Hi I am interested in modeling the optical properties of semiconductor nano-clusters. By nano-clusters, I mean crystalline structures of a binary semiconductor with varying numbers of atoms ranging from a few tens to more than a thousand atoms. I have gone about this by taking the equilibrium structure and generating nanoclusters either by editing by hand or using the nanocluster generation tool in Materials Studio. To model the structure and optical properties of these clusters, I have placed them into a large (periodic) vacuum box allowing for approximately a nanometer between cells to allow the wavefunctions to fall off before the next peridoic cell starts. I am then allowing the structures to relax at 0K using the BFGS algorithm and DFT in CASTEP. I am curious to know the optical properties of the clusters as well so I have computing these using the optical constant calculation flag in CASTEP. While I find that the free energy gradually falls with increasing cluster size as I would expect, I have also found that the optical constants, in particular the absorption appear to fall with increasing cluster size. I am normalizing the optical constants by the density of the computational cell to account for the presence of vacuum -- to get an "average" optical response. I am worried that my observed fall in both n and k is not physical and that this procedure is flawed in some way and am looking for comments as to what is correct or incorrect about this procedure. I have made clusters in several different shapes, but so far the trend of decreasing optical constant with size seems to be intact. Of course this could stem from my assumption that renormalization of the derived optical constants using the density is flawed, but I do not understand why this would be the case. My question is thus: Is the above procedure a valid way to calculate the optical constants for clusters correct? Or perhaps is it necessary to use a more sophisticated approach such as TDDFT to model the optical properties accurately? Any thoughts would be greatly appreciated. Thanks in advance.