[Wien] WARNING : Please restart anew from final position

Fri Jan 21 18:22:42 CET 2011

One general comment: be aware that what you are seeing in both cases
is (deliberately) a "WARNING", not an "ERROR". You are correct that
you need to be careful if you see the step size being reduced and make
sure that you do not have touching or too close to touching spheres.
In the worst case you will see many of these and the optimization will
keep trying smaller and smaller steps and not be able to get anywhere.

You also have to be careful if you see the curvature warning, but only
if this keeps on coming up.  This is one of those subtle things which
is not described in elementary books on optimization, only in the more
advanced ones. In 20% (maybe more) of optimizations the curvature
condition will fail once or twice. This is not a problem, the code is
smart enough to know what to do. (As the tutorial implies, this often
happens if the Hessian is changing, e.g. new "bonds" are being
created.) If it keeps on occurring then (and only then) should you
start to worry that the forces and energy could be inconsistent.

What you say is reasonable, and probably reducing the RMT of the N a
little (1.5 ?) would avoid the problem (use clminter, it is designed
for this). Keep the partially relaxed positions. It is probably safest
to remove .min_hess and re-run pairhess. (Only do this in case of
problems, in general a good .min_hess will save a lot of time.)

N.B., RKMAX comes in as RKMAX/min(RMT), so a value of 5 with your
smallest RMT 1.3 should be more than good enough and I would not go to
an RKMAX of 7 (too slow, overkill).

On Fri, Jan 21, 2011 at 10:07 AM, Kakhaber Jandieri
<kakhaber.jandieri at physik.uni-marburg.de> wrote:
> Dear Laurence,
>
> Thank you very much.
>
> Unfortunately, after some iteration step  I missed the warnings :
> WARNING: Step size reduced due to overlapping spheres
> and
> WARNING: CURVATURE CONDITION FAILED
>
> According to your tutorial, the last one means that the forces and energy
> are not consistent and one of the possible reasons can be the small RKMAX. I
> guess that in my case the reason is different since I see the same warning
> for increased value (7 instead of 5) of RKMAX. May be the reason is given in
> the first warning about the overlapping spheres. Below I will try to justify
> my suggestion.
>
> I am trying to relax the GaAs slab containing two substitutional nitrogen
> atoms in surface  layer. I constructed the supercell containing 72
> non-equivalent atoms: 5 layers with 12 atoms in each, appr. 10 Angstr.
> Vacuum and the passivation  layer of hydrogen atoms. I fixed only the
> hydrogen layer and the first GaAs layer. I remember your letter where you
> had some doubts about such consideration, but I want to compare the
> simulated Scanning Tunneling Microscopy images with that obtained in Phys.
> Rev. B75, 035338 (2008), where the similar approach was used (fixed layer,
> pasivation by hydrogen). So on current stage I decided to stay in frames of
> this approach.
>
> I have the following values of RMT:
> RMT(Ga)=2, RMT(N)=1.68, RMT(H)=1.3
>
> I think that RMT(N)=1.68 is the reason for my problems. The supercell was
> relaxed correctly up to some iteration step. Then the Ga and N spheres began
> to touch each other and the problems appeared. I have the following ground
> for such suggestion. The Ga-N distance in my “partially relaxed” supercell
> is 1.95 Angstr. and this value is very close to the sum of Ga and N sphere
> radii.
>
> I think that my supercell is not very far from the  full relaxation, since
> the distances between Ga and N atoms as well as the distance between surface
> Ga atoms (next neighbors to N atom) and their inward shifts are very close
> to the values given in Phys.Rev. paper.
>
> In spite of that, I still have the great desire to obtain the fully relaxed
> supercell and accordingly as precise STM image as possible.
>
> Could you give me an advice:
> 1) are my suggestions reasonable?
> 2) Will the reduction of RMT(N) solve the problem?
> 3) Is it reasonable to renew the calculations with “partially relaxed”
> supercell as a starting structure (but of course with reduced value of
> RTM(N))?
>
> Excuse me for so long letter.
> Beforehand thankful,
>
> Kakhaber Jandieri
>
>
>
> Quoting Laurence Marks <L-marks at northwestern.edu>:
>
>> Do grep -e :MIN *.ouptputM -- you need to keep going.
>>
>> On Thu, Jan 20, 2011 at 12:30 PM, Kakhaber Jandieri
>> <Kakhaber.Jandieri at physik.uni-marburg.de> wrote:
>>>
>>> Dear Wien2k users,
>>>
>>> Using "mini" I see the following message:
>>>
>>>>  stop
>>>>  mini        (00:43:49) 0.880u 0.060s 0:01.25 75.2%  0+0k 0+4024io
>>>> 0pf+0w
>>>
>>>  :WARNING Please restart anew from final position
>>>
>>>>  stop forces small, minimization stopped
>>>
>>>>  relaxation steps summarized in Surface-GaAs-2N-b.scf_mini
>>>
>>>>  relaxed structure and results in Surface-GaAs-2N-b.struct, scf, clmsum
>>>
>>>>  please use   save_lapw save_file_name   to save the results
>>>
>>>
>>> Could you explain what is the meaning of
>>> "WARNING Please restart anew from final position"?
>>>
>>> I found a similar question in mailing list (Ghosh SUDDHASATTWA, /Sat Jul
>>> 24 05:54:54 CEST 2010) but unfortunately without any replied suggestions.
>>>
>>> Beforehand thankful,
>>>
>>> Kakhaber Jandieri
>>> /
>>>
>>>
>>> _______________________________________________
>>> 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 and imaging to study the structure of matter.
>> _______________________________________________
>> Wien mailing list
>> Wien at zeus.theochem.tuwien.ac.at
>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>
>>
>
>
>
> Dr. Kakhaber Jandieri
> Department of Physics
> Philipps University Marburg
> Tel:+49 6421 2824159 (2825704)
>
>
> _______________________________________________
> 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 and imaging to study the structure of matter.