Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences
You have accessResearch articles

Methanol photo-reforming with water on pure titania for hydrogen production


    The behaviour of titania for the photo-reforming of methanol with water at ambient temperature has been examined. It is shown that the reactivity is very poor, compared with metal-loaded catalysts at low methanol levels in solution, but the rate becomes much higher at high methanol levels, such that the difference from metal-loaded samples is much less. The optimum yield is with approximately a 1 : 1 methanol/water solution. The reaction also proceeds well in the gas phase. During all such catalysis, the titania becomes blue, due to light absorption increasing across the range 400–800 nm. However, this does not result in visible range activity for the photo-reforming and is due to the reduction of the material in the presence of light and the formation of anion vacancies and Ti3+ centres. These anion vacancies are only very slowly re-oxidized in air on P25 titania, taking days to recover the original whiteness of the oxide. The performance of anatase, rutile and the mixed phase is compared.

    This article is part of a discussion meeting issue ‘Science to enable the circular economy'.

    1. Introduction

    Titania has traditionally been an important material in terms of its optical properties and has been added as a white pigment to paint (the main use for titania) for a long time. However, it has become of much more modern technological significance since it is now used for a wider variety of applications. Of most importance and again related to its interaction with light is photocatalysis. Titania is used for cleaning up the environment (e.g. NOx removal) via embedding in, for instance, pavements and roadways [1], for removing aqueous pollution, for cleaning window surfaces when used as a coating [2]. However, another important area for the future is its use in anaerobic conditions for producing hydrogen [36].

    One of the first papers to suggest this was the much quoted work of Fujishima & Honda [7], although they did actually show the amount of any, if any, hydrogen, just the photocurrent generation in an electrolyte. Such materials have evolved considerably over the years to give much enhanced yields of hydrogen, especially when doped at the surface with metal nanoparticles [3,810], and when used in combination with a hole scavenger, such as methanol [3,810]. There are also many reports of more complex systems which can operate in the visible region of the spectrum, though rates are usually lower than for metal-doped titania when applied over the whole solar spectrum at ground level.

    Although there has been considerable work on titania itself, it is mostly in aerobic conditions, and mainly for the uses described in the first paragraph. Much less has been done in pure inert or vacuum conditions.

    In more recent times, titania has become a material of some fascination to the surface science community, largely because it is one of the few bulk oxides which can have enough conductivity in ultra-high vacuum conditions to be imaged at atomic resolution by scanning tunnelling microscopy (STM) [1114]. Here, it is seen that reduction has significant effect on the colour of the material, its conductivity and also its surface structure.

    In the work here, we report results for the photocatalytic reforming of water with methanol in anaerobic conditions and show that titania alone is capable of carrying out this reaction, but only at high methanol concentrations. There is a very strong dependence of the rate on the methanol concentration, which is very different from the much more often reported rates for metal-doped titania. In concert with this, there are changes in the colour of the titania with the material becoming intense blue during the reaction. The relevance of these changes for the photocatalysis is discussed.

    2. Experimental

    All experiments were carried out in a 250 ml round-bottom Pyrex flask equipped with a rubber septum for sample extraction using a gas-tight syringe (1 ml, Hamilton). The required amount of catalyst, typically 150 mg, was placed in the reactor with the desired quantity of methanol and solution made up to 200 ml with MilliQ water. The initial reaction mixture was sonicated for 30 min to disperse the catalyst and provide thorough mixing of the components. The solution was then purged with argon for 30 min to remove air from the headspace and any dissolved oxygen from the system. Illumination of the reaction mixture was performed by a 150 W Xe arc lamp (LOT-Oriel) with constant stirring using a magnetic bar. Headspace samples were taken periodically (every 30 min) by gas-tight syringe and were used to calculate hydrogen concentration. Analysis was performed on a Shimadzu GC (mode 2014), containing Haysep-N and molecular sieve (Shimadzu 80–100 mesh) columns arranged in series. The reproducibility of photocatalytic data, over four sets for the same system, is a mean deviation of ± 3% of the hydrogen yield.

