[Wien] 'LOPW' - Plane waves exhausted

Laurence Marks L-marks at northwestern.edu
Thu May 3 16:55:56 CEST 2012 

Doing "make all" and "cp lapw1* ../" updates the executables; they are
built in SRC_lapw1 but need to be in $WIENROOT. If you did not copy
them then you are still using the old version which does fail for your
problem.

It may be that RKMAX=7.5 would be better as your RMT's are about 10%
larger than 2.0. It will help with the LOPW issue, although of course
it is slower and you are only using a single machine so would have to
be patient.

My suggestion is to check that you do have the most recent lapw1
properly installed, make a fresh directory and redo the initialization
then test. It should work fine...

Good luck.

On Thu, May 3, 2012 at 9:40 AM, Florian Meirer
<fmeirer.wien2k.mlist at gmail.com> wrote:
> I understand - many thanks for following up on my problem!
> Here is the info:
> a) NMATMAX=13000 (seemed sufficient so far, I have only 64 atoms;
> anyway I have 24GB RAM available)
> b) I did 'make -all' in SRC_lapw1/, but not 'cp lapw1* ../' (but this
> was done after the error occurred using the 'old' lopw.f file)
> c) I always used RKmax=7 (:RKM  : MATRIX SIZE 5348LOs: 256  RKM= 7.00
> WEIGHT= 8.00  PGR:)
> d) no mpi - I just use k-point parallelization (.machines file)
> Thanks!
>
> 2012/5/3 Laurence Marks <L-marks at northwestern.edu>:
>> N.B., to see what value it is using do "grep -e :RKM *.scf"
>>
>> On Thu, May 3, 2012 at 8:37 AM, Laurence Marks <l-marks at northwestern.edu> wrote:
>>> This is not quite a solution, it instead verifies that there are
>>> perhaps no other errors in your files.
>>>
>>> I downloaded your file this morning and tested it myself, and I do not
>>> reproduce the problem. I have an idea what the issue is, but I am not
>>> certain so some questions:
>>>
>>> a) What is the value of NMATMAX in SRC_lapw1/param.inc ? I have this
>>> set at 13000. It looks like you have plenty of memory so you may want
>>> to raise it.
>>> b) When you added the lopw.f file, did you use siteconfig to update or
>>> just do "make". If the later, did you do "make all" then "cp lapw1*
>>> ../" ?
>>> c) In the case that is working, what RKMAX is the program using (as it
>>> can reduce it when there is not enough space)
>>> d) Are you running mpi (I suspect not)?
>>>
>>> On Thu, May 3, 2012 at 5:31 AM, Florian Meirer
>>> <fmeirer.wien2k.mlist at gmail.com> wrote:
>>>> [solved] 'LOPW' - Plane waves exhausted
>>>>
>>>> Many thanks for the help!
>>>> It seems I was able to solve the problem by removing the LOs AND
>>>> setting the d-levels to LAPW for all atoms:
>>>> ----------------------------------
>>>> WFFIL  EF= 0.50000       (WFFIL, WFPRI, ENFIL, SUPWF)
>>>> 7.00       10    4 (R-MT*K-MAX; MAX L IN WF, V-NMT
>>>> 0.30    3  0      (GLOBAL E-PARAMETER WITH n OTHER CHOICES, global APW/LAPW)
>>>> 2   -1.83      0.002 CONT 0
>>>> 0    0.30      0.000 CONT 1
>>>> 1    0.30      0.000 CONT 1
>>>> ----------------------------------
>>>> I am not sure if I fully understand the reasoning, however, it seems to work...
>>>>
>>>>
>>>> 2012/5/3 Laurence Marks <L-marks at northwestern.edu>:
>>>>> Apologies for the slow response.  Try removing the first d LAPW, having two
>>>>> for the same level can lead to problems (i think). I may be wrong, check
>>>>> what is in case.output1_* (it is too late for me).
>>>>>
>>>>> ---------------------------
>>>>>
>>>>>
>>>>> 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
>>>>>
>>>>> On May 2, 2012 11:29 AM, "Florian Meirer" <fmeirer.wien2k.mlist at gmail.com>
>>>>> wrote:
>>>>>>
>>>>>> Thanks for the fast reply!
>>>>>> As suggested I tested using LAPW instead of APW+lo first (changed all
>>>>>> switches to 0 in case.in1_st) but it didn't work.
>>>>>> I have LO terms for d-levels and tested changing to LAPW only for them
>>>>>> (see below) - no success either...
