The ABINIT Users Mailing List ( CLOSED )

[Wissam Al-Saidi <alsaidi@jasmine.physics.wm.edu>]

I want to call your attention to this discussion on the Abinit  list.


Wissam



Xavier Gonze wrote:

> Dear Nuno,
>
> I think you should simply make up your own experience :
> do some representative MD run, with some toldff and ecut, and
> then do it again, with different toldff and ecut, and select the one
> that are convenient for your system and the property you are looking  
> for ...
>
> Xavier
>
>
>
>
>
> On 30 Mar 2006, at 11:16, Nuno Galamba wrote:
>
> > Dear Abinit users
> >
> > Thanks for your replys on my question about force convergence.
> > I am aware that if you set toldff 10-4 Ha/Bohr then the energy will  
> > converge
> > up to the 10-12 Ha. That clearly demonstrates that energy converges  
> > a lot
> > faster than the forces for a given tolerance. My question was: Is  
> > there a
> > toldff and an ecut that make my forces not to change if I increase  
> > ecut?
> > The implications of this are for example on the dynamics of a  
> > system. If for a
> > given ecut and toldff the forces are not converged although the  
> > energy is,
> > then on a MD simulation for example my dynamics can be  "compromised" 
> > even
> > though my energy is fully converged.
> >
> > When using local basis functions, unless you use an extrapolation  
> > method for
> > infinite basis sets, you are otherwise always working with non- conveged
> > energies and forces. The great advantage of plane waves is exactly  
> > to avoid
> > this problem, i.e., there are no Pulay forces and one can, in  
> > principle, work
> > with converged energy (and forces ?).
> >
> > I started doing some basic runs to get some insight on how fast  
> > energy and
> > forces converge and was a bit surprised on that I cannot realy see a
> > convergence pattern for the forces.
> >
> > The answer of Dr Michel Côté "If you have convergence up to 10-5 (5  
> > digits),
> > then it is normal that forces on atoms that are smaller still  change 
> > with
> > increasing the cutoff" unfortunately makes a lot of sense.
> >
> > Following this, if the forces on my system are of the order of 10-6  
> > Ha/Bohr
> > then I need at least a toldff, say 10-9 Ha/Bohr, for a given value  
> > of ecut
> > for which energy is converged, to avoid seeing my forces changing when
> > increasing ecut. Is this what you meant ????
> >
> > If so, then for most systems this is simply not possible. Then my  
> > question is:
> > Does anybody know what the effect is on the dynamics of a system of  
> > working
> > with non-converged forces??????
> >
> > Thanks a lot once again
> >
> > N. Galamba
> >
> >
> >
> >
> >
> >
> >
> > On Wednesday 29 March 2006 19:11, Michel Côté  wrote:
> >
> > > I do not understand your question very well. Toldff converge the  
> > > largest
> > > force. If you have convergence up to 10-5 (5 digits), then it is  
> > > normal
> > > that forces on atom that are smaller still change with increasing the
> > > cutoff. In other words, it is not the number of significant digit  
> > > that is
> > > converged but the absolute number.
> > >
> > > Michel
> > >
> > > Le 29/03/06 10:41, « Nuno Galamba » <ngalamba@cii.fc.ul.pt> a écrit :
> > >
> > > > Dear Abinit users
> > > >
> > > > I have a simple question regarding the choice of ecut for an  
> > > > arbitrary
> > > > system.
> > > >
> > > >  By doing a calculation for different values of ecut (say,
> > > > 10,15,20,25,30,35,40,45,50 Hartrees) for a given value of toldff  (say
> > > > 10-4 Ha/Bohr) one can  observe a plateau at which energy  converges 
> > > > to.
> > > >
> > > > Now the question is, say the energy is converged for ecut=30 Ha  
> > > > (meaning
> > > > it does not change up to the first 5 digits), in principle the
> > > > Hellmann-Feynman forces on every particle of the system should be
> > > > converged too.
> > > >
> > > > I am not quite sure this is the case. I observe that the force on  the
> > > > particles of the system are not the same. These are of the order  
> > > > of 10-6
> > > > Ha/Bohr. Souldn't one expect some kind of convergence behavior  for 
> > > > the
> > > > force on every particle when increasing systematically the number of
> > > > plane waves for a given calculation?
> > > >
> > > > Thanks in advance