[Wien] Cholesky in supercell

Tue Apr 2 03:54:49 CEST 2024

Did you follow my suggestion and not use -it? It appears that you did not.
I would not use it until :DIS is less than 0.05, and perhaps not even then.
You are still being incomplete about the information you provide:

On Tue, Apr 2, 2024 at 1:34 AM Natalia Andreeva <nvandr at gmail.com> wrote:

> Dear Professor Laurence Marks,
> Thank you very much for your reply. I will try to reveal the question
> of what I use for the calculation.
>
> * The error occurs immediately when the first calculation cycle has
> not yet completed.
> * No, I don't use runsp, the vacuum value is 25au and RMT=5.20
> * I have a supercell without a defect and correctly calculated SCF
> with parameters 1x1x15, for which I performed the calculation without
> runsp. For this structure, the parameters were chosen to be
> RMT=5.09, Mt(Ba)=2.25, Rt(Ti)=1.74, Rmt(O)=1.58. This structure is an
> insulator, I also minimized the parameters for it.
>
Are you using the same parameters? What k-points are you using? Did you
optimize the positions then use them for your supercell?

> * I use WIEN2k version 23.2 with ELPA for calculations
> * So far I do not use minimization, but I plan to do it after
> receiving the correct calculation of this structure.
> * I use iterative diagonalization using the -it flag, on the second loop.
>
> I tried to reduce the Ba RMT to 2.1 using -ecut .998, however,
> this resulted in the error 'no energy limits found'. I have reduced
> the mixing parameter to case.in up to 0.1, and the Cholesky error
> appeared again on the third calculation cycle.
>
Did you do a fresh initialization with the new RMTs? It should not have led
to a problem in the first iteration.

DO NOT REDUCE THE MIXING PARAMETER, this is fake information. I do not know
how many times I have had to repeat this, please read the "Mixing for
Dummies" pdf.

In a few cases it may be appropriate to do "echo 0.005 > .pratt" for the
very first iteration only. The mixer uses prior information, and when there
is none it can be too aggressive. (Starting parameters for optimization and
fixed-point code are always guesswork.)

>
> Thank you in advance,
> With Best Regards,
> Natalia.
>
> On Wed, Mar 27, 2024 at 8:21 PM Laurence Marks <laurence.marks at gmail.com>
> wrote:
>
>> You have not provided enough information for a full diagnosis. When does
>> this occur, right at the beginning? Are you using runsp? How much vacuum?
>> What RKMAX? Do you have ELPA? What version of Wien2k? K-points?
>> Minimisation? Which iterative? When are you using the iterative, in the 2nd
>> cycle?
>>
>> Some suggestions:
>> * Reduce the Ba RMT to 2.1. You may need to use -ecut .998 during
>> initialization.
>> * Use a vacuum of at most 1/3 of full unit cell. 20au is often enough, 30
>> certainly, 40 is excessive.
>> * Minimize first you 1x1x15 cell with no vacancy & P4mm symmetry. Check
>> it is a sensible insulator. If not then everything is wrong.
>> * Remove, manually the Ti 3d search, as it is less stable.
>> * Use ELPA not iterative. While iterative may be faster per cycle, it can
>> lead to more iterations and less stability so the nett is slower. For
>> certain I would not use it for a surface until I have somewhat converged.
>> (You have to be in the space of densities where you have ions as against
>> the initial neutral atoms.)
>>
>> --
>> Professor Laurence Marks (Laurie)
>> Northwestern University
>> www.numis.northwestern.edu
>> https://scholar.google.com/citations?user=zmHhI9gAAAAJ&hl=en
>> "Research is to see what everybody else has seen, and to think what
>> nobody else has thought" Albert Szent-Györgyi
>>
>> On Wed, Mar 27, 2024, 21:22 Natalia Andreeva <nvandr at gmail.com> wrote:
>>
>>> Dear Win2k users,
>>>
>>>  I am trying to calculate a sufficiently large 2x3x15 supercell  based
>>> on a tetragonal barium titanate (BaTiO3) cell. In this calculation, I
>>> create an oxygen vacancy and a surface. The calculation is successfully
>>> initialized with the parameters of atomic spheres equal to Rmt(Ba)=2.30,
>>> Rmt(Ti)=1.82, Rmt(O)=1.65. However, when I run the calculation with PBE,
>>> iterative diagonalization, an error appears:
>>>
>>> ** Error in Parallel LAPW1
>>> ** LAPW1 STOPPED
>>> ** check ERROR FILES!
>>> Cholesky INFO = 9896
>>> 'SECLR4' - POTRF (Scalapack/LAPACK) failed.
>>>
>>> I tried to find an answer in WIEN2k's letters – among the suggestions
>>> was rerun the calculation after setting other initial electron density
>>> parameters (dstart), changing the radii of atomic spheres, and others, but
>>> they did not help. What could I do to run this calculation correctly?
>>>
>>>
>>> Thank you in advance,
>>> Best Regards,
>>> Natalia
>>>
>>> Add reaction
>>> _______________________________________________
>>> Wien mailing list
>>> Wien at zeus.theochem.tuwien.ac.at
>>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>> SEARCH the MAILING-LIST at:
>>> http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>>>
>> _______________________________________________
>> Wien mailing list
>> Wien at zeus.theochem.tuwien.ac.at
>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>> SEARCH the MAILING-LIST at:
>> http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>>
>
>
> --
> Best Regards,
> Natalia Andreeva.
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> SEARCH the MAILING-LIST at:
> http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>

-- 
Professor Laurence Marks (Laurie)
Northwestern University
Webpage <http://www.numis.northwestern.edu> and Google Scholar link
<http://scholar.google.com/citations?user=zmHhI9gAAAAJ&hl=en>
"Research is to see what everybody else has seen, and to think what nobody
else has thought", Albert Szent-Györgyi
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20240402/9221c6d9/attachment.htm>