>>>>>> here is the content of my .in1_st file (lines 4-7 are the same for all
>>>>>> 10 inequivalent atoms - all Ge, one Sn):
>>>>>> ----------------------------------
>>>>>> WFFIL  EF= 0.50000       (WFFIL, WFPRI, ENFIL, SUPWF)
>>>>>>  7.00       10    4 (R-MT*K-MAX; MAX L IN WF, V-NMT
>>>>>>  0.30    4  0      (GLOBAL E-PARAMETER WITH n OTHER CHOICES, global
>>>>>> APW/LAPW)
>>>>>>  2   -1.83      0.002 CONT 0
>>>>>>  2    0.30      0.000 CONT 0
>>>>>>  0    0.30      0.000 CONT 1
>>>>>>  1    0.30      0.000 CONT 1
>>>>>> ...
>>>>>> ...
>>>>>> K-VECTORS FROM UNIT:4   -9.0       2.5  898   emin/emax/nband #red
>>>>>> ----------------------------------
>>>>>> What could be wrong with the struct file? Is there something I do
>>>>>> fundamentally wrong ... should I try to reduce symmetry?
>>>>>> Many thanks!
>>>>>>
>>>>>> 2012/5/2 Laurence Marks <L-marks at northwestern.edu>:
>>>>>> > It could be that there is some error in your file. One test might be
>>>>>> > to use LAPW instead of APW+lo and see if that runs (a single
>>>>>> > iteration). If it does then your structure is OK. Please check the UG
>>>>>> > if you are unclear what to change in case.in1
>>>>>> >
>>>>>> > Some comments that might be helpful:
>>>>>> > I believe that changing the number of k-points only helps by chance,
>>>>>> > i.e. avoiding a specific k-point that leads to problems, so that is
>>>>>> > not the issue.
>>>>>> > Similarly changing the number of parallel jobs should not matter,
>>>>>> > I would not reduce it to P, but you might be able to reduce the
>>>>>> > symmetry to something slightly lower than what you started with (e.g.
>>>>>> > P4mm + inversion or simple cubic). I use Cryscon for this, but I am
>>>>>> > sure that there are other ways.
>>>>>> >
>>>>>> > Last, maybe not least, the issue comes about for high-symmetry
>>>>>> > structures because the APW terms within the muffin-tins need to be
>>>>>> > orthogonal. With many identical atoms for higher L levels this can
>>>>>> > become complicated as many of the PW's are not unique (e.g. (001) and
>>>>>> > (002)). Hence I would check whether you have an LO terms for d-levels
>>>>>> > in your calculation by looking at case.in1. I don't think you should
>>>>>> > have, but maybe. You can also look in case.output1_* to see where they
>>>>>> > stop. It may be that you can change to LAPW for the d in case.in1 (I
>>>>>> > expect by default it is APW+lo).
>>>>>> >
>>>>>> > On Wed, May 2, 2012 at 9:59 AM, Florian Meirer
>>>>>> > <fmeirer.wien2k.mlist at gmail.com> wrote:
>>>>>> >> Dear Wien2k users,
>>>>>> >>
>>>>>> >> I am having a problem with a supercell calculation:
>>>>>> >>
>>>>>> >> - I am running wien version 11.1 (Release 14/6/2011) on a Intel XEON
>>>>>> >> X5650 @ 2.67Ghz (2 Processors), 24Gb RAM with operating system Ubuntu
>>>>>> >> 64bit, fortran compiler ifort 11.1.073 and math libraries R_LIB
>>>>>> >> (LAPACK+BLAS): -lmkl_lapack -lmkl_intel_lp64 -lmkl_intel_thread
>>>>>> >> -lmkl_core -openmp -lpthread -lguide
>>>>>> >>
>>>>>> >> - The purpose of my calculations is to perform a structural relaxation
>>>>>> >> (PORT) around an impurity (Sn dopant in Germanium)
>>>>>> >> - I successfully performed the same calculation for an As impurity in
>>>>>> >> Silicon and achieved good agreement with literature and experimental
>>>>>> >> data (XANES).
>>>>>> >>
>>>>>> >> - I started with the struct:
>>>>>> >> ----------------------
>>>>>> >> Germanium
>>>>>> >> F   LATTICE,NONEQUIV.ATOMS:  1 227 Fd-3m
>>>>>> >> MODE OF CALC=RELA unit=ang
>>>>>> >>  10.691735 10.691735 10.691735 90.000000 90.000000 90.000000
>>>>>> >> ATOM   1: X=0.12500000 Y=0.12500000 Z=0.12500000
>>>>>> >>           MULT= 2          ISPLIT= 2
>>>>>> >>        1: X=0.87500000 Y=0.37500000 Z=0.37500000
>>>>>> >> Ge1        NPT=  781  R0=0.00005000 RMT=    2.2300   Z: 32.0
>>>>>> >> ----------------------
