[Wien] Large charge fluctuation in hybrid calculations

[Laurence Marks]

In addition, be aware that monitoring :DIS is not enough, this only
shows the convergence of the L=M=0 density inside the muffin-tins,
While this is often a good indicator, often it is not particularly
when you have non-convergence in the plane waves. A small script I use
is:

grep -e :DIR -e :ENE -e :CHARG -e PRATT -e Pratt -e :DIS -e "MIXING
SC" -e PLANE *.scf $1 | \
grep -v -e SS -e scheme | tail -40

(remove line breaks caused by the emailer).

On Tue, Aug 11, 2009 at 7:24 AM, Laurence Marks<L-marks at northwestern.edu> wrote:
> Assuming that you are using MSEC1, the fundamental reason that
> "standard tricks" don't work is, simply, that these tricks don't work
> with MSEC1! (In fact they don't really work the way people think even
> with BROYD, and never have, although they might with PRATT.)
>
> The normal reason why a calculation does not converge is because the
> physics is badly posed in the model being used. You say that the
> problem occurs even in cycle 1 for NSO calculations, so is not related
> to -so or -eece. I would go back and look very carefully at how the
> 8-layer model is setup; are the atoms right, do you have inadequate
> k-points, is there something wrong in the symmetry? (99% certain that
> you will find something).
>
> N.B., I would be very cautious (skeptical) about applying a on-site
> hybrid method to a metal. This may be both incorrect physics and the
> problem may be unfeasible, i.e. there may be not self-consistent
> solution. (Nothing in fact says that a self-consistent solution has to
> exist!)
>
> 2009/8/11 Atta-fynn, Raymond <attafynn at uta.edu>:
>> Dear Users,
>>
>> I am performing non-magnetic calculations on a metal slab using hybrid DFT
>> (with TEMP smearing). I have been able to converge 1-7 layers and 9-layer
>> thick slabs with no serious problems. For some reason the 8-layer slab does
>> not want to converge. The charge density oscillates with large
>> amplitudes and eventually the job crashes. Calculations were performed the
>> standard way:
>>
>> runsp_c_lapw -ec 0.00001 -cc 0.001 -i 300 -p
>> runsp_c_lapw -ec 0.00001 -cc 0.001 -i 300 -p -so
>> runsp_c_lapw -ec 0.00001 -cc 0.001 -i 300 -p -so -eece
>>
>> The problem begins in step 1 (NSO calculations). I've pulled all  the
>> standard tricks (reduce mixing, etc.) but the problem will just not go away.
>> The first few cycles for the charge distance looks like:
>>
>> :DIS  :  CHARGE DISTANCE       ( 3.0623129 for atom    2 spin 2)
>> 4.8323633
>> :DIS  :  CHARGE DISTANCE       ( 3.7525968 for atom    8 spin 2)
>> 4.8109130
>> :DIS  :  CHARGE DISTANCE       ( 4.5335108 for atom    4 spin 2)
>> 4.7597241
>> :DIS  :  CHARGE DISTANCE       ( 5.1545071 for atom    8 spin 2)
>> 4.9702637
>> :DIS  :  CHARGE DISTANCE       ( 5.0675077 for atom    1 spin 2)
>> 4.8569369
>> :DIS  :  CHARGE DISTANCE       ( 4.9684119 for atom    1 spin 2)
>> 4.7391195
>> :DIS  :  CHARGE DISTANCE       ( 4.3727780 for atom    1 spin 2)
>> 4.3237943
>> :DIS  :  CHARGE DISTANCE       ( 3.7769948 for atom    4 spin 2)
>> 4.8503262
>> :DIS  :  CHARGE DISTANCE       ( 3.3436987 for atom    4 spin 2)
>> 3.8495739
>> :DIS  :  CHARGE DISTANCE       ( 3.2228522 for atom    1 spin 2)
>> 5.1590833
>> :DIS  :  CHARGE DISTANCE       ( 2.7148248 for atom    7 spin 2)
>> 3.7982418
>> :DIS  :  CHARGE DISTANCE       ( 3.3313579 for atom    7 spin 2)
>> 4.5302253
>> :DIS  :  CHARGE DISTANCE       ( 3.3598285 for atom    4 spin 2)
>> 5.0925813
>> :DIS  :  CHARGE DISTANCE       ( 2.9529033 for atom    4 spin 2)
>> 5.2109088
>> :DIS  :  CHARGE DISTANCE       ( 3.3483133 for atom    4 spin 2)
>> 5.0412665
>> :DIS  :  CHARGE DISTANCE       ( 2.9641391 for atom    7 spin 2)
>> 4.8027516
>> :DIS  :  CHARGE DISTANCE       ( 3.1663445 for atom    4 spin 2)
>> 4.7292585
>> :DIS  :  CHARGE DISTANCE       ( 3.2054421 for atom    5 spin 2)
>> 3.9307946
>> :DIS  :  CHARGE DISTANCE       ( 3.1672886 for atom    5 spin 2)
>> 4.4866535
>> :DIS  :  CHARGE DISTANCE       ( 3.2654575 for atom    1 spin 2)
>> 5.0048377
>> :DIS  :  CHARGE DISTANCE       ( 3.1100926 for atom    5 spin 2)
>> 4.2558609
>> Any suggestions would be appreciated.
>>
>> Thank you.
>> _______________________________________________
>> Wien mailing list
>> Wien at zeus.theochem.tuwien.ac.at
>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>
>>
>
>
>
> --
> Laurence Marks
> Department of Materials Science and Engineering
> MSE Rm 2036 Cook Hall
> 2220 N Campus Drive
> Northwestern University
> Evanston, IL 60208, USA
> Tel: (847) 491-3996 Fax: (847) 491-7820
> email: L-marks at northwestern dot edu
> Web: www.numis.northwestern.edu
> Chair, Commission on Electron Crystallography of IUCR
> www.numis.northwestern.edu/
> Electron crystallography is the branch of science that uses electron
> scattering to study the structure of matter.
>



-- 
Laurence Marks
Department of Materials Science and Engineering
MSE Rm 2036 Cook Hall
2220 N Campus Drive
Northwestern University
Evanston, IL 60208, USA
Tel: (847) 491-3996 Fax: (847) 491-7820
email: L-marks at northwestern dot edu
Web: www.numis.northwestern.edu
Chair, Commission on Electron Crystallography of IUCR
www.numis.northwestern.edu/
Electron crystallography is the branch of science that uses electron
scattering to study the structure of matter.