[Wien] Unable to achieve required force convergence in structure optimization
Peter Blaha
pblaha at theochem.tuwien.ac.at
Mon Aug 25 07:47:09 CEST 2014
If this is true, you should not be able to do any optimizations (not just limited
to a few mRy/bohr).
Anyway, for the moment:
grep inM *_lapw
grep inM SRC_*/*.f
(and also in w2web if you are using it)
and change inM to inmini. (You need at least to modify x, run and mixer)
An alternative was discussed in the mailing list, and you can modify your OSX
so hat it becomes cases sensitive.
k-mesh: FIRST do the optimization with a VERY lousy k-mesh. Later on, you can still increase it.
and: For what purpose do you need 1 meV absolute convergence ? You are always looking for
E-differences, thus you have to test the convergence with E-differences !
and: change RKmax, RMTs, ... and you will get MUCH LARGER changes in E-tot !
It is simply "stupid" to do the optimization with such a k-mesh.
Am 25.08.2014 04:14, schrieb Hu, Wenhao:
> Dear Peter:
>
> I just realized a problem. Because the operation system of my cluster is OSX, it's unable to distinguish case.inm and case.inM. You mentioned I have to change the convergence parameter to a small value in inM file for ending MSR1a mode. Everytime when I do structure optimization, I'm unable to modify the parameter in .inM file. I'm not very familiar with WIEN2K. Do you have any suggestions on solving this problem?
>
> About the kmesh, I want to get an accuracy of 1meV in the total energy, and I compared the convergence of the total energy in unrelaxed supercell with different sizes of kmesh. I found a 7x7x7 kmesh is required to reach that accuracy.
>
> I appreciate any suggestions from you.
>
> best regards,
> Wenhao
> ________________________________________
> From: Peter Blaha [pblaha at theochem.tuwien.ac.at]
> Sent: Sunday, August 24, 2014 6:15 AM
> To: Hu, Wenhao
> Subject: Re: [Wien] Unable to achieve required force convergence in structure optimization
>
> This is even a case which converges very nicely.
>
> init -b -sp -numk 100 -ecut -8 -rkmax 5
> runsp -fc 5 # indicates huge forces (150)
> runsp -fc 1 with MSR1a # converges in lt. 60 cycles.
>
> My forces are even below 0.2, although in case.inM I used the default of
> 2.0 for ending the MSR1a mode. (In any case, if you want small forces, you must set
> this value to a very small value too (eg. 0.1).
>
> After a first relaxation you will see that the Ni-Si distances increased a lot
> and you should restart with larger RMTs for Si.
>
> Why would you use a 7x7x7 mesh ? SiC is an insulator and together with the
> 2x2x2 supercell, at least for a first structural optimization you can do
> it with eg. 4 k-points in the IBZ (maybe even 1). Of course, one should always
> check his final results by continuing with a finer k-mesh (and a larger RKmax).
> But most likely, the small forces will stay pretty small.
>
> When you change Ni vs. C manually, you should also change R0 !
>
> PS: Once the Ni-Si distances have increased, one should test if spin-polarization
> (maybe even with LDA+U) results in a magnetic moment on Ni !
>
> Am 24.08.2014 01:44, schrieb Hu, Wenhao:
>> Dear professor Laurence Marks:
>>
>> The structure file of my calculation is:
>> Title
>> P 10
>> RELA
>> 16.476906 16.476906 16.476906 90.000000 90.000000 90.000000
>> ATOM 1: X=0.00000000 Y=0.00000000 Z=0.00000000
>> MULT= 1 ISPLIT= 2
>> C 1 NPT= 781 R0=0.00010000 RMT= 1.86 Z: 6.0
>> LOCAL ROT MATRIX: 1.0000000 0.0000000 0.0000000
>> 0.0000000 1.0000000 0.0000000
>> 0.0000000 0.0000000 1.0000000
>> ATOM -2: X=0.50000000 Y=0.00000000 Z=0.00000000
>> MULT= 3 ISPLIT=-2
