The ABINIT Users Mailing List ( CLOSED )

[Vincent Chevrier <vincent.chevrier@dahn.phys.dal.ca>]

Hi Christian,

Did you do a position optimization before the cell size optimization?
It is recommended to start with:
ionmov 2
optcell 0
and then take the result and do
ionmov 2
optcell 2

if this still gives you significant fluctuations you could try in the
first step:
ionmov 7
optcell 0
it usually does not converge as fast as ionmov=2, but I have found it to
work in cases where ionmov=2 gave me fluctuation issues.

good luck,
Vincent

> Hi,
>
> 1.0 d-3 is really too large to accept for abinit calculation, since the
> unit
> for tolmxf is Ha/a.u rather than eV/Ang.
>
> On Wed, Apr 29, 2009 at 9:54 PM, Ludwig, Christian
> <ludwigc@uni-mainz.de>wrote:
>
>> I am using 1.0d-3 for tolmxf. I found this value in published papers, so
>> I
>> figured it should be enough for my case.
>> In principle I agree, of course one should relax the forces, but in my
>> case
>> it took 74 Broyden steps for convergence. And the forces did not
>> converge
>> nicely. I had the impression that the fluctuate pretty randomly between
>> 1.0d-2 and 1.0d-3 until incidentally there are two consecutive low
>> values.
>> The development of the forces does not look systematic.
>> ________________________________________
>> Von: Josef Zwanziger [jzwanzig@gmail.com]
>> Gesendet: Mittwoch, 29. April 2009 15:32
>> An: forum@abinit.org
>> Betreff: Re: [abinit-forum] alternative to tolmxf
>>
>> Dear Christian,
>>
>> what would it mean to relax a structure without relaxing the forces?
>> What value of tolmxf are you using? If you really don't care about
>> getting a highly accurate structure, you could simply set tolmxf to a
>> coarse value (1.0D-4 say).
>>
>> On Wed, Apr 29, 2009 at 10:13 AM,  <ludwigc@uni-mainz.de> wrote:
>> > I am trying to relax a structure and observe the following: while the
>> total
>> > energy is stable pretty quickly, the forces fluctuate for some time,
>> which
>> > prevents a quick convergence. Since the total energy is the only value
>> I
>> am
>> > interested in, it would be nice to set a convergence criterion
>> analogous
>> to
>> > toldfe for the Broyden steps. Is this somehow possible? I found
>> nothing
>> in the
>> > list of input variables.
>> >
>> > Christian
>> >
>>
>>
>>
>> --
>> Josef W. Zwanziger
>> Professor of Chemistry
>> Canada Research Chair in NMR Studies of Materials
>> Director, Atlantic Region Magnetic Resonance Centre
>> Dalhousie University
>> Halifax, NS B3H 4J3 Canada
>> tel: +1 (902) 494-1960
>> fax: +1 (902) 494-1867
>> web: http://jwz.chem.dal.ca
>> jzwanzig@gmail.com, jzwanzig@dal.ca
>>
>>
>
>
> --
> Hai-Ping Lan
> Department of Electronics ,
> Peking University , Bejing, 100871
> lanhaiping@gmail.com, hplan@pku.edu.cn
>