on solving "Error EDDDAV: Call to ZHEGV failed. Returncode = xx"

[jcconesa]

Hi,
I have suffered from time to time the error in the subject. In my case, appearing at the beginning of the scf procedure, in particular as third line after

--- Iteration 1( 1) ---

I've tried several solutions proposed earlier in the forum:

commenting out the line #define USE_ZHEEVX in rmm-diis.F, davidson.F, subrot.F, and wavpre_noio.F and recompiling

using IALGO = 48 (this just changes the error)

using a different lapack

adding "LSCALAPACK = .FALSE." in INCAR

all of them unsuccessful.
Then I found something surprising: with exactly the same input files, I find the problem when using 32 cores (i.e. 4 nodes having each two quad-core Opterons) but not if using 24 cores (i.e. three nodes of the said type). The problem (in my case) seems thus to be related with the way in which the work is distributed among nodes. Then I found that the problem is also avoided if I suppress the line
LPLANE = .FALSE.
which I had in my INCAR.
Hopefully this information can help others. Maybe there is also something to change in the code for future versions.

All the best,

[admin]

thank you very much!

[hacioz]

Hi,
I have also suffered from this error for some calculations.

INCAR
_______________________________________________________________________
ISIF = 3
IBRION = 2
EDIFF = 1E-08 #stopping-criterion for ELM
EDIFFG = -1E-06 #stopping-criterion for IOM
NSW = 100
PREC = H
ENCUT= 450
LREAL = F
LWAVE = F
LCHARG = F
LMAXMIX = 6
_______________________________________________________________________
POSCAR
_______________________________________________________________________
Bulk
1.00000000000000
1.5581813806900786 -2.6988493187630396 0.0000000000000000
1.5581813806900786 2.6988493187630396 0.0000000000000000
0.0000000000000000 0.0000000000000000 4.0471377090623681
2 1
Direct
0.6666666666666714 0.3333333333333286 0.5000000000000000
0.3333333333333286 0.6666666666666714 0.5000000000000000
0.0000000000000000 0.0000000000000000 0.0000000000000000
__________________________________________________________________________
many lapack (netlib(3.1.1, 3.4.0), goto, vasp.5.lib tested, but no solition
also the above methods unsuccessful.

Thanks for your reply in advance.
Haci Ozisik

[admin]

No error message appears using your input files.
Probably the installation of the LAPACK on your machine was not done properly.

[SKM]

hi
i also got this error.
surprisingly, the first run for geometry optimisation completed (means wall time over) without such error.
but then when continued the next-run with CONTCAR copied to POSCAR, then the error occured after 2nd electronic step.
..... ......
WAVECAR not read
entering main loop
N E dE d eps ncg rms rms(c)
DAV: 1 0.981480160256E+04 0.98148E+04 -0.37255E+05 8496 0.985E+02
DAV: 2 0.508965503211E+03 -0.93058E+04 -0.90198E+04 11664 0.210E+02
Error EDDDAV: Call to ZHEGV failed. Returncode = 23 1 24

Can anyone help is sorting this out quickly? this problem eating away CPU time without success.

Regards
~

[martin.schlipf]

This is an error of the diagonalization routine. What you can try is switching to a different method (i.e. ALGO) or using the positions of the second to last step instead of the very last one.

Also helpful is to look at the structure (e.g. in VESTA) to make sure it still looks like what you expect. Occasionally it can happen that two atoms get to close to each other during the relaxation. In particular check, whether all your electronic iterations within the ionic loop converged.

[SKM]

Thanks Martin.
I tried with different IALGO and then the error is gone for this task.

but another point you mentioned rightly, which i was wondering for quite some time, as you mentioned:-

..... or using the positions of the second to last step instead of the very last one.

How do i pickup the CONTCAR values in between steps? May be its from OUTCAR file? if so, then how, as its in a bit disturbed structure, and that too if i use 'selective dynamics' then its further hard. Is there any bash script?

If there is any other way, kindly let me know.

[martin.schlipf]

You would need to do that manually, the structure is written to the output, but not in a format that is compatible with a POSCAR file.

[alex]

Hi there,

XDATCAR contains all structures right from the beginning to the end. The header is printed just once, but it's much easier to process than the OUTCAR file. ;-)

Cheers,

Alex

[SKM]

Thanks a lot Martin and Alex.
This helped me well to pickup the intermittent CONTCAR.
Regards.

