Solar Fuels by Heterogeneous Photocatalysis : From Understanding Chemical Bases to Process Development

The development of sustainable yet efficient technologies to store solar light into high energy molecules, such as hydrocarbons and hydrogen, is a pivotal challenge in 21st century society. In the field of photocatalysis, a wide variety of chemical routes can be pursued to obtain solar fuels but the two most promising are carbon dioxide photoreduction and photoreforming of biomass-derived substrates. Despite their great potentialities, these technologies still need to be improved to represent a reliable alternative to traditional fuels, in terms of both catalyst design and photoreactor engineering. This review highlights the chemical fundamentals of different photocatalytic reactions for solar fuels production and provides a mechanistic insight on proposed reaction pathways. Also, possible cutting-edge strategies to obtain solar fuels are reported, focusing on how the chemical bases of the investigated reaction affect experimental choices.


•
to increase efficiency in energy production and consumption processes; • to improve the ability to capture and sequester CO 2 from the atmosphere and its utilisation; • to decrease the carbon intensity of the economic system.
While the first strategy relies on optimising already existing technologies for the production and utilisation of traditional fuels, the other two rely on finding either new strategies to obtain alternative energy products or new energy sources.In other words, they require the replacement of fossil fuels, which the world depends on for 80% of energy production [19].This would mean a complete change in the paradigm of energy and transportation technologies, and thus a transition from Anthropocene to Sustainocene, an era in which economic and social wealth meets wise and responsible use of environmental resources [20].Considering these last two strategies more closely, one of them relies on CO 2 as an alternative and sustainable source for the same C-based molecules already used as fuels.Such an approach to CO 2 utilisation could be a significant opportunity for innovative industrial processes and also fit Circular Economy dictamens, providing a concrete integration of economic activities and social and environmental wellbeing in a sustainable way [21].
Conversely, reduction of carbon intensity in the economy represents a more radical approach, which pursues the use of alternative C-free energy vectors and intrinsically unties energy production from carbon dioxide emissions.Nowadays, several C-free energy technologies are available (such as hydropower, geothermal, wind, nuclear, ocean energy and photovoltaics [22]), however, they suffer from being efficient for electricity only, limiting their application in the automotive industry.Recently, among the alternative non-hydrocarbon fuels, hydrogen is an emerging fuel for the future for several reasons: no harmful combustion products are produced; it can be obtained through several technologies and from several sources, and its high diffusivity and high heat capacity make hydrogen generally safer than other fuels [23,24].Moreover, it can be directly used in highly efficient devices, i.e., fuel cells [25][26][27].
Nonetheless, several issues make hydrogen uncompetitive with traditional fuels such as hydrocarbons: insufficient supply for a worldwide market, severe and energy demanding production processes, and last but not least, lack of transportation networks and fuel cells' cost [25,28].Due to the issues related to direct utilisation of hydrogen as a fuel, it can be used indirectly to produce new attractive fuels from different sources, especially those that are bio-derived [29][30][31][32].Nowadays hydrogen is produced mainly from fossil fuels which account for 96% of its production capacity [33,34], but an appealing method to couple sustainable hydrogen production and renewable biomass valorisation is oxygenate compounds reforming [35][36][37][38], which can be used as a hydrogen-storage medium converting into hydrogen on-board through reforming processes, then fed into a fuel cell [39].This kind of energy source is readily available: Sims et al. assessed that 100 EJ/year of raw lignocellulose could be available worldwide at 2-3 $/GJ [40], while Long et al. estimated 251-1272 EJ/year of biomass from surplus cultivated land [41].
Both CO 2 reduction and hydrogen formation are uphill processes, since the products' energy is higher compared to reagents.In fact, if we consider Gibbs free energy of formation for most simple compounds, CO 2 and water are characterised by the lowest value of all C-based molecules (−394.0 and −228.4 kJ•mol −1 ), indicating that any transformation requires energy [48].
Required energy to perform this reaction can be provided in different ways, generally either as heat or electricity.Promising results were obtained both in CO 2 reduction by methanation and in steam reforming of biomass-derived alcohols [49][50][51]; however, these processes require temperatures higher than 400 • C, increasing the final cost of the alternative fuel [52][53][54][55].
