On the Limits of Photocatalytic Water Splitting

The major drawbacks on the limited H 2 and O 2 evolution activities of one-step photocatalytic water splitting systems are given here with the emphasis on charge recombination, back-oxidation reactions, and mass transfer limitations. Suppression of these unwanted phenomena is shown to be possible with the usage of small crystal-sized photocatalysts with low defect concentrations, presence of phase junctions, selection of co-catalyst that would be active for H 2 evolution but inactive for O 2 reduction, coating of the co-catalyst or the whole photocatalyst with selectively permeable nanolayers, and usage of photocatalytic systems with high solid – liquid and liquid – gas surface areas. The mass transfer limitations are shown to be important especially in the liquid – gas interfaces for agitated and suspended systems with estimated H 2 transfer rates in the range of (cid:1) 200 – 8000 μ mol/h.


Introduction
Hydrogen gas is one of the best alternatives to fossil fuels since it has a high gravimetric energy density (142 MJ/kg) and it produces zero carbon upon combustion. Hydrogen is also used as a major reactant in environmentally important reactions such as carbon dioxide hydrogenation to methanol [1] or ammonia production (Haber-Bosch reaction) [2]. For hydrogen to be used as a clean energy source, its production via renewable ways is of great importance. It is conventionally produced via steam reforming of methane and fossil fuels (energy intensive, ΔH 0 rxn = 206 kJ/mol, 700-1100°C [3]) and coal gasification, which results in significant amounts of carbon dioxide production. The renewable ways for carbon-free production include biological sources (microalgae and cyanobacteria) and electrolysis of water using wind energy and photovoltaic cells as electricity generation sources. In addition to the mentioned renewable ways, photocatalytic water splitting/oxidation is a promising alternative, in which solar energy is used as the driving force to split water molecules to hydrogen and oxygen on the surface of a catalyst. This renewable production method of hydrogen is advantageous over other renewable methods due to the free source of energy and lower cost of the photocatalysts when compared to that of photovoltaic cells or wind turbines. Solar-driven catalytic (photocatalytic) reactions are considered to be of fundamental importance to the catalysis community since the solar energy is inexhaustible; i.e., the solar energy absorbed by the lands and oceans on an hourly basis (432 EJ/h or 120,000 TW [4]) is comparable to the Earth's yearly energy consumption (reaching 575 EJ/year or 18 TW in 2017). However, the solar-to-hydrogen energy conversion efficiency value for photocatalytic water splitting systems is much lower (targeted to be 10%, currently reaching 1% [5]) than that of photovoltaic-assisted electrolysis (reaching 30% [6]) due to the major drawbacks in the one-step photocatalytic water splitting systems. Herein, we firstly introduce photocatalytic water splitting systems and give the major developments in materials such as visible light utilization and corresponding H 2 and O 2 production activity values (in Section 2). Then in Section 3, we discuss the causes of the low efficiencies in photocatalytic water splitting systems and the recent approaches in preventing energy efficiency-lowering factors such as inefficient visible light utilization, charge recombination, back-oxidation reactions, and mass transfer limitations.

Photocatalysis and water splitting
The first report on water splitting via harvesting photon energy is authored by Fujishima and Honda using a photoelectrochemical cell with a TiO 2 photoelectrode [7]. Following this first report suggesting the oxidation of water molecule via photogenerated holes on TiO 2 surface with the aid of small electrical voltage, photocatalytic water splitting on powder photocatalyst particles is demonstrated by other authors in the late twentieth century [8][9][10][11][12][13][14][15]. Metal-loaded semiconductors (such as Pt/TiO 2 ) are described as "short-circuited photoelectrochemical cells" that provide both the oxidizing centers and the reduction centers on the same catalyst (see Figure 1) [16].
