EDWAV: internal error, the gradient is not orthogonal
Posted: Wed Jul 11, 2018 10:58 pm
Dear all,
I keep getting the following error after the first NELMDL steps of the SCF loop for my system:
EDWAV: internal error, the gradient is not orthogonal
No matter how I change the INCAR. I tried to modify the number of bands, the cutoff, the NPAR/NCORE values, to change the VdW correctiong method (from none to TS to vdw-df) and to run on a different number of cores/nodes but with no luck. I checked the POTCAR to verify that the order of the atomic species follow the one of the pseudos listed in the POTCAR file.
The geometry cannot be "wrong" because of two reasons: 1) rearranging the atoms doesn't solve the issue and 2) the same geometry is the result of several relaxation steps taken by QuantumESPRESSO, so it cannot be unreasonable, as usually the replies given to this problem state.
The INCAR file is as follows:
I wanted to upload the outfile (not the OUTCAR as it does not report the warning) but the portal doesn't allow me to do that (it says the the extension is not allowed, despite I tried several of them).
I run the code on Marconi@Cineca, where each node has 48 cpus. I have already asked them for some support, but a part from playing out with the parallelization tags, they could not suggest any other solutions.
Any help would be really appreciated.
Thanks in advance,
I keep getting the following error after the first NELMDL steps of the SCF loop for my system:
EDWAV: internal error, the gradient is not orthogonal
No matter how I change the INCAR. I tried to modify the number of bands, the cutoff, the NPAR/NCORE values, to change the VdW correctiong method (from none to TS to vdw-df) and to run on a different number of cores/nodes but with no luck. I checked the POTCAR to verify that the order of the atomic species follow the one of the pseudos listed in the POTCAR file.
The geometry cannot be "wrong" because of two reasons: 1) rearranging the atoms doesn't solve the issue and 2) the same geometry is the result of several relaxation steps taken by QuantumESPRESSO, so it cannot be unreasonable, as usually the replies given to this problem state.
The INCAR file is as follows:
Code: Select all
SYSTEM = relax
ISTART = 0
ICHARG = 2
ISPIN = 1
PREC = Accurate
NWRITE = 2 #med-high verbosity
ADDGRID = .TRUE.
# controlling scf
EDIFF = 0.001 #electronic accuracy
EDIFFG = -0.001 #threshold in forces corresponding to 4D-4 Ry/Bohr(QE)
NELM = 250 # max number of scf steps
NELMDL = -3
NBANDS = 1950
ENCUT = 320
ISMEAR = 1
SIGMA = 0.2
LREAL = A
IMIX = 1 #mixing scheme of old-new KS solutions: linear
AMIX = 0.15 #weight of new KS solutions in the next iteration
BMIX = 0.0001
ALGO = All
#WEIMIN = 0
# setting up dipole correction
!IDIPOL = 3
!LDIPOL = .TRUE.
!DIPOL = 0.0 0.0 0.73
# controlling relaxation
NSW = 0 # number of ionic step in relax
IBRION = -1 #cg relax
ISIF = 0 #not calculating the stresses
# controlling dos plot
LORBIT = 12
NEDOS = 301
# output files settings
LWAVE = .TRUE. #wfcs for dos and bands
LCHARG = .TRUE. #charge for dos, bands, postprocess
!LVHAR = .TRUE. #potential for checking
# parallel settings
LPLANE = .TRUE.
#LSCALU = .FALSE.
NSIM = 4
NPAR = 12
KPAR = 1
# optB88 functional settings
GGA = BO
PARAM1 = 0.1833333333
PARAM2 = 0.2200000000
LUSE_VDW = .TRUE.
AGGAC = 0.0000
I run the code on Marconi@Cineca, where each node has 48 cpus. I have already asked them for some support, but a part from playing out with the parallelization tags, they could not suggest any other solutions.
Any help would be really appreciated.
Thanks in advance,