Skip to content

CASSCF Calculation

CASSCF

The majority of calculations conducted with HUMMR will be based on a CASSCF or MCSCF calculation as it is the central module of the program. The following is an example for a CASSCF input file.

casscf.inp
General
  CalcType CASSCF
  Charge 0
  Mult 1
  OrcaJSONName orca.json
  Ints RI
  Basis def2-svp
  AuxBasis def2-JK
End

CASSCF
  NEl 4
  NOrb 4
  NRoots 1
  CISolver FCI
  MaxIter 35
  OrbStep SuperCIPTDIIS
  SwitchOrbStep DIIS
End

Geom
  N 0 0 0
  N 0 0 1.1
End

Most keywords in the General and CASSCF blocks are self-explanatory. For example, the calculation above will feature 4 active electrons in 4 active orbitals and is allowed to run over 35 orbital optimization cycles. Note, that in HUMMR, both the total energy and orbital rotation gradient are required to reach convergence. The choice of orbital optimization algorithm will be critical in this regard. In the sample input, the perturbative approach of Kollmar et al. was chosen as standard approach and will be changed to a combination of Pulay’s DIIS algorithm and Staemmler's first-order method once the norm of the orbital gradient has fallen below a threshold (cf. SwitchOrbStepThresh).1,2,3 Other options are given above and we strongly advise to make use of different combinations when convergence is difficult to achieve.

An important aspect of CASSCF, MCSSCF and related calculations (e.g. SC-NEVPT2) concerns the generation and handling of two-electron repulsion integrals. HUMMR utilizes the open-source implementation of Frank Neese's SHARK4 integral generation and digestions system provided by Lible5 to evaluate two-electron integrals. To reduce the fast-growing computational cost related to the straightforward calculation of four-center integrals, it is possible to employ the density fitting or resolution-of-the-identity (RI) approximation.6 It is invoked by Ints RI in the General block of the input file (see example above). Note, that you have to provide a suitable auxiliary basis set if the RI approximation is used (see BASIS SETS).


Starting Orbitals

At this point it should also be noted that the quality of the starting orbitals is another critical ingredient of any CASSCF calculation and will greatly influence its convergence behavior. Hence, it is required to read a set of molecular orbital coefficients from a previous HUMMR or ORCA6 calculation as guess. If orbitals from an ORCA calculation should be utilized the name of the .json file needs to be provided after the keyword OrcaJSONName while in case of HUMMR orbitals the name of the orbital file needs to be precluded by OrbGuessName.

If the orbital coefficients that are read in as guess refer to a different basis set than the one in the current calculation the original basis set needs to be provided after the InputOrbitals keyword. The program will then project the orbital coefficients on the current atomic orbital basis. With the OrbGuessRotation keyword the order of the guess orbitals can be changed.

If desired, the set of orbitals that results from a HUMMR calculation (stored in the calcname.C0 file) can be read into HUMMR as guess for any subsequent calculation. In addition, HUMMR produces a file calcname.orca.json that can be converted to a .gbw file by the orca_2json program.


Implicit Solvation

As many chemical reactions occur not in the gas phase but in solution, HUMMR offers the possibility to simulate solvation with an conductor-like polarizable continuum model (C-PCM or COSMO). If you wish to include implicit solvation you simply need to add

1
2
3
ImplicitSolvation
   Epsilon 4.81          
End

to your input file. It will invoke C-PCM with a dielectric constant of 4.81 which corresponds to chloroform. Of course, one may change some technical setting for the implicit solvation model, i.e.

casscf_w_solv.inp
ImplicitSolvation
   Solver PCG            # solve cpcm-equations with pcg, or explicit inversion 
   Solvent CHCL3         # uses preset epsilon for solvent (see manual)
   Epsilon 4.81          # manually set epsilon for solvent
   LebedevOrder 8        # defines lebedev order for grid generation
   Model CPCM            # CPCM or COSMO
   RadialScaling 1.2     # The VdW radii are scaled up by this factor. 
   UpdatingScheme DIRECT # solve cpcm equations (direct) or second order opt (gradient)
   ChargeRepresentation pointlike # pointlike or Gaussian charge distribution
End
A more detailed description of the solvation-related keywords can be found at IMPLICIT SOLVATION.


  1. P. Pulay, Chem. Phys. Lett. 1980, 73, 393–398. 

  2. P. Pulay, J. Comput. Chem. 1982, 3, 556–560. 

  3. U Meier, V Staemmler, Theor. Chim. Acta 1989, 76, 95–111. 

  4. F. Neese, J. Comput. Chem 20223, 44, 381–396 

  5. https://github.com/MihkuU/Lible 

  6. F. Neese, F. Wennmohs, U. Becker, C. Riplinger, J. Chem. Phys. 2020, 152, 224108.