Photocatalytic reactions are initiated by absorption of light having an energy higher than (or equal to) the bandgap of the photocatalysts that consist of semiconductor materials. This bandgap energy should be larger than 1.23 V for overall water oxidation reaction, for which the maximum of the valence band and the minimum of the conduction band should be located at proper potentials for the oxygen and hydrogen evolution reactions to occur. To illustrate, the minimum of the conduction band energy level should be located at a more negative potential than 0 V vs. NHE, at pH = 0 for H 2 evolution (Eq. (1)), and the maximum of the valence band should be at a more positive potential than 1.23 V vs. NHE at pH = 0 for oxygen evolution reaction (Eq. (2)): Following the light absorption, photoexcited electrons are transferred to the conduction band, while a positively charged charge carrier (hole) is generated at the valence band. These charge carriers are then transferred to the catalyst surface (step 2 in Figure 1) to be utilized in surface redox reactions, unless they recombine in the bulk or on the surface (step 4). Ultimately, electrons and holes reduce/oxidize the adsorbed species on the catalyst surface (step 3), the products of which should then be desorbed from the surface to complete the overall process.

Semiconductors
TiO 2 , having a large bandgap (anatase: 3.2 eV), is the most commonly used photocatalyst due to its photostability, nontoxicity, and high activity (upon UV radiation λ < 387 nm). Following the report on water oxidation reaction  [36][37][38]) are shown to possess large bandgap energies (>3 eV) due to the maximum valence band levels consisting O 2p orbitals located near 3 V (vs. NHE at pH = 0). These d 0 and d 10 metal oxide catalysts are reported to show remarkable one-step photocatalytic water splitting activity under UV light irradiation [39] reaching 71% quantum yield with photocatalysts such as Al-doped SrTiO 3 [40] or Zn-doped Ga 2 O 3 [41]. The H 2 and O 2 evolution activity under UV radiation and the apparent quantum yields of some of these materials are given in Table 1. The apparent quantum yield is defined as the number of reacted electrons and holes divided by the number of incident photons on the photocatalysts. Table 1 is not intended to cover the whole range of particulate catalysts in literature but rather to give a selection of examples. A wider selection of d 0 and d 10 metal oxide particulate catalysts' one-step water oxidation activity and apparent quantum yields can be found in the works of Kudo et al., Chen et al.,and Domen et al. [39,42,43].
The most remarkable upgrades in the apparent quantum yields are achieved by material engineering such as (i) doping the metal oxides/perovskites with cations having lower valences, (ii) decreasing the crystal sizes to submicron levels, and (iii) loading with H 2 /O 2 evolution co-catalysts.

Co-catalysts
An important addition to the light-harvesting semiconductors is H 2 evolution/O 2 evolution co-catalysts on the surface. The early co-catalysts that have been widely used included the noble metals and transition metal oxides such as Pt [12,13] , and IrO 2 [54] that accelerate the oxygen evolution. These metals are considered to act as charge carrier sinks that suppress electron-hole pair recombination as well as increasing the reaction kinetics by lowering the activation energy of the redox reactions. Co-catalysts are also known to inhibit photodegradation of the photocatalysts such as oxysulfides and oxynitrides by generated holes due to the effective extraction of these holes by the co-catalysts [55,56].
Following the works of noble metal co-catalysts, Domen et al. showed water splitting activity on SrTiO 3 photocatalyst together with the co-catalyst NiO [57,58], which became the choice of H 2 evolution co-catalyst for many d 0 and d 10 metal oxides such as La 2 Ti 2 O 7 :Ba [44], La 4 CaTi 5 O 17 [59], Rb 4 Nb 6 O 17 [60], NaTaO 3 [46], and Ga 2 O 3 :Zn [47]. The photocatalyst stability of NiO-loaded K 2 La 2 Ti 3 O 10 is reported to increase by addition of a second co-catalyst, Cr, using a coimpregnation method [61]. Based on the promoting effect of Cr, a systematic study of Cr and various transition metals (such as Fe, Co, Ni, Cu, Ru, Rh, Pd, Ag, Pt) on (Ga 1-x Zn x )(N 1-x O x ) has been conducted [62], from which core-shell structures of core Rh nanoislands and shell Cr 2 O 3 structures (10-30 nm in size) are found to promote H 2 and O 2 evolution reactions to significant levels [63].