[mark_wistey1]

Just to add to the list, this error also can arise from NBANDS being too large. Try commenting it out.

[ian_brewis]

I've been having a similar problem. I've been getting the same error but even after changing ALGO my simulation ends early, albeit without the error message. Strangely though the same input files will run the simulation fine however its on another HPC where I've been encountering this error.

[henrique_miranda]

@ian_brewis You have started a separate thread for your problem: https://www.vasp.at/forum/viewtopic.php?f=4&t=18014
I believe that the reason you are seeing this error in your case is because of a mistake in your POSCAR (there are two overlapping atoms).

There are different reasons why this error might occur besides the ones suggested above.
To track down the source of error one needs to check carefully the inputs (INCAR, POSCAR, KPOINTS).

[alok_shukla1]

Hi, I am getting this error "Error EDDDAV: Call to ZHEGV failed. Returncode = 7 1 8" in the second ionic iteration. Can I request suggestions to fix this error??
Plz find INCAR and a couple of the last line of error below:

INCAR--

Code: Select all

 SYSTEM =  GRPH                            
 Electronic Relaxation 1
  ISTART  = 0
  ICHARG  = 2
  ISPIN   = 2
  LWAVE   =.False.
# NELMIN  = 8
  PREC    = High
# NBANDS  = 60
# ENCUT   = 500
# ENAUG   = 2275
  LREAL   = A
  EDIFF   =  1E-5
  EDIFFG  = -0.005
  NSW     = 1500
  IBRION  =  2
  ISIF    =  2
  IVDW = 11
  VDW_S6 = 1.0
  VDW_SR = 1.176
# VDW_S8 = 0.722
  LVDW= .TRUE.
  VDW_C6 =  1.75 24.67 
  VDW_R0 =  1.452 1.639

# POTIM   = 0.3
# PSTRESS = -22.2
   DOS related values
   ISMEAR = 0;
   SIGMA  = 0.005  
#  NGX =64 NGY = 64  NGZ = 98 
#  NGXF = 128 NGYF =128 NGZF = 196

error--

Code: Select all

 bond charge predicted
       N       E                     dE             d eps       ncg     rms          rms(c)
DAV:   1     0.108192034478E+04    0.15007E+04   -0.23078E+04 18288   0.344E+02    0.499E+01
DAV:   2     0.831675609721E+03   -0.25024E+03   -0.31723E+03 20192   0.139E+02    0.616E+01
DAV:   3     0.823418167251E+03   -0.82574E+01   -0.14896E+02 20896   0.330E+01    0.660E+01
DAV:   4     0.824118487087E+03    0.70032E+00   -0.14678E+01 19648   0.977E+00    0.684E+01
DAV:   5     0.740936615364E+03   -0.83182E+02   -0.76067E+01 17728   0.202E+01    0.977E+01
DAV:   6     0.727521251193E+03   -0.13415E+02   -0.45806E+01 17920   0.101E+01    0.105E+02
DAV:   7    -0.199629089842E+06   -0.20036E+06   -0.14552E+06 28656   0.214E+04    0.265E+02
DAV:   8    -0.449766254500E+05    0.15465E+06   -0.27191E+05 26040   0.506E+03    0.276E+02
DAV:   9    -0.183518626167E+04    0.43141E+05   -0.88465E+03 15560   0.333E+02    0.257E+02
DAV:  10    -0.868961735116E+06   -0.86713E+06   -0.13083E+06 26260   0.948E+03BRMIX: very serious problems
 the old and the new charge density differ
 old charge density:   196.00005 new  203.56769
    0.347E+02
 WARNING: Sub-Space-Matrix is not hermitian in DAV           10
  1.793901674087011E-005
DAV:  11    -0.835653412222E+05    0.78540E+06   -0.23990E+05 18140   0.548E+02BRMIX: very serious problems
 the old and the new charge density differ
 old charge density:   203.56769 new  210.46349
    0.314E+02
Error EDDDAV: Call to ZHEGV failed. Returncode =   7 1   8
Application 245234 resources: utime ~6873s, stime ~835s, Rss ~17704360, inblocks ~1412, outblocks ~410224
 

[merzuk.kaltak]

Please attach a full error report. That is include POSCAR and KPOINTS file.