PBE paw used in PBE calculation of PdO

Message

swen · #1 Post by **swen** » Fri May 13, 2011 2:50 pm

I have tried a optimisation calculation on PdO system using PBE+U with PBE paw data. I have tested it with U values from 0 - 6. But the resulting energy seems to be rubbish with U values >=2. But this didn't happen using PW91+U method with PW91 paw data and also PBE+U method with PW91 paw data.

I have used both the Pd_new and Pd_pv_new PAW data for the calculation in PBE+U.

The incar i used for the PBE+U calculation is as follow and it is similar to the PW91+U calculation

SYSTEM PdO_bulk
ENCUT = 500
PREC = High
ICHARG = 2 !Start of with atomic densities
IBRION = 2 !Conjugate-gradient
ISIF = 2 !Relax atoms not cell
ISPIN = 2 !Spin polarised
NSW = 40 !Nr of ionic steps
GGA = PE !PBE
EDIFFG = -0.01 ! check the convergence gradient is less than that if yes then stop
LDAU = .TRUE.
LDAUTYPE = 2 !dudarev et al method Ueff=(U-J)
LDAUL = 2 -1
LDAUU = 6.0 0.0
LDAUJ = 0.0 0.0
LDAUPRINT = 1

The summary of converged energy extracted from OUTCAR
for U=1; Final converged energy= -22.275926;
for U=2; Final converged energy= -8720.997228;
for U=3; Final converged energy= -23596.246960;
for U=4; Final converged energy= -30557.299718;
for U=5; Final converged energy= -38486.763015;
for U=6; Final converged energy= -46418.012462;

from the OUTCAR for U=2,
I have notice the 'eigenvalues EBANDS = 13.62680413' which then in the next step jumped to 'eigenvalues EBANDS = -143.58904424' and became more and more negative in subsequent steps.

Has anyone encounter this problem before? Does anyone know what causes this?
Thanks in advance

Scott

alex · #2 Post by **alex** » Sat May 14, 2011 2:56 pm

Hi Scott,

did your wavefunction converge at all? Check the OSZICAR for that.

Cheers,

alex

razworks · #3 Post by **razworks** » Sat May 14, 2011 8:36 pm

Its always a good idea to keep a copy of the OSZICAR file, as it may provide important information in such situtations.

swen · #4 Post by **swen** » Mon May 16, 2011 2:05 pm

Will check the OSZICAR file, thanks Alex and Razworks for the reply.
But do you mean the system reached convergence. If yes, then my system has reached convergence as written in the OUTCAR.

Will update the thread here once I have checked the OSZICAR.

Thank you

swen · #5 Post by **swen** » Wed May 18, 2011 1:36 pm

The dE in the OSZICAR do not show to convergence, but in the OUTCAR it states there 'reached required accuracy - stopping structural minimisation'

forsdan · #6 Post by **forsdan** » Wed May 18, 2011 4:41 pm

Check your std.out and std.err files to check whether or not you have a failure e.g. in the diagnolization routine.

Cheers,
/Dan

swen · #7 Post by **swen** » Fri May 20, 2011 3:54 pm

Forsdan,

I have checked that as well, no error reported in the std.out and std.err files

alex · #8 Post by **alex** » Fri May 20, 2011 5:05 pm

Did you remove your old CHG, CHGCAR or WAVECAR (if existent)

Cheers,

alex

swen · #9 Post by **swen** » Mon May 23, 2011 1:38 pm

no I didn't. they are still there

cheers
scott