Visible light utilization
The alterations to the semiconductors such as doping with low-valence cations, reducing particle sizes to submicron levels and obtaining a high degree of crystallinity help with the overall water splitting activity. However, activation of these photocatalysts uses a narrow portion of the solar spectrum (4%); i.e., the UV light sustains as a problem due to the large bandgap energies of these materials. To enable visible light utilization of the d 0 -and d 10 -type oxide semiconductors and to split water into H 2 and O 2 via one-step excitation, the valence band levels should be *0.1 g of photocatalyst is used instead of 1 g. shifted upward without changing the conduction band potentials. One approach to do this is to use oxynitrides to make use of N 2p states that lie at a more negative potential than O 2p states. Emerging LaMgTa 1Àx O 1+3x N 2À3x [64] and Ga 1Àx Zn x N 1Àx O x [65] oxynitrides are representatives of visible light active overall water splitting catalysts. Using N doping, the absorbed light wavelength can be increased up to 500 nm on solid solutions of GaN:ZnO (Ga 1Àx Zn x N 1Àx O x ) [66] and up to 600 nm on solid solutions of LaTaON 2 and LaMg 2/3 Ta 1/3 O 3 (LaMgTa 1Àx O 1+3x N 2À3x ) [64]. Other examples include LaSc x Ta 1Àx O 1+2x N 2À2x [67] and CaTaO 2 N [68] in which La or Ta sites are replaced by Ca and Sc that alters O/N ratios due to charge compensation, which in turn results in valence band energy level shift. Some examples of visible light active photocatalyst and their H 2 and O 2 evolution activity are given in Table 2. As it can be seen from the table, one-step water splitting quantum yields are quite lower when compared to those of the UVactivated photocatalysts ( Table 1). The exceptions to the low activity are reported by Rh 2-y Cr y O 3 (Rh 1.0 wt%, Cr 1.5 wt%)-loaded (Ga 1Àx Zn x )(N 1Àx O x ) photocatalyst [63], multiband InGaN/GaN nanowire arrays [69], and monodisperse 4 nm graphite nanoparticle-deposited C 3 N 4 catalysts [70].
An alternative way to cover both oxidation and reduction reactions with semiconductors that could be activated under visible light radiation is to utilize two individual photocatalysts with an electron transfer mediator to obtain two-step excitation known as the two-step water oxidation ("Z-scheme system," see Figure 2). In this system, O 2 evolution photocatalysts oxidize the water molecules to O 2 , while the photo-generated electron is transferred to the mediator to reduce the electron acceptor (such as Fe 3+ ions or IO 3 À ions). Then, the reduced mediator is oxidized by donating its electron to the H 2 evolution photocatalyst. At the same time, the photo-generated electrons in the H 2 evolution photocatalyst reduce H + s to H 2 .
The semiconductors used in this two-step water splitting process should be selected based on the energy levels of their corresponding valence or conduction band maximum/minimum that would enable O 2 /H 2 O oxidation and H + /H 2 reduction. As H 2 evolution and O 2 evolution reactions are realized at separate photocatalysts, these semiconductors could have bandgap energy values lower than 3 eV that would enable visible light utilization such as Pt-or RuO 2 -loaded WO 3

Semiconductor
Cocatalyst (E g $ 2.8 eV) or oxynitrides such as TaON (E g $ 2.4 eV) or Rh-doped SrTiO 3 (E g $ 2.4 eV). Examples of these materials and systems can be seen in Table 3.
The detailed reviews on two-step photocatalytic water splitting can be found elsewhere [83].

Drawbacks on photocatalytic activity
There are numerous and challenging processes that need to be realized for photocatalytic evolution of H 2 and O 2 ( Table 4) via a thermodynamically unfavorable reaction (Eq. (3)): These processes include (i) excitation of the semiconductor photocatalyst with photon having higher energies than the bandgap energy of the material, (ii) transfer of the photo-generated electrons and holes to the reaction sites on the surface, (iii) utilization of these charge carriers in the oxidation/reduction reactions, and (iv) desorption of the products from the surface of the photocatalyst to the liquid/gas medium.
