[CPMD-list] combination of PBE USPP with BLYP DFT

[Yi Liu]

Sorry. I just forgot to paste contents. Please delete the previous one.

Hi,

I had ever calculated the binding energy (BE) of water dimmer using
different PPs and xc functionals (they were always consistent in my test
though). I did realize that PBE PP/xc combination gives larger BE(closer to
experimental value) than BLYP PP/xc. Moreover, BE is much more sensitive to
xc functional than PP, though Martin-Troullier type PP always gives slightly
larger value than Godecker type.

My results below support this statement,

   PP               xc          BE(KJ/mol)
Godecker     BLYP         18
MT               BLYP         19
Godecker      PBE            21
MT                PBE           23

      Expt.                          22

I also found the same trends in Benzene-Water complex.

Several points should be noted:

1. DFT do has difficulty to predict BE very accuratedly for weak bound
system(eg. hydrogen bonded system), then DFT BE is not always satisfied for
hydrogen bonding.
2. Even for same type PP, the parameters are somewhat different. So above
comparison might depend on details of PPs.
3. Though BLYP has poor BE, it did give a better OH bond length and HOH
angle for water than PBE.

My experience shows that PBE is better than BLYP in describing weak binding
energy, however, BLYP is better in geometry structure (or dipole moment
etc.). I am not sure if this conclusion is general. I'd like to listern to
other experience about that.

Therefore, above discussion may answer your question, that is, PBE PP/xc
will still give larger dissociation energy than BLYP. In addition, it is not
necessary to use different xc functionals in generating PP and DFT
calculation.

Hope this helps!

Yi

----- Original Message -----
From: "Erik Emilio Santiso" <eesantis at unity.ncsu.edu>
To: "Jung Goo Lee" <jglee at nemo.physics.ncsu.edu>
Cc: <cpmd-list at cpmd.org>
Sent: Thursday, October 30, 2003 9:14 PM
Subject: Re: [CPMD-list] combination of PBE USPP with BLYP DFT


>
> Hi! Well, this looks more like a coincidence than anything else. I don't
> think it makes any sense to find a PP with one xc functional and then do
> the calculation with another one. What happens if you use the PBE PP's
> with the PBE functional?
>
> Erik.
>
> On Thu, 30 Oct 2003, Jung Goo Lee wrote:
>
> > Hi everyone,
> >
> > I would like to know whether the combination of PBE uspp in ATOMS option
> > and BLYP in DFT
> > option (the example of partial input is shown below)makes sense or not,
> > especially, for calculations
> > of free energy (constraint). This is because BLYP method underestimate
> > the dissociation free energy
> > compared with experiment. For example,  water dissociation in 32 water
> > molecules with MT PP/
> > BLYP method gave 17 kcal/mol in free energy while experimental value is
> > 21 kcal/mol(Trout and
> > Parrinello, Chem. Phys. Lett, 288(1998)343.
> > I tried the same calculation with PBE uspp and BLYP method.
> > I got about 20 kcal/mol althogh I tried with  a few hundred fs (instead
> > of 1 ps) MD at each point of
> > O-H bond constraint. So, this combination (PBE uspp/BLYP) seems to be
> > o.k. in free energy
> > calculations.
> >
> > Is this a reasonable and cheap(computationally) way to get out of the
> > problem with free energy
> > calculations with gbl uspp/BLYP?
> > Does any one have any rationalization or contradiction with reasonable
> > explanation?
> >
> > Jung-Goo Lee
> >
> > Example
> > --
> > --
> >  &ATOMS
> > *008-O-gpbe--bm.uspp NEWF BINARY
> >   LMAX=P
> >   1
> >   4.3728  4.5000  4.5000       1   8
> >
> > *001-H-gpbe--bm.uspp NEWF BINARY
> >   LMAX=S
> >   2
> >   5.5172  4.5000  5.9238       1   1
> >   5.5172  4.5000  3.0761       1   1
> > &END
> >
> > &DFT
> >  OLDCODE
> >  FUNCTIONAL BLYP
> >  GRADIENT CORRECTION BECKE88 LYP
> >  LDA CORRELATION LYP
> > &END
> > --
> > --
> >
> >
> > _______________________________________________
> > CPMD-list mailing list
> > CPMD-list at cpmd.org
> > http://www.cpmd.org/mailman/listinfo/cpmd-list
> >
> _______________________________________________
> CPMD-list mailing list
> CPMD-list at cpmd.org
> http://www.cpmd.org/mailman/listinfo/cpmd-list