[Pw_forum] Convergence test

[Huiqun Zhou]

Axel,

Your advices are valuable, thanks a lot!. Seeing most of my calculations
(geometry optimization) are still effective, I can set my heart at rest :-).

Dr. Huiqun Zhou
@Nanjing University

----- Original Message ----- 
From: "Axel Kohlmeyer" <akohlmey at cmm.chem.upenn.edu>
To: "Huiqun Zhou" <hqzhou at nju.edu.cn>
Cc: <pw_forum at pwscf.org>
Sent: Wednesday, November 14, 2007 1:31 PM
Subject: Re: [Pw_forum] Convergence test


> On Tue, 13 Nov 2007, Huiqun Zhou wrote:
>
> HZ> List users,
>
> dear huiqun zhou,
>
> HZ> I'm testing convergence on ecutwfc for chromite using USPP. Below are
> HZ> results of different ecutrho with different ecutwfc. I may be wrong, I 
> think
> HZ> I should watch the difference of total energy between successive runs 
> till
> HZ> it's less than 1-2 mRy. But the trend is not apparent in the results 
> below
> HZ> and it seems that the larger the ecutwfc, the lower the total energy.
>
> you are using an ultra-soft pseudopotential and for those
> actually the density cutoff is _much_ more important than
> the wavefunction cutoff. depending on the choice of the
> pseudopotential parameters, 35-40ry for the wavefunction
> cutoff should be a conservative choice, however a density
> cutoff of 250 is not. if you want to converge the energy
> very tightly, you may need to go up to 10x the wavefunction
> cutoff. you should also keep in mind that the total energy
> converges rather slowly, and that for most calculations
> you need energy differences or derivatives, which may converge
> faster, i.e. you may want to look at forces (use a non-symmetric
> input for that!!) or stress instead. what is sufficiently
> converged depends on your the kind of calculation you want
> to do. so for simple geometry optimization or MD 35/210
> or less may already be sufficient.
>
> HZ>
> HZ> I calculated E-V relations using two sets of ecutwfc and ecutrho:
> HZ> for ecutwfc = 35 ry and ecutrho = 240 ry, a = 8.405, B0 = 172.8, B0' = 
> 4.3
> HZ> for ecutwfc = 55 ry and ecutrho = 250 ry, a = 8.407, B0 = 178.6, B0' = 
> 3.6
> HZ>
> HZ> There are no big differences in unit cell parameter and elastic 
> properties.
> HZ> So, my questions are
> HZ> (1) usually, the ecutwfc of USPP should be around 35 ry or below, is 
> my
> HZ>   case an exception?
>
> no. it does not look like an exception.
>
> HZ> (2) how small the difference of total energy of successive runs should 
> be
> HZ>   in judging convergence against ecutwfc?
>
> total energy is not a good parameter. see above.
>
> cheers,
>  axel.
>
> HZ>
> HZ> Thank you in advance.
> HZ>
> HZ> Dr. Huiqun Zhou
> HZ> @Earth Sciences, Nanjing University, China
> HZ>
> HZ>
> HZ>
> HZ>
> HZ> ecutrho = 240 ry
> HZ> ecutwfc                     total energy
> HZ> 35                     -1461.54573299 Ry
> HZ> 40                     -1461.64206824 Ry
> HZ> 45                     -1461.71978835 Ry
> HZ> 50                     -1461.76587314 Ry
> HZ> 55                     -1461.76316597 Ry
> HZ> 60                     -1461.76692396 Ry
> HZ>
> HZ> ecutrho = 250 ry
> HZ> 40                     -1461.64351876 Ry
> HZ> 45                     -1461.70130510 Ry
> HZ> 50                     -1461.74724764 Ry
> HZ> 55                     -1461.76462943 Ry
> HZ> 60                     -1461.76838812 Ry
> HZ>
> HZ> ecutrho = 280 ry
> HZ> 50                     -1461.74778158 Ry
> HZ> 55                     -1461.76516515 Ry
> HZ> 60                     -1461.76892327 Ry
> HZ>
> HZ> _______________________________________________
> HZ> Pw_forum mailing list
> HZ> Pw_forum at pwscf.org
> HZ> http://www.democritos.it/mailman/listinfo/pw_forum
> HZ>
>
> -- 
> =======================================================================
> Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
>   Center for Molecular Modeling   --   University of Pennsylvania
> Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
> tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
> =======================================================================
> If you make something idiot-proof, the universe creates a better idiot.
>