[Pw_forum] problem in relaxation by using bfgs

Javad hashemifar jhashemifar at gmail.com
Fri Dec 22 11:52:20 CET 2006


Thanks Huiqun for your kind answer.
In fact I was using 2 processors for my job. But the interesting point
is that when about 2 hours ago I repeated calculation of one case on 4
processors (with exactly the same input parameters as before) the
mentioned error message did not appear any more and system relaxed
completely and correctly.
I will try to check the first_order extrapolation.
Thanks again
Javad Hashemifar


On 12/22/06, Huiqun Zhou <hqzhou at nju.edu.cn> wrote:
> Javad,
>
> I got the same problem several months before. In my case, I noticed that if
> I
> run my job on equal to or less than 2 processors (or CPU cores), it will be
> OK.
>
> Another way to get rid of it is to comment out the pot and wfc
> extrapolation.
> Of course, this will slow down your calculation to some extent.
>
> I didn't try to use 'first_order', you may test this first.
>
> Huiqun
>
>
> ----- Original Message -----
> From: "Javad hashemifar" <jhashemifar at gmail.com>
> To: <pw_forum at pwscf.org>
> Sent: Friday, December 22, 2006 4:20 PM
> Subject: [Pw_forum] problem in relaxation by using bfgs
>
>
> > Dear Pwscf users
> >
> > I am trying to relax a tetragonal interface supercell at different
> > values of vertical lattice parameter (celldm(3)). I am using
> > calculation=relax and the namelist IONS is as follows:
> >
> > ion_dynamic = bfgs
> > bfgs_ndim    = 3,
> > pot_extrapolation = "second_order",
> > wfc_extrapolation = "second_order",
> >
> > I have done this relaxation at several values of celldm(3) to find the
> > ones that has minimum energy and zero stress. Let me to note that I
> > can use of vc-relax calculations for this purpose because celldm(1) is
> > fixed. For some values of celldm(3) every thing is OK and the system
> > relax well and both etot_conv_thr and forc_conv_thr satisfies. But in
> > some other values of celldm(3) the calculation stops with the
> > following error:
> >
> > %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
> > from checkallsym : error #         1
> > some of the original symmetry operations not satisfied
> > %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
> >
> > After more accurate consideration of the cases that have above error,
> > I found that in that cases, the relaxation procedure has been done
> > correctly for a few steps (2 - 3 step) and then in the last steps the
> > position of atoms in x and y directions also has changed while there
> > is no force in these two direction. But after that program has tried
> > to symmetrize the position and change back the value of x and y
> > positions as follows:
> >
> > ATOMIC_POSITIONS (alat)
> > Ga       0.000000000   0.000000000   0.000000000
> > As       0.000000000   0.502872976   0.354710310
> > Cr        0.502872976   0.502872976   0.686421612
> > As       0.502872976   0.000000000   1.018132914
> >
> >
> >  SYMMETRIZED ATOMIC COORDINATES)
> >
> > ATOMIC_POSITIONS (alat)
> > Ga       0.000000000   0.000000000   0.000000000
> > As       0.000000000   0.500000000   0.350788698
> > Cr        0.500000000   0.500000000   0.682500000
> > As       0.500000000   0.000000000   1.014211302
> >
> > and then the above mentioned error occurs. It is the situation for all
> > the cases that relaxation has not been completed.
> > When I restarted one of this cases with damp method (instead of bfgs)
> > the above error did not appear and the system relaxed correctly.
> >
> > I would appreciate if some body explains why by using bfgs such errors
> > occurs and how I can solve that. Otherwise I should use the damp
> > method for relaxation that usually needs to more steps to reach to the
> > minimum energy.
> >
> > thanks in advance
> > Javad Hashemifar
> > _______________________________________________
> > Pw_forum mailing list
> > Pw_forum at pwscf.org
> > http://www.democritos.it/mailman/listinfo/pw_forum
> >
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>


-- 
======================================
Seyed Javad  Hashemifar,    Ph.D.
[current:]
Tel:+49-203-3794743        Fax:+49-203-3794742
Fachbereich Physik (AG Kratzer), Universitat Duisburg-Essen
47048 Dusiburg, Germany
[permanent:]
Tel:  +98-311-3912375       Fax: +98-311-3912376
Physics Department, Isfahan University of Technology
84156 Isfahan, Iran
---------------------------------------------------------------------------



More information about the users mailing list