Photocatalytic Reforming of Glycerol to H2 in a Thin Film Pt‐TiO2 Recirculating Photo Reactor

BACKGROUND: Withanincreaseinglobalbioenergyproduction, the ‘biorefineryconcept’hasnowbecomeasignificantfocusofresearch. Thedesire toachieveefficient conversionofbiomassmaterial intobothenergyandvalue-addedproducts requiresa combinationof technologies andprocesses. As such, the photocatalytic reformingof feedstocks such as glycerol to hydrogen (H2) has a lot of potential. RESULTS: Reported here is the first example of a thin film-based photocatalytic system capable of achieving H2 evolution using a glycerol feedstock. Using a titania (TiO2) sol–gel, glass columns were coated with a thin TiO2 layer before using photodeposition to add platinum (Pt) as a co-catalyst. The coated columns were assembled into a simple yet effective recirculating system which used low power UV irradiation. Under optimum conditions (two coated columns and a 40 mL min flow rate), a steady state of 0.9 ∼mol min H2 with a photonic efficiency [ηphoton (%)] of 10.22 % was achieved. Furthermore, only one column showed flaking and loss of coating whereas the remaining columns were stable for the duration of the study, which equated to > 100 h of experimental testing including replicates and determining optimal parameters. CONCLUSION: H2 evolution via photocatalytic glycerol reforming in a Pt-TiO2 thin film catalyst recirculating system has been demonstrated under UV irradiation and ambient conditions. The system developed highlights that it is both catalyst development and reactor engineering that are required to continue to advance the field of photocatalysis. © 2020 The Authors. Journal of Chemical Technology & Biotechnology published by JohnWiley & Sons Ltd on behalf of Society of Chemical Industry. Supporting information may be found in the online version of this article.


INTRODUCTION
The reforming of waste compounds and biomass for the generation of hydrogen (H 2 ) has been shown to be a viable and potentially costeffective method for alternative energy. [1][2][3][4][5] In recent years, photocatalytic applications have turned their attention towards bioenergy and reforming processes with a view towards both energy production and value-added products. [6][7][8] Photocatalytic formation of H 2 has been an area of intense research since the initial publication from Fujishima and Honda 9 in 1972; however, with the stringent requirements of pure water splitting, a number of alternative sources have been investigated. Sacrificial electron donors (SEDs) such as methanol, oxalic acid and trimethylamine are among the most commonly reported in the literature; [10][11][12] however, the cost-effectiveness of such a system is low. A more viable approach is biomass feedstocks which are both readily available and of relatively low commercial value. Cellulose-derived substances have been reported for reforming to H 2 by Zhang et al., 13 Caravaca et al. 14 and more recently by Wakerley et al. 6 Although cellulose conversion can be limited due to its stable structure, glycerol as a feedstock for reforming to H 2 presents a number of advantages including: high potential stoichiometric yield of H 2 , low commercial value due to overproduction and good solubility in water.
The process of photocatalytic reforming of biomass-based substances over a single photocatalyst can be considered a combination of both water splitting and biomass oxidation. Upon electron-hole generation, two electrons in the conduction band are involved with the reduction of H + to H 2 via a co-catalyst such as platinum (Pt), whereas C x H y O z based compounds are oxidized to H 2 O and CO 2 at the valence band. This process can occur directly through photogenerated holes or indirectly via hydroxyl radicals, to irreversibly oxidize the substrate. Such a process has the advantage over traditional water splitting of supressing recombination to increase efficiency and avoiding the back reaction of H 2 with O 2 . Furthermore, the reforming of glycerol is a favourable reaction in regard to thermodynamics as the heat content of H 2 is theoretically higher than that of glycerol. H 2 has a lower heat content of ΔH c H 2 = 285.8 kJ mol −1 which corresponds to 2000 kJ mol −1 (7 × ΔH c H 2 ) based on the stoichiometry of the reaction, whereas glycerol is less at 1660 kJ mol −1 .
