[Wien] Problem with LAPWSO + ORB(nmod=3)

Peter Blaha pblaha at theochem.tuwien.ac.at
Mon Mar 8 17:20:34 CET 2010


I could verify the problem.

It is in SRC_orb/main.f

Find the line:
         thetaloc=ACOS(xloc(3))

and put just BEFORE this line:
         if(xloc(3).gt.1.d0) xloc(3)=1.d0

and recompile.

Reason: numerical rounding problems. After rotations from (111) direction into local (001),
the vector was (0,0,1.000000005) and the ACOS(1.000000005) of a number gt. 1. leads to NANs ....


pieper schrieb:
> Using WIEN2k_09.1 (Release 5/2/2009) on a dual-core PC with SuSE Linux11.2 I apparently do something wrong when I try to calculate the effect ofan external field on a system with SO interaction: the SCF crashes inlapwso.
> I switched back to the fccni example file to localize the problem and didthe following step by step, using w2web and its defaults in sessions set tospin-polarized and k-parallel:
> 1) dir /fccniso created, fccni.struct copied there and renamed tofccniso.struct 2) StructGen, adapt RMT's to touching spheres struct saved3) initialize calc using defaults all the way, 3000 k-pointsspin-polarized4) scf-cycle, spin-polarized, k-parallel: ok after 8 cycles, save_lapw5) initso setting .inso toWFFIL 4  1  0                      llmax,ipr,kpot  -10.0000   1.50000           emin,emax (output energy window)   1.  1.  1.                 direction of magnetization (lattice vectors) 0                           number of atoms for which RLO is added 0 0 0 0 0                    number of atoms for which SO is switch off;atoms
> .in1 and kgen using the defaults.7) scf-cycle, spin-pol, k-parallel, so: ok after 6 cycles, save_lapw8) scf-cycle, spin-pol, k-parallel, so, orb: w2web conveniently presents.inorb for changes9) .inorb set to :  3  1  0                     nmod, natorb, iprPRATT  1.0                    BROYD/PRATT, mixing  1 1 2                          iatom nlorb, lorb  10.                          Bext in T  1. 1. 1.                     field direction
> 10) .indm and .indmc set to:-9.                      Emin cutoff energy 1                       number of atoms for which density matrix iscalculated 1  1  2      index of 1st atom, number of L's, L1 0 0           r-index, (l,s)index  
> (here I did not rely on w2web presenting the template since it seems thatthe script misses it in complex cases)11) scf-cycle, spin-pol, k-parallel, so, dm, orb: crash in lapwso with.dayfile:
> Calculating fccniso in /home/pieper/WIEN2k/Fccni/fccnisoon tschucktschuck with PID 19259
>     start 	(Fr 5. Mär 14:39:08 CET 2010) with lapw0 (40/99 to go)
>     cycle 1 	(Fr 5. Mär 14:39:08 CET 2010) 	(40/99 to go)
>>   lapw0 -p	(14:39:08) starting parallel lapw0 at Fr 5. Mär 14:39:08 CET2010-------- .machine0 : processorsrunning lapw0 in single mode2.272u 0.024s 0:02.29 100.0%	0+0k 0+472io 0pf+0w>   lapw1  -up -p 	(14:39:10) starting parallel lapw1 at Fr 5. Mär14:39:10 CET 2010->  starting parallel LAPW1 jobs at Fr 5. Mär 14:39:10 CET 2010running LAPW1 in parallel mode (using .machines)2 number_of_parallel_jobs     localhost(140) 0.920u 0.004s 0:00.92 100.0%	0+0k 0+2232io 0pf+0w     localhost(140) 0.912u 0.012s 0:00.87 105.7%	0+0k 0+2208io 0pf+0w   Summary of lapw1para:   localhost	 k=280	 user=1.832	 wallclock=1.791.884u 0.092s 0:04.20 46.9%	0+0k 0+4864io 0pf+0w>   lapw1  -dn -p 	(14:39:15) starting parallel lapw1 at Fr 5. Mär14:39:15 CET 2010->  starting parallel LAPW1 jobs at Fr 5. Mär 14:39:15 CET 2010running LAPW1 in parallel mode (using .machines.help)2 number_of_parallel_jobs     localhost(140) 0.908u 0.020s 0:00.89 103.3%	0+0k 0+2232io 0pf+0w     localhost(140) 0
