[Pw_forum] Problem in SCF convergence with larger unit cells
Dear all, I'm trying to optimize the atomic structure of GaP using QE. When I use a unit cell only containing two atoms, I have no problem in SCF convergence and the structure get optimized in the first step. But when I make my unit cell bigger (let's say a unit cell with 64 atoms) just by repeating my smaller optimized unit cell in real space and run relax calculation, the SCF calculation never get converged. Does anyone know where the problem is? Following is my input: &CONTROL calculation = "relax", dt = 30.D0, prefix = 'gap', pseudo_dir = './', outdir='./', / &system ibrav= 6, celldm(1) =20.52, celldm(3) =1, nat= 64, ntyp= 2, ecutwfc = 12.0, degauss = 0.05D0, occupations = "fixed", smearing= "methfessel-paxton", / &ELECTRONS conv_thr= 1.D-7, mixing_beta = 0.3D0, electron_maxstep= 200, / &IONS pot_extrapolation = "second_order", wfc_extrapolation = "second_order", / ATOMIC_SPECIES Ga 69.72 Ga.blyp-d-hgh.UPF P 30.97 P.blyp-hgh.UPF ATOMIC_POSITIONS (angstrom) Ga 2.714678 2.714678 0.00 Ga 0.00 0.00 0.00 Ga 2.714678 0.00 2.714678 Ga 0.00 2.714678 2.714678 P 4.072017 4.072017 1.357339 P 1.357339 1.357339 1.357339 P 4.072017 1.357339 4.072017 P 1.357339 4.072017 4.072017 Ga 2.714678 2.714678 5.429356 Ga 0.00 0.00 5.429356 Ga 2.714678 0.00 8.144034 Ga 0.00 2.714678 8.144034 P 4.072017 4.072017 6.786695 P 1.357339 1.357339 6.786695 P 4.072017 1.357339 9.501373 P 1.357339 4.072017 9.501373 Ga 2.714678 8.144034 0.00 Ga 0.00 5.429356 0.00 Ga 2.714678 5.429356 2.714678 Ga 0.00 8.144034 2.714678 P 4.072017 9.501373 1.357339 P 1.357339 6.786695 1.357339 P 4.072017 6.786695 4.072017 P 1.357339 9.501373 4.072017 Ga 2.714678 8.144034 5.429356 Ga 0.00 5.429356 5.429356 Ga 2.714678 5.429356 8.144034 Ga 0.00 8.144034 8.144034 P 4.072017 9.501373 6.786695 P 1.357339 6.786695 6.786695 P 4.072017 6.786695 9.501373 P 1.357339 9.501373 9.501373 Ga 8.144034 2.714678 0.00 Ga 5.429356 0.00 0.00 Ga 8.144034 0.00 2.714678 Ga 5.429356 2.714678 2.714678 P 9.501373 4.072017 1.357339 P 6.786695 1.357339 1.357339 P 9.501373 1.357339 4.072017 P 6.786695 4.072017 4.072017 Ga 8.144034 2.714678 5.429356 Ga 5.429356 0.00 5.429356 Ga 8.144034 0.00 8.144034 Ga 5.429356 2.714678 8.144034 P 9.501373 4.072017 6.786695 P 6.786695 1.357339 6.786695 P 9.501373 1.357339 9.501373 P 6.786695 4.072017 9.501373 Ga 8.144034 8.144034 0.00 Ga 5.429356 5.429356 0.00 Ga 8.144034 5.429356 2.714678 Ga 5.429356 8.144034 2.714678 P 9.501373 9.501373 1.357339 P 6.786695 6.786695 1.357339 P 9.501373 6.786695 4.072017 P 6.786695 9.501373 4.072017 Ga 8.144034 8.144034 5.429356
Re: [Pw_forum] Problem in SCF convergence with larger unit cells
Dear Reza Vatan > ecutwfc = 12.0, > Ga 69.72 Ga.blyp-d-hgh.UPF > P 30.97 P.blyp-hgh.UPF It is very strange that you converge at all with a so minimal cutoff and so hard hgh norm-conserving pseudopotentials. I would bet that ecutwfc = 100.0 is barely sufficient for Ga with semicore d shell in valence. I suggest that you raise ecutwfc and find convergence for GaP in the 2 atoms cell. HTH Giuseppe Quoting reza vatan : > Dear all, > > I'm trying to optimize the atomic structure of GaP using QE. When I use a > unit cell only containing two atoms, I have no problem in SCF convergence > and the structure get optimized in the first step. But when I make my unit > cell bigger (let's say a unit cell with 64 atoms) just by repeating my > smaller optimized unit cell in real space and run relax calculation, the > SCF calculation never get converged. Does anyone know where the problem is? > > Following is my input: > > > &CONTROL > calculation = "relax", > dt = 30.D0, > prefix = 'gap', > pseudo_dir = './', > outdir='./', > / > &system > ibrav= 6, celldm(1) =20.52, celldm(3) =1, nat= 64, ntyp= 2, > ecutwfc = 12.0, > degauss = 0.05D0, > occupations = "fixed", > smearing= "methfessel-paxton", > / > &ELECTRONS > conv_thr= 1.D-7, > mixing_beta = 0.3D0, > electron_maxstep= 200, > / > &IONS > pot_extrapolation = "second_order", > wfc_extrapolation = "second_order", > / > ATOMIC_SPECIES > Ga 69.72 Ga.blyp-d-hgh.UPF > P 30.97 P.blyp-hgh.UPF > ATOMIC_POSITIONS (angstrom) > Ga 2.714678 2.714678 0.00 > Ga 0.00 0.00 0.00 > Ga 2.714678 0.00 2.714678 > Ga 0.00 2.714678 2.714678 > P 4.072017 4.072017 1.357339 > P 1.357339 1.357339 1.357339 > P 4.072017 1.357339 4.072017 > P 1.357339 4.072017 4.072017 > Ga 2.714678 2.714678 5.429356 > Ga 0.00 0.00 5.429356 > Ga 2.714678 0.00 8.144034 > Ga 0.00 2.714678 8.144034 > P 4.072017 4.072017 6.786695 > P 1.357339 1.357339 6.786695 > P 4.072017 1.357339 9.501373 > P 1.357339 4.072017 9.501373 > Ga 2.714678 8.144034 0.00 > Ga 0.00 5.429356 0.00 > Ga 2.714678 5.429356 2.714678 > Ga 0.00 8.144034 2.714678 > P 4.072017 9.501373 1.357339 > P 1.357339 6.786695 1.357339 > P 4.072017 6.786695 4.072017 > P 1.357339 9.501373 4.072017 > Ga 2.714678 8.144034 5.429356 > Ga 0.00 5.429356 5.429356 > Ga 2.714678 5.429356 8.144034 > Ga 0.00 8.144034 8.144034 > P 4.072017 9.501373 6.786695 > P 1.357339 6.786695 6.786695 > P 4.072017 6.786695 9.501373 > P 1.357339 9.501373 9.501373 > Ga 8.144034 2.714678 0.00 > Ga 5.429356 0.00 0.00 > Ga 8.144034 0.00 2.714678 > Ga 5.429356 2.714678 2.714678 > P 9.501373 4.072017 1.357339 > P 6.786695 1.357339 1.357339 > P 9.501373 1.357339 4.072017 > P 6.786695 4.072017 4.072017 > Ga 8.144034 2.714678 5.429356 > Ga 5.429356 0.00 5.429356 > Ga 8.144034 0.00 8.144034 > Ga 5.429356 2.714678 8.144034 > P 9.501373 4.072017 6.786695 > P 6.786695 1.357339 6.786695 > P 9.501373 1.357339 9.501373 > P 6.786695 4.072017 9.501373 > Ga
Re: [Pw_forum] Problem in SCF convergence with larger unit cells
Dear Giuseppe, I had tried the calculations with higher ecutwfc before and as you said it perfectly worked. But the problem is that the calculations take much longer. Is there any way to keep the ecutwfc as small as possible and use different pseudopotentials and still get convergence? Please advise. Many thanks in advance. Regards, Reza On Sat, Jun 11, 2016 at 2:21 PM, Giuseppe Mattioli < giuseppe.matti...