>> -2: X=0.00000000 Y=0.50000000 Z=0.00000000
>> -2: X=0.00000000 Y=0.00000000 Z=0.50000000
>> C 2 NPT= 781 R0=0.00010000 RMT= 1.86 Z: 6.0
>> LOCAL ROT MATRIX: 0.0000000 0.0000000 1.0000000
>> 1.0000000 0.0000000 0.0000000
>> 0.0000000 1.0000000 0.0000000
>> ATOM -3: X=0.50000000 Y=0.50000000 Z=0.00000000
>> MULT= 3 ISPLIT=-2
>> -3: X=0.50000000 Y=0.00000000 Z=0.50000000
>> -3: X=0.00000000 Y=0.50000000 Z=0.50000000
>> C 3 NPT= 781 R0=0.00010000 RMT= 1.86 Z: 6.0
>> LOCAL ROT MATRIX: 1.0000000 0.0000000 0.0000000
>> 0.0000000 1.0000000 0.0000000
>> 0.0000000 0.0000000 1.0000000
>> ATOM 4: X=0.50000000 Y=0.50000000 Z=0.50000000
>> MULT= 1 ISPLIT= 2
>> Ni4 NPT= 781 R0=0.00010000 RMT= 2.08 Z: 28.0
>> LOCAL ROT MATRIX: 1.0000000 0.0000000 0.0000000
>> 0.0000000 1.0000000 0.0000000
>> 0.0000000 0.0000000 1.0000000
>> ATOM -5: X=0.25000000 Y=0.25000000 Z=0.00000000
>> MULT=12 ISPLIT= 8
>> -5: X=0.75000000 Y=0.25000000 Z=0.00000000
>> -5: X=0.25000000 Y=0.75000000 Z=0.00000000
>> -5: X=0.75000000 Y=0.75000000 Z=0.00000000
>> -5: X=0.25000000 Y=0.00000000 Z=0.25000000
>> -5: X=0.75000000 Y=0.00000000 Z=0.25000000
>> -5: X=0.25000000 Y=0.00000000 Z=0.75000000
>> -5: X=0.75000000 Y=0.00000000 Z=0.75000000
>> -5: X=0.00000000 Y=0.25000000 Z=0.25000000
>> -5: X=0.00000000 Y=0.75000000 Z=0.25000000
>> -5: X=0.00000000 Y=0.25000000 Z=0.75000000
>> -5: X=0.00000000 Y=0.75000000 Z=0.75000000
>> C 5 NPT= 781 R0=0.00010000 RMT= 1.86 Z: 6.0
>> LOCAL ROT MATRIX: 0.0000000-0.7071068-0.7071068
>> 0.0000000-0.7071068 0.7071068
>> -1.0000000 0.0000000 0.0000000
>> ATOM -6: X=0.25000000 Y=0.25000000 Z=0.50000000
>> MULT=12 ISPLIT= 8
>> -6: X=0.75000000 Y=0.25000000 Z=0.50000000
>> -6: X=0.25000000 Y=0.75000000 Z=0.50000000
>> -6: X=0.75000000 Y=0.75000000 Z=0.50000000
>> -6: X=0.25000000 Y=0.50000000 Z=0.25000000
>> -6: X=0.75000000 Y=0.50000000 Z=0.25000000
>> -6: X=0.25000000 Y=0.50000000 Z=0.75000000
>> -6: X=0.75000000 Y=0.50000000 Z=0.75000000
>> -6: X=0.50000000 Y=0.25000000 Z=0.25000000
>> -6: X=0.50000000 Y=0.75000000 Z=0.25000000
>> -6: X=0.50000000 Y=0.25000000 Z=0.75000000
>> -6: X=0.50000000 Y=0.75000000 Z=0.75000000
>> C 6 NPT= 781 R0=0.00010000 RMT= 1.86 Z: 6.0
>> LOCAL ROT MATRIX: 0.0000000-0.7071068-0.7071068
>> 0.0000000-0.7071068 0.7071068
>> -1.0000000 0.0000000 0.0000000
>> ATOM -7: X=0.12500000 Y=0.12500000 Z=0.12500000
>> MULT= 4 ISPLIT= 4
>> -7: X=0.87500000 Y=0.87500000 Z=0.12500000
>> -7: X=0.87500000 Y=0.12500000 Z=0.87500000
>> -7: X=0.12500000 Y=0.87500000 Z=0.87500000
>> Si7 NPT= 781 R0=0.00010000 RMT= 1.46 Z: 14.0
>> LOCAL ROT MATRIX: 0.4082483-0.7071068 0.5773503
>> 0.4082483 0.7071068 0.5773503
>> -0.8164966 0.0000000 0.5773503
>> ATOM -8: X=0.62500000 Y=0.12500000 Z=0.12500000
>> MULT=12 ISPLIT= 8
>> -8: X=0.12500000 Y=0.62500000 Z=0.12500000
>> -8: X=0.12500000 Y=0.12500000 Z=0.62500000
>> -8: X=0.87500000 Y=0.37500000 Z=0.12500000
>> -8: X=0.37500000 Y=0.87500000 Z=0.12500000
>> -8: X=0.87500000 Y=0.87500000 Z=0.62500000
>> -8: X=0.87500000 Y=0.12500000 Z=0.37500000
>> -8: X=0.37500000 Y=0.12500000 Z=0.87500000
>> -8: X=0.87500000 Y=0.62500000 Z=0.87500000
>> -8: X=0.12500000 Y=0.87500000 Z=0.37500000
>> -8: X=0.12500000 Y=0.37500000 Z=0.87500000
>> -8: X=0.62500000 Y=0.87500000 Z=0.87500000
>> Si8 NPT= 781 R0=0.00010000 RMT= 1.46 Z: 14.0
>> LOCAL ROT MATRIX: -1.0000000 0.0000000 0.0000000
>> 0.0000000-0.7071068-0.7071068
>> 0.0000000-0.7071068 0.7071068
>> ATOM -9: X=0.62500000 Y=0.62500000 Z=0.12500000
>> MULT=12 ISPLIT= 8
>> -9: X=0.62500000 Y=0.12500000 Z=0.62500000
>> -9: X=0.12500000 Y=0.62500000 Z=0.62500000