    Pellets of pure titania (P25, anatase and rutile) were made using a die set to compress 300 mg into a 13 mm diameter disc with approximately 1.5 tonnes of pressure. These pellets were placed into a 250 ml Pyrex round-bottom flask supported on a three-dimensional printed stick and purged with inert gas (N2, Ar or He) to remove oxygen. Depending on the test, the round-bottom flask was either empty, containing 20 ml of water or 20 ml of 50% methanol/water solution.

    3. Results

    (a) Photocatalytic reforming of methanol

    The reaction of a high concentration of methanol in water over P25 titania was measured over time and is shown in figure 1. Here, we can see that there is an induction time before hydrogen production begins, but which then increases with time and is near linear after around 150 min. It is clear that a significant amount of hydrogen is produced. We have varied the methanol concentration in solution and the dependence of hydrogen evolution rate on concentration is shown in figure 2 (the time evolution curves are shown in electronic supplementary material, figure S1). Here, a maximum in rate is observed somewhere in the region 20–50% methanol. We also compare this with the more usual catalysts used for this reaction in figures 3 and 4 and electronic supplementary material, figure S2—namely those loaded with precious metals. The major differences are that


    the rates are always higher when metal is present,


    the rates show the most difference when the methanol concentration is low, but the difference is reduced significantly at high methanol concentrations as is the case in figure 3,


    there are colour changes in the catalyst under irradiation which are quite different with and without metal loading, as discussed below.

    Figure 1.

    Figure 1. Hydrogen production as a function of time from 200ml of 50% methanol in water solution and 50mg of TiO2 (P25).

    Figure 2.

    Figure 2. Rate of hydrogen produced from 150mg of TiO2 (P25) and 200ml of various volume concentrations of methanol in water.

    Figure 3.

    Figure 3. Average hydrogen production rates in ml min−1 from 200 ml of 50% methanol in water solution. For 150 mg of catalysts TiO2 (P25), 0.5%Pd-TiO2 (P25), TiO2-anatase, 0.5%Pd-anatase and a mixture of anatase + rutile (0.075 mg of each).

    Figure 4.

    Figure 4. Hydrogen production as a function of time. A 150 mg of TiO2 (P25) and 200 ml of MilliQ water with 160 µl of formaldehyde, methanol or formic acid added. (Online version in colour.)

    We have previously shown that rates are very high for metal-loaded catalysts even at very low methanol concentrations, for which rates are immeasurably small without metal. However, for the Pd-loaded catalysts, the rate is near zero order already at very low methanol concentrations (0.02%) and increases relatively little between 0.02% and high levels [15,16]. Clearly metal loading has a very positive effect on the reaction, which is probably a combination of the extended lifetime of the excited electron by trapping on the metal particle [17,18] together with direct involvement of the metal in the chemical conversion pathway [3].

    We also examined the reaction of two different polymorphs of titania under these high methanol concentrations in figure 3, which shows that P25 is far better than either anatase or rutile individually, though the combination of the two gives significantly higher rates, as reported by others [8,19,20]. Thus, P25 (a combination of approximately 80% anatase and 20% rutile made by air pyrolysis of TiCl4) has the highest rate, while anatase and rutile physically mixed shows approximately 4× the rate which could be obtained from the sum of the individual components. Clearly, then, there is the synergistic effect of the combination which has been reported by others.

    (b) Reaction of related molecules

    A number of publications refer to various intermediates involved in methanol photoreactions on TiO2—such as formaldehyde and formic acid for instance. So, to determine whether these were endpoints for the reaction, we examined the reactions of these two molecules. The results are shown in figure 4 and what this makes clear is that the reaction rate of these molecules is significantly higher than for methanol, showing that they cannot be endpoints, but may perhaps be short-lived intermediates in solution.

