Dear all,

Jose C. Conesa wrote:

>Dear ADF users,
>I have tried to use SICOEP in ADF. I applied it to a paramagnetic Ni complex, with the
>following keywords:
>....................................
>BASIS
>type TZP
>core none
>END
>
>CHARGE -1 1
>
>RELATIVISTIC SCALAR ZORA
>UNRESTRICTED
>
>SINGULARFIT FAST
>SICOEP
> LOCALIZE
>END
>
>XC
> lda vwn
> gga rpbe
>END
>....................................
>I find the following error message (occurring after the 2nd SCF cycle has initiated and the
>energies of spin-up and spin-down orbitals computed in it have been output):
>
>sicpar: assumption failed: npelem = 39555 but (levec*(levec+1))/2 = 79003
>
> (and, later:)
>
>sicpar: npmat FUBAR
>
> The program then aborts. Can someone tell the reason for the error? There is in the Users'
>Guide an indication that SICOEP does not work well with d or f valence shells present. Is this
>the reason? Can someone tell, anyway, me which calculation result(s) (are) most affected by
>this inadequacy of SICOEP for such systems: total energy, energy level positions,
>electron/spin density distributions, ... ?
> Also, I found an error in a previous attempt to use SICOEP together with the
>RELATIVISTIC SPINORBIT option. Are these two options incompatible in ADF? Will they be
>always so?
> Finally: the Users' Guide cautions about the inadequacy of standard fitting sets when
>working with SICOEP. Does this mean that I should attempt to use this latter option only with
>fitting sets different from those in the standard ADF2004.1 database? If so, how could I
>obtain or build such larger fitting sets?
>
>

As described in the ADFUsersGuide SICOEP can not be used parallel.
The error that adf gives (sicpar: assumption failed:.....)
suggests that you tried to use SICOEP in a parallel run.

ADF gives inaccurate results if SICOEP is used for d and f shells if the
fit set is not large enough.
You should check the papers by S. Patchkovskii:
Patchkovskii, S., J. Autschbach, and T. Ziegler, Journal of Chemical
Physics, 2001. *115*: p. 26-42.
Patchkovskii, S. and T. Ziegler, Journal of Chemical Physics, 2002.
*116*: p. 7806.
Patchkovskii, S. and T. Ziegler, Journal of Physical Chemistry, 2002. *A
106*: p. 1088.
to get more on the accuracy of the results.

Indeed spin-orbit coupling and SICOEP are not compatible.
I do not think that someone is planning to implement this, but that
might change.

Building larger fit sets is difficult if one wants to avoid numerical
problems like dependent fit sets.
One possible suggestion to improve the fit set is to use the relatively
large fit sets that are used in
$ADFHOME/atomicdata/ET or $ADFHOME/atomicdata/ZORA/QZ4P atomic data files,
to replace the fit set in the standard atomic data files.

SICOEP is not yet an option that can be used without thorough testing.

Best regards,
Erik van Lenthe
----------------------------------------------------------------------
Dr. Erik van Lenthe, Scientific Computing & Modelling NV
Vrije Universiteit, Theoretische Chemie, De Boelelaan 1083
1081 HV Amsterdam, The Netherlands
Phone: +31 20 44 47625 FAX: +31 20 44 47629
e-mail: vanlenthe@scm.com WWW: http://www.scm.com
----------------------------------------------------------------------
Received on 2005-01-07 14:27:32