[gpaw-users] Convergence using GLLBSC

[Hongliang Xin]

Thanks Marcin. It is great to see it converges. I guess it is the
compromise we have to take for now.
I will check in PBE how smearing affects optical spectra that is what we
are interested in.

Thanks again.
Hongliang


On Tue, Jul 22, 2014 at 5:47 AM, Marcin Dulak <Marcin.Dulak at fysik.dtu.dk>
wrote:

>  Hi,
>
>
> On 07/21/2014 12:35 AM, Hongliang Xin wrote:
>
>   Thanks for reply. I was using gpaw 0.10.0.9447 with
> gpaw-setups-0.8.7929. I tried to converge only the occupied band but still
> have the problem. I attached here the python script with output file.
>
>   i see that this example converges with PBE.
> I can make it converge with GLLBSC with width=0.1, but not with a low
> temperature, like your 0.01 - i'm not sure this
> is acceptable for your purpose.
> Your GPAW version is old - in the test.py you find what resembles the
> latest GPAW settings (eigensolver, mixer) -
> all standard systems should converge with those.
>
> Best regards,
>
> Marcin
>
>  Thanks,
>  Hongliang
>
>
> On Mon, Jul 14, 2014 at 1:48 AM, Marcin Dulak <Marcin.Dulak at fysik.dtu.dk>
> wrote:
>
>>  Hi,
>>
>>
>> On 07/14/2014 08:02 AM, Hongliang Xin wrote:
>>
>>   Dear All,
>>
>>  I am testing a small Ag cluster with 56 atoms using GLLBSC for better
>> description of electronic structure. The convergence is really bad. I am
>> just wondering anyone has some tips for using GLLBSC functional in general.
>> I paste here the calculator and convergence.
>>
>>  please provide a full script.
>> Which version of GPAW and setups are you using?
>>
>>
>>  Thanks,
>>  Hongliang
>>
>> Here is calculator:
>> calc = GPAW(mode = 'fd',
>>             basis = 'dzp',
>>             nbands = -int(len(atoms)*5*0.15),
>>             xc = 'GLLBSC',
>>             kpts = (1,1,1),
>>              #poissonsolver=PoissonSolver(relax='GS', eps=1e-10, nn='M'),
>>             occupations=FermiDirac(0.01),
>>             convergence={#'energy': 1.0e-3,
>>                          #'density': 1.0e-4,
>>                          #'eigenstates': 1.0e-8,
>>                          'bands': -30},
>>
>> if you are converging unoccupied bands then you need
>> eigensolver=Davidson(N)
>> or eigensolver=CG(N), with N being ~3. Davidson is few times faster but
>> may perform more steps.
>> Do not decrease convergence thresholds, and start with defaults
>> (i have put a # mark for things to be commented out).
>>
>> Best regards,
>>
>> Marcin
>>
>>              #mixer=Mixer(beta=0.005,nmaxold=7,weight=50.0),
>>
>>             spinpol = False,
>>             usesymm = False,
>>             maxiter = 1000,
>>             gpts=h2gpts(0.18, atoms.get_cell(), idiv=8),
>>             txt='out.txt',
>>             verbose=False)
>>
>> Here is the convergence:
>> iter: 988  19:29:53  -2.9   -2.3     -133.856014  2      5
>> iter: 989  19:30:22  -2.7   -2.2     -133.543733  3      8
>> iter: 990  19:30:50  -3.1   -2.6     -133.450748  2      3
>> iter: 991  19:31:17  -3.3   -2.7     -133.489621  3      2
>> iter: 992  19:31:46  -3.2   -2.5     -133.613931  3      4
>> iter: 993  19:32:14  -3.2   -2.4     -133.650107  3      4
>> iter: 994  19:32:42  -3.2   -2.3     -133.608913  2      4
>> iter: 995  19:33:12  -3.4   -2.3     -133.687230  3      5
>> iter: 996  19:33:40  -3.7   -2.2     -133.720904  2      3
>> iter: 997  19:34:08  -3.4   -2.2     -133.658227  3      5
>> iter: 998  19:34:37  -3.4   -2.3     -133.594052  2      5
>> iter: 999  19:35:07  -3.1   -2.3     -133.577819  3      5
>> iter: 1000  19:35:36  -3.1   -2.4     -133.606794  3      5
>>
>>
>> --
>>  Hongliang Xin
>> Postdoc
>> SUNCAT
>> SLAC National Accelerator Laboratory
>> Stanford University
>> Phone: (734) 474-3629 <%28734%29%20474-3629>
>> E-mail: hxin at slac.stanford.edu <hxin at umich.edu>
>>              hongliang06 at gmail.com
>>
>>
>> _______________________________________________
>> gpaw-users mailing listgpaw-users at listserv.fysik.dtu.dkhttps://listserv.fysik.dtu.dk/mailman/listinfo/gpaw-users
>>
>>
>>
>>
>
>
> --
>  Hongliang Xin
> Postdoc
> SUNCAT
> SLAC National Accelerator Laboratory
> Stanford University
> Phone: (734) 474-3629
> E-mail: hxin at slac.stanford.edu <hxin at umich.edu>
>              hongliang06 at gmail.com
>
>
>
> --
> ***********************************
>
> Marcin Dulak
> Technical University of Denmark
> Department of Physics
> Building 307, Room 229
> DK-2800 Kongens Lyngby
> Denmark
> Tel.: (+45) 4525 3157
> Fax.: (+45) 4593 2399
> email: Marcin.Dulak at fysik.dtu.dk
>
> ***********************************
>
>


-- 
Hongliang Xin
Postdoc
SUNCAT
SLAC National Accelerator Laboratory
Stanford University
Phone: (734) 474-3629
E-mail: hxin at slac.stanford.edu <hxin at umich.edu>
            hongliang06 at gmail.com
<hxin at umich.edu>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.fysik.dtu.dk/pipermail/gpaw-users/attachments/20140722/24294c4b/attachment.html>