Si is an excellent absorber material for use in 2-photon photoelectrochemical hydrogen production. So far nearly all studies of silicon photoelectrodes have employed frontal illumination despite the fact that in most water-splitting 2-photon device concepts the silicon is the “bottom” cell in the tandem stack and therefore illuminated from the back with respect to the electrolyte. In the present work, we investigate back-illuminated Si photoelectrodes experimentally, as well as by modelling, the dependence of induced photocurrent on various parameters, such as carrier diffusion length (Le) and surface recombination velocity (vs) to quantify their relative importance. A bifacial light absorbing structure (p+pn+ Si) is tested under back-illumination conditions which mimic the actual working environment in a tandem water splitting device. The thickness of the absorbing Si layer is varied from 30 to 350 μm to assess the impact of the diffusion length/thickness ratio (Le/L) on photocatalytic performance. It is shown how the induced photocurrent (JL) of a back-illuminated sample increases as wafer thickness decreases. Compared to the 350 μm thick sample, a thinned 50 μm thick sample shows a 2.7-fold increase in JL, and consequently also a higher open circuit voltage. An analytical model is developed to quantify how the relative Le/L-ratio affects the maximum JL under back-illumination, and the result agrees well with experimental results. JL increases with the Le/L-ratio only up to a certain point, beyond which the surface recombination velocity becomes the dominant loss mechanism. This implies that further efforts should to be focused on reduction of surface recombination. The present study is the first experimental demonstration of a Si wafer based photocathode under back-illumination. Moreover, the comparative experimental and theoretical treatment also highlights which photoabsorber properties merit the most attention in the further development towards full tandem water splitting devices.