There's been a lot of news about artificial photosynthesis lately (here and here), and while I think it's probably more useful to generate electricity via solar photovoltaic and solar thermal technology, there will no doubt be some cases where it'll make sense to use the sun to split water into hydrogen and oxygen directly.
Researchers at the Lawrence Berkeley National Laboratory has shown what they call the "first fully integrated nanosystem for artificial photosynthesis".
A key to the system is its structure on the nanoscale:
“Similar to the chloroplasts in green plants that carry out photosynthesis, our artificial photosynthetic system is composed of two semiconductor light absorbers, an interfacial layer for charge transport, and spatially separated co-catalysts,” says Peidong Yang, a chemist with Berkeley Lab’s Materials Sciences Division, who led this research. “To facilitate solar water- splitting in our system, we synthesized tree-like nanowire heterostructures, consisting of silicon trunks and titanium oxide branches. Visually, arrays of these nanostructures very much resemble an artificial forest.” (source)
But while it works, the system so far only has a sunlight-to-fuel efficiency of 0.12%. That's not quite as low as it might seem because natural photosynthesis isn't that efficient to begin with, but to be commercially viable the efficiency would have to be much higher. Maybe that's possible, but it'll be hard to compare with solar PV and solar thermal which already have efficiencies that are orders of magnitude better (see this chart), so it could actually make more sense to generate electricity and then produce hydrogen via electrolysis rather than use artificial leaves.
Via Berkeley Lab