A promising alternative energy source is sunlight.The big advantage of sunlight exploitation is that it is available almost anywhere on the Earth's surface, it can be considered renewable on a human time scale.and most importantly, it is inexpensive [56][57][58].In the last couple of decades, the exploitation of sunlight for energy production has gained even more attention.At the moment, the only available technologies in the open market are photovoltaics (PV), which are used for electricity production [59,60].For this reason, the scientific community is actively looking for innovative, sustainable yet efficient processes to convert solar energy into high energy molecules, which are generically called solar fuels.Moreover, Lewis and Nocera have explained why photocatalysis for solar fuels production, can also be thought of as a reliable solution to sunlight intermittency which is an issue in solar energy utilization [61].In recent years, a great number of papers have been published on these topics, providing either an insight on physicochemical properties of these reactions or technologically advanced materials and photoreactor designs, although the connection between the two topics has not yet been sufficiently highlighted.
For this reason, this review is aimed at summarizing the advancement of the research on the two most promising solar fuel production technologies (i.e., CO 2 photoreduction and photoreforming), providing a comprehensive overview on the most innovative solutions to overcome physicochemical limitations by process design.

Proposed Reaction Pathways
The first study on photoelectrocatalytic carbon dioxide reduction using semiconductors was reported by Inoue and co-workers in 1979 [62].Irradiation was used to activate a semiconductor anode which was coupled with a platinum cathode, allowing water oxidation and carbon dioxide reduction.It took only a year from this pioneering result for a published paper on purely photocatalytic CO 2 reduction with water on common semiconductors (such as WO 3 , SrTiO 3 and TiO 2 ) [63].However, the latter approach only became of interest from the late 1990s onwards, when many studies were published on the topic.In all cases, the design of an effective process requires a deep physicochemical knowledge of the reaction.
As with any heterogeneous catalytic process, CO 2 photoreduction also consists of three different steps: 1.
Reactants' adsorption and photons absorption on the photocatalyst; 2.
Heterogeneously catalysed chemical reaction; and

Products' desorption.
The first step is fundamental because it allows both reactants to interact with each other with a suitable orientation for the redox reaction to happen.
When the reaction is performed in water, both as a reductant and a reaction medium, CO 2 solubility in the aqueous phase system must be considered as a possible limiting step for the reaction in the liquid phase whereas in the gas phase, mixability is not an issue to be considered.Gaseous CO 2 adsorption can occur with different geometries, affecting overall process efficiency.In the literature, it is well reported that CO 2 adsorption on titanium oxide is considerably weaker than vapour [64][65][66].In fact, CO 2 adsorption on TiO 2 follows a Freundlich model, which is generally used for non-ideal sorption processes [67]: where q is adsorbed gas (mmol of gas/g of adsorbent), p is pressure ad equilibrium, while k f and n are Freundlich constants.The calculated 1/n value, 0.4, indicates that the predominant CO 2 adsorption mechanism is chemical adsorption rather than physical adsorption.Unfortunately, Krischok et al. observed that when CO 2 and water competitive adsorption occurs, CO 2 adsorption was blocked by the presence of pre-adsorbed water on the titanium dioxide surface while weakly adsorbed CO 2 was displaced by post-dosed H 2 O, and there was little evidence of bicarbonate formation in either case [68].Tan and co-workers calculated adsorption constants for CO 2 and H 2 O on graphene oxide modified TiO 2 according to a Langmuir-Hinshelwood model and calculated a value of 0.0193 and 8.070 bar −1 respectively, confirming findings from previous research [69].
As observed by Henderson and co-workers for TiO 2 based materials [70,71], oxygen from water coordinates with surface titanium sites, whilst protons interact with bridging oxygen, yielding surface hydroxyl groups, which act as adsorption sites for molecular water [72].He et al. described several possible configurations of CO 2 and TiO 2 interactions, calculating binding energies by computational models [73].CO 2 can be linearly adsorbed by Ti-O bonds (Figure 1a) or bidentate chelate-like structures can be observed by the interaction of oxygen atoms in CO 2 with adjacent Ti centres (Figure 1b) or carbonate-like structures can be found when C and O atoms from CO 2 interact with surface Ti and O sites (Figure 1c).According to their calculations, the last two configurations, which involve the weakening of C=O bond, seem to be the most likely intermediate in the reduction route.The first step is fundamental because it allows both reactants to interact with each other with a suitable orientation for the redox reaction to happen.
