[Wien] MSR1a: mixer.error ('MIXER' - No feasible step small enough, check RMT and model)

[Laurence Marks]

This is tricky. It is dangerous to try it during a run_lapw, and the
results will be unpredictable.

If you are running on dedicated computers and not using pbs or some
other queue control then I suggest that you do "touch .stop" to
terminate them, then reduce the RMTs.

If you are running under pbs or similar then it is possible (maybe) to
amend the script being used and add to the end something like

save -f -d Safe_Hold
setrmt case -r 10
cp *set* case.struct_new
x clminter
mv case.clmsum_new case.clmsum
run_lapw ....

And then do a "touch .stop". However, this may not work!

N.B., remember that if you are comparing different options for a
structure you have to use the same RMT, RKMAX etc. In that case the -a
option to setrmt is useful.


On Tue, Sep 17, 2013 at 9:09 AM, Salman Zarrini
<salman.zarrini at tu-darmstadt.de> wrote:
>
> Dear Professor Marks,
>
> Thank you for your informative answer, I followed your suggestion and
> I did "setrmt_lapw" and "x clminter" for jobs stopped by this error. I
> have some other jobs which still running with out this error but with
> warning "step was reduced due to touching spheres" means the same
> story is going to happen for them as well, I was wondering if I can
> "setrmt_lapw" and after that "x clminter" for the jobs while thay are
> still running but with warning, or, I have to be waiting till they
> stop by the error and then treat them?
>
> Sincerely yours,
>
> Salman Zarrini
> AK Professor Berger
> TU Darmstadt, Germany
> Clemens-Schöpf-Institut
> Petersenstr. 22
> 64287 Darmstadt
>
> Quoting "Laurence Marks" <L-marks at northwestern.edu>:
>
>> You need to continue to make the RMTs smaller (using clminter, don't
>> redo dstart), perhaps 10% smaller than when they would touch. Later
>> when the positions are optimized you can increase the RMT.
>>
>> To explain, the "No feasible step small enough" occurs when MSR1a
>> believes that to reduce the forces it needs to move the atoms to
>> positions where the spheres would overlap. In general you cannot use
>> MSR1a with the default nearly touching spheres that setrmt creates,
>> you have to go smaller. How small will depend upon the system. With a
>> slab calculation, if you are using a simple 1x1 cell then in general
>> the distances between atoms at the surface will be substantially
>> smaller than those in the bulk.
>>
>> The message also says "and model". If you are using an unrealistic
>> model then you may well get distances between atoms which are very
>> small. GIGO -- Garbage In, Garbage Out. While a simple 1x1 slab model
>> may be reasonable for many metals, for most oxides (for instance) and
>> many semiconductors it is unreasonable.
>>
>> Note: MSR1a may well move the atoms a long way, e.g. 0.5 au or even
>> much more than this (2.0 au), if the initial positions are far from
>> the equilibrium positions, since as it runs some atoms can come quite
>> close to each other although they will end up with reasonable spacings
>> if you have a viable physical model.
>>
>>
>> On Mon, Sep 16, 2013 at 10:08 AM, Salman Zarrini
>> <salman.zarrini at tu-darmstadt.de> wrote:
>>>
>>> Dear Professor Marks,
>>>
>>> In minimizing the internal positions of an slab in Wien2k, all the
>>> steps were followed correctly
>>> and "MSR1a" was used as mixing factor in case.inm with a normal
>>> scf-scripts run_lapw.
>>> However, mixing.error ('MIXER' - No feasible step small enough, check
>>> RMT and model)
>>> showed up after 23 cycles. The RMTes were adjusted  to %4 smaller,
>>> however, the same
>>> story was happened again.
>>> So, I would be really appreciated if you could provide me a piece of
>>> recommendation to
>>> overcome the problem.
>>> By the way, Wien2k_12 is hired in the all calculations as well.
>>>
>>> Sincerely yours,
>>>
>>> Salman Zarrini
>>> AK Professor Berger
>>> TU Darmstadt, Germany
>>> Clemens-Schöpf-Institut
>>> Petersenstr. 22
>>> 64287 Darmstadt
>>
>>
>>
>> --
>> Professor Laurence Marks
>> Department of Materials Science and Engineering
>> Northwestern University
>> www.numis.northwestern.edu 1-847-491-3996
>> "Research is to see what everybody else has seen, and to think what
>> nobody else has thought"
>> Albert Szent-Gyorgi
>>
>>
>
>



-- 
Professor Laurence Marks
Department of Materials Science and Engineering
Northwestern University
www.numis.northwestern.edu 1-847-491-3996
"Research is to see what everybody else has seen, and to think what
nobody else has thought"
Albert Szent-Gyorgi