Comment on "Efficient Photochemical Water Splitting by a Chemically Modified n-TiO2" (II)

+ See all authors and affiliations

Science  19 Sep 2003:
Vol. 301, Issue 5640, pp. 1673
DOI: 10.1126/science.1085119

Khan et al. (1) recently reported quite extraordinary results regarding a new carbon-doped rutile photoanode for the photosplitting of water and conversion of photon energy to chemically stored energy in the form of hydrogen (1). The article was commented upon as “a good lead in a good direction” with respect to the U.S. Department of Energy's 10% benchmark in efficiency for a commercially viable catalyst (2).

Efficiency is, together with material stability, a major issue for candidate water photolysis materials to be used for the highly desirable and important goal of economical solar hydrogen production, and we acknowledge that modification of TiO2-based materials in the direction proposed by Khan et al. is of considerable interest. This comment, however, concerns the reported 8.35% efficiency for the solar energy conversion system, which, based on the data presented by Khan et al., we believe is misleading and shows an insufficient concern on the part of the study's authors about the spectrum of real terrestrial sunlight.

Considering a chemically unbiased photo-electrochemical system, producing hydrogen from water through the half-cell reaction 2H++2e→H2, the conversion efficiency can be determined from the magnitude of the voltage externally applied to support the reaction, Eapp and the spectral dependence of the quantum yield ϕ(λ,Eapp)—that is, the number of electrons participating in the hydrogen producing reaction per incident photon at wavelength λ. A quantitative estimate of the conversion efficiency should be calculated for an appropriate spectral distribution of photon flux density Nph(λ), such as that obtained from the widely used global air mass 1.5 (AM 1.5) solar spectrum (3) referred to by Khan et al. The conversion efficiency is then given by $Math$ $Math$(1) where E0rev=1.23 V is the standard-state reversible potential for the water-splitting reaction, e is the elementary charge, and PAM1.5 is the total power density of AM 1.5 sunlight, normalized to 100 mW/cm2 (3).

Because the quantum yield and its spectral dependence (the action spectrum) for the chemically modified n-TiO2 (CM-n-TiO2) material investigated by Khan et al. were not reported, we derived an ultimate limit to the conversion efficiency by assuming a 100% quantum yield up to the declared absorption threshold at 535 nm. Performing the integration in Eq. 1 up to this threshold by means of the trapezoidal rule (3) with the tabulated global AM 1.5 data, the efficiency top limit at 0.3 V electrode bias was found to be 8.1%, which is lower than the 8.35% efficiency reported by Khan et al.

An average quantum yield even close to 100%, however, is not realistic. To find a better estimate, we used the optical absorption spectrum reported for the CM-n-TiO2 (Fig. 1). Although the resulting calculated efficiency of 2.7% is subject to uncertainty, the large deviation from the value reported by Khan et al. calls for an explanation.

Khan et al. appear to have compared the overall intensity of the experimental light source with that of standard solar illumination, without compensating for the differing spectral photon flux distributions (1). Failure to take the spectral photon distribution into account generally produces misleading efficiency numbers for real solar applications—color really matters. Overestimation of the efficiency results when the laboratory lamp used in experiments emits a photon distribution that differs from the real solar spectrum, in such a way that it favors the material being studied. This is frequently the case for ultraviolet- to visible light–absorbing materials and infrared-filtered emission from xenon arc lamps.

In an earlier work by Khan and Akikusa (4), a strikingly similar preparation of n-TiO2 was characterized under very similar conditions as those in the more recent paper (5). Using the action spectrum measured at –0.24 volt/SCE (saturated calomel electrode) and the corresponding photocurrent density of about 3 mA/cm2 presented in this earlier work, we estimated the current generation efficiency of the 150-W xenon arc lamp to be at least six times higher than that of standard AM 1.5 sunlight (6). Additional infrared light filtering argues for an even higher spectral mismatch in the work reported in (1), and we are therefore inclined to believe the reported value should be reduced by at least a factor of six to obtain the true AM 1.5 solar conversion efficiency of the CM-n-TiO2, which implies a real efficiency below 1.4%.

In summary, we have estimated the solar conversion efficiency of the reported CM-n-TiO2 material in three different ways, each of which results in values below or far below the reported value. Efficiency is of fundamental importance for the viability of the material in real solar applications. Color matters.