When the reaction is performed in water, both as a reductant and a reaction medium, CO2 solubility in the aqueous phase system must be considered as a possible limiting step for the reaction in the liquid phase whereas in the gas phase, mixability is not an issue to be considered.Gaseous CO2 adsorption can occur with different geometries, affecting overall process efficiency.In the literature, it is well reported that CO2 adsorption on titanium oxide is considerably weaker than vapour [64][65][66].In fact, CO2 adsorption on TiO2 follows a Freundlich model, which is generally used for non-ideal sorption processes [67]: where q is adsorbed gas (mmol of gas/g of adsorbent), p is pressure ad equilibrium, while kf and n are Freundlich constants.The calculated 1/n value, 0.4, indicates that the predominant CO2 adsorption mechanism is chemical adsorption rather than physical adsorption.
Unfortunately, Krischok et al. observed that when CO2 and water competitive adsorption occurs, CO2 adsorption was blocked by the presence of pre-adsorbed water on the titanium dioxide surface while weakly adsorbed CO2 was displaced by post-dosed H2O, and there was little evidence of bicarbonate formation in either case [68].Tan and co-workers calculated adsorption constants for CO2 and H2O on graphene oxide modified TiO2 according to a Langmuir-Hinshelwood model and calculated a value of 0.0193 and 8.070 bar −1 respectively, confirming findings from previous research [69].
As observed by Henderson and co-workers for TiO2 based materials [70,71], oxygen from water coordinates with surface titanium sites, whilst protons interact with bridging oxygen, yielding surface hydroxyl groups, which act as adsorption sites for molecular water [72].He et al. described several possible configurations of CO2 and TiO2 interactions, calculating binding energies by computational models [73].CO2 can be linearly adsorbed by Ti-O bonds (Figure 1a) or bidentate chelate-like structures can be observed by the interaction of oxygen atoms in CO2 with adjacent Ti centres (Figure 1b) or carbonate-like structures can be found when C and O atoms from CO2 interact with surface Ti and O sites (Figure 1c).According to their calculations, the last two configurations, which involve the weakening of C=O bond, seem to be the most likely intermediate in the reduction route.Very recently, Olivo et al., by means of CO2 adsorption studies using FTIR analyses, observed that all three species are present on metal modified TiO2, but samples with the highest photocatalytic activity proved to be characterised by the highest share of adsorbed CO2 in bidentate structures [74].
From all provided evidence, it is clear that CO2 adsorption on commonly used semiconductors, and titanium dioxide in particular, represents one of the most challenging issues in the development of a photocatalytic process for CO2 reduction.
Considering the sequent step, i.e., photocatalytic electron transfer, without any catalyst or photocatalyst, reduction cannot happen due to the high energy barrier in the gas phase (>200 kJ•mol −1 ) [75].As reported by Nguyen and Ha from theoretical and spectroscopic studies, CO2 can interact with water vapour [76]; oxygen lone pairs in H2O can interact with anti-bonding 2πu at central carbon atoms in CO2, leading to the formation of van der Waals complexes [77].Very recently, Olivo et al., by means of CO 2 adsorption studies using FTIR analyses, observed that all three species are present on metal modified TiO 2 , but samples with the highest photocatalytic activity proved to be characterised by the highest share of adsorbed CO 2 in bidentate structures [74].
From all provided evidence, it is clear that CO 2 adsorption on commonly used semiconductors, and titanium dioxide in particular, represents one of the most challenging issues in the development of a photocatalytic process for CO 2 reduction.
Considering the sequent step, i.e., photocatalytic electron transfer, without any catalyst or photocatalyst, reduction cannot happen due to the high energy barrier in the gas phase (>200 kJ•mol −1 ) [75].As reported by Nguyen and Ha from theoretical and spectroscopic studies, CO 2 can interact with water vapour [76]; oxygen lone pairs in H 2 O can interact with anti-bonding 2π u at central carbon atoms in CO 2 , leading to the formation of van der Waals complexes [77].
Despite the great efforts in understanding CO 2 photoreduction by using conventional catalysis tools, the effect of photons has not been successfully described, and due to substantial differences in experimental procedures and reaction regimes, it is difficult to compare data and assumptions from different teams.UV light is more commonly used than visible light.In fact, bare titanium dioxide, the most common catalyst, requires in the case of anatase phase, a radiation wavelength equal to or lower than 388 nm [78].On the Earth's surface, the average irradiance provided by the sun is 1000 W•m −2 , but only ca. 5% is UV irradiation [79].In most reported CO 2 photoreduction tests, energy input is considerably higher that available solar energy.In fact, irradiance value is usually in the range of 1000-3000 W•m −2 , which is much higher than the UV light fraction in sunlight [80][81][82][83].Results from tests performed at lower irradiances are reported in only a few papers; for instance, Woolerton et [85].Nonetheless, at the moment, papers reporting CO 2 photoreduction tests using an irradiance below this value are very rare [86,87].
