[Wien] lapw0 using TB-mBJ potential hangs for slab calculations

[Yoshiyuki Ohtsubo]

Dear Professor Blaha

I replaced the ordinary vxclm2.f routine with the attached one.
It works all right for the slab attached before (the test case with a small 
unit cell).
The SCF cycle was converged well and the obtained electronic structure was 
consistent with the
experimental results.
Lapw0 hanged with the default cutoff value, but this could be avoided with 
the increased cutoff value.
I hope that it works well for various surfaces.

Thank you very much !!

///////////////////////////////////////////
Yoshiyuki Ohtsubo
Ph.D. student,
Department of Chemistry, Graduate School of Science, Kyoto University,
Kyoto 606-8502, Japan

>mBJ was not meant for surfaces, where the density (and correspondingly tau, 
>...) can come
>close to zero.
>
>Try the attached vxclm2.f routine, which for option 28 (mBJ) would not call 
>the brj subroutine
>when the density is below a certain value. Eventually you have to 
>"experiment" and find a
>good cutoff value, so that the calculation does not hang. Eventually you 
>could also add
>similar traps for small tau or g2rho.
>
>Please remember: the default mBJ calculations use an "average of grad 
>rho/rho" over the unit cell.
>Such an average does NOT MAKE SENSE for a surface, since it will depend on 
>the amount of vacuum
>in your slab-supercell.
>
>Thus when you use mBJ for a surface, I'd recommend to use the case.grr file 
>from the bulk, and
>do not provide a case.in0_grr file in the surface calculation !!!!!
>
>
>>Am 03.08.2011 09:18, schrieb Yoshiyuki Ohtsubo:
>> Dear Professor Blaha and WIEN users,
>>
>> I am using WIEN2k_11.1 to calculate the surface electronic structure on
>> semiconductor (Ge) using TB-mBJ potential.
>> I have found a problem described below and ask for your assistance or 
>> advice.
>> I am using the slab geometry to calculate the surface states. The 
>> detailed structure is in the attached
>> case.struct file. (It is a test case to compare the well-known surface 
>> band structure.)
>>
>> During the SCF cycle, lapw0 seems to hang: never END after one day, while 
>> it requires only a few minites
>> for other exchange and correlation potentials, such as PBE96, with the 
>> same case.struct file.
>> I know that the similar trouble has been reported for WIEN10.1 in this 
>> mailing list.
>> ("LAPW0 hangs in mBJ calculation" Tue Dec 14 23:28:10 CET 2010)
>> However, I could not find how to fix this problem.
>>
>> My environment for WIEN2k calculation is below.
>> Machine: Linux (CENTOS 5) with intel fortran compiler 10.1.021
>> Input files: case.struct file is attached. Other input files are made by 
>> w2web and I modified
>> case.in0/inm for mBJ calculation, according to the userguide.
>> The case.grr value is fixed at that obtained from Ge bulk (~1.44) during 
>> the slab calculation.
>> I am not using the lapw0_mpi parallel option.
>>
>> I have already checked these points.
>> 1.
>> This problem seems to occur due to some delicate conditions. It occurs 
>> after tens of iterations of SCF
>> cycles, and slight change of the conditions can avoid this "hanging".
>> I confirmed that the problem did not occur with a rewind for a few cycles 
>> and restart the SCF cycle with
>> another mixing factor. (However, this cycle also hanged a few tens of 
>> cycles later...)
>> 2.
>> In case.dayfile, the messages as written below are shown ten times before 
>> the hang-up.
>> The same type of messages also appears even when lapw0 finished without 
>> the hang-up.
>> ===================
>> int:rho,tauw,grho,g2rho 9.018537139432929E-003 3.896353988848424E-003
>> 1.185570127100159E-002 3.336162841227647E-003 tauwrong=
>> 3.190123470208489E-003
>> =================== (values change gradually)
>> 3.
>> I inserted the line "print*, tau,g2rho" in the brj.f file to check these 
>> values.
>> The hang-up seems to occur when tau and g2rho are very small.
>> (for example, ~2.3E-007 and 5.6E-009, respectively, for a case I have 
>> checked)
>> 4.
>> In my environment, this hang-up problem occurs only for the slab geometry 
>> with TB-mBJ potential.
>> TB-mBJ calculation for TiC and bulk semiconductor crystal (Si and Ge) 
>> showed no hanging-up in lapw0
>> and the obtained band structure seems to be correct.
>> The values tau and g2rho were orders of magnitude larger than those 
>> written above.
>>
>> If you required any other information in order for me to get some help, 
>> please ask me. I would send it.
>> Thank you very much.
>>
>> ////////////////////////////////////////////
>> Yoshiyuki Ohtsubo
>> Ph.D. student,
>> Department of Chemistry, Graduate School of Science, Kyoto University,
>> Kyoto 606-8502, Japan
>>
>>
>> _______________________________________________
>> Wien mailing list
>> Wien at zeus.theochem.tuwien.ac.at
>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien