Changing NPAR results in changes of LOPTICS=.TRUE. output
Posted: Thu Jul 18, 2013 4:57 pm
Hey all,
While checking NPAR efficiencies, I noticed that my dielectric function in the OUTCAR of my LOPTICS calculations differ noticeably at low EVs. Above 10 eV the convergence is pretty much omplete, but I start seeing mild divergence below 10 eV, large divergence at below 5 ev, and orders of magnitude divergence at below 2 eV.
I do the calculation in two steps, first I do a self consistent calculation to get the WAVECAR and CHGCAR files, then I use those for the initial starting point of the LOPTICS electronic convergence. The INCAR files for both steps are identical between the different runs except for the NPAR values which are consistent in each. (i.e. For NPAR=4, I do an scf run with NPAR=4, then use that WAVECAR for my LOPTICS run with NPAR=4).
The total energies were reasonably consistent between the calculations, all within .3meV of each other. Just not the dielectric function.
Here is the INCAR file from my LOPTICS and scf calculations.
System = Slab: NiSi2 ~ c-Si - Self Consistent Calculation
Starting Parameters
ISTART = 0 : Wave Functions: New - 0, Continuation - 1, Same Basis = 2
ICHARG = 2 : Charge Density: Wavecar - 0, Chgcar - 1, New - 2
LREAL = Auto : Projection Operators: Large Cell - Auto, Small Cell - .FALSE.
LWAVE = .TRUE. : Write out WAVECAR? Yes - .TRUE., No - .FALSE.
LCHARG = .TRUE. : Write out CHGCAR? Yes - .TRUE., No - .FALSE.
NPAR = 4 : Bands Parallelization Parameter
## KPAR = 4 : K-Point Parallelization Parameter
Electronic Relaxation Parameters
ALGO = N : Electronic Algorithm: SCF - N, LOPTICS - Exact, w/fields - CHI
NELM = 250 : Electronic Convergence Steps
EDIFF = 1e-2 : Electronic Convergence Loop Break Condition
Ionic Relaxation Parameters
IBRION = 2 : Ionic Relaxation Algorithm
ISIF = 3 : Ionic Parameters Allowed to Relax
NSW = 0 : Ionic Convergence Steps: No Relaxation - 0
EDIFFG = 1e-3 : Ionic Convergence Loop Break Condition
Job Specific Parameters
ISMEAR = -5 : Kpoint Broadening Method: Tetr - -5, Gauss - 0, MP(N) - 1 + N
# SIGMA = 0.1 : Kpoint Broadening Parameter
NBANDS = 512 : Number of Kohn Sham Bands
NEDOS = 2001 : Controls Frequency Grid Density for DOS, LOPTICS
# CSHIFT = 0.1 : Complex Broadening Parameter
ENCUT = 300 : Energy Cut-off for Plane Wave Basis
System = Slab: NiSi2 ~ c-Si - Independent Particle Optics Calculation
Starting Parameters
ISTART = 2 : Wave Functions: New - 0, Continuation - 1, Same Basis = 2
ICHARG = 1 : Charge Density: Wavecar - 0, Chgcar - 1, New - 2
LREAL = Auto : Projection Operators: Large Cell - Auto, Small Cell - .FALSE.
LOPTICS = .TRUE. : Calculate Frequency Dependent Dielectric Matrix
LWAVE = .FALSE. : Write out WAVECAR? Yes - .TRUE., No - .FALSE.
LCHARG = .FALSE. : Write out CHGCAR? Yes - .TRUE., No - .FALSE.
NPAR = 4 : Bands Parallelization Parameter
# KPAR = 4 : K-Point Parallelization Parameter
Electronic Relaxation Parameters
ALGO = Exact : Electronic Algorithm: SCF - N, LOPTICS - Exact, w/fields - CHI
NELM = 250 : Electronic Convergence Steps
EDIFF = 1e-2 : Electronic Convergence Loop Break Condition
Ionic Relaxation Parameters
IBRION = 2 : Ionic Relaxation Algorithm
ISIF = 3 : Ionic Parameters Allowed to Relax
NSW = 0 : Ionic Convergence Steps: No Relaxation - 0
EDIFFG = 1e-4 : Ionic Convergence Loop Break Condition
Job Specific Parameters
ISMEAR = -5 : K-Point Broadening Method: Tetr - -5, Gauss - 0, MP(N) - 1 + N
SIGMA = 0.1 : K-Point Broadening Parameter
NBANDS = 512 : Number of Kohn Sham Bands
NEDOS = 2001 : Controls Frequency Grid Density for DOS, LOPTICS
CSHIFT = 0.1 : Complex Broadening Parameter
ENCUT = 300 : Energy Cut-off for Plane Wave Basis
While checking NPAR efficiencies, I noticed that my dielectric function in the OUTCAR of my LOPTICS calculations differ noticeably at low EVs. Above 10 eV the convergence is pretty much omplete, but I start seeing mild divergence below 10 eV, large divergence at below 5 ev, and orders of magnitude divergence at below 2 eV.