As the timescale of these processes varies, recombination of the electrons and holes in the bulk or on the surface happens more frequently than the rate of the    Table 3. Z-scheme-type photocatalysts for water splitting without sacrificial agents.
chemical oxidation/reduction reactions. Recombination is therefore considered to be one of the main reasons limiting the photocatalytic activity. Together with the recombination events, realization of back-oxidation reactions (Eq. (4)) on noble metals and the rate-limiting mass transfer events are the major drawbacks in an efficient photocatalytic process: Natural photosynthesis yields a much higher rate of O 2 evolution (see Table 4) when compared to artificial water splitting due to improved charge carrier and mass transfer events. From this comparison, it is clear that the photocatalytic systems still need to be perfected to compete with the nature's intricate design.

Charge recombination
Due to the presence of the multiple processes, the overall photocatalytic reactions are extremely complicated. In order to obtain an efficient photocatalytic performance, the photo-generated charges must be transferred to the surface reaction sites as rapidly as possible while preventing recombination or trapping of these charge carriers. It is reported by Leytner and Hupp that 60% of the trapped electron-hole pairs recombine with a timescale of about 25 ns while releasing heat of 154 kJ/mol [85]. As the defects such as vacancies and dislocations are considered as recombination sites, higher crystallinity of the photocatalysts is often aimed to decrease the recombination rates. From diffusion point of view, the shorter distances for the charge carriers to the surface reaction centers are also aimed to prevent the recombination. Shorter pathways are achieved via smaller crystal/particle sizes of the photocatalysts. More than two times of increase in the H 2 and O 2 evolution rates on Al-doped SrTiO 3 photocatalyst (reaching an apparent quantum yield of 56% [45]) as the particle size drops from few micrometers to 200 nm is a direct evidence of the effect of the particle size. Another method for reducing the charge recombination is to make use of phase junctions. One example is the α-βphase junction of Ga 2 O 3 , which results in enhanced interfacial charge transfer, charge separation, and therefore enhanced water splitting activity [86]. Loading the photocatalysts with co-catalysts such as noble metals or transition metal oxides to accelerate the reduction/oxidation reactions is a commonly employed method.   [41], assuming 10 m 2 /g surface area and 10 15 sites/cm 2 site density. These co-catalysts are known to enhance the charge migration from the semiconductor depending on the alignment of the potentials of the semiconductor and the co-catalyst. As these co-catalysts accelerate the desired H 2 evolution and O 2 evolution reactions, they can also increase the rates of undesired secondary reactions such as hydrogen oxidation or oxygen reduction to water reactions.

Back-oxidation reactions
Introduction of one-step photocatalysts for overall water splitting combined the H 2 evolution and O 2 evolution sites on the same catalyst surface. This design of a photocatalytic system that realizes both charge trapping and reduction/oxidation reactions on the same surface not only accelerated the charge recombination but also allowed secondary reactions on these reduction/oxidation centers. When fast removal of the products, i.e., H 2 and O 2 , is not provided and there are no barriers that prevent interaction of these products with highly active sites, the reaction of H 2 and O 2 on the photocatalyst surface to produce H 2 O (2H 2 + O 2 ! H 2 O) is highly probable. And back-oxidation of the produced H 2 is considered to be one of the main reasons for observed low photocatalytic water splitting activity values.
As early as 1985, Sato and coworkers realized the importance of back-oxidation of H 2 with O 2 to produce H 2 O. They have realized that the metal-loaded photocatalysts, mainly Pt-or Pd-loaded TiO 2 , can oxidize H 2 with O 2 easily under the same photocatalytic water oxidation conditions. They have reported first-order reaction rate constants in the range of 0.23-0.51 h À1 for Pt, 0.32-1.8 h À1 for Pd, and 0.2-0.3 h À1 for Rh, suggesting the least active metal for back-oxidation reaction to be Rh [10]. Later in 2000, Anpo and coworkers investigated back-oxidation reaction on Pt/TiO 2 systems under dark conditions and observed increased backoxidation rate with increasing Pt loading (up to 0.1 wt.% [87]). While Pt is active for H 2 evolution (Eq. (1)), it is also notoriously active for dark H 2 -O 2 recombination reaction (Eq. (4)) even at room temperature [88]. In order to prevent H 2 -O 2 recombination reaction, the Pt surface is modified with F ions for Pt/TiO 2 catalyst, and the reaction rate decreased from 2 to 0.3 h À1 upon F À modification [89]. The inhibition mechanism is suggested to be due to the occupation of the H 2 surface adsorption sites on Pt by F atoms.
