Re: CCL:B3LYP functional specification of constants

[CCL <ccl(at)kressworks.com>]

My original question:
 > While investigating prior to doing some extensive B3LYP calculations, I've
 > noticed that GAMESS and MOLPRO use 0.08 times the Slater Exchange in the
 > B3LYP functional while NWCHEM and GAUSSIAN use 0.8.  Which is correct?
 Received one reply.  Thanks to Huub van Dam for his response (given at the end
 of this note).  In addition, from the GAMESS manual:
 >             Note that B3LYP in GAMESS is based in part on the VWN5
 >          correlation functional.  Since there are five formulae
 >         in the VWN paper for local correlation, other programs
 >          may use other formulae, and therefore generate different
 >          B3LYP energies.  For example, NWChem's manual says it uses
 >          the "VWN 1 functional with RPA parameters as opposed to the
 >          prescribed Monte Carlo parameters", but NWChem can be made
 >          to use the VWN5 formula by
 >              xc HFexch 0.20 slater 0.80 becke88 nonlocal 0.72
 >                   lyp 0.81 vwn_5 0.19
 >          If you use some other program, its B3LYP energy will be
 >          different from GAMESS if it does not employ VWN5.
 So we see there is no unique definition for the b3lyp functional.  This seems
 a recipie for confusion but hey, what do I know?
 Note: MOLPRO and GAMESS (as of their current incarnations) use the same
 definition for the b3lyp functional.
 Jim
 > I think MOLPRO uses 0.8 times the Slater Exchange functional as well. The
 > confusion stems from what people call the Becke '88 functional. In MOLPRO
 > the Becke '88 functional is defined as
 >
 >    B88 = rho^(4/3)*(c+dB88)
 >
 > where dB88 is the gradient enhancement factor and c*rho^(4/3) is the Slater
 > functional. In other codes the functional is defined as
 >
 >    B88' = rho^(4/3)*dB88
 >
 > without the Slater functional term. Of course the parameters you have to
 > use to arrive at the correct B3LYP expression depend on which definition of
 > the Becke '88 you use.
 >
 > A different matter is that different programs are actually using different
 > expressions for B3LYP, because of a lot of confusion about which VWN
 > functional was implied in the correlation part. Some codes use VWN5 (e.g.
 > TURBOMOLE, GAMESS and MOLPRO), whereas GAUSSIAN claims to be using VWN3 and
 > NWCHEM follows the GAUSSIAN definition. As a result B3LYP energies from
 > GAMESS and MOLPRO will differ from those obtained from GAUSSIAN and NWCHEM.
 >
 > Best wishes,
 >
 >   Huub van Dam
 >   CCLRC Daresbury Laboratory
 >   h.j.j.vandam(at)dl.ac.uk