The photocatalytic valorization of glycerol to potential green fuels has been reported in the literature, with the focus primarily on H 2 evolution. [15][16][17][18] The initial research into photocatalytic glycerol reforming was first described in the work by both Bowker et al. andKondarides et al. in 2008 and2009, respectively. The work by Bowker et al. identified that H 2 evolution from glycerol was four times greater over palladium-titania (Pd-TiO 2 ) than gold (Au)-TiO 2 , whereas Kondarides and colleagues investigated Pt loading, photocatalyst content, glycerol concentration, pH and temperature for optimal H 2 evolution under ambient conditions. Subsequent papers by Kondarides et al. also investigated the kinetics and mechanism of glycerol reforming focusing on both reduction and oxidation to a range of liquid-phase products along with the activity of copper oxide (CuO x) /TiO 2 for the photoreforming of glycerol. To achieve a high photocatalytic activity in H 2 evolution, a co-catalyst is often used in conjunction with TiO 2 . Although to date a number of metals have been reported including Pt, 19 Pd, 15 Cu, 20 nickel (Ni), 21 Au 15 and tungsten (W), 22 Pt and Pd are frequently favoured. In 2008, Fu et al. 23 showed Pt to be the most active co-catalyst for H 2 evolution from glucose with the activity decreasing in the order of Pt > Au > Pd > Rh > Ag > Ru. Recently, Estahbanati et al. 18 reported on the impact of operating parameters such as glycerol concentration, catalyst loading, Pt % and pH for H 2 evolution over Pt-TiO 2 . Their work identified 50 % glycerol (v/v), 3.9 g L -1 catalyst, 3.1% Pt and pH 4.5 to be optimum conditions. Previously, Jiang et al. 19 investigated the role of the co-catalyst Pt and its deposition route on the formation of H 2 , finding that the nature of the light harvesting of P25 and the chemical state and size of Pt along with its interaction with P25 are fundamental for glycerol reforming. Slamet et al. 17 focused on the activity of Pt-N-TiO 2 nanotube materials, reporting a maximum quantum efficiency of 37.36 % for Pt (1%)-N-TiO 2 nanotubes.
Despite the increasing number of publications on glycerol reforming, there are currently none that utilize a thin film catalyst-based system. Within the field of photocatalysis, there has always been a strong research focus on the synthesis of new materials; however, the deployment of such materials is also paramount. The advantages of thin film-based systems including reduced downstream processing and increased viability for scale-up, have been documented previously, 24 yet there are a number of limitations which must be overcome. The primary obstacle is eliminating mass transfer to ensure increased interaction between the catalyst and substrate. Moreover, promoting sufficient light penetration and distribution coupled with complete catalyst surface irradiation is equally important. Previously this has been achieved by coating the catalyst onto a range of supports including glass rods and spheres, 24,25 optical fibres, 26,27 activated carbon, 28 rotating discs 29 and stainless steel. 30 A drawback to a number of thin film-based systems is the reduced catalyst surface area which often equates to reduced activity. Therefore, it is essential that reactor design maximizes the catalyst coating and irradiation characteristics with a view towards future scale-up. This paper reports on the development of a simple yet effective, small-scale thin film-based system capable of glycerol reforming to H 2 . A Pt-TiO 2 film was coated onto the inner wall of thin borosilicate glass columns before being irradiated by low power UV sources. To date there have been no examples in the literature of such a unit being deployed for glycerol reforming. The unit was operated as a closed recirculating batch system with samples taken from the feed-tank headspace for H 2 detection. The development of such a system reported here demonstrates that immobilized reactor designs are not just suitable for glycerol reforming to H 2 but present a potential method for feasible scale-up.

