Examples and Explanations of PC GAMESS Input Files
Back
to Modeling Molecules and Reactions
PC
GAMESS Web-Page
Let's start now by repeating the simple PC GAMESS input example for formaldehyde:
! This PC GAMESS input file is OK for RAM >
100MB
!
$CONTRL SCFTYP=RHF RUNTYP=OPTIMIZE COORD=CART
NZVAR=0 MULT=1 ICHARG=0 $END
$SYSTEM TIMLIM=20000 MEMORY=10000000 $END
$STATPT NSTEP=1000 $END
$BASIS GBASIS=STO NGAUSS=3 $END
$GUESS GUESS=HUCKEL $END
$DATA
Test...HCHO molecule - RHF/STO-3G (a comment line)
Cn 1
C 6.0 0.6084782705
-0.0000011694 0.00000000000
O 8.0 -0.6082418894
-0.0000002093 0.00000000000
H 1.0 1.2040919862
-0.9264398115 0.00000000000
H 1.0 1.2040973125
0.9264340484 0.00000000000
$END
One may easily make out that a PC-GAMESS input-file's text consists of several groups: each group begins with a $GroupName and ends with a suffix $END (even DATA is a group that contains the nuclear-framework specification and symmetry for the molecule). Please note that the first column of any line other than a comment line is kept blank (a comment line within a PC-GAMESS input file generally begins with a ! sign), and that in the input file there must be no blank line either above or below the input text. The groups and their constituent keywords are explained in details within an in-package documentation file named INPUT (click here for its on-web version).
Within the group CONTRL, the keywords found above are:
SCFTYP: Its possible values are RHF (Restricted Hartree-Fock),
ROHF (Restricted Open-Shell Hartree-Fock), UHF (Unrestricted Hartree-Fock),
MCSCF (Multi-Configuration Self-Consistent Field) etc.
RUNTYP: Its possible values are ENERGY (for fixed-geometry single-point
calculation), OPTIMIZE (to optimize nuclear-framework geometry), GRADIENT (to
calculate gradients around the single-point geomtery) etc.
COORD:
Its possible values are CART (Cartesian coordinates), ZMT etc.
NZVAR:
Gives choices for way of geometry-optimization. Generally NZVAR=0 is used: in PC-GAMESS
it means that Cartesian coordinates will be considered for
geometry-optimization.
MULT:
Its value is the spin-multiplicity of the molecular-electronic state (1
for ground-states of H2,
N2
etc., 3 for O2
ground-state etc.).
ICHARG: The net charge of the molecule/ion (0 for HCHO as given above, +1 for NH4+
ion).
Within the other groups, the keywords found above are:
TIMLIM: Maximum
limit of time in minutes for the calculation (so, give it big).
MEMORY: For a >100 MB computer, may make it ten million as shown above, this
value works in most cases.
NSTEP: The number of steps in a geometry-optimization program: if you don't give
it a big value (say 1000), a geometry-optimization calculation may get forcibly stopped before it has
actually completed.
GBASIS: It is the name of the Gaussian basis set, and the allowed options for it
are STO, N21, N31, N311, DZV etc.
NGAUSS: This means the number of Gaussians (N), and pertains to only GBASIS=STO, N21, N31, or N311.
For GBASIS=STO, NGAUSS may be 2 or 3 meaning the STO-2G or STO-3G basis set
respectively. The
combination GBASIS=N31 NGAUSS=6 means the 6-31G basis set. To mean
the 6-31G** [i.e., 6-31G(d,p)] basis set add NDFUNC=1 NPFUNC=1 to it (meaning one each
of d-type & p-type polarization functions respectively).
GUESS: It gives the type of initial guess for the molecular orbital. GUESS=HUCKEL
generally works, but for MCSCF calculations should generally use GUESS=MOREAD,
and these MO-initial-guesses be given within an additional $VEC .... $END group,
copied from the punch-file output for the similar-parameter CI-run that preceded
the MCSCF run.
However, the above example deals wholly within the Hartree-Fock (HF) concepts. HF computations do not take care of the phenomenon of electron correlation and so are quite approximate. The beyond-HF methods that we may discuss just within this space here are 2nd-order Møller-Plesset perturbation (MP2), configuration-interaction (CI) and Multi-Configuration Self-Consistent-Field (MCSCF) methods.
Implementing MP2 in PC-GAMESS is a rather simple job -- we may just add MPLEVL=2 within the CONTRL group, and it'll run. Even geometry-optimization is possible with MP2 calculation with SCFTYP=RHF, though such a computation takes time. Below is an example of MP2/RHF geometry-optimization for Li2 molecule using the 6-31G** basis set, starting with a bond length of 3.0 Å, finally arriving at the bond length of 2.78156619860 Å.
! This PC-GAMESS input file is OK for RAM >
100MB
!