Reactions involved in all semiconductors' light harvesting indicate that a high photons input, from a kinetic point of view, enhances the formation of photocatalytically active sites (Equation ( 2)) [88].
However, the reverse reaction can also happen, and according to Liu and Li [89], the recombination rate of e − -h + pairs is nearly two or three orders of magnitude faster than the rate of charge separation/transport and chemical reaction itself.Therefore, at high irradiance conditions, where the surface is saturated with photons, the rate of charge carrier recombination becomes higher than both light adsorption and surface reaction rates [90].Despite these general considerations, there are no specific guidelines for photons' absorption for CO 2 photoreduction.In fact, only Tan et al. have reported the effect of irradiance on methane formation [69].In their paper, they reported that between 650 W•m −2 and 1800 W•m −2 , which is a much higher range than the one considered in this part of the work, methane production increased with irradiance.However, the growth was not linear and the subject was not further investigated.
Electronic transfer is of great importance in CO 2 photoreduction, since desired product formation requires more than two electrons transfer.In addition, according to Rasko and Solymosi, this step determines overall kinetic rates of the process [91].On the contrary, from a thermodynamic point of view, considering reduction potentials, reported in Table 1, methane formation is the most favoured among the C 1 products due to its less negative potential [92].However, kinetic insights on this process are not unanimous; different reaction mechanisms have been proposed for this reaction according to different experimental parameters, such as developed catalyst, irradiation source and reaction medium [64,88,89,94].All proposed mechanisms describe similarly what occurs in oxidative sites on the photocatalyst surface; valence band's (VB) electron deficiency, induced by sufficiently energetic incident irradiation, is responsible for water oxidation to molecular oxygen [95,96].On the contrary, the phenomena involved in the reduction process in the conduction band (CB) have not been clarified yet.
In  [98,99].In the literature, several authors confirmed that the first step is the formation of peroxocarbonate species, which are reduced to formic acid, formaldehyde, and methanol afterwards [89,100,101].Nonetheless, over the last decade, due to the growth in the processes and catalysts developed, a single mechanism cannot be assumed because a variety of reaction pathways can occur according to reaction conditions and catalysts [89,95].
According to CO 2 photoreduction tests performed in aqueous phase, as reported by Liu and Li [89], the presence of peroxide species in water were hypothesised; but, being characterised by a high redox potential, they are extremely unstable, and thus, unreliably detectable.Therefore, peroxocarbonate species are the first intermediates, then, they are directly converted into formate ions, formic acid, and further reduction products.However, formaldehyde is the first product formed when basic aqueous solutions are adopted, and further reduction to methanol or oxidation to formic acid may occur.
According to Galli et al., wide product distribution in liquid phase tests indicates that CO 2 photoreduction occurs by consecutive reaction steps at neutral pH [102] (Equations ( 3)-( 8)).

VB : 2H
CB : Due to the high H 2 O/CO 2 ratio in liquid phase systems, CO 2 undergoes hydrogenation faster than deoxygenation, leading to all oxygenated products found in reaction medium.However, it was observed that TiO 2 catalyst modification with metal electron traps increases hydrogenation rate and improving methane yield by enhancing the stability of electron-hole couples [103,104].
Furthermore, the accumulation of organic compounds in the liquid phase can be followed by hydrogen formation by PR; produced organic compounds act as hole scavengers themselves in a PR reaction (Equation ( 9)).This reaction is very likely to happen with methanol due to its reactivity compared to other C 1 compounds [105,106].
On the contrary, CO 2 photoreduction can also be performed in the gas phase, feeding water as vapour, leading to slightly different results in observed mechanisms.For example, Karamian and co-workers [88] reported that in most cases, in gaseous systems CO is the first intermediate product of CO 2 photoreduction by water vapour [74].
Generally speaking, in the gas phase deoxygenation is faster than hydrogenation and it leads to •C species that are reduced afterwards, whereas in the liquid phase, peroxo species undergo hydrogenation preferentially [107], yielding to possible intermediate products.However, reaction conditions, and in particular, temperature, irradiance, and reaction time modify the reaction pathway and product distribution as a consequence.In detail, if CO 2 deoxygenation is faster than dehydrogenation, methane production is favoured with respect to oxygenated compounds [94].This is the case for the vapour phase reaction, characterised by CO 2 excess.This mechanism involves the formation of •C radicals that recombine with •H originated from water [108].