>> -9: X=0.37500000 Y=0.37500000 Z=0.12500000
>> -9: X=0.87500000 Y=0.37500000 Z=0.62500000
>> -9: X=0.37500000 Y=0.87500000 Z=0.62500000
>> -9: X=0.37500000 Y=0.12500000 Z=0.37500000
>> -9: X=0.87500000 Y=0.62500000 Z=0.37500000
>> -9: X=0.37500000 Y=0.62500000 Z=0.87500000
>> -9: X=0.12500000 Y=0.37500000 Z=0.37500000
>> -9: X=0.62500000 Y=0.87500000 Z=0.37500000
>> -9: X=0.62500000 Y=0.37500000 Z=0.87500000
>> Si9 NPT= 781 R0=0.00010000 RMT= 1.46 Z: 14.0
>> LOCAL ROT MATRIX: 0.0000000-0.7071068-0.7071068
>> 0.0000000-0.7071068 0.7071068
>> -1.0000000 0.0000000 0.0000000
>> ATOM -10: X=0.62500000 Y=0.62500000 Z=0.62500000
>> MULT= 4 ISPLIT= 4
>> -10: X=0.37500000 Y=0.37500000 Z=0.62500000
>> -10: X=0.37500000 Y=0.62500000 Z=0.37500000
>> -10: X=0.62500000 Y=0.37500000 Z=0.37500000
>> Si10 NPT= 781 R0=0.00010000 RMT= 1.46 Z: 14.0
>> LOCAL ROT MATRIX: 0.4082483-0.7071068 0.5773503
>> 0.4082483 0.7071068 0.5773503
>> -0.8164966 0.0000000 0.5773503
>> This is a nickel doped SiC. The central carbon atom is substituted with a nickel atom. Besides, I removed a NN silicon of the nickel from the lattice. A 2x2x2 supercell was implemented. And a 7x7x7 k mesh was applied. I appreciate any suggestions from you.
>>
>> Best,
>> Wenhao
>>
>>
>> Date: Sun, 27 Jul 2014 00:16:39 -0500
>> From: Laurence Marks <L-marks at northwestern.edu>
>> To: A Mailing list for WIEN2k users <wien at zeus.theochem.tuwien.ac.at>
>> Subject: Re: [Wien] Unable to achieve required force convergence in
>> structure optimization
>> Message-ID:
>> <CANkSMZCmNj+1a4mYKiLf9qaFw=PWUGfZym4RULt5UkFE5tGxeg at mail.gmail.com>
>> Content-Type: text/plain; charset="utf-8"
>>
>> There is a file called README in SRC_mixer, and a copy of this is also in
>> the notes section of the WIEN2k web page.
>>
>> Beyond this nobody can help you as you have not provided any details, e.g.
>> case.structure, k-points etc.
>>
>> ___________________________
>> Professor Laurence Marks
>> Department of Materials Science and Engineering
>> Northwestern University
>> www.numis.northwestern.edu1-847-491-3996
>> Co-Editor, Acta Cryst A
>> "Research is to see what everybody else has seen, and to think what nobody
>> else has thought"
>> Albert Szent-Gyorgi
>> On Jul 26, 2014 7:18 PM, "Hu, Wenhao" <wenhao-hu at uiowa.edu> wrote:
>>
>>> Dear all:
>>>
>>> I met a problem when I use MSR1a method to determine equilibrium structure
>>> of my lattice. The convergence I want is 0.5mRy/au. But when I check the
>>> scfm file, the force on some specific atoms can be as large as 2 mRy/au.
>>> This problem remained even after I repeatedly ran my calculation for
>>> several times and lower the convergence from 0.5 to 0.2. Can anyone give me
>>> any suggestions?
>>>
>>> Thanks in advance.
>>> Wenhao
>>> _______________________________________________
>>> 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
>>>
>> -------------- next part --------------
>> An HTML attachment was scrubbed...
>> URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20140727/63578221/attachment-0001.htm>
>>
>
> --
> -----------------------------------------
> Peter Blaha
> Inst. Materials Chemistry, TU Vienna
> Getreidemarkt 9, A-1060 Vienna, Austria
> Tel: +43-1-5880115671
> Fax: +43-1-5880115698
> email: pblaha at theochem.tuwien.ac.at
> -----------------------------------------
>
--
-----------------------------------------
Peter Blaha
Inst. Materials Chemistry, TU Vienna
Getreidemarkt 9, A-1060 Vienna, Austria
Tel: +43-1-5880115671
Fax: +43-1-5880115698
email: pblaha at theochem.tuwien.ac.at
-----------------------------------------
More information about the Wien
mailing list