El-Ph of the conduction bands
Posted: Fri Apr 07, 2017 3:06 am
Dear all!
I am currently trying to reproduce the ncomms11755 paper "Electron-phonon coupling in hybrid lead halide perovskites" by Wright, Verdi et al. by using CsPbBr3 (in the cubic 5-atoms unit cell).
22 filled and 22 empty bands were used for the nscf step on 216 k-points (6x6x6 MP). Phonon dispersion comes out nicely and calculated Raman peaks match the experiment very well.
I worked out the valence bands just fine and using a lower number of random points on the fine grid (sufficient for my purpose) values agree nicely. But for some reasons the conduction bands are not working out for me. The wannier90 interpolated bands agree very well with the ones calculated from QE and the spread of each MLWF is between 1-2 A.
For the valence bands the el-ph calculation on a 12x12x12 fine k/q grid took about 2 days (I only have 6 processors/16 GB ram on the testing node) but for the CB after 2 days the calculation is still crunching numbers on the first irreducible q-point. In detail, it shows
whilst epw.x is running at full steam on all cores. I have browsed the forum and saw that random q/k points were recommended at some point so I tried that one next, but the results were similar (here, the calculation progressed to the 3rd point and remained there for a day or so). Is it significantly more difficult to calculate the CB compared to the VB (assuming the wannier interpolation was equally successful for both)?
here is my poor-man's version of the el-ph Im(Sigma) (meV) overlayed on the (DFT) band-structure: https://drive.google.com/file/d/0B4jAwk ... sp=sharing
any comment from the experts is, as always, greatly appreciated!
With best regards,
Chris
PS: I cannot edit my profile... maybe an admin would unlock that function for me?
the epw.in follows
I am currently trying to reproduce the ncomms11755 paper "Electron-phonon coupling in hybrid lead halide perovskites" by Wright, Verdi et al. by using CsPbBr3 (in the cubic 5-atoms unit cell).
22 filled and 22 empty bands were used for the nscf step on 216 k-points (6x6x6 MP). Phonon dispersion comes out nicely and calculated Raman peaks match the experiment very well.
I worked out the valence bands just fine and using a lower number of random points on the fine grid (sufficient for my purpose) values agree nicely. But for some reasons the conduction bands are not working out for me. The wannier90 interpolated bands agree very well with the ones calculated from QE and the spread of each MLWF is between 1-2 A.
For the valence bands the el-ph calculation on a 12x12x12 fine k/q grid took about 2 days (I only have 6 processors/16 GB ram on the testing node) but for the CB after 2 days the calculation is still crunching numbers on the first irreducible q-point. In detail, it shows
Code: Select all
1 42 42 T T
1 43 43 T T
1 44 44 T T
1 45 45 T T
1 46 46 T T
1 47 47 T T
1 48 48 T T
Estimated size of gmap: ngxx =14374
whilst epw.x is running at full steam on all cores. I have browsed the forum and saw that random q/k points were recommended at some point so I tried that one next, but the results were similar (here, the calculation progressed to the 3rd point and remained there for a day or so). Is it significantly more difficult to calculate the CB compared to the VB (assuming the wannier interpolation was equally successful for both)?
here is my poor-man's version of the el-ph Im(Sigma) (meV) overlayed on the (DFT) band-structure: https://drive.google.com/file/d/0B4jAwk ... sp=sharing
any comment from the experts is, as always, greatly appreciated!
With best regards,
Chris
PS: I cannot edit my profile... maybe an admin would unlock that function for me?
the epw.in follows
Code: Select all
--
&inputepw
prefix = 'cspbbr3'
amass(1)=207.20000,
amass(2)=132.90500,
amass(3)=79.90400,
outdir = './'
iverbosity = 1
num_iter = 1000
elph = .true.
ep_coupling = .true.
kmaps = .false.
epbwrite = .true.
epbread = .false.
epwwrite = .true.
epwread = .false.
etf_mem = .false.
lpolar =.true. !large LO-TO splitting at G!
nbndsub = 17
nbndskip = 22 !skip the VBs
wannierize = .true.
num_iter = 1000
dis_win_max = 12.00
dis_win_min = -6.0
dis_froz_min= -6.0 !this excludes Pb:d states
dis_froz_max= 5.0 !Ef=4.8357 so states below are frozen
proj(1) = 'Br:p'
proj(2) = 'Pb:s;p'
proj(3) = 'Cs:p;s' !with this projections spread ~1-2A
wdata(1) = 'bands_plot = .true.'
wdata(2) = 'begin kpoint_path'
wdata(3) = 'G 0.00 0.00 0.00 X 0.00 0.50 0.50'
wdata(4) = 'X 0.00 0.50 0.50 M 0.50 0.50 0.00'
wdata(5) = 'M 0.50 0.50 0.00 G 0.00 0.00 0.00'
wdata(6) = 'G 0.00 0.00 0.00 R 0.50 0.50 0.50'
wdata(7) = 'R 0.50 0.50 0.50 X 0.00 0.50 0.50'
wdata(9) = 'end kpoint_path'
wdata(10) = 'bands_plot_format = gnuplot'
elecselfen = .false.
phonselfen = .false.
a2f = .false.
specfun = .false.
wmin_specfun = -4
wmax_specfun = 1
nw_specfun = 20
parallel_k = .true.
parallel_q = .false.
fsthick = 3.0 ! eV
eptemp = 300 ! K (same as PRB 76, 165108)
degaussw = 0.1 ! eV
degaussq = 0.1 ! meV
efermi_read = .true.
fermi_energy = 4.8357 !from the nscf calculation
dvscf_dir = './save'
filukk = './cspbbr3.ukk'
nqf1 = 12
nqf2 = 12
nqf3 = 12
!rand_q=.true.
!rand_k=.true.
! rand_nq= 1000
!rand_nk= 1000
nkf1 = 12
nkf2 = 12
nkf3 = 12
!filkf='cspbbr3.qpt'
!filqf='cspbbr3.qpt'
nk1 = 6 !same or multiple of nq
nk2 = 6
nk3 = 6
nq1 = 6
nq2 = 6
nq3 = 6
/
20 cartesian
0.000000000 0.000000000 0.000000000 1
0.000000000 0.000000000 0.166666667 1
0.000000000 0.000000000 0.333333333 1
0.000000000 0.000000000 -0.500000000 1
0.000000000 0.166666667 0.166666667 1
0.000000000 0.166666667 0.333333333 1
0.000000000 0.166666667 -0.500000000 1
0.000000000 0.333333333 0.333333333 1
0.000000000 0.333333333 -0.500000000 1
0.000000000 -0.500000000 -0.500000000 1
0.166666667 0.166666667 0.166666667 1
0.166666667 0.166666667 0.333333333 1
0.166666667 0.166666667 -0.500000000 1
0.166666667 0.333333333 0.333333333 1
0.166666667 0.333333333 -0.500000000 1
0.166666667 -0.500000000 -0.500000000 1
0.333333333 0.333333333 0.333333333 1
0.333333333 0.333333333 -0.500000000 1
0.333333333 -0.500000000 -0.500000000 1
-0.500000000 -0.500000000 -0.500000000 1