[molpro-user] The smallest exponent that can be used to specify a basis function
Tatiana Korona
tania at tiger.chem.uw.edu.pl
Sat May 11 18:59:59 BST 2013
Dear Gerald,
A gaussian with the exponent equal to 1.E-50 has been used by Stanton and Gauss
to obtain IP-EOM-CC3 energies directly from an EE-EOM-CC3 program, see
J.Chem.Phys.,111,8785,1999, and ref. 18 there. If it were excluded by the
program, then they wouldn't be able to calculate ionization potentials and make
numbers for that paper. So, such a small exponent is certainly possible and
_active_ in the program they used (probably Aces). I repeated this trick to
obtain IP and EA with Molpro, but could use only the exponents like 1.E-12 for a
continuum orbital. However, one can verify that this exponent is enough to
provide a sufficient number of accurate digits for ionization potentials and
electron attachments, e.g. I reproduced all digits given in papers on EA-EOM-CC
methods.
You can test the following simple input
***, IP-EOM-CCSD (simulated with EE-EOM-CCSD)
memory,20,m
gprint,basis,orbital=10
zsymel=nosym
basis={
default,H=avdz,O=avdz
s,O,1.e-12
}
geometry={H1
O,H1,R, !Z-matrix for water
H2,O,R,H1,THETA}
Theta=104 !start bond angle
R=0.96 Ang !start bond distance
hf
ccsd
eom,-5.1,vir_orb=6.1 ! fix on excitations TO this orbital
This input will produce first 4 IPs for water. 6.1 is our zero-energy continuum
orbital. However, if you change an additional exponent on oxygen from 1.e-12 to
e.g. 1.e-15 or smaller, than you will get:
ITERATION DDIFF GRAD ENERGY 2-EL.EN.
DIPOLE MOMENTS DIIS ORB.
1 0.000D+00 0.000D+00 32025.50226135 64244.652599 NaN
NaN NaN 0 start
2 0.000D+00 0.232D+02 -16.95078177 84.371664 NaN
NaN NaN 1 diag
3 0.187D+02 0.165D+00 32657.17607612 65469.453071 NaN
NaN NaN 2 diag
4 0.209D+02 0.106D+02 32655.12447108 65465.460273 NaN
NaN NaN 3 diag
5 0.433D-01 0.108D+02 32649.67192410 65454.845120 NaN
NaN NaN 4 diag
?TOTAL ENERGY UNREASONABLE, ETOT = 0.32650D+05, ENEST =
-0.75852D+02
in Hartree-Fock.
Best wishes,
Tatiana
On Sat, 11 May 2013, Gerald Knizia wrote:
>
> On Wed, 2013-05-08 at 11:41 +0200, Evgeniy Gromov wrote:
>> Gamess-US can converge SCF "hands down" with an exponent 1.E-50 .
>> Molpro however could do the same with 1.E-10. 1.E-15 or smaller
>> didn't work :(
>
> Unlike other programs, Molpro never throws away basis functions if they
> become unreasonable. Are you sure that your 1e-50 exponent basis
> function actually ended up the calculation? Because for such a function
> even handling its normalization coefficient would be a serious problem.
>
> While it is perfectly possible to calculate integrals over such
> functions[1] (it is even done routinely with zero exponent functions for
> implementing simpler integrals in terms of more complicated integrals),
> using them as basis functions is very much not recommended. If you want
> to try it anyway, be sure to turn off all integral screening,
> compression, and approximation options.
>
> [1] See JCTC 7 2387 (2011)
> --
> Gerald Knizia
>
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user
>
Dr. Tatiana Korona http://tiger.chem.uw.edu.pl/staff/tania/index.html
Quantum Chemistry Laboratory
University of Warsaw
Pasteura 1, PL-02-093 Warsaw, POLAND
`The man who makes no mistakes does not usually make anything.'
Edward John Phelps (1822-1900)
More information about the Molpro-user
mailing list