“…(69) We observe that in the long-wavelength approximation, when ω 1 (k → 0) ω p , this expression (69) reduces to the generalized Drude-Lorentz model (19) if we choose Q 1 (k,w) = Q 1 (0,ω) = iν(ω) and that this model is unable to incorporate the asymptotic form (8), i.e., we have that −1 MM1 (k,w → ∞) 1 + ω 2 p /w 2 + · · · . Certainly, for a constant collision frequency or with Q 1 (k,w) = Q 1 (k,w = 0) = ih(k), h(k) > 0, the fact that the loss function L 1 (k,ω) = − Im −1 MM1 (k,ω)/ω has finite moments {C 0 (k),0,C 2 } can be easily checked by direct (analytic) integration.…”