[Pw_forum] convergence

Emine Kucukbenli kucukben at sissa.it
Sun Oct 23 10:37:47 CEST 2011


Hi Bhabya,

Well, I would reduce the conv_thr instead, probably to 1d-6 or smaller..
But in principle you can increase the forc_conv_thr=1d-2
and still get rid of the warning you have and probably of the  
convergence problem too.

It all depends on what you want:
=If this is a production run, you want accuracy, reduce conv_thr (try  
the default 1d-6 or smaller)

=If this is a test system, a toy problem for you to learn how to run  
the code, or get some basic idea on the properties of the system that  
you have and timings of computation, and if your computational  
resources are tight, then increase forc_conv_thr.

Basically, arrange the parameters of your calculations to match your needs.
Perform some small tests when you are in doubt.
cheers,
emine kucukbenli, phd student, sissa, italy.

ps: add your affiliation pls. was it Bhabha Atomic Research Centre, Mumbai?


Quoting bhabya sahoo <bdslipun at gmail.com>:

> shall i reduce to force con-thershold to 10^-2
>
> On Sat, Oct 22, 2011 at 2:36 AM, Emine Kucukbenli <kucukben at sissa.it> wrote:
>
>> Hello Bhabya Darsan,
>>
>> > but at the end the warning comes to reduce conv-threshold
>> > shall there may be some modification to the file
>>
>> Yes this is exactly the modification you need:
>> Reduce the convergence threshold conv_thr.
>> The force threshold you are using is the default 1d-3
>> You should not expect precise forces when the energy is calculated
>> almost within
>> the same precision.
>>
>> Also, this issue was previously discussed in the mailing list:
>> http://www.democritos.it/pipermail/pw_forum/2009-March/012104.html
>>
>> cheers
>> emine kucukbenli, phd student, sissa, italy
>> _______________________________________________
>> Pw_forum mailing list
>> Pw_forum at pwscf.org
>> http://www.democritos.it/mailman/listinfo/pw_forum
>>




More information about the Pw_forum mailing list