[Pw_forum] Conergence Delay

Mike Marchywka marchywka at hotmail.com
Mon May 19 01:10:42 CEST 2014






----------------------------------------
> From: paolo.giannozzi at uniud.it
> To: pw_forum at pwscf.org
> Date: Sun, 18 May 2014 22:25:05 +0200
> Subject: Re: [Pw_forum] Conergence Delay
>
> On Sat, 2014-05-17 at 15:35 +0000, Elliot Menkah wrote:
>
>> I have this relax calculation running on 16 cores.
>> Its been running for the past 73 days.
>
> ... and you don't find anything strange in it?!?
>
>> Can anyone please help me understand the interpretation
>> as to when it would converge
>
> nobody can help you. Convergence depends upon many different factors.
> Ill-chosen input data or even well-chosen input data for nasty systems
> may never converge. Each self-consistency step should converge in no
> more than a few tens of iterations. If it doesn't, it is pointless
> to insist.

As I understand the situation currently, convergence is pretty much empirical 
there is not a lot of theory on what will and will not converge if you play with
the mixing parameter. Is this more or less accurate or are there specific
references that you could suggest for more details? this question is not so
much directed at the immediate point of this thread but more general curiosity.

Thanks.



>
> P.
>> . Especially the non-human-readable iteration number.
>>
>> iteration #*** ecut= 60.00 Ry beta=0.20
>> Davidson diagonalization with overlap
>> ethr = 4.68E-11, avg # of iterations = 3.4
>>
>> negative rho (up, down): 0.229E-04 0.227E-04
>>
>> total cpu time spent up to now is 5498418.9 secs
>>
>> total energy = -3906.86845896 Ry
>> Harris-Foulkes estimate = -3906.86845896 Ry
>> estimated scf accuracy < 0.00000077 Ry
>>
>>
>> Kind Regards,
>>
>> Elliot
>>
>
> --
> Paolo Giannozzi, Dept. Chemistry&Physics&Environment,
> Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
> Phone +39-0432-558216, fax +39-0432-558222
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
 		 	   		  



More information about the users mailing list