It has been known for over thirty years that hydrogen can be produced directly from sunlight and water by making use of a photo-electrochemical cell. The most promising material for the key component of the cell, the photo-anode, is titanium dioxide, largely on account of its exceptionally high corrosion resistance and reasonable electronic conductivity. However, because bulk titanium dioxide has a band gap of 3eV (the optimal band gap is approximately 2eV) the cell's efficiency (roughly 2%) is much too low for commercialization (estimated to require an efficiency of about 10%). Much research has been directed to finding ways to reduce the band gap of titanium dioxide whilst retaining its corrosion resistance and basic stability. In this paper, the various strategies for reducing the band gap of titanium dioxide are reviewed including cation and anion doping, departure from the stoichiometric composition, particle size and shape, near surface segregation and defect disorder are reviewed and evaluated. The special role of multi-scale computations to direct the experimentation and the need for a multi-variate approach to optimize the required properties are specifically commented on.