$CONTRL SCFTYP=RHF MPLEVL=2 RUNTYP=OPTIMIZE COORD=CART
NZVAR=0 MULT=1 ICHARG=0 $END
$SYSTEM TIMLIM=20000 MEMORY=10000000 $END
$STATPT NSTEP=1000 $END
$BASIS GBASIS=N31 NGAUSS=6 NDFUNC=1 NPFUNC=1 $END
$GUESS GUESS=HUCKEL $END
$DATA
Li2 MODEL - Total energy -14.8854500393, Li |x| is 1.3907830993
Cn 1
Li 3.0 -1.500000 0.000000 0.000000
Li 3.0 1.500000 0.000000 0.000000
$END
Implementing CI in PC-GAMESS needs some understanding of the CASSCF (Complete Active Space Self-Consistent-Field) concept, according to which there are assumed some frozen, low-lying core-MOs which are never ever vacated to contribute electrons for configuration-interaction. So these core orbitals are always doubly occupied in all configurations: the number of these core MOs is called NCORE, and they contain a total of (2*NCORE) electrons. The rest number of electrons is considered as the number of active electrons (called NELS in PC-GAMESS). Naturally, the total number of electrons (NE) in the molecule must equal (2*NCORE+NELS). Further, as per the CASSCF concept, the number of active, upper-energy MOs taking part in the CI process is not assumed to be unlimited, but is rather to be specified by the user (called NACT). A judicious choice for NCORE, so as to suitably lessen it from (NE/2), would come from chemical intuition and computational experience. Same must be said about the choice for NACT -- here, however, the question becomes: how much more should be NACT, compared to (NELS/2)? The name CASSCF further implies that within these limits bound by NCORE and NACT, the complete set (figuratively speaking, space) of molecular electronic configurations will be considered in computations (and no more configuration beyond that).
After deciding about NCORE & NACT (and calculating NELS), if one choose the computationally more exhaustive ALDET option (instead of the less easily comprehensible GUGA option) for CITYP, one may directly go for a CI-computation, as shown in the example below. Here Li2 has six electrons in total; let us consider its core MOs, namely sg 1s and su* 1s, to be the 'frozen core' ones, thus choosing NCORE to be 2. NELS is so, obviously, 2 (i.e., 6-2*2). Now considering the not-so-high energy 'additional' MOs su* 2s and sg 2p to be 'reachable' by the two 'active' electrons, in addition to its lowest-configuration 'home-MO' sg 2s, we take NACT to be 3 (i.e., 1+2). We need to put these values NCORE=2 NACT=3 NELS=2 within the additional CIDET group, as shown in the example below.
! This PC-GAMESS input file is OK for RAM >
100MB
!
$CONTRL SCFTYP=ROHF RUNTYP=ENERGY COORD=CART NZVAR=0
MULT=1 ICHARG=0 CITYP=ALDET $END
$SYSTEM TIMLIM=20000 MEMORY=10000000 $END
$STATPT NSTEP=1000 $END
$BASIS GBASIS=N31 NGAUSS=6 NDFUNC=1 NPFUNC=1 $END
$GUESS GUESS=HUCKEL $END
$CIDET NCORE=2 NACT=3 NELS=2 $END
! The Li1 & Li2 (x,y,z)-data were obtained from MP2/RHF optimization
$DATA
Li2 MODEL - CI Run - Total energy is -14.8743054508
Cn 1
Li 3.0 -1.3907830993 0.0000000 0.0000000
Li 3.0 1.3907830993 0.0000000 0.0000000
$END
Implementing the MCSCF method
that optimizes the constituent MOs for the multi-configuration situation,
which method is naturally more accurate than the corresponding limited-CI (i.e.,
same-active-space CI) calculation, now needs only a few minor steps ahead to be taken by the user.
At first we need to make SCFTYP=MCSCF. Now, here, GUESS=HUCKEL
is a bad choice for the initial-guess MOs: instead the MOs obtained from the preceding
CI calculation (with same values for NCORE, NACT
& NELS)
should be used as the initial-guess, so make GUESS=MOREAD.
Also, to read the guess MOs, the $VEC .... $END part, containing the
simple-CI-obtained MOs*, is to be copied from the bottom part (yes, be careful
not to copy from the upper part of the file) of the 'Punch'
output file obtained by the CI run, and pasted in the input file as exemplified
in the MCSCF input-file below. We need to also put the value of NORB,
equaling it to the value of (NCORE+NACT)
-- in this example, 5.
Finally, the option CITYP=ALDET is to be replaced with CISTEP=ALDET
and transferred to the new MCSCF group, and the CIDET group in the
CI input file above is to be renamed as the DET group
in the MCSCF input file below.
* Note the small $OCCNO .... $END portion
just above this $VEC .... $END part in the CI-calculation's Punch file -- this contains
the (electron) occupation numbers of the forcibly-frozen-core MOs and the
active MOs as per the CI calculation results, offering justification for
the CI paradigm. Thus in this example we would see that the five MO occupation
numbers are 2.000,
2.000, 1.932, 0.040 & 0.028, implying
that the two 'upper' MOs (su*2s & sg2p)
have a combined occupation number of 0.068,
which is a significant 3.4% of the two active
electrons.
! This PC-GAMESS input file is OK for RAM >
100MB
!