Another modification to the noble metal surfaces is reported by Lercher et al., in which CO is chemisorbed on the Rh co-catalyst for GaN:ZnO semiconductor. Chemisorbed molecular layer of CO suppressed the back-oxidation reaction by selective metal poisoning of the back-oxidation sites by CO. While H 2 evolution rates of 28 μmol/h are achieved (75 mg photocatalyst, 300 W Xe lamp [90]), significant CO oxidation to CO 2 is also observed.
The back-oxidation reaction-inhibiting effects of the nanolayer coating on noble metals are shown on Rh/Cr 2 O 3 -loaded GaN:ZnO photocatalysts. Rh/Cr 2 O 3 coreshell structure [91] is formed by photodeposition of Rh and reduction of CrO 4 2À by electrons coming from Rh upon radiation, resulting in few nanometer thickness of Cr 2 O 3 layer (2-3 nm, see Figure 3). Hydrated Cr 2 O 3 nanolayer is reported to selectively permeate protons for H 2 evolution reaction [92], whereas it hinders O 2 permeation from the layer inhibiting O 2 reduction reaction (Eq. (5)) on Rh sites [93]. The same effect is also valid for Cr 2 O 3 -coated Pt catalyst (GaN:ZnO). Backoxidation rates on Pt-loaded GaN:ZnO photocatalyst decreased significantly from $105 Â 10 À12 molecules/s to $8 Â 10 À12 molecules/s, while photocatalytic H 2 evolution rate increased from $5 Â 10 À12 molecules/s to $30 Â 10 À12 molecules/s upon Cr 2 O 3 coating [93]. Apart from the oxygen-blocking role of the Cr 2 O 3 nanolayer, much lower back-oxidation rate of Rh-loaded GaN:ZnO when compared to Pt-loaded GaN:ZnO (11 Â 10 À12 molecules/s vs. 105 Â 10 À12 molecules/s) explains the significant photocatalytic H 2 evolution activity on Rh/Cr 2 O 3 -loaded GaN:ZnO (130 Â 10 À12 molecules/s). Lower back-oxidation rate of Rh-loaded GaN: ZnO could be related to the low-oxygen reduction reaction (Eq. 5) activity of Rh when compared to Pt [94]: Similar selective permeability concept is considered to be the case for Ni/NiO core-shell structures deposited on various photocatalysts such as SrTiO 3 or NaTaO 3 [34,57]. In these systems, in addition to the back-oxidation reaction impeding effect of NiO layer on Ni [58], low-oxygen reduction activity of NiO x catalysts when compared to Pt can also be considered to be effective for improved water splitting activity.
Coatings of the whole photocatalyst instead of the co-catalyst by oxyhydroxides of Ti, Nb, and Ta are reported on Rh-loaded SrTiO 3 :Sc photocatalyst. Surface nanolayer not only suppressed back-oxidation reactions but also prevented the access of sacrificial agents such as ethanol to the photocatalyst surface, resulting in nearly stoichiometric H 2 /O 2 ratios [95]. Surface nanolayer coatings on the whole photocatalysts have proven to also prevent photodecomposition (N 2 evolution) of oxynitride photocatalysts while increasing the overall water splitting activity [64,68].
Prevention of the secondary reactions such as H 2 oxidation or O 2 reduction reaction to H 2 O is found to be essential for improving the overall water splitting activity and the apparent quantum yield values (reaching apparent quantum yield value of 69% under irradiation at 365 nm [40]). In addition to the reduced backoxidation rates, complementary measures such as decreasing the charge recombination rates and enhancing the product transfer rates away from the surface (increasing the mass transfer rates) are necessary for increased photocatalytic water splitting activity.

