Diffusion through different grain boundaries (GBs) in thin films is analyzed by solving the diffusion equations numerically. Supposing a bimodal grain structure, i.e. boundaries with low and high diffusivities, the thin film diffusion problem is solved using the following boundary conditions: one surface is maintained at unit concentration, at the back surface zero flux or fast surface diffusion and the formation of a secondary diffusion source is supposed. In the case of a C kinetic regime, the contributions of the fast and slow boundaries can be separated: the appearance time of the secondary diffusion source is determined by the high diffusivity boundaries, on the other hand, the diffusion broadening of the region in which the grain boundaries are saturated is controlled by the slow GBs. Similarly, the out-diffusion from the secondary diffusion source is also controlled by the slow GBs. From the time evolution of the concentration profiles measured experimentally in the Ta/Co/Si system, Si diffusivities are determined by fitting our numerical solutions. Our results underline the importance of the fitting procedure proposed in this article, i.e. the fitting of the central part of the film can be free of errors due to instrumental broadening around the interfaces. (C) 2011 Elsevier B.V. All rights reserved.