The importance of validating codes and best practice for GW calculations. A community effort involving Yambo, Abinit and BerkeleyGW provides a framework for users and developers to validate the precision of new applications and methodological improvements relating to GW codes. Plane-wave GW codes give the same results provided everything is very carefully converged but can differ dramatically when not. Innocent looking k and q sampling settings and PP models can do real damage. Our results comprise an important verification of codes using the GW method for systems in the condensed phase, showing that different implementations can agree numerically at a level much greater than the known accuracy of the GW approximation and the underlying approximate Kohn-Sham eigensystem.
The paper can be found here.