.916u 0.012s 0:01.00 92.0%	0+0k 0+2208io 0pf+0w   Summary of lapw1para:   localhost	 k=280	 user=1.824	 wallclock=1.891.872u 0.092s 0:04.32 45.3%	0+0k 0+4848io 0pf+0w>   lapwso -up -orb -p	(14:39:19) running LAPWSO in parallel mode      localhost 0.148u 0.032s 0:00.18 94.4% 0+0k 0+9640io 0pf+0w      localhost 0.148u 0.044s 0:00.19 94.7% 0+0k 0+10208io 0pf+0w   Summary of lapwsopara:   localhost	 user=0.296	 wallclock=0.370.332u 0.140s 0:02.08 22.5%	0+0k 0+20672io 0pf+0w>   lapw2 -c -up -so -p 	(14:39:21) running LAPW2 in parallel mode      localhost 0.760u 0.016s 0:00.77 100.0% 0+0k 0+344io 0pf+0w      localhost 0.704u 0.044s 0:00.74 100.0% 0+0k 0+352io 0pf+0w   Summary of lapw2para:   localhost	 user=1.464	 wallclock=1.511.604u 0.128s 0:03.61 47.6%	0+0k 0+1904io 0pf+0w>   lapw2 -c -dn -so -p 	(14:39:25) running LAPW2 in parallel mode      localhost 0.748u 0.028s 0:00.77 98.7% 0+0k 0+344io 0pf+0w      localhost 0.720u 0.036s 0:00.72 104.1% 0+0k 0+352io 0pf+0w   Summary of lap
w2para:   localhost	 user=1.468	 wallclock=1.491.608u 0.128s 0:03.19 53.9%	0+0k 0+1904io 0pf+0w>   lapwdm -up -p -so -c	(14:39:28) running LAPWDM in parallel mode      localhost 0.064u 0.024s 0:00.08 100.0% 0+0k 0+80io 0pf+0w      localhost 0.072u 0.012s 0:00.07 114.2% 0+0k 0+80io 0pf+0w   Summary of lapwdmpara:   localhost	 user=0.136	 wallclock=0.150.180u 0.088s 0:03.09 8.4%	0+0k 0+568io 0pf+0w>   lcore -up	(14:39:31) 0.012u 0.004s 0:00.01 100.0%	0+0k 0+464io 0pf+0w>   lcore -dn	(14:39:31) 0.016u 0.000s 0:00.01 100.0%	0+0k 0+464io 0pf+0w>   mixer 	(14:39:31) 0.028u 0.000s 0:00.02 100.0%	0+0k 0+600io 0pf+0w:ENERGY convergence:  0 0.0001 0:CHARGE convergence:  0 0.0000 0
>     cycle 2 	(Fr 5. Mär 14:39:31 CET 2010) 	(39/98 to go)
>>   lapw0 -p	(14:39:31) starting parallel lapw0 at Fr 5. Mär 14:39:31 CET2010-------- .machine0 : processorsrunning lapw0 in single mode2.272u 0.020s 0:02.30 99.5%	0+0k 0+456io 0pf+0w>   orb -up -p	(14:39:34) 0.000u 0.000s 0:00.00 0.0%	0+0k 0+32io 0pf+0w>   orb -dn -p	(14:39:34) 0.000u 0.000s 0:00.00 0.0%	0+0k 0+32io 0pf+0w>   lapw1  -up -p 	(14:39:34) starting parallel lapw1 at Fr 5. Mär14:39:34 CET 2010->  starting parallel LAPW1 jobs at Fr 5. Mär 14:39:34 CET 2010running LAPW1 in parallel mode (using .machines)2 number_of_parallel_jobs     localhost(140) 0.900u 0.016s 0:00.91 100.0%	0+0k 0+2232io 0pf+0w     localhost(140) 0.912u 0.008s 0:00.92 98.9%	0+0k 0+2208io 0pf+0w   Summary of lapw1para:   localhost	 k=280	 user=1.812	 wallclock=1.831.876u 0.088s 0:04.10 47.5%	0+0k 0+4864io 0pf+0w>   lapw1  -dn -p 	(14:39:38) starting parallel lapw1 at Fr 5. Mär14:39:38 CET 2010->  starting parallel LAPW1 jobs at Fr 5. Mär 14:39:38 CET 2010running LAPW1 in parallel mo
de (using .machines.help)2 number_of_parallel_jobs     localhost(140) 0.924u 0.008s 0:01.05 87.6%	0+0k 0+2232io 0pf+0w     localhost(140) 0.904u 0.028s 0:00.92 100.0%	0+0k 0+2208io 0pf+0w   Summary of lapw1para:   localhost	 k=280	 user=1.828	 wallclock=1.971.880u 0.100s 0:04.14 47.8%	0+0k 0+4864io 0pf+0w>   lapwso -up -orb -p	(14:39:42) running LAPWSO in parallel mode**  LAPWSO crashed!0.048u 0.048s 0:02.06 3.8%	0+0k 0+384io 0pf+0werror: command   /opt/Wien2k_09/lapwsopara -up lapwso.def   failed
>>   stop error
> ------------------
> I checked if I did something wrong in .inorb by creating a new dir, caseand session, doing the same initialization up to step 5). There I did
