Metals present special challenges to computation and the problems are directly related to the plasmon modes they support. In a dielectric insulator details of the electromagnetic fields are confined to the scale of the wavelength of light in the dielectric. This limit is very helpful when planning a computation particularly if we use a real space description of the fields. Insulators are easy targets for computation in another respect: their dielectric functions tend to be independent of frequency and therefore we are free to use a variety of techniques including time domain methods.
Metals belong to the awkward squad when it come to calculation. The length scale of the wavefield can vary from the wavelength of light in free space down to the smallest detail of the metallic structure. Extreme concentrations of radiation into a few cubic nanometres are not uncommon and must be described in detail if an accurate overall response is to be calculated.
To handle these difficulties the transfer matrix method is commonly used as it computes at a fixed frequency. Hence dispersion of the dielectric function is not a problem. Another technique that is employed for metals is the adaptive mesh whereby the grid of points on which the field is specified expands or contracts according to the details expected in the fields. As it happens there are particularly easy ways of implementing this potentially tricky technique without rewriting computer code. Finally it is possible to adapt the 'FDTD', finite difference time domain, methods to work for dispersive materials whilst retaining the advantageous 'order N' scaling of this methods with system size.