MATERIALS AND METHODS
Thin film Pt-TiO 2 coated onto glass column Thin film Pt-TiO 2 catalyst coated on to a glass column was prepared in two steps. At first the titania films were coated onto glass columns using a modified titania paste method shown by Mills et al., 31 followed by subsequent photodeposition of Pt nanoparticles on titania films. Twenty millilitres of titanium (IV) isopropoxide were added in a round-bottomed flask containing 4.65 g glacial acetic acid under an inert N 2 atmosphere. To this, 120 mL of 0.1 mol L −1 nitric acid was added and, after mixing the reaction solution, was heated rapidly to 80°C and maintained for 8 h. The reaction solution was allowed to cool to room temperature and any remaining aggregate particles were removed using a 0.45 m syringe filter. The resulting colloidal solution was hydrothermally treated in a teflon pot with lid in an autoclave at 220°C for 12 h. This hydrothermal step was used to grow the particles from 5 to 10-15 nm. Upon removal of the solution from the autoclave the separated out colloidal particles were re-dispersed using ultrasound. The reaction solution was then concentrated to~9-11 wt% using a rotary evaporator, followed by the addition of polyethylene glycol (50 wt% Carbowax 20 mol L -1 ) as a binder to help prevent the formation of small surface cracks when the paste is cast and allowed to dry. The titania paste was cast on the glass column using the doctor-blade technique, which allows a homogeneous spread of the paste. After drying at ambient conditions, the titania thin films were annealed at 450°C, for 2 h.
The coated tubes were weighed obtaining the value for the weight of TiO 2 on one tube (c. 0.03417 g), this value was necessary for determining how much of the platinum precursor solution is required. The amount of precursor was determined from the weight of TiO 2 present to obtain a 1 wt% loading of Pt. Titania thin film-coated glass columns were then filled with the required amount of dihydrogenhexachloroplatinate (IV) hydrate in methanolic solution (75:25 v/v MeOH:H 2 O). Subsequently, the films were irradiated with a UVA lamp for ≈20 min. As a result, the~1 wt% Pt-TiO 2 films had a dark brown-black appearance. The thin filmcoated glass columns were washed with distilled water and dried in air.
Catalyst characterisation X-ray diffraction (XRD) measurements were conducted on unused samples (i.e. not irradiated or used in any photocatalytic experiments) along with samples that had been used extensively under experimental conditions. In addition, scanning electron microscopy (SEM) analysis was conducted only using the latter sample. Samples for both XRD and SEM were prepared by cutting a section of the quartz tube and further breaking it into smaller pieces suitable for analysis. For XRD, a Pt-TiO 2 coated tube that remained stable and active throughout the experiments was used along with an unused coated sample and a sample that had the lost the catalyst coating as a result of flaking. For SEM, a small piece of the Pt-TiO 2 coated sample was mounted onto a specimen stub equipped with a screw to secure the sample in place.
The XRD measurements in this work were carried out on a PANanalytical X'Pert Pro X-ray diffractometer. The X-ray source was Cu with a wavelength of 1.5405 Å. All measurements were carried out ex situ using a spinning stage. The diffractograms were recorded from 4°to 75°with a step size of 0.017°.
The SEM analysis in this work was carried out on a Quanta FEG 250 Scanning Electron Microscope. SEM was then carried out under high vacuum and images of the sample recorded in the magnification range of ×100-7000.

Immobilized reactor design concept
The immobilized recirculating reactor comprised a reactor feed tank and an illumination unit, which is detailed in Fig. 1. The feed tank utilized the Propeller Fluidised Photo Reactor (PFPR) as a controllable air-tight environment in which the glycerol feedstock was mixed and gas samples were removed. The full details of the PFPR can be found in our previous publication. 11 The illumination unit comprised coated tubes with a bespoke half-moon structured design which held up to six 8 W black lamps (Koninklijke Philips, Netherlands. Supplied by RS Components, UK).
The 8 W black lamps emitted light in the UV-A region and was measured to have a peak wavelength at 370 nm using a Stella-Net Black Comet spectrometer (StellaNet INC, Tampa, Florida, USA). To accurately monitor the spectral output of the lamp, the photon flux was determined using the potassium ferrioxolate actinometrical method 32 and Eqn (1). To calculate the flux in the reactor, an uncoated glass column was filled with actinometry solution and irradiated for a dedicated period of time before analysis by UV-visible spectroscopy.
where Fe 2+ (mol) were determined based on the potassium ferrioxalate method (7.68 × 10 −6 ), ⊞Fe 2+ was set at 0.97 and t was the time (min) the actinometry solution was irradiated for. The photonic efficiency was then determined based on the calculated photon flux (1.75 × 10 −5 mol min −1 ) and Eqn (2).
where η photon (%) is the photonic efficiency, rH 2 is the H 2 formation rate (mol min -1 ) and photon flux is the rate of photons entering the reactor (mol min -1 ), as determined by actinometry. As H 2 formation is a 2-electron step, the rH 2 was multiplied by 2.

