Dear Dr. Yimin Ding,
As explained by Daniele, the Drude model in some versions of yambo has been looked when combined with PPA. I am not sure if it is available in the version you are using, however the variables you need to add in your GW/PPA input in order to check it are:
PPAPntXp
DrudeWXp
note the last character 'p' stands for PPA, while within the full frequency approach it is replaced by 'd'. You can ignore any possible warning and check if the inclusion of the Drude term has any effect in the GW quasiparticles.
Just to point out some features of GW calculations for metals in yambo, they need two contributions to the polarizability, coming from intra and inter-band transitions. In a standard calculation (e.g. without the inclusion of the Drude model), both contributions are computed for finite q, but in the limit q -> 0 only the inter-band term is numerically evaluated and the Drude model is a simple way to add the missing intra-band limit. There are two practical consequences: first, the effects of the intra-band limit are not always relevant, for example the contribution of the inter-band limit could be much more larger than the intra-band one and it can be ignored (inter+intra ~ inter), for instance this is the case of Al. Second, even when the intra-band limit is important, the results can be improved systematically by increasing the number of k-points, which results in the smaller finite q closer to q=0, of course, this convergence is cumbersome as mentioned by Daniele and is preferable to use other approaches like the Drude model in order to accelerate it.
So, I would suggest you trying to evaluate the importance of the intra-band limit in each of your systems. If the Drude model is not available with PPA, you can try it with the full frequency approach, as suggested by Daniele. This is a much more expensive method and there will be an efficient alternative, called the multipole approach (MPA), in new releases of yambo (
https://link.aps.org/doi/10.1103/PhysRevB.104.115157). Fortunately, to test the effects of the intra-band limit there is not need to use converged GW parameters, you can set low fixed values for the energy cutoff and number of bands in the polarizability, and change only the number of k-points. The difference in the computed quasiparticles with and without the Drude model are more or less independent of the fixed parameters and can be use as an estimate of the error in your fully converged calculations.
Regarding the case of metals for which there is not an available Drude frequency, the intra-band limit can be numerically evaluated at the independent particle or the RPA level, as suggested by Daniele, for instance by integrating eq. 17 of
https://link.aps.org/doi/10.1103/PhysRevB.64.195125. You can also read these other papers:
https://doi.org/10.1103/PhysRevB.77.035117
https://doi.org/10.1103/PhysRevB.82.035104
https://doi.org/10.1103/PhysRevB.86.035120
In the near future there may be other alternative approaches to address the intra-band limit in yambo. I can only comment on the ones I have worked on. There will be a dedicated paper for metals particularizing the MPA method for this kind of systems and proposing a new simple and general ab initio method to evaluate the intra-band limit.
I hope I have been somehow helpful.
Best regards,
Dr. Dario A. Leon
Norwegian University of Life Sciences