Why is there an optimal thickness for doped nanocrystalline silicon layers in heterojunction solar cells, and how can it be predicted?