[Wien] mBJ run stops w/o error message in lapw0

[tran at theochem.tuwien.ac.at]

Hello,

In case you don't know, the mBJ potential can not be applied as such
to systems with infinite vacuum (e.g., isolated molecule, surfaces),
because the average of |grad(rho)|/rho (and therefore the parameter c
in mBJ) is infinite in this case. You have to fix the parameter c
(to the value of bulk ZnO for instance) to avoid this problem
(you can do it by creating manually case.in0abp, see the UG).

For your problem, lapw0 is probably hanging in the brj.f subroutine
because of Newton method. In the next WIEN2k release, brj.f will be
replaced by a new one (attached to this email) without such problem.

So, replace brj.f in SRC_lapw0 and recompile to have new executables
lapw0 and lapw0_mpi. Also, fix the parameter c.

F. Tran


On Tue, 29 Apr 2014, Kamil Klier wrote:

Dear Colleagues,

I am seeking your advice for solving the following problem exemplified
on a model of a Cu-ZnO slab (40 ZnO with one Cu(0) on the slab surface):

After a successful completion of parallel sp -so -orb, the subsequent
mBJ run stops in the lapw0 step within (after completing lapw0_grr
with *.output0_grr completed with no error) which is "frozen" at the
stage listed below between asterisks:

**********************************************************************************************************
   lapw0 -grr -p       (12:24:46) starting parallel lapw0 at Tue Apr 29
12:24:46 EDT 2014
-------- .machine0 : processors
running lapw0 in single mode
404.681u 60.392s 7:45.39 99.9%  0+0k 0+0io 0pf+0w
>   lapw0 -p    (12:32:31) starting parallel lapw0 at Tue Apr 29
> 12:32:32 EDT 2014
-------- .machine0 : processors
running lapw0 in single mode
   int:rho,tauw,grho,g2rho  2.824350877551655E-002  6.103772359501373E-004
    8.304021873750874E-003 -5.082098170048767E-002 tauwrong=
   -4.495911213141860E-003
   int:rho,tauw,grho,g2rho  2.825356952954721E-002  6.095285785363750E-004
    8.299724832528738E-003 -5.083964511044604E-002 tauwrong=
   -4.491001032794553E-003
   int:rho,tauw,grho,g2rho  2.706250413726011E-002  1.941298184091310E-003
    1.449639819246514E-002 -4.116294968078159E-002 tauwrong=
   -5.376820894514224E-003
   int:rho,tauw,grho,g2rho  2.707222807780667E-002  1.940275366823992E-003
    1.449518227059034E-002 -4.118138094334088E-002 tauwrong=
   -5.374249039619829E-003
   int:rho,tauw,grho,g2rho  2.509516009055439E-002  3.708033157499589E-003
    1.929286766746054E-002 -2.311418448605970E-002 tauwrong=
   -9.069682710427986E-004
   int:rho,tauw,grho,g2rho  2.510432669330462E-002  3.706885067692818E-003
    1.929340340674948E-002 -2.313235104252417E-002 tauwrong=
   -9.063351511093207E-004
   int:rho,tauw,grho,g2rho  8.029845099838655E-004  5.176735132921631E-004
    1.289470918481737E-003  1.548234958760512E-003 tauwrong=
    4.963361214083710E-004
   int:rho,tauw,grho,g2rho  8.026535824516776E-004  5.174160846724353E-004
    1.288884593717297E-003  1.547070346152785E-003 tauwrong=
    4.958243436286073E-004
   int:rho,tauw,grho,g2rho  6.726180563432650E-004  4.323545213751433E-004
    1.078535039428152E-003  1.083803658602449E-003 tauwrong=
    3.060876913841199E-004
   int:rho,tauw,grho,g2rho  6.723607160887903E-004  4.321348107839394E-004
    1.078054675469817E-003  1.082698790040087E-003 tauwrong=
    3.055804491221129E-004
055804491221129E-004
************************************************************************************************************

There is no error message - the job just hangs in this state, sometime
in the first mBJ scf cycle, but other times in the n-th mBJ scf cycle
(e.g. n = 7, or even 21).

I have tried to alleviate possible linearization error by adding 1.0
Ry LOs for Cu3d, but with no success.

I am using Wien2k-13.1, running the job at the Brookhaven National Lab
CFN cluster, where we installed and successfully ran many Wien2k jobs,
and can supply additional files for your analysis as needed.  Previous
successful completion of this job prior to mBJ indicates no problem
within the CFN cluster and with our installation parameters. Our use
of -orb is based on experience with pure ZnO, but the present problem
persists with or without running LDA+U.

Any hint/advice/experience would be highly appreciated.

Regards,

Kamil Klier
Lehigh University
e-mail kk04 at lehigh.edu
-------------- next part --------------
A non-text attachment was scrubbed...
Name: brj.f
Type: text/x-fortran
Size: 3515 bytes
Desc: 
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20140429/72e52b0a/attachment.f>