Photocatalytic procedure
In a typical experiment, 100 mL of a predetermined concentration of glycerol feedstock was added to the reactor (unless otherwise stated, the initial concentration was 1 mol L -1 ) and stirred at 400 rpm (2.5 dcV). The entire reactor system was purged with N 2 at 150 mL min −1 for 20 min before any illumination. Upon completing purging, the pump was started and recirculated the solution through the reactor system for ≈30 min before illumination was started. During the photocatalytic procedure, gas samples were taken periodically from the headspace of the PFPR and analyzed by gas chromatography-thermal conductivity detection (GC-TCD) for the formation of H 2 . The first sample was taken after purging to ensure that no H 2 or O 2 was present. An Agilent Technologies 7280 A GC system was used coupled with a packed column (RESTEK, 2 mm inner diameter) and a TCD. The injector was operated at temperature 150°C, pressure 26.1 psi and flow rate 22.9 mL min −1 . The flow rate in the column was 20 mL min −1 with an oven temperature of 50°C, whereas the detector was maintained at 200°C with a flow rate of 5 mL min −1 . Argon was used as the carrier gas. The detection of H 2 was determined by comparison to a standard injection of pure H 2 , whereas quantification was determined from a calibration of known concentrations.

Catalyst characterization
The XRD pattern of the synthesized Pt-TiO 2 films coated onto glass columns is shown in Fig. 2. Fig. 2(a) and (b) show the XRD diffractograms of Pt-TiO 2 films on glass columns before and after experimental use, respectively. The latter was a sample that remained stable and active throughout the experimental investigation, which also showed no signs of catalyst loss through flaking. The crystal structures of these fully coated samples were confirmed to be anatase TiO 2 based on the 2⊔ peaks that correspond to standard peaks for the compound. This also is in agreement with data published previously in the literature, which confirmed the sol-gel synthesis route produced a pure anatase thin film. 31 The samples show minimal difference before and after use, with anatase peaks identified in both, confirming the stability of the coating. By contrast, the XRD diffractogram shown in Fig. 2 (c) was for a sample that had lost the catalyst coating as a result of flaking during experimental testing. Visually, the sample appeared clear and showed no activity for H 2 formation from glycerol reforming when under irradiation. The sample showed only a single broad peak around 23°corresponding to quartz with no other peaks detected, which confirmed that anatase TiO 2 was not present. Fig. 3 shows SEM images of the Pt-TiO 2 film coated onto the glass column supports, which corresponds to the sample analyzed by XRD in Fig. 2(b). Although this sample had been used under experimental conditions, the XRD and H 2 evolution data confirmed that it remained active and stable. Fig. 3(a) is a low magnification image showing the glass column sample coated with the thin film secured by a screw on the specimen stub. Fig. 3 (b) and (c) shows the surface of the glass column at a higher magnification. The Pt-TiO 2 thin film was seen as a roughened surface with thickness ≈5 μm.

Photocatalytic H 2 production
The production of H 2 was monitored in the immobilized recirculating unit to assess the overall efficiency of the system with a view towards stability, reusability and the effect of feedstock flow rate. The stability and reusability of the Pt-TiO 2 catalyst coating was a key consideration in determining the efficiency of the reactor. Although an immobilized system is often mass transport limited, it has the advantage of being easily recycled with fewer downstream processing requirements than traditional suspended units. The reusability of the current unit is demonstrated in Fig. 4, showing the evolution of H 2 over six reaction cycles under light and dark conditions along with N 2 purging and the addition of fresh glycerol. Before the final cycle, fresh glycerol was added to the system to further confirm the stability of the catalyst coating.
