abinit / abipy

Open-source library for analyzing the results produced by ABINIT
http://abinit.github.io/abipy
Other
117 stars 96 forks source link

New version of ConducWork #210

Closed Olivier7017 closed 3 years ago

Olivier7017 commented 3 years ago

Summary

New version of ConducWork

Olivier7017 commented 3 years ago

Thanks for the reply, I've done a new commit to follow your recommandations. In particular,

Notes :

gmatteo commented 3 years ago

Added boxcutmin and mixprec, with default value of respectively 1.1 and 1. Removed the whole DVDB interpolation step to make sure my implementation works in the next months Removed symsigma=1 since it is not necessary Changed variable name DDB to ddb_path, DVDB to dvdb_path and withKerange to with_kerange. Solved the merge conflict

Ok, many thanks for the modifications. :

I suppose boxcutmin and mixprec are only meant to be used in the conductivity task.

These variables can be used in every eph_task that needs to compute H1|Psi0> where H1 is the DFPT Hamiltonian. so they affect eph_task 1, 4, -4 and, to some extent, the numerical values stored in the netcdf files (e.g. SIGEPH.nc) The calculation of mobilities/conductivities done in the rta_routine is essentially a post-processing of the data stored in the SIGEPH.nc file. In a nutshell, rta_driver does not use mixprec and boxcutmin but it will use lifetimes computed with these two variables.

I used a dict for the dependency since I couldn't make it work as a list. The following line didn't work deps=[(DDB, "DDB"), (DVDB, "DVDB")]

My bad. deps is indeed a dictionary so

deps={new[taskNumber-1]: "WFK", phwork: ["DDB","DVDB"]}

is fine (and you don't need to import Dependency directly)

Thanks for contributing to the project. I'm gonna merge this PR.