> 5') scf-cycle, spin-pol, k-parallel, orb: w2web conveniently presents.inorb for changes6') .inorb changed to:  3  1  0                     nmod, natorb, iprPRATT  1.0                    BROYD/PRATT, mixing  1 1 2                          iatom nlorb, lorb  10.                         Bext in T  1. 1. 1.                    field direction
> which is the same as above.7') scf-cycle, spin-pol, k-parallel, orb: w2web conveniently presents.indm (not .indmc, which is not necessary at that point) for changes8') .indm set to:-9.                      Emin cutoff energy 1                       number of atoms for which density matrix iscalculated 1  1  2      index of 1st atom, number of L's, L1 0 0           r-index, (l,s)index  
> which is again the same as above (there .indmc = .indm)10') scf-cycle, spin-pol, k-parallel, dm, orb: ok after 6 cycles,save_lapw This seems to indicate, that .inorb and .indm(c) should be ok ...?11') initso setting .inso toWFFIL 4  1  0                      llmax,ipr,kpot  -10.0000   1.50000           emin,emax (output energy window)   1.  1.  1.                 direction of magnetization (lattice vectors) 0                            number of atoms for which RLO is added 0 0 0 0 0                    number of atoms for which SO is switch off;atoms
> that was ok above ...12') scf-cycle, spin-pol, k-parallel, so, dm, orb: crashed in lapwso withdayfile:
> Calculating fccniB in /home/pieper/WIEN2k/Fccni/fccniBon tschucktschuck with PID 1615
>     start 	(Fr 5. Mär 13:52:06 CET 2010) with lapw0 (40/99 to go)
>     cycle 1 	(Fr 5. Mär 13:52:06 CET 2010) 	(40/99 to go)
>>   lapw0 -p	(13:52:06) starting parallel lapw0 at Fr 5. Mär 13:52:06 CET2010-------- .machine0 : processorsrunning lapw0 in single mode2.272u 0.036s 0:02.31 99.5%	0+0k 0+472io 0pf+0w>   orb -up -p	(13:52:08) 0.000u 0.000s 0:00.00 0.0%	0+0k 0+32io 0pf+0w>   orb -dn -p	(13:52:08) 0.000u 0.000s 0:00.00 0.0%	0+0k 0+32io 0pf+0w>   lapw1  -up -p 	(13:52:08) starting parallel lapw1 at Fr 5. Mär13:52:08 CET 2010->  starting parallel LAPW1 jobs at Fr 5. Mär 13:52:08 CET 2010running LAPW1 in parallel mode (using .machines)2 number_of_parallel_jobs     localhost(140) 0.912u 0.008s 0:00.91 100.0%	0+0k 0+2224io 0pf+0w     localhost(140) 0.912u 0.012s 0:00.95 96.8%	0+0k 0+2192io 0pf+0w   Summary of lapw1para:   localhost	 k=280	 user=1.824	 wallclock=1.861.896u 0.092s 0:04.13 47.9%	0+0k 0+4840io 0pf+0w>   lapw1  -dn -p 	(13:52:12) starting parallel lapw1 at Fr 5. Mär13:52:12 CET 2010->  starting parallel LAPW1 jobs at Fr 5. Mär 13:52:12 CET 2010running LAPW1 in parallel mode (using 
.machines.help)2 number_of_parallel_jobs     localhost(140) 0.924u 0.020s 0:01.03 91.2%	0+0k 0+2224io 0pf+0w     localhost(140) 0.904u 0.024s 0:00.87 105.7%	0+0k 0+2192io 0pf+0w   Summary of lapw1para:   localhost	 k=280	 user=1.828	 wallclock=1.91.864u 0.100s 0:04.29 45.6%	0+0k 0+4832io 0pf+0w>   lapwso -up -orb -p	(13:52:16) running LAPWSO in parallel mode**  LAPWSO crashed!0.040u 0.044s 0:02.06 3.8%	0+0k 0+376io 0pf+0werror: command   /opt/Wien2k_09/lapwsopara -up lapwso.def   failed
>>   stop error
> lapwso.error contains two lines:
> **  Error in Parallel LAPWSO**  Error in Parallel LAPWSO
> So what am I doing wrong (besides using a Wien2k version from last springand not finding the answer in the mailing list)?
> Martin_______________________________________________Wien mailing listWien at zeus.theochem.tuwien.ac.athttp://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien

-- 

                                       P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------



More information about the Wien mailing list