The ABINIT Users Mailing List ( CLOSED )

[lan haiping <lanhaiping@gmail.com>]

Dear Matteo, Thank you for this detailed instruction.
Another naive question about ecutwfn, 
when we do convergence testing on ecutwfn,  
what is the proper initial setting for it ?
I once used half of ecut , but it seems very large computations are required, even exhausted the physical memory about 16 GB for about 100 electrons.
Is there any trick to find a proper guess value for starting convergence test ?

Thanks againd for your time and help.
regards,

On Wed, Mar 18, 2009 at 2:55 AM, Matteo Giantomassi <Matteo.Giantomassi@uclouvain.be> wrote:

> we can obviously have  that ecuteps  controls the dimension of
> $G\times G$ for  $\chi_0$ in reciprocal space.
> My question is what does ecutwfn  account for $\chi_0$ ?  From the
> equation
> for $\chi_0$ i donot find any extra parameters  to reach final result.
> This
> setting really makes me confusion, any hints and comments are appreciated.

 ecutwfn defines the number of plane waves used to represent the wave
functions
 entering the definition of the oscillator matrix elements

> \begin{equation}
> M_{kij}(q+G)=<k-qi|e^{-i(q+G)r}|kj>
> \end{equation}

 The oscillators are calculated using a two-step procedure.

 1) First one performs a FFT, from G to r space, of the two kets
   |k-q,i>  and |kj>  where the number of employed G-vectors
  is equal to npwwfn.

 2) The product u^*_{k-q,i}(r) u_{k,j}(r)
  is constructed on the real space FFT mesh,
  and Fourier transformed to reciprocal space to obtain M_{kij}(q,G)

 Regards,
 Matteo Giantomassi

-- 
Hai-Ping Lan 
Department of Electronics ,
Peking University , Bejing, 100871
lanhaiping@gmail.com, hplan@pku.edu.cn