BUG: ANADDB electron-phonon coupling, LOVA with use_k_fine

MULTIBINIT, aTDEP, ANADDB, optics, cut3d, mrg* postprocessors

Moderators: MMNSchmitt, gonze

Locked
etea
Posts: 2
Joined: Mon Mar 24, 2014 9:38 am

BUG: ANADDB electron-phonon coupling, LOVA with use_k_fine

Post by etea » Tue Dec 13, 2016 6:34 pm

Hello,

There is a bug in the LOVA transport calculation in ANADDB. The bug only applies to the use of a denser mesh for electron velocities (use_k_fine).

In the file 77_ddb/integrate_gamma_tr_lova.F90, the transport functions are calculated as a sum over (k,k') points and (m,n) bands involving the electron-phono coupling elements and Fermi velocities. When using the denser mesh for electron velocities (use_k_fine), some arrays whose size normally follows kptrlatt, account for the finer mesh kptrlatt_fine. These include the velocity table (elph_tr_ds%el_veloc). The transport sum is performed over the coarse k-point grid (ikpt_phon), but the velocity table is defined on the fine grid. So, a table mapping the coarse grid to the fine grid is missing.

Attached is a modified file with the quick fix: allocate a table "kphon2kfine", and make the mapping onto fine velocities "ikpt_phonv,ikpt_phonqv".
I worked on abinit-7.10.5. The only difference with abinit 8.0.8 concern the ABI_STAT_ALLOCATE and ABI_CHECK routines, so the attached correction can easily be ported to abinit 8.0.8.

Best,
Eric
Attachments
integrate_gamma_tr_lova.F90
modified ANADDB source file from ABINIT 7.10.5
(8.31 KiB) Downloaded 297 times

Locked