In recent years, research on epitaxial growth for photovoltaics became more important due to the increasing interest in thin-film silicon solar cells. Two significant challenges need to be resolved before this technique can become a competitive industrial alternative to the current dominating technology of bulk silicon solar cells: (i) the availability of a high-throughput and cost-effective epitaxial CVD reactor and (ii) efficiencies approaching those of bulk Si solar cells. In this paper, two CVD systems are studied: an AP-CVD commercial reactor, as a reference system, and an experimental LP-CVD system for optimization of a low-cost semi-industrial process. For low growth rates, an LP-CVD process is realized with a defect density around 5 x 10(3) defects/cm(2), comparable with the layers grown in the commercial reactor with a growth rate of 3.9 mu m/min. First solar cells, grown in the LP-CVD reactor show an efficiency of 8.2% on mono-crystalline samples. Cells on various low-cost substrates, grown in the reference reactor, show efficiencies between 12% and 13% with IMEC's industrial screen-printing process. The short-circuit current of epitaxial cells is limited to 28 mA/cm(2) (typically 5 mA/cm(2) less than for bulk Si cells). Therefore, the thin epitaxial cell concept requires optimal light trapping, increasing the optical path length. Experiments show that a porous silicon (PS), giving an internal reflectance up to 80%. However, the intermediate layer as an internal reflector can fulfill this role adequately. effectiveness of this reflector depends on its influence on the quality of the epi-layer. Measurements show a lower-quality epi-layer for samples with a PS intermediate layer, but indicate that further optimization of the pre-deposition bake could lead to a compromise between current gain by internal reflectance and losses caused by the increased defect density. (c) 2005 Published by Elsevier B.V.