    (c) Blue titania and UV/Vis spectra

    As often reported, titania can become blue in the presence of reducing agents such as methanol and other alcohols [2128], and this is indeed the case here during photocatalytic reactions with pure TiO2 (P25) with 50% methanol in water (electronic supplementary material, figure S3). It is important to note here that metal-loaded titania does not attain this blue colour during photocatalysis. To investigate this further, the blue TiO2 powder was recovered (by filtering and drying) at the end of a photo reaction and analysed by UV/Vis spectroscopy. Electronic supplementary material, figure S4 shows the UV/Vis analysis of TiO2 (P25) before reaction, after reaction, one month later (after being exposed to air) and one month later after being calcined in air (400 C for 3 h).

    We have also carried out experiments in the gas phase with the disc form of P25, as described in the experimental section. Figure 5 shows the absorption spectrum of the disc after removal from the reactor. The initial colour is shown in the inset and it is, of course, much more intense here since the titania is not diluted by solution (as it is in electronic supplementary material, figure S4). The pelleted material is an intense blue, and the UV/Vis spectrum reflects this with absorption increasing in the region between 400 nm and 800 nm. However, it is not only in the presence of a reductant that this blue colour is observed, it also occurs after exposure to the lamp in an inert gas—He or N2, as is the case in figure 5 (exposure for 1 h in N2). As can be seen this absorption diminishes over time after removal from the reactor into air, but very slowly, having much higher absorption at 800 nm than the initial white P25 even after 120 h exposure. Note that exposure of the titania in air to the lamp did not produce the blue colour.

    Figure 5.

    Figure 5. UV/Vis spectra of a 13mm TiO2 (P25) pellet after irradiation for 1 h in an atmosphere of N2. Scanned at various time intervals after being exposed to air, as indicated. The lowest curve is for P25, and the highest for 2 min after irradiation (Online version in colour.)

    Of course, it may be possible that there are very small amounts of a reductant present, and it is this which reduces the titania. We can estimate how much this might be. It is known that pure titania changes colour upon reduction [14] Both Diebold [11] and ourselves [1214] have associated this colour change with the presence of Ti3+ interstitials and reduction of the material, and have also associated it with changes of colour during reduction/re-oxidation. The blue colour is thought to be due to the presence of these interstitials at very low levels (TiO2−x, where x is less than 0.001), with higher levels resulting in the darkening of titania and the formation of black titania, at which point Magneli phases/shear planes are formed. Aona & Hasiguti [29] carried out detailed measurements by electron paramagnetic resonance and estimate that these phases appear above approximately x = 3 × 10−4. Note that the blue colour seen in figure 5, electronic supplementary material, figures S5, S7 and S8 seems to be the limit of reduction in these experiments; the sample never goes grey or black.

    So, on this basis, we can give a rough estimate of how little of a reductant is required to form a level of vacancies with, say, x = 10−4, which seems to be associated with the level we would expect for the blue colour to be present. The pellets weigh approximately 0.3 g representing 0.004 mol of titanium and O2, and twice that the number of O atoms. Thus, at x = 10−4, we have approximately 8 × 10−7 vacancies, which is the same as the amount of reductant required, if we assume 100% efficiency of reaction and removal of one oxygen atom per reductant (and it may of course be more oxygen than that per reductant, depending on the nature of the reaction). So that represents 8 × 10−7 mol of the reductant, or approximately 0.01% in the 250 ml flask. We cannot be certain that that kind of level of reductant was not present in the flask.

    In that sense, electronic supplementary material, figureS5 also shows the result from exposure of the disc form of the pellet to a gas phase methanol/water mix, and the same kind of blue colour is formed, and very quickly.

