From mbdtsnm {*at*} hpf.ch.man.ac.uk  Tue Oct  7 08:16:40 1997
Received: from nessie.mcc.ac.uk  for mbdtsnm ":at:" hpf.ch.man.ac.uk
	by www.ccl.net (8.8.3/950822.1) id HAA24997; Tue, 7 Oct 1997 07:10:02 -0400 (EDT)
Received: from yama.mcc.ac.uk by nessie.mcc.ac.uk with SMTP (PP);
          Tue, 7 Oct 1997 12:09:31 +0100
Received: from serenity.mcc.ac.uk by yama.mcc.ac.uk with SMTP (PP);
          Tue, 7 Oct 1997 12:06:36 +0100
Received: from hpf.ch.man.ac.uk [130.88.12.28] by serenity.mcc.ac.uk 
          with smtp (Exim 1.651 #21) id 0xIXTF-0005Zk-00;
          Tue, 7 Oct 1997 12:06:33 +0100
Date: Tue, 7 Oct 1997 12:05:20 +0100
Message-Id: <28762.9710071105 ^%at%^ hpf.ch.man.ac.uk>
From: "Nathaniel (noj) Malcolm" <mbdtsnm {*at*} hpf.ch.man.ac.uk>
To: chemistry ^at^ www.ccl.net
In-Reply-To: <3439DEDC.41C6 #at# h1tw0036.hoechst.com> (schiffer #at# h1tw0036.hoechst.com)
Subject: Re: CCL:DFT and H2O Dimer
Reply-To: Noj.Malcolm {*at*} man.ac.uk


>Hi CCL'ers,
>I am not d'accord with Rudolf Herrmann's statement, that DFT is
>not suitable for the water dimer, because "DFT methods are not meant
>to treat long range interactions such as van der Waals forces". The
>electrostatic part of the intermolecular interaction ist really the
>most long range part and is treated the same way in DFT as in
>conventional ab initio methods. What is really missing is Dispersion,
>which is important for the interaction of non-polar molecules like
>hydrocarbons and esp. for aromatic systems. Therefore the interactions
>in the water dimer, which are dominated by the electrostatic and
>polarization part of the interaction (i.e. H-bonds), are very well
>described by DFT methods, like BP, BLYP, B3LYP, etc. (not LDA !, but
>LDA is not DFT). On the other hand, the benzene dimer ( and of
>course all noble gas dimers ) is only poorly described by DFT methods.
>Best (!!!!) references :


just a brief comment - on this comment to an earlier statement - 
i guess that one would
generally be keen to use DFT (rather than HF say) due to its inclusion
of correlation effects. As has also been discussed on the CCL recently the
'dispersion' effects are mainly driven by correlation effects 
therefore it seems to me that to say that DFT methods are 'good' for
the water dimer
problem because it gets most of the effects correct (except maybe
dispersion, which should not be the main force)
is placing too much faith in DFT - as it should really be able to deal
with dispersion as well (if it got the correlation right).


I guess my point is why not just use HF rather than DFT if its 
inclusion of correlation is not good enough to give a reasonable
physical description of dispersion forces?

noj

p.s. i'm sure that DFT does in practice offer improvement over HF its just that
i don't feel that Dr. Schiffer's argument explained why one should
consider it a 'good' method for these interactions.

(and i'm sure you could find the odd physicist out there who would be
able to argue that LDA IS DFT - maybe just not 'good' DFT for chemists)

-- 
Dr. Noj Malcolm			
Theory Group,			
Dept. of Chemistry,		
University of Manchester,	
Oxford Road,			
Manchester.			
M13 9PL			

e-mail noj.malcolm -A_T- man.ac.uk