Mass transfer limitations
Mass transfer limitations especially in the slurry photocatalytic systems can be the most overlooked problem in the photocatalytic field. To complete the photocatalytic reaction cycle, adsorption of the reactants, reduction/oxidation of the reactants, desorption of the products, and transfer of the products from the photocatalyst surface to the gas phase need to be realized. When the rates of the mass transfer of the products from the surface are slower than the reduction/ oxidation rates, produced H 2 and O 2 would stay longer on the surface, resulting in promotion of back-oxidation reactions. Moreover, when the mass transfer rates are slower than the reaction kinetics, the apparent H 2 and O 2 evolution rates in the gas phase will be limited by the mass transfer rates.
Experimental evidence for mass transfer limitations in agitated systems is presented in a previous publication [96]. In a batch slurry reactor, where the catalyst particles are suspended via agitation, observed H 2 evolution rates for UVirradiated Pt/TiO 2 photocatalyst showed improvement with increasing stirring rates up to 900 rpm (Figure 4a). This improvement is a direct indication of mass transfer limitations on the solid-liquid and gas-liquid interfaces as the turbulence in the liquid and therefore boundary layers are affected by increasing stirring rates. In another experiment, the effect of liquid volume is investigated by varying catalyst weight and liquid volumes (keeping the catalyst concentration constant). H 2 evolution rates on an hour basis (μmol H 2 /h) are found the same regardless of the liquid volume (or catalyst weight) above 62.5 ml (Figure 4b) as the H 2 evolution rate per gram and hour basis decreased as liquid volume increased. Similar H 2 evolution rates regardless of the catalyst weight indicate significant mass transfer limitations in the liquid-gas interface.
Mass transfer limitations for different photocatalytic reaction systems are analyzed by different groups. For immobilized photocatalyst systems, the importance of internal mass transfer resistance is emphasized [97]. In another investigation, severe mass transfer limitations are observed in the product separator (liquid-gas interface) for a fluidized bed/separator system, in which modification of the liquidgas surface area enhanced the H 2 evolution rates by 350% [98].
To prevent mass transfer limitations in the photocatalytic tests and to report actual kinetic rates; stirring rates, liquid levels, and mass transfer areas should be designed carefully. To design these parameters, approximate mass transfer rates should be known. Here, we present a sample calculation for H 2 mass transfer rate in a slurry reactor containing 0.5 g TiO 2 photocatalyst having a surface area of 40 m 2 /g inside an agitated glass reactor having 200 ml liquid volume and a tank diameter of 7 cm.
Mass transfer resistances in a gas-liquid-solid multiphase photocatalytic systems involve the internal mass transfer, mass transfer from the solid catalyst particles to liquid (Eq. 6), transfer from the liquid bulk to the liquid interface (Eq. 7), and transfer from the liquid-gas interface to the gas phase (Eq. 8). Photocatalysts such as perovskites and TiO 2 are known to be nonporous (unless mesoporous versions are prepared on purpose [99,100]) and have surface area values between 5 and 50 m 2 /g. For nonporous photocatalysts, the internal mass transfer limitations can be discarded (Eq. 9). Hence, the H 2 mass transfer rate equation will have a form containing the mass transfer resistances from the solid-liquid and liquid-gas interfaces as seen in Eq. 8: The mass transfer limitations coming from the solid-liquid and liquid-gas interfaces may play important role depending on the photocatalytic reactor type. The most often used photocatalytic reactor systems such as slurry reactors have solidliquid and liquid-gas phase interfaces that suspend its catalysts by agitation using an impeller or a magnetic stirrer. The convection mass transfer coefficient for solidliquid interface of such a system could be estimated using Eq. 10 suggested by Armenante and Kirwan for agitated tanks using Kolmogorov's theory for Reynold's number calculation to consider the effect of solid particle size [101]: where k s is the convection mass transfer coefficient from solid to liquid in (m/s), d p is the particle diameter in (m), D H2ÀH2O is the diffusion coefficient of H 2 in liquid water (m 2 /s), and Sc is the Schmidt number. A rough estimation for k s for such system can be found in Table 5.
