A key step in order to realise photo-electrochemical (PEC) water splitting to produce hydrogen sustainably, is reactor design. Good engineering will minimise energy losses (both optical and ohmic) due to reactor construction, whilst ensuring the H2 and O2 produced are separated, and this can subsequently relax the requirements on the photo-absorber material and/or electro-catalysts. In this paper we show that separation of the products through hydrodynamic flow alone would negate the need for the conventionally used membrane, which has an associated ohmic drop and cost. This is demonstrated to be possible using a ‘laminar flow between two parallel plates’ reactor design and AR/Pe and AR are found to be the two key dimensionless numbers that predict product cross-over (where AR, Pe are aspect ratio and Péclet number respectively). Supersaturation was used as an indicator of bubble formation, which disrupts the laminar flow required for separation and it is shown that by increasing the reactor pressure, higher current densities can be tolerated before supersaturation occurs. Removal of the dissolved hydrogen and oxygen from electrolyte is discussed. A multi-physics model, which employs an optical transfer matrix method, is used to validate the previous conclusions. Experimental data for hematite and Pt deposited on FTO was used as the anode and cathode respectively. Parasitic optical losses and efficiency improvement with stacking are shown for the example reactor configuration. Additionally, the concept of stacking this reactor design in order to absorb light in multiple passes is introduced. This approach relaxes a classical constraint on photo-absorber materials: large absorption length compared to small diffusion length of charge carriers in the semiconductor.