A linear profile of H 2 can be seen under irradiation, whereas production plateaued under dark conditions. Under dark conditions the photocatalytic formation of H 2 stopped with only a small increase observed as a result of the pump continuing to recirculate, thereby transporting H 2 formed in the irradiation chamber to the feed tank gas headspace. These results indicate that the reaction was induced by the adsorption of photons onto the catalyst surface. Furthermore, all blank experiments conducted in the absence of light and/or catalyst showed no H 2 evolution (see- Fig. S1). In the absence of glycerol and/or Pt, and therefore under pure water splitting conditions, no significant H 2 (or O 2 ) was recorded showing that the H 2 evolution detected was a result of glycerol photoreforming coupled with electron trapping at Pt (see Fig. S1). Upon purging the system with N 2 in later cycles, the evolution of H 2 can be seen to be reproducible indicating the coated columns were stable. An average reaction rate (rH 2 ) of 1.03 μmol H 2 min −1 was recorded with a standard deviation of only 0.02. The slight variation that can be observed between individual time points could be a result of the appearance of partial oxidation products contributing towards H 2 formation. The photocatalytic reforming of glycerol, via oxidation and reduction reactions, can lead to both H 2 and a range of liquid phase products with the initial compounds expected to be dihydroxyacetone and glyceraldehyde. From these parent intermediates, compounds such as glyceric acid, glycolaldehyde, glyoxylic and oxalic acid were expected to be formed. These products are likely to be generated directly or indirectly from both OH• radical attack and direct hole interaction at the valence band which also was discussed by Minero et al. 33,34 They showed that the formation of dihydroxyacetone and glyceraldehyde were formed via H-abstraction as a result of OH• attack, whereas glycolaldehyde was thought to occur via direct hole interaction and ⊎-fragmentation. Based on the products likely to be formed, however, it can be assumed that they can subsequently undergo further oxidation rapidly, leading to proton reduction to H 2 . The identification and quantification of such products is currently the focus of future work with a view to better understanding the mechanism and attempting to capture these high value compounds.
Reduced activity through the loss of catalyst thin films is currently a limitation with immobilized photocatalytic designs. Parameters such as mechanical stress, sheering, heat and viscosity can all significantly impact the strength and adhesion of the thin film to the support material. In the current system, the viscosity of glycerol and flow rate were two key parameters which impacted the activity of the catalyst film. After completing a series of stability experiments (as shown in Fig. 4), it was observed that the coating on one of the columns was beginning to 'flake' with particles becoming suspended in the system. The initial flaking of the catalyst coating was found to be minimal, with no significant impact on activity detected. Upon prolonged  use and washing with distilled H 2 O, however, more substantial flaking was observed along with a corresponding drop in photocatalytic activity. After a series of washing stages, it was found that the column became transparent. The inactivity was confirmed by results shown in Fig. 5(a) (supported by Fig. S2), which showed that recirculating the glycerol feed through one coated and uncoated column [ Fig. 5(b)(ii)] generated a rH 2 equal to a single coated column [ Fig. 5(b)(i)]. Using two Pt-TiO 2 coated columns, which showed no catalyst loss, produced a rH 2 = 0.62 μmol min −1 , which was almost double that of a single column (rH 2 = 0.35 μmol min −1 ). Interestingly, the remaining columns were stable throughout the duration of the study with only minor flaking observed, which showed no impact on photocatalytic activity. To ensure that no suspended particles were present, thorough washing steps were carried out between runs.
The effect of flow rate The primary challenge in any immobilized film-based photocatalytic reactor is overcoming the mass transfer limitations which can significantly restrict the efficiency of a system. Moreover, as the mechanical stability of catalyst films can often be an additional issue, the method of eliminating mass transfer limitations should ensure the coating is not removed in the process. This is particularly challenging given the viscosity of glycerol solutions. In an immobilized system, it is ideal to have a thin film of solution passing over a fully irradiated catalyst surface. In the literature, this has often been achieved by creating a thin falling film by using coated plates positioned at different angles 35 and by using a pump to create a 'fountain-like' effect. 36 In the unit investigated here, a peristaltic pump was used to recirculate the glycerol solution through narrow coated columns. As the columns had an internal diameter of 5 mm, which was fully coated with catalyst, the equivalent of a thin film of solution was created which had increased residual contact with the activated surface.