The creation of air/inert bubbles in the continuous phase (water) due to agitation could be considered as the transfer mechanism of produced H 2 from the liquid phase to the gas phase. In such systems, comparing the mass transfer resistance from liquid to interface and interface to gas, it can be assumed that nearly all of the mass transfer resistance comes from the liquid side of the interface [102], leaving Eq. 9 as Eq. 11: The liquid side mass transfer coefficient for such a system could then be calculated using Calderbank and Moo-Young correlation for rising small bubbles of gas in continuous liquid phase (Eq. 12) [103]: The first term on Eq. 12 is the molecular diffusion term, whereas the second term is for the rise of the bubbles due to gravitational forces independent of the agitation. With estimations on the bubble size and gas holdup of such a system (given in Table 6), the mass transfer coefficient and k L a L term are calculated to be 2.5 Â 10 À4 m/s and 2.2 Â 10 À6 m 3 /s, rendering liquid-gas mass transfer resistance way more important than solid-liquid resistance.
The overall mass transfer coefficient and the mass transfer rate from solid to the gas phase can be calculated with the estimated k s a S and k L a L values. As the concentration of H 2 in the gas phase will be negligible (C G $0), the liquid phase interface can also be assumed to be equal to zero (C H2,Li $0) with negligible gas phase resistance. Therefore, from Eq. 11, the rate of H 2 mass transfer can be calculated by assuming H 2 concentration at the catalyst surface and the gas holdup ratio in the liquid. The rate of H 2 mass transfer values for 200 ml of water and 0.5 g of catalyst are calculated in the range of $200-8000 μmol/h (see Figure 5) for a surface H 2 adsorption capacity range of 50-400 μmol/g (H 2 chemisorption on 0.1% Pt/TiO 2 is reported to be $400 μmol/g at room temperature [104]) and gas holdup ratio between 0.001 and 0.005. The H 2 concentration on the solid surface and the gas-liquid contact area are not easy to estimate. However, for the limited gas-liquid area, the photocatalytic reaction rates above the calculated mass transfer rate will be suppressed due to the limiting mass transfer rates. Therefore, special care must be given for the UVirradiated photocatalytic systems, in which observed H 2 and O 2 evolution rates are found to be close to the calculated mass transfer rates here (see Table 1).
These studies show that for each type of photocatalytic system that contains limited gas-liquid contact area or immobilized photocatalyst, mass transfer limitations should not be underestimated, and not only the materials but also the systems should be improved for better photocatalytic efficiencies.

Parameter
Unit Value Bubble size, d b μm 700 Density of water, ρ L kg/m 3 997 at 25°C Density of air, ρ G kg/m 3 1.18 at 25°C Viscosity of water, μ L Pa s 890 Â 10 À6 at 25°C Diffusion coefficient of H 2 in water m 2 /s 6.3 Â 10 À9 at 25°C Convection mass transfer coefficient for liquid, k L m/s 2.5 Â 10 À4 Liquid-gas bubble contact area, a L ¼ 6VG db ¼ 6ϕ db V L where ϕ ¼ VG= VL m 2 0.008 (gas holdup, ϕ, assumed to be 0.005) k L a L m 3 /s 2.2 Â 10 À6 Table 6. Convection mass transfer coefficient calculation for liquid-gas transfer and parameters used in the calculation.