    4. Discussion

    In previous publications [3,15,16,3033], we have noted how much better metal-loaded titania is than the unloaded material for hydrogen production by photo-reforming at ambient temperature. The significance of the metal was its importance in activating methanol (and other similar molecules [10,20,32]), especially its ability to perform dehydrogenation/decarbonylation of the alcohol [33,34]. However, most of that work was carried out at very low methanol concentrations in solution and it is clear from the data here that titania becomes much better at the reaction at high alcohol concentrations. Thus at a methanol concentration of 160 µl (0.08%) there is a factor of 50 difference between titania and 0.5 wt% Pd/TiO2, whereas at 50% methanol in solution the rate difference has reduced to a factor of 3, figure 3.

    Obviously then the titania is itself capable of decomposing and photo-reforming the methanol at these high concentrations, though at the highest concentrations it becomes negative order (figure 2). It is likely from the infrared work of Arana et al. [35] that the major intermediate present at the surface is the formate under these circumstances. It is also likely that the adsorption constant on titania alone is such that the surface does not compete well with water at low solution concentrations and that the coverage increases at higher concentrations, as suggested in other work [28].

    However, the observation that a blue colour is seen after a short time of irradiation begs the question—is it only blue titania which can carry out the reaction? This is supported by the observed induction time for the blue colour during irradiation. Maybe the defects induced by the irradiation are important for the reaction. However, at low methanol, the samples still go blue after a short time of irradiation, but the rate does not increase significantly at this point. It is our feeling that the blue colour is not required for hydrogen production, as described below, and is due to very much minority species which may anyway reside in the bulk of the sample [14].

    What is the nature of this blue colour? Well, it is most likely to be due to the formation of electron–hole (E–H) pairs, but these are normally associated with a very short lifetime before recombination, with a half life in inert of the order of nanoseconds [17], or a hundred nanoseconds or so [18] for very thin colloidal films. As shown in figure 6 the lifetime of these states on P25 here is very long when formed in inert gas; it does not go blue in air or oxygen. Clearly, then, this is a different process than that reported by Tang et al. [18]. The loss appears to be in two stages, one in the first hour or so, which is rapid compared with the second, which takes place over days. Even after 5 days, the colour has not returned to the white colour of the original P25. Electronic supplementary material, figure S6 is a plot of ln(absorbance) against time (as log scale to clarify the short time change) and clearly shows the two different orders associated with this process of decolourisation. The half lives are around 200 min for the fast process and around 4000 min for the slower one.

    Figure 6.

    Figure 6. The decay of absorption in the visible as a function of exposure time to air, from the data of figure 5. (Online version in colour.)

    It is likely that the longevity of the blue colour is due to the loss of holes (O) to the gas phase as oxygen (see below), leaving anion vacancies and (probably localized electrons), as described for single crystals of titania [12,14].

    The very slow nature of this second process indicates that it is not due to normal E–H recombination since it is so slow. Our only explanation for this occurring in the presence of inert gas is that a process occurs by the electron (effectively Ti3+) remaining trapped in the solid and the hole (effectively O) is somehow lost perhaps as follows. However, as described above, we also cannot discount a very small amount of a reductant being present during the experiment, as the cause of oxygen loss from the material.

    Ti4+O2+ hvTi3++OO12O2+e
    It is likely that the latter also results in Ti3+ formation. Work on single crystals of titania [11] has shown that high temperature treatment in vacuum results in reduction and the production of interstitial Ti3+, and variations of colour from transparent pale blue to deep blue to black, depending on the extent of reduction. Although the colour can be deep, and the effect on electrical conductivity high [11,13,14], nevertheless this all occurs with very low levels of non-stoichiometry, with x in TiO2−x being very low, typically less than 0.001 [13,29]. Such defects are induced by high temperature treatment, whereas here they are produced by photo-excitation at ambient temperatures. Under these conditions, it is most likely that the colour centres are simply a combination of Ti3+ sites, and oxygen vacancies with itinerant 3d electron states of some considerable mobility in the conduction band. That they have considerable mobility is evident from investigating the discs above after irradiation. The disc is around 1 mm thick, so the rear of the disc is well outside the penetration depth of the light and yet it becomes blue relatively quickly (within 5 min) on the front and within 30 min at the back. Electronic supplementary material, figure S7 shows that the middle of the disc can still be white if irradiation is stopped after 5 min.