@ism.cnr.it> wrote: > > Dear Reza Vatan > > > ecutwfc = 12.0, > > > Ga 69.72 Ga.blyp-d-hgh.UPF > > P 30.97 P.blyp-hgh.UPF > > It is very strange that you converge at all with a so minimal cutoff > and so hard hgh norm-conserving pseudopotentials. I would bet that > ecutwfc = 100.0 is barely sufficient for Ga with semicore d shell in > valence. I suggest that you raise ecutwfc and find convergence for GaP > in the 2 atoms cell. > HTH > Giuseppe > > Quoting reza vatan : > > > Dear all, > > > > I'm trying to optimize the atomic structure of GaP using QE. When I use a > > unit cell only containing two atoms, I have no problem in SCF convergence > > and the structure get optimized in the first step. But when I make my > unit > > cell bigger (let's say a unit cell with 64 atoms) just by repeating my > > smaller optimized unit cell in real space and run relax calculation, the > > SCF calculation never get converged. Does anyone know where the problem > is? > > > > Following is my input: > > > > > > &CONTROL > > calculation = "relax", > > dt = 30.D0, > > prefix = 'gap', > > pseudo_dir = './', > > outdir='./', > > / > > &system > > ibrav= 6, celldm(1) =20.52, celldm(3) =1, nat= 64, ntyp= 2, > > ecutwfc = 12.0, > > degauss = 0.05D0, > > occupations = "fixed", > > smearing= "methfessel-paxton", > > / > > &ELECTRONS > > conv_thr= 1.D-7, > > mixing_beta = 0.3D0, > > electron_maxstep= 200, > > / > > &IONS > > pot_extrapolation = "second_order", > > wfc_extrapolation = "second_order", > > / > > ATOMIC_SPECIES > > Ga 69.72 Ga.blyp-d-hgh.UPF > > P 30.97 P.blyp-hgh.UPF > > ATOMIC_POSITIONS (angstrom) > > Ga 2.714678 2.714678 0.00 > > Ga 0.00 0.00 0.00 > > Ga 2.714678 0.00 2.714678 > > Ga 0.00 2.714678 2.714678 > > P 4.072017 4.072017 1.357339 > > P 1.357339 1.357339 1.357339 > > P 4.072017 1.357339 4.072017 > > P 1.357339 4.072017 4.072017 > > Ga 2.714678 2.714678 5.429356 > > Ga 0.00 0.00 5.429356 > > Ga 2.714678 0.00 8.144034 > > Ga 0.00 2.714678 8.144034 > > P 4.072017 4.072017 6.786695 > > P 1.357339 1.357339 6.786695 > > P 4.072017 1.357339 9.501373 > > P 1.357339 4.072017 9.501373 > > Ga 2.714678 8.144034 0.00 > > Ga 0.00 5.429356 0.00 > > Ga 2.714678 5.429356 2.714678 > > Ga 0.00 8.144034 2.714678 > > P 4.072017 9.501373 1.357339 > > P 1.357339 6.786695 1.357339 > > P 4.072017 6.786695 4.072017 > > P 1.357339 9.501373 4.072017 > > Ga 2.714678 8.144034 5.429356 > > Ga 0.00 5.429356 5.429356 > > Ga 2.714678 5.429356 8.144034 > > Ga 0.00 8.144034 8.144034 > > P 4.072017 9.501373 6.786695 > > P 1.357339 6.786695 6.786695 > > P 4.072017 6.786695 9.501373 > > P 1.357339 9.501373 9.501373 > > Ga 8.144034 2.714678 0.00 > > Ga 5.429356 0.00 0.00 > > Ga 8.144034 0.00 2.714678 > > Ga 5.429356 2.714678 2.714678 > > P 9.501373 4.072017 1.357339 > > P 6.786695 1.357339 1.357339 > > P 9.501373 1.357339 4.072017 > > P 6.7866
Re: [Pw_forum] Problem in SCF convergence with larger unit cells
