[CPMD-list] TDDFT question

[Axel Kohlmeyer]

On Sat, 9 Dec 2006, Philip Shemella wrote:

[...]

PM>  From what I've understood most of the hybrid functionals are used so 
PM> they can include exact HF exchange.  With CPMD, HF exchange is not 
PM> usually calculated, unless you have too much computer time or a really 
PM> good reason to do so!

right, CPMD is a 'pure' DFT code. exact HF exchange (and thus hybrid
functionals like B3LYP or PBE0) is for most cases prohibitively 
expensive.

PM> If you don't want to go into the code -- one way to create a hybrid 
PM> functional manually would be to calculate the exchange-correlation 
PM> energy with separate jobs and then find the plug in the coefficients. 
PM> That is kind of a home-made hybrid functional.  Or... Perhaps the code 
PM> prints the Becke88, HF, and Slater exchange as well as the correlation 
PM> terms.  IF that is the case then you could just change the  exchange, 
PM> the HF exchange, and whatever other terms in the mixture that you like. 

i would say, that in this case changing the code, sound much simpler
to me.

PM>   If you look at the TOTAL ENERGY section after wavefunction convergence 
PM> you'll see the contributions from each term. I would test it for my own 
PM> sanity but I don't find any pseudopotentials for B3LYP.

there is currently no 'atomic' (i.e. pseudopotential generation) code 
for CPMD that supports hybrid functionals. people usually use the 
potentials generated for the 'pure' functional (i.e. BLYP for B3LYP, 
PBE for PBE0). hybrid functionals are not a commonly used feature of 
CPMD, and the improvements from them w.r.t. to the what people usually 
look at and what accuracy they need.

in general, it is problematic to see CPMD as 'yet another gaussian' type
of program. both packages are in fact somewhat orthogonal, i.e. what is
easy or straightforward to do with one, is expensive or complicated or
impossible to do with the other and vice versa.  transferring ideas,
strategies and concepts from one to the other can be misleading. my
recommendation is always, to start with a few typical, simple (and fast)
example applications and get a 'feeling' for how everything works before
you jump to your real (research) project. otherwise you'll have
difficulties to tell whether you made a technical mistake or whether
what you want to do simply does not work.

cheers,
    axel.

PM> 
PM> Take care, Phil
PM> 
PM> 
PM> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.