scaLAPACK: Routine ZPOTRF ZTRTRI failed! 17 1 1

[tanay11]

Hi,
I am trying to structure relaxation for Ruthenium on Compute canada cluster using VASP/6.1.2 .
But I am facing an error as :

scaLAPACK: Routine ZPOTRF ZTRTRI failed! 17 1 1

My INCAR file is as follows:
#Start
ISTART = 0
ICHARG = 2

#Electronic
ENCUT = 520
PREC = Accurate
ISMEAR=2
SIGMA = 0.2
NELM = 60
EDIFF=1E-6

#Spin related
#ISPIN = 2
#MAGMOM = 2*1.0 1*0.5 2*0.75

#Ionic
EDIFFG=-0.01
NSW = 20
IBRION=2
ISIF=3
LREAL=.FALSE.
ADDGRID=.TRUE.


can you suggest why this error is coming? Thanks in advance.

[martin.schlipf]

Without a complete set of input files, I cannot comment on the specific system. But as a general guide, what you see is that your Hamiltonian expressed in the basis of the orbitals is not an invertible matrix anymore.

Here are a couple of things you can try: Change the size of the orbital basis using NBANDS, use a different ALGO, change the convergence settings (mixing, threshold).
I also noticed that you didn't specify NPAR or NCORE. If you are using larger jobs you definitely want to optimize your process distribution.

[SKM]

hi
i got this error too.
when i am doing a facet dependent bulk models, its happening immediately after first ionic step (with default NELM of 60 steps).

Strangely its happening for (001) facet of bulk structure of BiOX (X=Br, Cl). Not sure what and why is it happening only for this facet (i.e. x-y plane is (001)). The INCAR is copied here. KPOINTS = [8x8x8]
Total 36 atoms.

The INCAR
--------------------------------
PREC = Accurate
NPAR=16
IBRION = 2
ISIF = 3
ENCUT = 500
EDIFF = 1.0e-05
EDIFFG = 1e-04
IVDW = 11
ISMEAR = 1; SIGMA = 0.01
IALGO = 48
LREAL = Auto
LWAVE = .FALSE.
LCHARG = .FALSE.
ADDGRID = .TRUE.
ISYM = 2
SYMPREC = 1e-07
AMIX = 0.01
NSW=500
NELMDL = -4
----------------------------------------


The error is ..............
..............
..............
RMM: 58 -0.262718368159E+03 0.81940E+01 -0.87383E-01 22372 0.728E-01 0.479E+01
RMM: 59 -0.266582176916E+03 -0.38638E+01 -0.15842E-01 22516 0.453E-01 0.473E+01
RMM: 60 -0.282859557698E+03 -0.16277E+02 -0.44278E+00 21837 0.287E+00
1 F= -.28707101E+03 E0= -.28707093E+03 d E =-.287071E+03
curvature: 0.00 expect dE= 0.000E+00 dE for cont linesearch 0.000E+00
trial: gam= 0.00000 g(F)= 0.138E+02 g(S)= 0.173E+06 ort = 0.000E+00 (trialstep = 0.100E+01)
search vector abs. value= 0.173E+06
bond charge predicted
scaLAPACK: Routine ZPOTRF ZTRTRI failed! 1 1 1
LAPACK: Routine ZTRTRI failed!
--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 0 in communicator MPI_COMM_WORLD
with errorcode 1.

NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
You may or may not see output from other processes, depending on
exactly when Open MPI kills them.
--------------------------------------------------------------------------

When i changed from IALGO=48 to 38 then no such error, but its very slow run.

[henrique_miranda]

@SKM
The electronic calculation for the first ionic step did not converge so you will probably get the wrong forces.
The structure of the second ionic step is generated based on those forces and might be completely unphysical.
I suggest that you first make sure that the electronic calculation is converging and only then start performing the ionic relaxation.

[martin.schlipf]

For future reference, in 6.3 this error message is updated to read

Orbital orthonormalization failed in the inversion of matrix
Possible solutions for this issue are:
- Please check whether the atoms are too close to each other.
- Decreasing POTIM might help during relaxations or MDs.
- Deleting CHGCAR or WAVECAR if incompatible with the structure.
- Try calculating without symmetrizing (ISYM = 0 or ISYM = -1).
- As last resort: using a different PAW for some of the atoms.

In your case (@SKM), you probably ended up in the first of these cases. The electronic scf didn't converge as Henrique pointed out, leading to unphysical forces which pushed the atoms to close to each other.