Re: CCL:Frozen core definition in Gaussian



A related issue is the definition of core orbitals in molecules such
 as GaF (where 3d "core" orbitals of Ga are energetically above the
 valence
 2p orbitals of F -> source of many errors).  This and the inclusion of Ga
 3d orbitals into valence space are discussed in two recent articles.  One
 is Published in Chem. Phys. Lett. last month (vol. 350,
 573--576) by Rassolov, Pople, Redfern, and Curtiss; the other one is
 appearing presently in Theor. Chem. Accounts by G. Petersson and
 coworkers. The original problem that prompted the inclusion of Ga 3d into
 valence space was discovered by Radom et al. and Bauschlicher et al. a
 while ago. It looks like we will be able to return 3d orbitals of Ga back
 to its core, without much loss of accuracy.
 On Mon, 7 Jan 2002, Kirk Peterson wrote:
 >
 > All:
 >
 > after stumbling across something today, I thought that it would
 > be of enough interest to bring it to people's attention.  Perhaps
 > this is common knowledge, but I'm guessing it's not.  When
 > carrying out ab initio calculations on molecules containing 3rd row
 > main group elements, i.e., Ga - Kr, the natural definition of the
 > core electrons for these 3rd row elements (at least in my opinion)
 > includes the following electrons: 1s, 2s, 2p, 3s, 3p, and 3d.  (Of course,
 > when doing calculations on Ga- or Ge-containing species, one should at
 > some point be interested in 3d correlation.)  Much to my surprise,
 > it appears that the default frozen core in Gaussian does NOT
 > include the 3d electrons (i.e., the 3d electrons are included in
 > the valence shell with the 4s and 4p electrons).  Perhaps this was
 > to make the definition consistent with the 3d transition metals where
 > they are obviously valence, but I would be willing to guess that not
 > too many people would guess this choice for an element such as Br.
 >
 > Besides probably doing much more work than desired (correlating 10's
 > of extra electrons usually adds nontrivial amounts of computational
 > expense), most basis sets for these elements do not describe 3d
 correlation.
 > In particular, the cc-pVnZ basis sets for these elements were developed for
 > valence correlation only, the definition of which did NOT include the 3d
 > electrons of these elements.  So in addition to the added computational
 expense,
 > you might actually do some harm since one may be introducing extra BSSE
 > > from correlating core-like electrons with a valence basis set (never a
 good idea).
 >
 > Of course, one can change the definition of the frozen core in Gaussian
 > by using the old ReadWindow (RW) option, but the quick FC (i.e.,
 > MP2=FC or just MP2) will result in the situation described above.  I guess
 > one should note that just mentioning in a publication that the frozen core
 > approximation was used is surprisingly ambiguous...
 >
 > best wishes,
 >
 > Kirk Peterson
 > Washington State University
 > EMSL/Pacific Northwest Nat'l Laboratory
 >
 > -= This is automatically added to each message by mailing script =-
 > CHEMISTRY' at \`ccl.net -- To Everybody  | CHEMISTRY-REQUEST' at \`ccl.net
 -- To Admins
 > MAILSERV' at \`ccl.net -- HELP CHEMISTRY or HELP SEARCH
 > CHEMISTRY-SEARCH' at \`ccl.net -- archive search    |    Gopher:
 gopher.ccl.net 70
 > Ftp: ftp.ccl.net  |  WWW: http://www.ccl.net/chemistry/   | Jan: jkl' at \`ccl.net
 >
 >
 >
 >
 --
 Vitaly Rassolov
 Dept. of Chemistry and Biochemistry         (803) 777-7811
 University of South Carolina            fax (803) 777-9521
 631 Sumter St
 Columbia, SC 29208
 rassolov' at \`mail.chem.sc.edu