    Clearly then the material has to re-oxidize, presumably by the reverse of the equations above, and this is apparently very difficult on P25, hence the slow loss of colour shown in figure 5. Note, however, that when the irradiation is carried out in air the sample remains white, so presumably the loss mechanism is countered by re-ad/absorption of oxygen in the sample before it can penentrate the material to any significant degree. It is interesting to note that it is only P25 which shows this behaviour: both anatase and rutile titania also go blue (and indeed rutile is a more intense blue, see electronic supplementary material, figure S8), but when exposed to air go white extremely quickly. Like P25 they do not go blue in air.

    The nature of the two different rates of re-oxidation on P25 probably relates to two different processes taking place. The faster process is re-oxidation of the surface region, followed by a slower bulk process, perhaps limited by inter-grain and inter-nanoparticle transport, especially between nanoparticles of anatase and rutile in the sample.

    But does this blue colour have any significant influence of the reaction rate? For instance, can the photons absorbed in the visible region induce photocatalysis? The answer appears to be no, because measuring the rates with various filters in place (electronic supplementary material, figure S9 shows the absorbance profiles of the filters) shows hydrogen production only in the range <∼ 400 nm, figure 7, in other words, it is still the normal band gap excitation of the titania which is the important energy storage mechanism for driving the catalysis, even though the sample absorbs light in the visible region.

    Figure 7.

    Figure 7. Rate of photocatalytic H2 evolution from a 0.5% Pd-TiO2 catalyst made by incipient wetness with a series of long band pass filters. Reactions was performed under continuous flow operation, 30 ml min−1 N2 flow, 150 mg catalyst, 1 ml of methanol and 200 ml water. (Online version in colour.)

    In conclusion, it appears that P25 titania is active for methanol photo-reforming in water, but only at relatively high levels of methanol. Metal loading in the catalyst results in much higher rates when methanol concentration is low, but the gain is much less at high methanol concentrations, though metals do still promote photo-reforming even then. The samples go blue when metal is absent, but this blue coloration does not result in visible light activity for the materials. The blue coloration is likely to be due to the formation of anion vacancies within the material.

    Data accessibility

    There is supplementary information on the journal webpage. There are seven figures there. Electronic supplementary material, figure S1 shows the time dependence of hydrogen evolution for different mixes of methanol/water while electronic supplementary material, figure S2 shows the same, but for several different materials. Electronic supplementary material, figure S3 shows the change in colour of the catalyst during irradiation in the liquid phase. Electronic supplementary material, figure S4 shows the UV/Vis spectra of P25 titania before and after use in the liquid phase and electronic supplementary material, figure S5 shows the colour change of a titania disc during irradiation. Electronic supplementary material, figure S6 gives the time dependence of air re-oxidation of a disc, while S7 shows the cross-section of a disc to show only partial reduction during irradiation at short times and S8 displays a deep blue colour for irradiated rutile. Electronic supplementary material, figure S9 presents the absorption characteristics of the filters used in the study. The DOI for the data underpinning the article ‘Methanol photo-reforming with water on pure titania for hydrogen production’ is

    Authors' contributions

    M.B. involved in supervising and author of this article. W.J., student, helped in performing experiments.

    Competing interests

    We declare we have no competing interests.


    We are grateful to EPSRC for support under the Catalysis Hub grant (grant nos. EP/K014854/1, EP/K014714/1) and for a studentship to W.J. (EP/K014714/1) and to RCAH Harwell for facilities support.


    One contribution of 14 to a discussion meeting issue ‘Science to enable the circular economy’.

    Present address: Department of Chemistry, UCL, 20 Gordon St., London WC1H 0AJ, UK.

    Electronic supplementary material is available online at

    Published by the Royal Society. All rights reserved.