As the field of dilute magnetic semiconductors continues to thrive in the slums of caveat emptor calculations, it appears that intermediate band gap solar cells are following a similar fate.
Building upon much earlier work in the field, in 1997 Luque and Marti calculated that introducing an intermediate band gap state (IB) into a semiconductor could offer a means to achieve solar conversion efficiencies of up to 63.1%, well beyond the single-junction limit. This could be achieved in many ways, but one of the most straightforward is through impurity doping.
The principal assumption of the high efficiency is that there exists a half-filled, localised defect band 0.7 eV above the valence band (as it is half-filled, valence -> IB and IB -> conduction band optical transitions are independent). Generally, a localised half-filled band will be subject to a Peierls distortion, so it is not an easy task. A perfect case for electronic structure theory to lead the way...
One of the early disasters was doping of the III-V materials GaAs and GaP, e.g. an excellent IB was predicted when Ti is substituted on the *anion* site (over a series of papers!). Much of the subsequent work focused on transition metal doping and alloying of oxides, II-VI and III-V compounds; however, it is well demonstrated that standard local/semi-local density functionals cannot deal with transition metals, especially when you are interested in quantitative band positions and optical absorption properties. In particular, it is difficult to see how a partially occupied 3d band could be stable with respect to a Jahn-Teller distortion. The same issue applies to recent work on Ti and Cr doping of the Ga site in CuGaS2, where the results will be highly sensitive to the treatment of the 3d states and to the form of ionic charge compensation.
The main danger here is that experimentalists read these papers, get their hopes up, and ultimately lose respect for the predictive role of material simulation.