$CONTRL SCFTYP=MCSCF RUNTYP=ENERGY COORD=CART NZVAR=0
MULT=1 ICHARG=0 $END
$SYSTEM TIMLIM=20000 MEMORY=10000000 $END
$STATPT NSTEP=1000 $END
$BASIS GBASIS=N31 NGAUSS=6 NDFUNC=1 NPFUNC=1 $END
$GUESS GUESS=MOREAD NORB=5 $END
$MCSCF CISTEP=ALDET $END
$DET NCORE=2 NACT=3 NELS=2 $END
! The $VEC... part below were copied from bottom of 'Punch' obtained by CI run!
$VEC
1 1 7.06134037E-01-6.50428610E-03-2.28551173E-03 0.00000000E+00 0.00000000E+00
1 2 2.02085213E-04 1.98541026E-03 0.00000000E+00 0.00000000E+00 6.08254185E-03
1 3 5.84959315E-03 5.84959315E-03 0.00000000E+00 0.00000000E+00 0.00000000E+00
1 4 7.06134037E-01-6.50428610E-03 2.28551173E-03 0.00000000E+00 0.00000000E+00
1 5 2.02085213E-04-1.98541026E-03 0.00000000E+00 0.00000000E+00 6.08254185E-03
1 6 5.84959315E-03 5.84959315E-03 0.00000000E+00 0.00000000E+00 0.00000000E+00
2 1-7.05997769E-01 1.11438941E-02 5.19371917E-03 0.00000000E+00 0.00000000E+00
2 2-4.36326511E-03-3.49292779E-03 0.00000000E+00 0.00000000E+00-5.63225802E-03
2 3-6.76106790E-03-6.76106790E-03 0.00000000E+00 0.00000000E+00 0.00000000E+00
2 4 7.05997769E-01-1.11438941E-02 5.19371917E-03 0.00000000E+00 0.00000000E+00
2 5 4.36326511E-03-3.49292779E-03 0.00000000E+00 0.00000000E+00 5.63225802E-03
2 6 6.76106790E-03 6.76106790E-03 0.00000000E+00 0.00000000E+00 0.00000000E+00
3 1-1.47945014E-01 1.07763486E-01 1.06697194E-01 0.00000000E+00 0.00000000E+00
3 2 3.98189453E-01-2.06344369E-02 0.00000000E+00 0.00000000E+00 3.40325516E-02
3 3 2.52458263E-02 2.52458263E-02 0.00000000E+00 0.00000000E+00 0.00000000E+00
3 4-1.47945014E-01 1.07763486E-01-1.06697194E-01 0.00000000E+00 0.00000000E+00
3 5 3.98189453E-01 2.06344369E-02 0.00000000E+00 0.00000000E+00 3.40325516E-02
3 6 2.52458263E-02 2.52458263E-02 0.00000000E+00 0.00000000E+00 0.00000000E+00
4 1 0.00000000E+00 0.00000000E+00 0.00000000E+00 7.13965401E-04 6.22023645E-02
4 2 0.00000000E+00 0.00000000E+00 5.85758530E-03 5.10326768E-01 0.00000000E+00
4 3 0.00000000E+00 0.00000000E+00-3.89066332E-05-3.38963846E-03 0.00000000E+00
4 4 0.00000000E+00 0.00000000E+00 0.00000000E+00 7.13965401E-04 6.22023645E-02
4 5 0.00000000E+00 0.00000000E+00 5.85758530E-03 5.10326768E-01 0.00000000E+00
4 6 0.00000000E+00 0.00000000E+00 3.89066332E-05 3.38963846E-03 0.00000000E+00
5 1 9.95267964E-02 1.86019503E-02 3.99347699E-02 0.00000000E+00 0.00000000E+00
5 2 1.40675163E-01 7.42714128E-01 0.00000000E+00 0.00000000E+00-3.13185272E-02
5 3-3.05295144E-02-3.05295144E-02 0.00000000E+00 0.00000000E+00 0.00000000E+00
5 4-9.95267964E-02-1.86019503E-02 3.99347699E-02 0.00000000E+00 0.00000000E+00
5 5-1.40675163E-01 7.42714128E-01 0.00000000E+00 0.00000000E+00 3.13185272E-02
5 6 3.05295144E-02 3.05295144E-02 0.00000000E+00 0.00000000E+00 0.00000000E+00
$END
$DATA
Li2 MODEL - MCSCF Run - Total energy is -14.8852107606
Cn 1
Li 3.0 -1.3907830993 0.0000000 0.0000000
Li 3.0 1.3907830993 0.0000000 0.0000000
$END
Forming electron-density (cube) files using PC GAMESS, and their further analysis
References:
[1] Pentium/Win32 PC-GAMESS version 7.0 (Dragon), build number 3970
(optimized for Pentium IV), Granovsky, A.A., Laboratory of Chemical Cybernetics (Moscow State Univ., Moscow),
1994-2006.
[2] Schmidt, M.W., Baldridge, K.K., Boatz, J.A., Elbert, S.T., Gordon, M.S., Jensen, J.H., Koseki, S., Matsunaga, N., Nguyen, K.A., Su, S.J., Windus, T.L., with Dupuis, M. and Montgomery, J.A., J. Comput. Chem., 14 (1993) 1347