Dear Reza You can try to use ultrasoft pseudopotentials you may try these ones you find in the online library Ga.pbe-dn-rrkjus_psl.0.2.UPF P.pbe-n-rrkjus_psl.0.1.UPF But you should check the convergence, and I expect something like ecutwfc=40.0 ecutrho=320.0~480.0 Best Regards Giuseppe Quoting reza vatan : > Dear Giuseppe, > > I had tried the calculations with higher ecutwfc before and as you said it > perfectly worked. But the problem is that the calculations take much > longer. Is there any way to keep the ecutwfc as small as possible and use > different pseudopotentials and still get convergence? Please advise. > > Many thanks in advance. > > Regards, > Reza > > On Sat, Jun 11, 2016 at 2:21 PM, Giuseppe Mattioli < > giuseppe.matti...@ism.cnr.it> wrote: > >> >> Dear Reza Vatan >> >> > ecutwfc = 12.0, >> >> > Ga 69.72 Ga.blyp-d-hgh.UPF >> > P 30.97 P.blyp-hgh.UPF >> >> It is very strange that you converge at all with a so minimal cutoff >> and so hard hgh norm-conserving pseudopotentials. I would bet that >> ecutwfc = 100.0 is barely sufficient for Ga with semicore d shell in >> valence. I suggest that you raise ecutwfc and find convergence for GaP >> in the 2 atoms cell. >> HTH >> Giuseppe >> >> Quoting reza vatan : >> >> > Dear all, >> > >> > I'm trying to optimize the atomic structure of GaP using QE. When I use a >> > unit cell only containing two atoms, I have no problem in SCF convergence >> > and the structure get optimized in the first step. But when I make my >> unit >> > cell bigger (let's say a unit cell with 64 atoms) just by repeating my >> > smaller optimized unit cell in real space and run relax calculation, the >> > SCF calculation never get converged. Does anyone know where the problem >> is? >> > >> > Following is my input: >> > >> > >> > &CONTROL >> > calculation = "relax", >> > dt = 30.D0, >> > prefix = 'gap', >> > pseudo_dir = './', >> > outdir='./', >> > / >> > &system >> > ibrav= 6, celldm(1) =20.52, celldm(3) =1, nat= 64, ntyp= 2, >> > ecutwfc = 12.0, >> > degauss = 0.05D0, >> > occupations = "fixed", >> > smearing= "methfessel-paxton", >> > / >> > &ELECTRONS >> > conv_thr= 1.D-7, >> > mixing_beta = 0.3D0, >> > electron_maxstep= 200, >> > / >> > &IONS >> > pot_extrapolation = "second_order", >> > wfc_extrapolation = "second_order", >> > / >> > ATOMIC_SPECIES >> > Ga 69.72 Ga.blyp-d-hgh.UPF >> > P 30.97 P.blyp-hgh.UPF >> > ATOMIC_POSITIONS (angstrom) >> > Ga 2.714678 2.714678 0.00 >> > Ga 0.00 0.00 0.00 >> > Ga 2.714678 0.00 2.714678 >> > Ga 0.00 2.714678 2.714678 >> > P 4.072017 4.072017 1.357339 >> > P 1.357339 1.357339 1.357339 >> > P 4.072017 1.357339 4.072017 >> > P 1.357339 4.072017 4.072017 >> > Ga 2.714678 2.714678 5.429356 >> > Ga 0.00 0.00 5.429356 >> > Ga 2.714678 0.00 8.144034 >> > Ga 0.00 2.714678 8.144034 >> > P 4.072017 4.072017 6.786695 >> > P 1.357339 1.357339 6.786695 >> > P 4.072017 1.357339 9.501373 >> > P 1.357339 4.072017 9.501373 >> > Ga 2.714678 8.144034 0.00 >> > Ga 0.00 5.429356 0.00 >> > Ga 2.714678 5.429356 2.714678 >> > Ga 0.00 8.144034 2.714678 >> > P 4.072017 9.501373 1.357339 >> > P 1.357339 6.786695 1.357339 >> > P 4.072017 6.786695 4.072017 >> > P 1.357339 9.501373 4.072017 >> > Ga 2.714678 8.144034 5.429356 >> > Ga 0.00 5.429356 5.429356 >> > Ga 2.714678 5.429356 8.144034 >> > Ga 0.00 8.144034 8.144034 >> > P 4.072017 9.501373 6.786695 >> > P 1.357339 6.786695 6.786695 >> > P 4.072017 6.786695 9.501373 >> > P 1.357339 9.501373 9.501373 >> > Ga 8.144034 2.714678 0.00 >> > Ga 5.42935