The ABINIT Users Mailing List ( CLOSED )

[Ronald Cohen <rcohen@ciw.edu>]

This does not work. Even if I set nnsclo to a large number, it does not 
iterate the wave functions to  make them self-consistent:

iscf     3
tolwfr   1.0d-15
nstep    0
nnsclo   100
ecut     100.0
getden  -1
ngkpt    3 3 3

gives

----iterations are completed or convergence reached----

Mean square residual over all n,k,spin=   2.0770E+01; max=  2.3954E+01
  0.1667  0.1667  0.1667    1  2.21166E+01 kpt; spin; max resid(k); 
each band:
2.18E+01 2.15E+01 2.20E+01 2.05E+01 2.05E+01 2.06E+01 1.98E+01 2.10E+01
2.08E+01 2.04E+01 2.08E+01 2.03E+01 2.03E+01 2.05E+01 2.21E+01 2.03E+01
2.13E+01 2.04E+01




Fabien Bruneval wrote:
> Dear Ronald,
>
> Once you have calculated the ground-state density, you can diagonalize 
> the corresponding hamiltonian using "iscf -2" using the conjugate 
> gradient algorithm of abinit. The number of optimization steps of the 
> CG is governed by the variable "nstep".
> In your input file, you set nstep to 0. This means that your are not 
> calculating the kinetic energy of the Kohn-Sham wavefunctions, but the 
> kinetic energy of the starting wavefunctions (which are in fact random).
>
> If you want some more information, check the variable "nnsclo" in the 
> input variable list.
>
>
>
> Fabien
>
>
>
> Ronald Cohen wrote:
> > The Kohn-Sham kinetic energy is defined from the occupied wave 
> > functions for a given Kohn-Sham potential. It does not require 
> > self-consistency to compute. That is the value I need.
> >
> > Ron
> >
> >
> > Pawel Scharoch wrote:
> >
> > >    Dear colle agues,
> > >
> > >      there is one think I do not  quite understand. I would think 
> > > that  kinetic energy is the density functional and assumes a 
> > > particular  value for the ground state density. The ground state 
> > > density, once  it has been calculated in the SC procedure, with the 
> > > use of special k-points,
> > >  does not depend on a particular choice of k-points any more, 
> > > neither  the kinetic energy does! What is the problem then ?
> > >
> > >    With best regards,
> > >
> > >     Pawel
> > >
> > > ----- Oryginalna wiadomos'c' -----
> > > Od: "Allan, Douglas C Dr" <AllanDC@corning.com>
> > > Data: Sroda, Czerwiec 13, 2007 3:23 pm
> > > Temat: RE: [abinit-forum] non self-consistent kinetic energy
> > > Do: forum@abinit.org
> > >
> > >  
> > >
> > > > Hi Ron,
> > > >
> > > > I am no longer a user of ABINIT although I was in the past.  It may be
> > > > necessary to save a self-consistent potential at the same time that 
> > > > the
> > > > charge density is saved, I am not sure.  I know that we routinely did
> > > > this kind of calculation to compute band structures -- obviously 
> > > > for the
> > > > band structure we wanted a new selection of k-points that had 
> > > > nothing to
> > > > do with the original self-consistent "special points" set.  I think I
> > > > usually saved the potential, but in principle it should be easy to
> > > > compute a potential non-self-consistently from an input density and 
> > > > then
> > > > iterate the wavefunctions to convergence, all non-self-
> > > > consistently.  I
> > > > can't tell you exactly what iscf and other choices you need but 
> > > > surely a
> > > > current user can provide this to you.  The kinetic energy you seek 
> > > > wouldhave to be computed (non-self-consistently) in order to 
> > > > iterate the
> > > > wavefunctions to convergence in that fixed potential, so you would 
> > > > havethe number you seek.
> > > >
> > > > It may help to look in the help file to see how to compute band
> > > > structures and follow that procedure.
> > > >
> > > > Best regards,
> > > > Doug -----Original Message-----
> > > > From: Ronald Cohen [mailto:rcohen@ciw.edu] Sent: Tuesday, June 12, 
> > > > 2007 9:34 PM
> > > > To: forum@abinit.org
