PC GAMESS/Firefly CIS MODULE DOCUMENTATION. $CIS group required when CITYP=CIS The CIS method (singly excited CI) is the simplest way to treat excited states. By Brillouin's Theorem, a single determinant reference such as RHF will have zero matrix elements with singly substituted determinants. The ground state reference therefore has no mixing with the excited states treated with singles only. Reading the references given in Section 4 of this manual will show the CIS method can be thought of as a non-correlated method, rigorously so for the ground state, and effectively so for the various excited states. Some issues making CIS rather less than a black box method are: a) any states characterized by important doubles are simply missing from the calculation. b) excited states commonly possess Rydberg (diffuse) character, so the AO basis used must allow this. c) excited states often have different point group symmetry than the ground state, so the starting geometries for these states must reflect their actual symmetry. d) excited state surfaces frequently cross, and thus root flipping may very well occur. The CIS code in the PC GAMESS/Firefly is based on heavily modified sources of the original GAMESS (US) AO-basis CIS code written by Dr. Simon P. Webb (Ref. 3 below). The implementation allows the use of only RHF references, but can pick up both singlet and triplet excited states. Nuclear gradients are available, as are properties. NCORE = n Omits the first n occupied alpha and beta orbitals from the calculation. The default for n is the number of chemical core orbitals. NSTATE = Number of states to be found (excluding the ground state). ISTATE = State for which properties and/or gradient will be calculated. Only one state can be chosen. HAMTYP = Type of CI Hamiltonian to use. = SAPS spin-adapted antisymmetrized product of the desired MULT will be used (default) = DETS determinant based, so both singlets and triplets will be obtained. Note that this option will disable use of FASTINTS/GENCON code during direct CIS runs. MULT = Multiplicity (1 or 3) of the singly excited SAPS (the reference is necessarily single RHF). Only relevant for SAPS based run. DIAGZN = Hamiltonian diagonalization method. = DAVID use Davidson diagonalization. (default) = FULL construct the full matrix in memory and diagonalize, thus determining all states (not recommended except for small cases). DGAPRX = Flag to control whether approximate diagonal elements of the CIS Hamiltonian (based only on the orbital energies) are used in the Davidson algorithm. Note, this only affects the rate of convergence, not the resulting final energies. If set .FALSE., the exact diagonal elements are determined and used. Default=.TRUE. NGSVEC = Dimension of the Hamiltonian submatrix that is diagonalized to form the initial CI vectors. The default is the greater of NSTATE*2 and 10. MXVEC = Maximum number of expansion basis vectors in the iterative subspace during Davidson iterations, before the expansion basis is truncated. The default is the larger of 8*NSTATE and NGSVEC. NDAVIT = Maximum number of Davidson iterations. Default=50. DAVCVG = Convergence criterion for Davidson eigenvectors. Eigenvector accuracy is proportional to DAVCVG, while the energy accuracy is proportional to its square. The default is 1.0E-05. CISPRP = Flag to request the determination of CIS level properties, using the relaxed density. Relevant to RUNTYP=ENERGY jobs, although the default is .FALSE. because additional CPHF calculation will be required. Properties are computed as a normal byproduct of runs involving the CIS gradient. CHFSLV = Chooses type of CPHF solver to use. = CONJG selects an ordinary preconditioned conjugate gradient solver. This is the default. = DIIS selects a diis-like iterative solver. RDCISV = Flag to read CIS vectors from a $CISVEC group in the input file. Default is .FALSE. MNMEDG = Flag to force the use of the minimal amount of memory in construction of the CIS Hamiltonian diagonal elements. This is only relevant when DGAPRX=.FALSE., and is meant for debug purposes. The default is .FALSE. MNMEOP = Flag to force the use of the minimal amount of memory during the Davidson iterations. This is for debug purposes. The default is .FALSE.
The additional PC GAMESS/Firefly specific keywords are: CPTOL - convergence criterion for CPHF equations, default is 1.0d-5 for GAMESS (US) compatibility. THRDII - threshold to turn on DIIS if it was selected to solve CPHF. Default is 0.05 MAXGC - maximum allowed number of trial vectors to be routed through GENCON engine, default is 1. If the number of trial vectors is greater than MAXGC, only FASTINTS will be used. The reason is that for moderately contracted GC basis sets like cc-pVXZ, gencon is faster than fastints only for relatively small number of trial vectors (this is by gencon design). On the other hand, for ANO-like basis sets, it is always better to set MAXGC to be equal the number of initial guess vectors, as fastints will be much slower. ONEDIM = .true./.false. Default is true for abelian groups, false otherwise. Option to select fast algorithm for excited state properties/gradients for systems with symmetry. Any state transforming by any pure one-dimension representation on the point group can be handled by this manner. It is safe to always turn this option on as the check is performed and this option disables itself if the symmetry is not appropriate. PRTTOL - threshold for CIS determinants/csf printout and also for states symmetry determination. Default is 0.05. ISTSYM - symmetry of states of interest. Default is zero, i.e., does not use any symmetry during calculations. Setting this to the desired index of irrep (according to PC GAMESS/Firefly numbering) will solve only for the states of the desired symmetry and exploiting full (including non-abelian) symmetry of molecule, thus significantly reducing computation time. ALTER - flag to modify internal logic of Davidson diagonalization code to use dynamic number of trial vectors. Default is .true. Setting it to .false. will slow-down calculations by forcing DAVIDSON diagonalization code to work exactly as in the GAMESS (US). ICUTCP - the effective value of ICUT to be used by CONJG CPHF solver while solving CPHF equations in direct SCF mode. The main purpose of this keyword is to avoid numerical instability problems causing conjugate gradient solver to diverge upon reaching near-convergence. E.g., one may set ICUT to 9 or 10 while tighten ICUTCP to be 10 or 11. The default is the value of ICUF of $CONTRL group. The state-tracking feature of the PC GAMESS/Firefly' CIS code can be activated by selecting negative value of istate in the $cis group. It is intended for geometry optimization of the excited states in the case of root flipping. Note that oscillator strengths printed in the CIS summary table are calculated using transition dipoles length form only. ========================================================== $CISVEC group required if RDCISV in $CIS is chosen This is formatted data generated by a previous CIS run, to be read back in as starting vectors. ========================================================== Below is the sample input file. $CONTRL SCFTYP=RHF CITYP=CIS $END $SYSTEM TIMLIM=3000 MEMORY=3000000 $END $BASIS GBASIS=n31 ngauss=6 NDFUNC=1 $END $CIS NSTATE=3 ISTSYM=0 ISTATE=1 $END $DATA H2O CNV 2 O 8.0 0.0000000000 0.0000000000 0.7205815395 H 1.0 0.0000000000 0.7565140024 0.1397092302 $END
Selected CIS references:
J.B.Foresman, M.Head-Gordon, J.A.Pople, M.J.Frisch J.Phys.Chem. 96, 135-149(1992)
R.M.Shroll, W.D.Edwards Int.J.Quantum Chem. 63, 1037-1049(1997)
S.P.Webb Theoret.Chem.Acc. 116, 355-372(2006)
Last updated: March 18, 2009