Island coalescence during the early stages of polycrystalline film formation is thought to lead to the large values of tensile stress observed in experiments. Continuum models have been developed which are in semi-quantitative agreement with in-situ experimental observations of stress evolution during coalescence. However, the size of coalescing particles is in the nanometer range, so that atomistic treatments might be expected to more accurately reveal the mechanisms leading to coalescence stresses. We have performed molecular dynamics (MD) computational experiments in which small metallic clusters both in free space and on weakly bound substrates are allowed to coalesce. These atomistic simulations complement the finite element modeling (FEM) and analytical work on the tensile strains developed during grain boundary formation by coalescence. Additionally, the atomistic simulations implicitly capture the effects of anisotropy and special orientations. The MD simulations also provide atomic-level detail of the boundary structure that is absent in continuum models. We have focused on simulations of Ag as a high mobility material at and above room temperature and have begun studying Ni as a material that can behave either as a high or low mobility material depending on the deposition temperature(i.e. Ni has a higher activation energy for adatom motion than Ag.). We find that for small islands, the initial boundary formation occurs in times on the order of nanoseconds. For nanometer-sized clusters in free space and on traction-free substrates, the cluster can undergo large rotations prior to boundary formation. The simulated boundary heights fit well with both a FEM model and an analytical model. Work is ongoing to further quantify the results of the simulations and relate the atomistic forces and positions to the stress state. We are pursuing methods for including the effects of traction at the interface, both through continuum and discrete methods.