[molpro-user] Basis Set Confusion

[Frank Uhlig]

Dear Kirk,

I do (partially) accept that point of view. But then maybe there are two
ways that people look at this. Because I remember emphasis being exacty
on "nonzero contraction coefficients for all basis function". No matter
of extensive arguments for me but some kind of preliminary clarification
on the terms seems to be necessary sometimes.
Anyways, thank you very much again.

Best regards,

Frank



On Mon, 2009-10-26 at 09:52 -0700, Kirk Peterson wrote:
> Dear Frank,
> 
> one way to think about it is that only the Hartree-Fock part of the  
> correlation consistent basis sets are
> generally contracted, i.e., all primitives contribute to each  
> contraction of a given angular
> symmetry.  The uncontracted s and p functions are simply correlating  
> functions like d, f, g, etc.  Generally only
> atomic natural orbital (ANO) basis sets use generally contracted  
> correlating functions, but even then the most
> diffuse primitives are often uncontracted to provide greater  
> flexibility.  Of course you can always consider an
> uncontracted function to be generally contracted -  the contraction  
> coefficients of all the other primitives are just zero   :^)
> 
> best regards,
> 
> Kirk
> 
> On Oct 26, 2009, at 8:40 AM, Frank Uhlig wrote:
> 
> > Dear Kirk,
> >
> > thanks for your quick and helping reply. Although I do have some
> > problems with the nomenclature "generally contracted basis set" in the
> > context of Dunnings aug-cc-pVXZ basis sets.
> > In my understanding generally contracted basis sets contain all the
> > primitives (nonzero contraction coefficients) in every basis function
> > (emphasis on all and every). So Dunning's basis sets wouldn't be
> > described by this terminology.
> > Would be really nice to get some "outside" clarification on the
> > terminology because for now I heard different opinions.
> >
> > Best regards,
> >
> > Frank
> >
> > On Fri, 2009-10-23 at 14:15 -0700, Kirk Peterson wrote:
> >> Dear Frank,
> >>
> >> no worries, these two "versions" of aug-cc-pV5Z are exactly  
> >> equivalent
> >> to each other. On the EMSL
> >> website there is a radiobox that by default is checked called
> >> "optimize general contractions". This exists
> >> for programs like Gaussian that can not exploit generally contracted
> >> basis sets. This option essentially
> >> just deletes from contractions primitives that are also uncontracted
> >> in the basis. Davidson discusses
> >> this strategy in  Chem. Phys. Lett. 260, 514 (1996). If desired, you
> >> can obtain the exact same basis set as Molpro
> >> by unchecking this box.
> >>
> >> On a related note, for second row atoms it is highly recommended to
> >> use the so-called +d basis sets,
> >> aug-cc-pV(n+d)Z.  I would guess for your AlCl3 case they will give
> >> much improved convergence at least
> >> for your structural parameters and probably your energetics as well.
> >> These of course are available
> >> within Molpro or from EMSL.
> >>
> >> best regards,
> >>
> >> Kirk
> >>
> >> On Oct 23, 2009, at 11:46 AM, Frank Uhlig wrote:
> >>
> >>> Dear Molpro-Users,
> >>>
> >>> I have a question concerning the by Molpro provided Dunning aug-cc-
> >>> pV5Z
> >>> basis sets for aluminum and chlorine (Molpro Version 2008). While
> >>> trying
> >>> to perform CBS evaluations for AlCl3 I found that there is a
> >>> difference
> >>> concerning the contraction of the first three s functions. While in
> >>> the
> >>> Molpro basis set library 20 of 21 primitives get contracted to
> >>> respectively one s-function the by the EMSL provided basis sets show
> >>> for
> >>> the same functions only 16/21 contracted primitives. The contraction
> >>> coefficients of each primitive contracted in both cases are the  
> >>> same.
> >>> Thus I am wondering what lead to this difference and would  
> >>> appreciate
> >>> any further information.
> >>>
> >>> Best regards,
> >>>
> >>> Frank Uhlig
> >>>
> >>> _______________________________________________
> >>> Molpro-user mailing list
> >>> Molpro-user at molpro.net
> >>> http://www.molpro.net/mailman/listinfo/molpro-user
> >>
> >
>