,0,0 

orbital-unrelaxed/orb-unr: turns the electric field after the converged HF solution. 

off: no electric field (default)

none: no excited state calculation (default)

ee: for EE-EOM

sf: for SF-EOM

ea: for EA-EOM

ip: for IP-EOM

cis: roots at the CIS level

cisd: performs a CISD calculation. Specify number-of-roots=1 if only CISD ground state energy is wanted.

Specifies the orbitals to freeze: divide single orbitals of the same irrep with a comma "," and the different irreps with a slash "/". To specify a range, use the dash  "-".  qp must specify the quasi-particle space ("occ" for the hole/occupied or "virt" for the particle/virtual space). spin must specify the α or β robitals using "a" or "b", respectively. Example 1: frozen-orbitals_occ_a=1,2/0/2/1,3. Example 2: frozen-orbitals_virt_b=1/0/1-4/0. Note that the enumeration of the irreps is given in the start of the output.

Options:
off/eneronly/energy-only/0: does not calculate analytic gradients. (Default)
cc/analytic-cc/1: calculates CC analytic gradients and writes them on a CFOUR interface data on the disk if the respective properties have been set.
eom/analytic-eom/2: calculates EOM-CC analytic gradients and writes them on a CFOUR interface data on the disk if the respective properties have been set for the EOM state specified in gradient-state.

Specifiy the EOM state to be optimized in the format IRREP.STATE starting from STATE=1. Example gradient-state=2.2 for the second state of the second IRREP

Εntries (0 or 1) are specified for each irrep, seperated by "/". E.g. 1/0/0/1. If one entry is "1", "guessval" is interpreted as double excitation from α to α and β to β orbitals and as a single excitation for "0". Only works for a closed-shell reference. Example: a double-excited vector as a guess for second irrep where both an α- and a β-electron have been excited from occupied orbital 5 to virtual orbital 1 with a coefficient of (1.0 + i 0.0) . guessdouble=0/1/0/0 guesspos=/5,1// guessval=/1.,0//. 

List of integers separated by "," for each irrep seperated by "/".E.g. 1,2/1,3. The numbers are interpreted as pairs of occ/virt indices to be given as manual guesses for an initial r vector. Only works for a closed-shell reference. No symmetry check is done here. Use carefully. For ip: only give occ orbital, for ea: only give virt orbital

List of numbers seperated by "," for each irrep seperated by "/". E.g. 1.0,0.0/0.0,1.0. The numbers are interpreted as the real and imaginary part of the manual guess for the position given in guesspos. Example: a vector with two single-excitations as a guess where an α- and a β-electron have been excited from occupied orbital 5 to virtual orbital 1 with a coefficient of (1.0 + i 0.0) for the EOM-EE variant: guessdouble=0 guesspos=5,1 guessval=1.0,0.0

Alternative keyword for program

Method to calculate the CC jacobian. To be used in combination with "newton-type=multilevel". At the moment following combinations are implemented. see J. Chem. Phys. 153, 014104 (2020) https://doi.org/10.1063/5.0010989

1. coupled-cluster = ccsd   && jacobian-method = ccsd/cc2

2. coupled-cluster = cc3   && jacobian-method = ccsd

Specifies the lock-criterion for the roots in the Davidson procedure. While root is not locked, the initial guess for the EOM-eigenvector r is used a reference to choose the eigenvector of the current iteration. When root is locked, the eigenvector of the previous iteration will be used instead. Note: if "davidson-method=multi", unlocked roots are chosen based on their relative energy ordering. Default 1e-1

Maximum number of basis vectors for the Davidson procedure. Default 30.

Maximum number of iterations for the reference state energy. Default 50.

Maximum number of iterations for the excited states calculation. Default 30

Maximum number of iterations for the calculation of properties. Default 30

Direction of magnetic field (x/X, y/Y or z/Z) needed for the output files of mcd. E.g. mcd = x . Default: none. MCD will only be calculated with Response Properties (Prop = 2)

d: dipole moments,

t: transition moments,

et: transition moments between excited states

Default none.

Combinations are allowed (dt, det). Single-state properties are always calculated. Transition properties are unavailable, when orbital-relaxed properties are asked for.

Choose the multiplicity of the EOM roots. 1 (singlets) or 3 (triplets). Only works for closed-shell references. In the CIS guess, the spin-expectation value is calculated and only the roots with the chosen multiplicity are used for the subsequent EOM-EE (or CISD) calculation. If nothing is specified, multiplicity of the roots is ignored (default). 

Specifies how the CC-amplitudes are formed. 

1. denominator: The amplitudes are formed via the orbital denominator (default)

2. multilevel: The amplitudes are formed in a newton type fashion using the gradient of the CC Lagrangian. If the same level it is used it is the exact jacobian if a lower method is used it is an approximate scheme. To select the method for the jacobian calculation the keyword jacobian-method has to be set. see J. Chem. Phys. 153, 014104 (2020) https://doi.org/10.1063/5.0010989

numbers separated by "/" for each irrep. E.g. 1/0/2/5. Specifies the number of roots for the excited-states calculation per irrep. Specify here the symmetry of the excitation (rather than of the final excited state).

default: -1, options: 3,4. Performs some contractions on-the-fly. The number indicates the number of virtual indices of the kind of intermediates which is to be calculated on the fly; 

4: intermediates with four virt. indices are calculated on the fly (abcd)

3:(abcd) and (abci) intermediates are calculated on the fly

This might save memory at the expense of computational speed. 

