Dear all and EPW developers,
we are using IFCs obtained from an SSCHA minimization for our polaron calculations via lifc = .true (EPW 5.9).
The electron–phonon matrix elements after running with wannierize = .true. and lifc = .true. look fine. We checked the trace of |g_mn nu|^2 along a high-symmetry q-path for k = Γ and k = R. The dispersion from phband.freq also perfectly matches the original phonon dispersion we plotted along the same high-symmetry path.
However, the polaron formation energies as a function of cell size (especially for a small hole polaron) show an erratic behavior instead of the usual smooth progression. As a result, the extrapolation becomes unreliable for the small polaron.
A similar fluctuating behavior is seen for a large electron polaron, but in that case the formation energy progression becomes smooth for larger cell sizes and in the end enters the expected linear regime.
Increasing the cell size further did not improve the situation for the small polaron. To give an idea of the issue’s severity: a linear fit of the formation energies for the larger cells yields an uncertainty of about 0.25 in the slope.
When using the same material but with DFPT phonons obtained from QE and lifc = .false., the small-polaron formation energies show a well-behaved linear regime. Although, formation energies calculated for even and odd cell sizes converge separately, respectively.
As a cross-check, we computed the IFCs from the DFPT phonons and used them with lifc = .true.. Against our expectations, this led to unusable electron–phonon matrix elements showing divergences at multiple high-symmetry points. Subsequent polaron calculations then again produced a non-smooth formation-energy progression (as expected for such problematic matrix elements).
Is there anything that requires special attention when using lifc = .true. for Wannierization and polaron calculations?
What could be the source of this problem, and are there any additional checks we can perform?
Best regards,
Markus
Erratic polaron formation energies when using IFCs from SSCHA with lifc = .true.
Moderator: stiwari