Thank you Santiago! It shows already a feature similar to a Drude model in the red and infrared region.

Thank you for the suggestion! I did a couple of calculations for dielectric function for Au and Ag these days, after volume optimization and k-grid optimization. The results I got are compared with dielectric functions from Johnson and Christy(1972 http：refractiveindex.info) . I found that the interband features of simulations are generally redshifted compared with experiments. First question is can this discrepancy be minimized in some way by adjusting parameters?

To include intraband features, I have used the q points as the following. The imaginary part is significantly larger than experiments. My second question is how should I chose q points for different metals to study the intraband transitions? Thanks you very much in advance!
<qpointset>
<qpoint> 0.0 0.0 0.001 </qpoint>
<qpoint> 0.0 0.001 0 </qpoint>
<qpoint> 0.001 0.0 0.0 </qpoint>
<qpoint> 0.001 0.001 0.001 </qpoint>
<qpoint> 0.0 0.0 0.002 </qpoint>
<qpoint> 0.0 0.002 0 </qpoint>
<qpoint> 0.002 0.0 0.0 </qpoint>
<qpoint> 0.002 0.002 0.002 </qpoint>
<qpoint> 0.0 0.0 0.005 </qpoint>
<qpoint> 0.0 0.005 0 </qpoint>
<qpoint> 0.005 0.0 0.0 </qpoint>
<qpoint> 0.005 0.005 0.005 </qpoint>
<qpoint> 0.0 0.0 0.01 </qpoint>
<qpoint> 0.0 0.01 0 </qpoint>
<qpoint> 0.01 0.0 0.0 </qpoint>
<qpoint> 0.01 0.01 0.01 </qpoint>
</qpointset>