EPW Forum

Dear developer,
I use EPW to calculate the hole polarons in Ga2O3. While the psir_plrn.xsf file given a wrong polaron condition. In my calculation, the p orbitals of O are used to do wannierization. And then, the polaron is viewed in Vesta. However, the vesta shows that the polaron is localized at two Ga ions rather than O ions. More interestingly, the number of Ga ions is 1.5 times that of O ions, which means that the psir_plrn.xsf exchanged the atom index. The important tags in the input files of EPW are listed below:
****************EPW1.in
elph = .true.
epwwrite = .true.
lpolar = .true.
nbndsub = 18 ### valence bands
bands_skipped = 'exclude_bands = 1:26,45-64'
wannierize = .true.
num_iter = 500
iprint = 2
proj(2) = 'O:p'
wannier_plot = .true.
wannier_plot_supercell = 6 6 6
nk1 = 6
nk2 = 6
nk3 = 6
nq1 = 2
nq2 = 2
nq3 = 4
band_plot = .true.
filkf = './kpoints'
filqf = './kpoints'
/
*****************EPW2.in
elph = .true.
epwread = .true.
lpolar = .true.
nbndsub = 18
plrn = .true.
restart_plrn = .false.
type_plrn = 1
init_plrn = 1
init_sigma_plrn = 2.0
niter_plrn = 500
conv_thr_plrn = 5E-4
ethrdg_plrn = 1E-5
nk1 = 6
nk2 = 6
nk3 = 6
nq1 = 2
nq2 = 2
nq3 = 4
nkf1 = 4
nkf2 = 4
nkf3 = 4
nqf1 = 4
nqf2 = 4
nqf3 = 4
/
****************EPW3.in
nbndsub = 18
plrn = .true.
restart_plrn = .true.
type_plrn = 1
cal_psir_plrn = .true.
interp_Ank_plrn = .true.
interp_Bqu_plrn = .true.
filkf = './kpoints'
filqf = './kpoints'

Dear jack,

From your input file I see that you are using a 6x6x6 k- and a 2x2x4 q-grid. The k- and q- grids need to be commensurate, and these are not, so epw1 should print an error message (as I have reproduced using your file). Please use commensurate k- and q- grids (e.g. 6x6x6k and 2x2x2q, and probably denser ones for convergence).

I have re-run the calculation with a 6x6x6k- 2x2x2q- grid, and I obtain a polaron whose wave function is formed by two p-like orbitals in two neighbor O atoms (similar to Fig.1 c here: https://pubs.acs.org/doi/10.1021/acs.jpcc.5c01741). I have also checked that the number of atoms coincide with N_{Ga} = nkf1xnkf2xnkf3x4 and N_{O} = nkf1xnkf2xnkf3x6.

If you keep encountering this issue, please share the psir_plrn.xsf file so that we can try to debug.

Best,
Jon

Dear Jon,
Thank you for your response. I find the reason why the polaron is located at Ga ions. In my input files of pwscf, the atom are randomly listed as below.
************
O 0.1684592136 0.1684592136 0.1091891520
O 0.4964570539 0.4964570539 0.2544974211
O 0.8275544503 0.8275544503 0.4396017480
Ga 0.9086276961 0.9086276961 0.2040510375
Ga 0.6592106719 0.6592106719 0.3139138852
O 0.1724455497 0.1724455497 0.5603982520
O 0.5035429461 0.5035429461 0.7455025789
O 0.8315407864 0.8315407864 0.8908108480
Ga 0.3407893281 0.3407893281 0.6860861148
Ga 0.0913723039 0.0913723039 0.7959489625
************
In this order, the polaron is located at Ga ions.
I rearranged the order of atoms as below, and obtained polaron similar to yours. The EPW I used is downloaded from gitlab, and it works well at a 6x6x6 k- and a 2x2x4 q-grid.
**********
O 0.1684592136 0.1684592136 0.1091891520
O 0.4964570539 0.4964570539 0.2544974211
O 0.8275544503 0.8275544503 0.4396017480
O 0.1724455497 0.1724455497 0.5603982520
O 0.5035429461 0.5035429461 0.7455025789
O 0.8315407864 0.8315407864 0.8908108480
Ga 0.9086276961 0.9086276961 0.2040510375
Ga 0.6592106719 0.6592106719 0.3139138852
Ga 0.3407893281 0.3407893281 0.6860861148
Ga 0.0913723039 0.0913723039 0.7959489625
********
At last, I want to ask whether the pseudopotential can influence the formation energy of polaron.
Thank you
Best

Dear jack,

Yes, the choice of the pseudopotential might affect the electronic structure, phonons, and the electron-phonon matrix elements; as a result it might indirectly influence the polaron formation energy as well. However, the differences should be quantitatively small, and no qualitative changes are to be expected, if the pseudopotentials are reasonably accurate.

It is always good to check that the pseudos you are using are accurate and transferable, and that the wave function and charge density cutoff are properly converged accordingly.

Best,
Jon Lafuente-Bartolome