I do the calculation in two steps, first I do a self consistent calculation to get the WAVECAR and CHGCAR files, then I use those for the initial starting point of the LOPTICS electronic convergence. The INCAR files for both steps are identical between the different runs except for the NPAR values which are consistent in each. (i.e. For NPAR=4, I do an scf run with NPAR=4, then use that WAVECAR for my LOPTICS run with NPAR=4).
The total energies were reasonably consistent between the calculations, all within .3meV of each other. Just not the dielectric function.
Here is the INCAR file from my LOPTICS and scf calculations.
System = Slab: NiSi2 ~ c-Si - Self Consistent Calculation
Starting Parameters
ISTART = 0 : Wave Functions: New - 0, Continuation - 1, Same Basis = 2
ICHARG = 2 : Charge Density: Wavecar - 0, Chgcar - 1, New - 2
LREAL = Auto : Projection Operators: Large Cell - Auto, Small Cell - .FALSE.
LWAVE = .TRUE. : Write out WAVECAR? Yes - .TRUE., No - .FALSE.
LCHARG = .TRUE. : Write out CHGCAR? Yes - .TRUE., No - .FALSE.
NPAR = 4 : Bands Parallelization Parameter
## KPAR = 4 : K-Point Parallelization Parameter
Electronic Relaxation Parameters
ALGO = N : Electronic Algorithm: SCF - N, LOPTICS - Exact, w/fields - CHI
NELM = 250 : Electronic Convergence Steps
EDIFF = 1e-2 : Electronic Convergence Loop Break Condition
Ionic Relaxation Parameters
IBRION = 2 : Ionic Relaxation Algorithm
ISIF = 3 : Ionic Parameters Allowed to Relax
NSW = 0 : Ionic Convergence Steps: No Relaxation - 0
EDIFFG = 1e-3 : Ionic Convergence Loop Break Condition
Job Specific Parameters
ISMEAR = -5 : Kpoint Broadening Method: Tetr - -5, Gauss - 0, MP(N) - 1 + N
# SIGMA = 0.1 : Kpoint Broadening Parameter
NBANDS = 512 : Number of Kohn Sham Bands
NEDOS = 2001 : Controls Frequency Grid Density for DOS, LOPTICS
# CSHIFT = 0.1 : Complex Broadening Parameter
ENCUT = 300 : Energy Cut-off for Plane Wave Basis
System = Slab: NiSi2 ~ c-Si - Independent Particle Optics Calculation
Starting Parameters
ISTART = 2 : Wave Functions: New - 0, Continuation - 1, Same Basis = 2
ICHARG = 1 : Charge Density: Wavecar - 0, Chgcar - 1, New - 2
LREAL = Auto : Projection Operators: Large Cell - Auto, Small Cell - .FALSE.
LOPTICS = .TRUE. : Calculate Frequency Dependent Dielectric Matrix
LWAVE = .FALSE. : Write out WAVECAR? Yes - .TRUE., No - .FALSE.
LCHARG = .FALSE. : Write out CHGCAR? Yes - .TRUE., No - .FALSE.
NPAR = 4 : Bands Parallelization Parameter
# KPAR = 4 : K-Point Parallelization Parameter
Electronic Relaxation Parameters
ALGO = Exact : Electronic Algorithm: SCF - N, LOPTICS - Exact, w/fields - CHI
NELM = 250 : Electronic Convergence Steps
EDIFF = 1e-2 : Electronic Convergence Loop Break Condition
Ionic Relaxation Parameters
IBRION = 2 : Ionic Relaxation Algorithm
ISIF = 3 : Ionic Parameters Allowed to Relax
NSW = 0 : Ionic Convergence Steps: No Relaxation - 0
EDIFFG = 1e-4 : Ionic Convergence Loop Break Condition
Job Specific Parameters
ISMEAR = -5 : K-Point Broadening Method: Tetr - -5, Gauss - 0, MP(N) - 1 + N
SIGMA = 0.1 : K-Point Broadening Parameter
NBANDS = 512 : Number of Kohn Sham Bands
NEDOS = 2001 : Controls Frequency Grid Density for DOS, LOPTICS
CSHIFT = 0.1 : Complex Broadening Parameter
ENCUT = 300 : Energy Cut-off for Plane Wave Basis