>>>>>> >> then created a supercell (2x2x2) and replaced one Ge by Sn (final
>>>>>> >> .struct file attached).
>>>>>> >> The parameters I used for testing:
>>>>>> >> R0's were adjusted to 0.00005 for Ge and to 0.00001 for Sn (as
>>>>>> >> recommended: http://www.wien2k.at/reg_user/faq/r0.html);
>>>>>> >> Default settings for the rest:
>>>>>> >> RKmax=7, k=200, PBE-GGA, -6.0Ry, not spin-polarized
>>>>>> >> SCF: run_lapw -p -it -ec 0.0001 -fc 1 -cc 0.001 -NI
>>>>>> >> I am running 10 parallel jobs and lapw1_10.error returns:
>>>>>> >> Error in LAPW1
>>>>>> >>  'LOPW' - Plane waves exhausted
>>>>>> >>
>>>>>> >> - I have already consulted the mailing list and tried to following:
>>>>>> >> 1) replaced SRC_lapw1/lopw.f with the one provided here:
>>>>>> >>
>>>>>> >> http://zeus.theochem.tuwien.ac.at/pipermail/wien/2011-August/015142.html
>>>>>> >> and recompiled in folder.
>>>>>> >> 2) increased RKmax (also suggested in the mailing list somewhere)
>>>>>> >> 3) changed number of parallel jobs
>>>>>> >> 4) broke symmetry of struct file completely; i.e. each atom labeled
>>>>>> >> individually, P-type struct -> LOPW does not crash but a) the
>>>>>> >> calculation takes forever and b) the forces did not converge
>>>>>> >> 5) finally, as it seems related to the k-points, I reduced the number
>>>>>> >> of k-points to 4 (k=100) and LOPW does not crash ... however, I am not
>>>>>> >> happy using only 4 k-points... this seems too low?
>>>>>> >> 6) I read that: "This is usually due to an error in your struct file.
>>>>>> >> (Specifying the same atom twice,....)"
>>>>>> >>
>>>>>> >> (http://www.wien2k.at/reg_user/mailing_list/wien-digest.archive/summary_2001),
>>>>>> >> but I cannot find my mistake...
>>>>>> >>
>>>>>> >> - The thing that puzzles me is the fact that the same calculation
>>>>>> >> worked perfectly fine for an As impurity in Si (same structure, same
>>>>>> >> parameters ...)
>>>>>> >> I am probably missing something but cannot find my mistake.
>>>>>> >>
>>>>>> >> Many thanks for any suggestions!
>>>>>> >> Best regards,
>>>>>> >> Florian Meirer
>>>>>> >>
>>>>>> >> ------------------------------------------------------------
>>>>>> >> Florian Meirer (PhD)
>>>>>> >> MiNALab - Center for Materials and Microsystems - irst
>>>>>> >> FBK - Fondazione Bruno Kessler
>>>>>> >> ------------------------------------------------------------
>>>>>> >> _______________________________________________
>>>>>> >> Wien mailing list
>>>>>> >> Wien at zeus.theochem.tuwien.ac.at
>>>>>> >> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>>>>> >
>>>>>> >
>>>>>> >
>>>>>> > --
>>>>>> > 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
>>>>>> > _______________________________________________
>>>>>> > Wien mailing list
>>>>>> > Wien at zeus.theochem.tuwien.ac.at
>>>>>> > http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>>>>> _______________________________________________
>>>>>> Wien mailing list
>>>>>> Wien at zeus.theochem.tuwien.ac.at
>>>>>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> Wien mailing list
>>>>> Wien at zeus.theochem.tuwien.ac.at
>>>>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>>>>
>>>> _______________________________________________
>>>> Wien mailing list
>>>> Wien at zeus.theochem.tuwien.ac.at
>>>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>>
>>>
>>>
>>> --
>>> 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
>>> _______________________________________________
>>> Wien mailing list
>>> Wien at zeus.theochem.tuwien.ac.at
>>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>
>>
>>
>> --
>> 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
>> _______________________________________________
>> Wien mailing list
>> Wien at zeus.theochem.tuwien.ac.at
>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien



-- 
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

More information about the Wien mailing list