I've a question about the convergence parameters of the G0W0 lifetimes (no on-mass shell approximation).
I attached you a typical input-file.
At the moment I try to converge the following parameters to the following limits:
- -FFTGvecs, EXXRLvcs -> "infinity"
-NGsBlkXd -> "infinity" #number of RL for the Kernel; usually INcreases the lifetiems?
-BndsRnXd -> "infinity" #number of bands for X; usually DEcreases the lifetiems?
-GbndRnge -> "infinity" #number of bands für G; usually DEcreases the lifetimes?
-GDamping -> "zero" # Gamma in W?; usually INcreases the lifetiems? I had problems when using the PPA, as decreasing GDamping perfectly scaled (linearly, decreasing GDamping by an arbitrary factor gave me the original lifetime times this factor) so that I was not able to find a converged value. Without PPA this was possible. Is this a commonly known fact that PPA is not suitable for lifetimes?
I'm a little bit confused about the influence of the following parameters:
-DmRngeXd: This parameter heavily increases the lifetimes when setting to a smaller value. Furthermore I thought that taking more bands into account will decrease the lifetimes because of additional possibilities to scatter. However, setting this parameter to 0.0001 this is not true any longer - shouldn't I touch this parameter?
-dScStep: Setting this parameter to a smaller value (i.e. 0.005) changes the lifetimes in a strange way - suddenly the lifetime of some holes is negative while I thought that only electrons will have a negative lifetime in the conduction bands? Or did I change this parameter in the wrong direction - so do I have to increase this parameter?
Thank you very much for some comments in advance and kind regards,
Andreas Lücke