For the symmetry recognition in the interface to LONDON, this number specifies the threshold for the orbital rotation recognition. Default 1e-2

Provide a name to write the vectors for the natural orbitals in case they were calculated. 

Provide a name to write the vectors for the natural transition orbitals in case they were calculated.

hdiag: use orbital differences for the preconditioner to be used by the davidson procedure (default). 

hbardiag: use the diagonal elements of the similarity transformed Hamiltonian matrix as the preconditioner. This option is not recommended and may have bugs (doesn't improve convergence) 

cfour: Assumes interface data from CFOUR. 

london: Assumes interface data from LONDON

For "coupled-cluster": 

0/none: No properties are calculated (default). 

1/lamda/unrelaxed: the Λ-amplitude-equations are solved and properties requiring the one-electron density matrix are calculated, with no inclusion of orbital relaxation. 

2/relaxed: the Λ-amplitude equations are solved and properties requiring the one- and two- electron density matrix are calculated with inclusion of orbital relaxation of the reference. 

For "equation-of-motion". 

0/none: no properties are calculated. 

1/eom/eom-unrelaxed: excited-state properties requiring the one-electron density matrix are calculated at the expectation value level with no orbital relaxation. 

2/response/response-unrelaxed: excited states requiring the one-electron density matrix are calculated at the response level (relaxation of the T-amplitudes) with no inclusion of orbital relaxation. 

3: non-functional. 

4/eom2/eom-relaxed: excited-state properties requiring the one- and two-electron density matrix are calculated at the expectation value level, with inclusion of orbital relaxation. 

5/response2/response-relaxed: excited state-properties requiring the one- and two-electron density matrix are calculated at the response level (inclusion of T-amplitude relaxation), with inclusion of orbital relaxation (full analytic derivatives). 

Option 4 and 5 do currently #not calculate transition properties.

Depending on the topic, it specifies the convergence criterion for the reference-state or the excited-state properties, respectively. Default 1e-7

Give the truncation for the cluster operator including photonic indices. Example qed=sd-1 includes one-photon singles and doubles and single photons.

off: Calculation with no photons (Default)
sd-1: Calculations with up to 1 photon inclusion with singles and double mixed amplitudes
sd-12 Calculations with up to 2 photon inclusion with singles and double mixed amplitudes

Give file names separated by "," for each irrep separated by "/". The files must contain eom vectors to be used as a guess for an excited-state calculation. Example: readguess=solvvec1.1, solvvec1.2/solvvec2.1

 RHF: For CCSD/CCSDT contractions onto CC-amplitudes skip spin-blocks bb, aaaa, bbbb, aaaaaa, bbbbbb. 

Reconstruct skipped blocks for DIIS.

UHF: (default)

on: for the "davidson-method=multi" option, in every iteration all basis vectors (rather than the most recent ones) are orthogonalized against each other by Gram-Schmidt orthogonalization. 

Default: off

 on: Performs contractions in single-precision (but switches back to double-precision for DIIS/Davidson). Also if convergence higher than single-precision (1e-7) is demanded the double-precision is recovered in the last iterations.

off: (default)

List of integers separated by "," for each irrep separated by "/".E.g. 1,5,6/0. The numbers are interpreted as CIS first guesses not to be used in the subsequent EOM (or CISD) calculation. 

on: detailed contributions are printed for the calculation of the spin-multiplicity.

off: shorter output for the multiplicity calculation (default).

Note: spin-expectation values require the two-electron density matrices and are hence calculated when properites=2 in coupled-cluster and properties=4 or 5 in equation-of-motion CC.

 Single-Value-Decomposition (Experimental)

on:

off: (default)

 Single-Value-Decomposition Threshold (see keyword svd)

Default: 1e-6

Regulates the symmetry recognition in the case of a calculation using the LONDON interface. 

off/c1: No symmetry is to be used (advised for non-abelian groups). 

auto: symmetry recognition will choose the group with the highest order. 

Give specific name to restrict the symmetry recognition to the symmetry specified.

For the symmetry recognition in the interface to LONDON, this number specifies the threshold for the recognition of symmetry elements for the molecular system (including the magnetic field). Default 1e-4

on: for the reference-state calculation in case of CCSDT or CC3, a CCSD calculation is performed in advance in order to be used as a guess for the subsequent calculation involving triple excitations.

off: CCSDT and CC3 calculations use an MP2 guess (default).

none: no CC calculation is performed. 

ccsd: CCSD calculation (default). 

ccsdt: CCSDT calculation. 

ccsd(t): CCSD(T) calculation. 

ccsdt-none: no CC calculation is performed, amplitudes are read from file, CCSDT assumed for EOM (may have bugs). 

ccd: CCD calculation. 

ucc3/2: UCC3/2 calculation. 

adc3: ADC3 calculation. uccsd4: non-functioning, not debugged. 

cc2/3: CC2/3 calculation. 

ccsd*: CCSD calculation for the reference state and EOM-CCSD* calculation for the excited states. 

ccsd(t)(a)*. CCSD(T)(a) calculation for the reference state and CCSD(T)(a)* for the excited states.

1/true: CC amplitudes are written to disk

0/false: the CC amplitudes are not written to disk (default)

writes the one-electron density on disk as an interface data file for COrbit19 according to the properties level asked if "on"

on: write the vectors of the natural orbitals in case they were calculated. 

off: default

on: write the vectors of the natural transition orbitals in case they were calculated. 

off: default

list of integers separated by "," for each irrep separated by "/". E.g. 1,5,6/0. The numbers are interpreted as converged EOM-vectors to be written to disk