Future of photocatalytic H 2 evolution
The literature examples of photocatalytic water splitting activities show improvements in visible light utilization, charge separation, and prevention of back-oxidation reactions via fine tuning of photocatalyst materials that enabled more efficient water splitting systems. The efficiency of these systems working under sunlight is better defined with solar-to-hydrogen energy conversion efficiency (STH), i.e., hydrogen production rate times the Gibbs free energy for generating 1 mole of H 2 divided by the power of incident sunlight (Eq. 13): The estimated STH required for the particulate photocatalytic systems to be economically compatible with current H 2 production technologies is 10% [104]. However, the highest STH values obtained with current developed photocatalysts are in the order of 1-2% (1.8% at 400-475 nm using Rh/Cr 2 O 3 -loaded GaN:Mg/ InGaN:Mg photocatalyst [69] and 2% at 420 nm using CDots-C 3 N 4 [70]). The STH conversion efficiency depends both on the catalytic activity and the extent of the utilization of sunlight that depends on the bandgap of the semiconductor. With current photocatalysts having absorption edges around 500 nm, even 100% apparent quantum yields would not guarantee 10% STH values [106]. For a photocatalyst to show 10% STH values, it should have absorption edges at least at 600 nm with apparent quantum yields around 60%. Under the light of these calculations, it can be said that the present photocatalysts having adsorption edge values around 450 nm and quantum yields around 10% are far from being utilized in commercial systems. In order to achieve targeted STH values, the photocatalysts with lower bandgap energies such as (oxy)nitrides and (oxy)sulfides should be improved for H 2 evolution activities while ensuring their thermal stability and photostability.
Large-scale photocatalytic water splitting reactors are implemented with current low STH values as of 2015. The first example of large-scale photocatalytic water splitting utilized Pt-loaded C 3 N 4 photocatalyst with sacrificial electron donor triethanolamine in a flat-panel-type photocatalytic reactor system in 2015 [107]. The solar-to-hydrogen conversion efficiency is reported to be 0.12%, for which the photocatalytic activity is monitored for 30 days. In such systems, where a sacrificial reagent such as triethanolamine or methanol is irreversibly oxidized at a more negative potential than water (thermodynamically more favorable) at the oxidation centers, the photo-generated charges can be more efficiently separated, thus increasing H 2 evolution rates. However, in those systems, hydrogen production is not solely due to the water splitting; as the carbon-and hydrogencontaining "sacrificial agents" are being oxidized at the oxidation centers, they produce hydrogen as well as aldehydes, carboxylic acid, and carbon dioxide [108].
Another large-scale photoreactor is reported by Domen et al., who used Aldoped and RhCrO x -loaded SrTiO 3 photocatalyst sheets in their 1 Â 1 m water splitting panel [45]. The achieved STH value under simulated sunlight is 0.6% at 331 K and limited to a maximum value of 1.4% due to the large bandgap energy of the photocatalyst (3.2 eV). Improved STH values (reaching 1.1%) are shown to be possible on a two-step excitation system in 2016 using photocatalyst sheets having smaller bandgap energy values such as Mo-doped BiVO 4 (2.4 eV) and La-and Rhdoped SrTiO 3 [5].

Conclusion
The developments in the photocatalytic water splitting reactions are explained here with the emphasis on the one-step photocatalysis systems. The early photocatalyst improvements with bandgap engineering, co-catalyst usage, and size reductions are shown to contribute to the increased visible light-driven H 2 evolution activity values. The main drawbacks in the present systems are discussed to be the charge recombination, back-oxidation reactions of the products into water, and mass transfer limitations especially in the three-phase systems. Using defect-free small crystals of photocatalysts and making use of phase junctions or metal cocatalysts are suggested to decrease charge recombination rates. Back-oxidation of H 2 into water or oxygen reduction reaction to water is expected in many noble metal-containing particulate photocatalyst systems. The prevention of these unwanted secondary reactions is shown to be possible to some extent by modification of the noble metal surfaces. Some examples of these modifications are anion coating, partial adsorption of a poison, or nanolayer coating of the co-catalyst or the whole photocatalyst. Selective permeation property of the nanolayer coatings such as Cr 2 O 3 is reported to suppress the back-oxidation rates, resulting in enhanced H 2 and O 2 evolution rates. Possible mass transfer limitations, limiting the observed rates in three-phase systems, are predicted especially in the liquid-gas interfaces.
The literature examples attracted attention for the liquid-gas interfaces in suspended systems and internal mass transfer limitations for the immobilized photocatalyst systems. It is concluded that, in addition to the required developments in activities with suppression of charge recombination, back-oxidation, and mass transfer limitations, future of the photocatalytic systems would necessitate active and stable photocatalysts with narrower bandgap energies (to be activated at >600 nm) for achieving targeted 10% solar-to-hydrogen energy conversion efficiency value.