The effect of flow rate was found to impact H 2 evolution in relation to both mass transfer between glycerol and the catalyst surface and movement of the photogenerated H 2 [ Fig. 6(a); and see Fig. S3]. It was found that an optimum recirculating flow rate of 40 mL min −1 was achieved, in which the residual contact time (R t = 0.06 min) between the catalyst and substrate was sufficient to allow for H 2 evolution (89 μmol after 100 mins irradiation), while ensuring transport of H 2 from the catalyst surface to the reactor gas headspace. Interestingly, the order of efficiency of glycerol flow rates was 40 > 30 > 50 > 20 > 10 > 0 mL min −1 [ Fig. 6(a)]. At 0 mL min −1 , the data confirms that the system was entirely mass transport limited as no evolution of H 2 was recorded in the time frame, however upon starting the pump and recirculating the feedstock, H 2 was detected in the headspace. Alternatively, increasing the flow rate to 50 mL min −1 showed a drop in H 2 production, which suggested that mass transfer limitation no longer predominated at flow rates > 40 mL min −1 . At flow rates ≥10 and ≤ 40 mL min −1 , Fig. 6(c) shows a linear relationship existed when Ln (rH 2 ) versus Ln (R t ) was plotted, suggesting firstorder kinetics and a rate order of 0.35. Fig. 6(b) also shows R t for the system at increasing flow rates. Although significantly increased at lower flow rates (R t = 0.23 min at 10 mL min −1 ), it is interesting to note that this did not equate to higher levels of H 2 ; R t decreased at higher flow rates and the data showed that H 2 evolution increased until 40 mL min −1 (R t = 0.06 min). Furthermore, Reynolds number calculations (Table 1) showed that at flow rates >30 mL min −1 , the flow regime was in the transition region and approaching turbulent flow, which is favourable for overcoming mass transfer limitations.
In relation to the reaction rates, a steady-state of H 2 evolution (c. 0.9 μmol min −1 ) was achieved after 40 min of irradiation (see SI- Fig. S3). Although this value can be compared to that reported previously in the literature, it is important to note that comparison of photocatalytic systems is challenging. There are a number of parameters which influence the activity of heterogenous photocatalysis such as irradiation positioning and intensity, substrate concentration, reactor geometry including catalyst deployment (i.e., suspended or immobilized) and the physical and chemical properties of the catalyst. Therefore, to allow for a more direct comparison with previous studies, the steady-state reaction rate achieved in this study has been calculated as a function of irradiation time (h −1 ) and quantity of catalyst present (gTiO 2 −1 ) for a single column. A H 2 formation rate of 1580 μmol H 2 g cat −1 h −1 per TiO 2 -coated column (based on 3.4 × 10 −2 g catalyst coated onto each tube) was determined for this study. This rate is within values reported in previous publications, which typically have a range of 1500-4300 μmol gTiO 2 −1 h −1 for Pt-loaded TiO 2 photocatalysts 16,17,19 Furthermore, the use of NiO (NiOx-TiO 2 ) in the place of Pt as a co-catalyst shows slightly lower rates of formation at 1230 and 900 μmol g-TiO 2 −1 h -1 . 37,38 The value reported here is higher than that observed when using NiO as a co-catalyst and, although lower than previous Pt-TiO 2 systems in the literature, remains comparable despite being an immobilized system.
A reaction rate as a function of g cat is often used as an equivalent unit in the literature and is a useful metric for comparison, yet the implications of using such a high loading are rarely considered. In the majority of suspended systems, without sufficient reactor design optimization, a loading of 1 g cat could result in significant light scattering and attenuation coupled with decreased light penetration which subsequently would impact photon-induced excitation and activity. Furthermore, this could have significant impact with a view towards any attempts at reactor scale-up. In 1998 Ray and Beenackers developed Eqn (3) to express how photocatalytic reactor volume can be determined. 39 Although the equation was directed primarily towards environmental remediation    Photocatalytic reforming of glycerol to H2 www.soci.org applications, it highlights the complexity and importance of irradiated catalyst surface.
