AM05 not working

[jsfilhol]

Hi,

Ihave some problem testing the AM05 functional in vasp5.2.2 for a CH polymer.
The calculation presently crashes without any iteration with
Error EDDDAV: Call to ZHEGV failed. Returncode = 25 2 8
Changing the blas version does not get any better results has it just go on for ever.
nothing very fancy and it is working with PBEsol
POSCAR
1.00000000000000
12.4960000000000004 -0.6855300000000000 -3.5186299999999999
0.0000000000000000 9.3654299999999999 -3.0431699999999999
0.0000000000000000 0.0000000000000000 13.5000000000000000
12 12
Selective dynamics
Direct
0.3538168112927141 0.4368780731830421 0.0142020223200472 T T T
0.3601742538651970 0.3116147662789637 0.0439188611342989 T T T
0.3715965630403063 0.5707976871062884 0.0915940995105790 T T T
0.3651789701537110 0.6961896544908366 0.0620386796789049 T T T
0.3988722919807354 0.5901826676606923 0.2113412900083890 T T T
0.3640301472295868 0.1964555970408803 0.0624111471153966 T T T
0.3264856685324024 0.4173924198766799 0.8944269430940714 T T T
0.3700508543072175 0.0711303793161478 0.0920405627525552 T T T
0.3929992708641653 0.0905571114669516 0.2114770151358950 T T T
0.3613456566849080 0.8117026004176213 0.0440729626915502 T T T
0.3553209621026133 0.9372382799130233 0.0147320919889459 T T T
0.3324027062077960 0.9180961055330648 0.8953293687933725 T T T
0.3282897122211606 0.0218687137916174 0.8805827644404111 T T T
0.3301079078487618 0.1501451580551864 0.2442269961714116 F F F
0.4007862908398656 0.4865826205305980 0.2261283845490767 T T T
0.3397556246704657 0.6546349359809992 0.2475239808839277 F F F
0.4773944699539071 0.6476929576479176 0.2534095821669738 T T T
0.3974790410476510 0.9869117466970415 0.2264395577253859 T T T
0.4690961810444652 0.1530000748070096 0.2566312785477032 T T T
0.3951833266562814 0.8582661077973194 0.8625204682942018 F F F
0.2478048246097770 0.3602610078507545 0.8525056892668339 T T T
0.3853757292999395 0.3525734422328310 0.8581106830213390 F F F
0.2561645004152078 0.8560536103080559 0.8500806932582849 T T T
0.3247882351311538 0.5209542820165148 0.8796048774504018 T T T

INCAR
PREC = Normal
ALGO =fast
GGA =AM
EDIFF = 1e-6
NELM = 60
ENCUT = 600
LREAL= A
Ionic relaxation
EDIFFG = -1e-2
NSW = 45
IBRION = 2
POTIM =0.5
LPLANE = .TRUE.
NPAR = 1
LSCALU = .FALSE.
NSIM = 4
DOS related values
ISMEAR = 0 ; SIGMA=0.05




I am using PAW pbe pseudo, if I had read well the associated paper, it should work, but a few lines about how vasp 5.2 version handle different XC functional with the same PAW was not so clear: Is there a problem coming from there ?

Thanks for your help

[forsdan]

I might be wrong, but isn't ZHEGV a LAPACK routine that calculates the eigenvalues? Have you tried to use the LAPACK routines supplied with VASP?

If you switch to ALGO=VeryFast, does it make a difference?

Also if you run one ionic step with PBEsol and then restart the calculation with AM05 using the WAVECAR from the PBEsol run as an input, does it work in that case?

Best regards,
/Dan

[jsfilhol]

Hi,
We have tried the supplied vasp lapack lib, but it is still going for ever without getting the first iteration done.
changing ALGO does not solve anything
Doing a start from a wavecar generated by a PBEsol does not solve the problem

As someone ever use AM05 in vasp with PBE PAW pseudo? Was it working

Thanks again

JS

[jsfilhol]

On that problem:
AM05 is now working on larger system in the same unit cell
if the cut off is change to 400 eV the diagonalization is done but we got
WARNING in EDDRMM: call to ZHEGV failed, returncode
and vasp just crashes

very strange behavior

[VASP001]

Can I try to set ADDGRID =.TRUE. IBRION =1 ?
Maybe it works.