As the emphasis on carbon-free energy sources grows, so does the pressure to produce higher efficiency renewable and non-polluting energy harvesting methods. In recent years, photoelectrochemical water splitting has become an attractive method for generating hydrogen directly from water using solar energy. This is because, in principle, no such carbon is involved in the combustion process, with the only product being water. However, device development has been slowed by issues concerning the properties of the electrode semiconducting materials. Such electrode materials must simultaneously satisfy a crucial set of criteria to be considered economical. To date, no system has satisfied all of these criteria simultaneously. Such a multivariate problem is ideally suited for study via a computational materials screening and/or inverse materials design procedure.

Here, three examples of low-cost computational screening procedures that consider these criteria are investigated. Firstly, a low-cost means to screen the bulk component of the ionisation potential, based on past work by Frensley and Kroemer, was assessed in light of modern computational tools. Secondly, the formation of self-trapped small polarons was considered within a high throughput framework. Thirdly, special quasi-random structure models were used to rapidly predict the homogeneity and catalytic activity of a range of rutile oxide alloys. Additionally, each of these three examples was reviewed and summarised in terms of five suggested criteria for an ideal computational materials screening procedure.

It is hoped that the results and discussion of this work will provide the reader with an understanding of the advantages and limitations of these computational materials screening tools. These tools may be used in a modular fashion, with or without machine learning protocols, to accelerate the prediction of suitable candidate electrode materials for photoelectrochemical cell development.