> > > > Subject: Re: [abinit-forum] non self-consistent kinetic energy
> > > >
> > > > Yes, exactly. That is what I want to do! So far I have not been 
> > > > able to
> > > > see how to do so.
> > > >
> > > > Sincerely,
> > > >
> > > > Ron
> > > >
> > > >
> > > >
> > > >
> > > > Allan, Douglas C Dr wrote:
> > > >   
> > > > > Dear Xavier and Ron,
> > > > >
> > > > > My reading of Ron's question suggests that he wants a kinetic       
> > > >
> > > > energy   
> > > > > corresponding with a given charge density that was generated       
> > > >
> > > > with   
> > > > > wavefunctions at k-points different from the new set.  That is,       
> > > >
> > > > he   
> > > > > does not have input wavefunctions for the case of interest, he       
> > > >
> > > > needs   
> > > > > to compute them.  If I am right, then he needs to read the       
> > > >
> > > > charge   
> > > > > density in and then do a non-self-consistent calculation that is 
> > > > > iterated to convergence in the wavefunctions.  The resulting       
> > > >
> > > > kinetic   
> > > > > energy should be what he seeks.  I no longer know the proper       
> > > >
> > > > iscf to
> > > > achieve this.
> > > >   
> > > > > Note that he is not going self-consistent again, but he is       
> > > >
> > > > iterating   
> > > > > to convergence in the wavefunctions.  This should be possible       
> > > >
> > > > for any   
> > > > > desired k-point grid whether it makes sense as a self-consistent 
> > > > > special points grid or not.  It is the same method one uses to       
> > > >
> > > > compute
> > > >   
> > > > > a bandstructure given a charge density.
> > > > >
> > > > > Best regards,
> > > > > Doug
> > > > >
> > > > > -----Original Message-----
> > > > > From: Xavier Gonze [mailto:gonze@pcpm.ucl.ac.be]
> > > > > Sent: Tuesday, June 12, 2007 9:12 AM
> > > > > To: forum@abinit.org
> > > > > Subject: Re: [abinit-forum] non self-consistent kinetic energy
> > > > >
> > > > > Dear Ron,
> > > > >
> > > > > On 12 Jun 2007, at 14:36, Ronald Cohen wrote:
> > > > >
> > > > >       
> > > > > > I am trying to figure out how to get ABINIT to compute and         
> > > >
> > > > print the   
> > > > > > kinetic energy that corresponds to a  given input  charge         
> > > >
> > > > density,   
> > > > > > generated with a different k-point set. This is needed to         
> > > >
> > > > correct QMC
> > > >
> > > >   
> > > > > > results. It is important to do this  without going self-         
> > > >
> > > > consistent   
> > > > > > again.With iscf=-3 I get a total energy, but not the kinetic         
> > > >
> > > > energy.>> In fact, if I print results%ek in the code it is zero. 
> > > > Setting nstep   
> > > > > > to 0 or 1 seems to give nonsense with iscf>0.
> > > > > >
> > > > > > Ronald Cohen
> > > > > >             
> > > > >
> > > > > Well, I would have said that you have to activate evaluation of       
> > > >
> > > > the   
> > > > > energy components by the routine energy.F90 (you might have a       
> > > >
> > > > look at   
> > > > > the header of that routine), which corresponds to nstep=0 . In       
> > > >
> > > > this   
> > > > > case, the energy is evaluated directly from the input wavefunctions.
> > > > > Moreover, there is a test case : Test v1#89 (some of the       
> > > >
> > > > datasets).    
> > > > > So, I am a bit surprised
> > > > > to see that nstep 0  gives nonsense for you. Could you send the       
> > > >
> > > > input   
> > > > > file ?
> > > > >
> > > > > Best regards,
> > > > > Xavier
> > > > >
> > > > >         
> > > >
> > > >
> > > >     
begin:vcard
fn:Ronald Cohen
n:Cohen;Ronald
org:Carnegie Institution of Washington;Geophysical Laboratory
adr:;;5251 Broad Branch Rd., N.W.;Washington;DC;20015;USA
email;internet:cohen@gl.ciw.edu
tel;work:2024788937
x-mozilla-html:FALSE
url:http://www.gl.ciw.edu/~cohen
version:2.1
end:vcard