Where Q is the volumetric flow rate (m 3 s -1 ), C in is the inlet pollutant concentration (mol m -3 ), 'X' is the desired fractional conversion, ICD is the illumination catalyst density and AMP is the average mass conversion rate.
Therefore, increasing the loading of catalyst while maintaining a desirable reaction rate is dictated by the ability of the irradiation source to deliver sufficient photons to the surface of the catalyst.
In the system presented here, to obtain a catalyst loading equal to 1 g would simply require either an increase in the length or number of glass rods used, which importantly could be achieved without impacting the irradiation pathway.
With light and reactor design clearly impacting overall photocatalytic activity, it is important to consider further measurements of efficiency that take these parameters into account. Moreover, although light penetration and distribution are key factors in any photocatalytic system, their role is crucial in an immobilized design due to limitations that the catalyst platform presents. In the current investigation the catalyst coating on the inside of the glass rods ensured that maximum light penetration could be achieved more readily than in a traditional slurry system. If the coating is considered in its most basic form, which is a thin layer of catalyst particles, then the ideal method of activation is to irradiate from one side, thus allowing the generation and movement of electrons which react with glycerol molecules as it passes through the inside of the column (Fig. 7). This reduces the number of 'objects' between the source of irradiation and the target which subsequently improves the efficiency. A similar approach was adopted by Peill and Hoffmann in their optical fibre photoreactor which improved catalyst activation through enhanced light transmission and distribution. 40 The quantum yield of a photocatalytic system can often be determined based on the ratio of 'moles of product formed' to the 'moles of photons absorbed by the catalyst'. This allows the reaction rate to be viewed in relation to the irradiation source as opposed to simply the value of H 2 formed. Accurately determining the number of photons absorbed or photon flux is a significant challenge in photocatalysis, however, as the calculations often involve making a number of assumptions. In this instance, the moles of product formed was taken as the maximum reaction rate for H 2 (rH 2 max) and the photon flux (determined via actinometry) was taken as the moles of photons 'absorbed by the catalyst'. Based on this, the efficiency of the system reported here can be referred to as the photonic efficiency [η photon (%)], which also is known as an apparent quantum efficiency. Although actinometry can provide an accurate photon flux for a reactor (by replacing the reaction solution with potassium ferrioxalate), it is often assumed that this value equates to the number of photons absorbed or 'used' by a catalyst particle. In a typical suspended system, this is an unfavourable assumption as light scattering and reflectance are difficult to account for. In the system presented here, however, the photon flux is determined based on the photons which penetrated through the glass reactor wall and reached the potassium ferrioxalate solution inside the column. This value therefore provides an accurate representation of the photons which were capable of reaching the thin film surface.
As photocatalytic H 2 production is a two-electron step (2H + + 2e − = H 2 ), the photonic efficiency [η photon (%)] in this study was defined as twice the number of H 2 moles generated divided by the photon flux and was given as a percentage [Eqn (2)] ( Table 2). The highest recorded η photon in this study was 10.22 % under irradiation from the 24 W black lamps. The η photon (%) reported here shows a high level of potential for an immobilized system for continued enhancement. The ability to scale the system up with additional coated columns along with the use of lower power irradiation source could produce a favourable system to traditional suspended slurry reactors. Coupling this with coatings that have enhanced stability could further improve the overall photocatalytic efficiency of the system.
Overall, H 2 evolution via photocatalytic glycerol reforming in a Pt-TiO 2 thin film catalyst recirculating system has been demonstrated under UV irradiation and ambient conditions. The system developed shows that despite immobilizing a catalyst onto a solid static support, significant H 2 production was still achieved; under optimum conditions a rate of 0.9 μmol H 2 min -1 was achieved along with a η photon = 10.22 %. Overall, the stability and reusability of the thin film coating was good with only one example of flaking observed, which indicates the potential for this design towards scaling-up and reduced downstream catalyst recovery requirements. Table 2 Calculated H 2 evolution rates (rH 2 max), photon flux and η photon (%) for the thin film system with 24 W black lamps Irradiation source rH 2 max aa (mol min −1 ) Photon flux (mol min −1 ) η photon (%)