From all the evidence reported in the literature, briefly summarized in Table 2, it is possible to assume that CO 2 reduction can undergo two different pathways, as shown in Figure 2.    From all the evidence reported in the literature, briefly summarized in Table 2, it is possible to assume that CO2 reduction can undergo two different pathways, as shown in Figure 2.
pressure improves CO2 solubility in reaction medium two parallel reaction pathways are hypothesized [113] Cu/TiO2 The first step is CO2 splitting into CO and O2 [81] Cu/TiO2  Once reagents are turned into products, they need to be desorbed from the catalytic surface so as to leave catalytic sites available for new reagents' molecules to be adsorbed.

Photoreactor Design
The choice of the reactor design and process conditions heavily affect substrate and photon delivery to the photocatalytic surface.While mass transfer is commonly considered in conventional catalysis processes, in photocatalysis, photon transfer must be addressed too.Due to the novelty of this technology and the increase in publications on this topic, different rig configurations are reported in the literature.In fact, if we consider reported photoactivity data for a worldwide benchmark photocatalyst, such as Evonik's P25, with the same conditions of temperature, pressure and light intensity, performances might change either in terms of activity and product distribution [92,116,117].In the first studies, three-phase photoreactors (solid catalyst, liquid reaction medium, gaseous CO 2 ) are generally reported [62].In this case, the solvent, which is water, acts as a reductant as well [118], therefore, several phenomena should be considered for mass transfer such as solubility, diffusion, convection and migration.CO 2 solubility in an aqueous phase system must be considered as a possible limiting step for the reaction in the liquid phase.
From an historical point of view, the first developed photoreactors for CO 2 reduction were batch reactors where the catalyst was suspended in a liquid medium, usually water, and carbon dioxide was bubbled through reaction medium and light reached the reaction medium through quartz windows [62,[119][120][121].This choice of three-phases slurry reactors was, and still is, widely used due to many advantages.These experimental rigs are similar to those used for carbon dioxide photoelectrocatalytic reduction [122,123], and most importantly, they have been widely used in photooxidation processes, which in some cases, are available on a commercial scale [124,125].
In addition, the catalyst does not require any casting or shaping procedure before its use so it can be used as a powder without further modification.Mass transfer is achieved by mixing, either magnetically or mechanically, to provide homogeneity in the reaction medium.
However, liquid phase setups are affected by experimental drawbacks that need to be overcome.The two main drawbacks to be overcome are extremely low carbon dioxide solubility in water (0.03 M at 25 • C and 1 atm) [126], and inefficient light harvesting by medium scattering [127,128].The most common strategy to solve the first problem is the introduction of a base to improve CO 2 solubility.Many authors use basic solutions as reaction media, which leads to the formation of more soluble and stable carbonates and bicarbonate species in the solution [129].Generally, the most used base is sodium hydroxide [130,131], but use of other bases is reported, such as sodium bicarbonate [132], potassium biphosphate [133] and triethanolamine (TEA) [134].
Quin and co-workers did not use water as a reaction medium, but used methanol, whose CO 2 solubility is much higher [135], and they observed that the main product was methyl formate which was obtained by esterification of the solvent and formic acid from CO 2 reduction.Alternatively, Liu and co-workers used isopropanol as a solvent, increasing CO 2 solubility and leading to similar results [136].However, it has been observed that these solvents can act as sacrificial agent and be oxidised in TiO 2 's valence band instead of water [111].This means minor sustainability of the process because in these photocatalytic systems, the reductant is more expensive and industrially derived from fossil sources (mainly natural gas or coal gasification).
Rossetti et al. reported that the use of a pressurised photoreactor increased CO 2 dissolution in aqueous media, yielding higher catalytic activity in methane production avoiding chemical absorption as carbonates in high pH conditions [87], which might lead to reactor walls corrosion.Indeed, the increase in pressure enhances carbon dioxide dissolution in reaction medium to increase reduction products and overall process productivity, but it was observed that reaction thermodynamics hindered photocatalytic performances for values higher than 10 bars [137].Even tuning temperature represents a challenge, since its increase positively influences kinetics and mass transfer, but it reduces CO 2 solubility and favours electron-hole recombination.Kaneco and co-workers performed photoreduction in liquid CO 2 , overcoming solubility problems [138] at 20 • C, but, compared to common photocatalytic reactions, pressure is extremely high (6.5 MPa) in order to maintain CO 2 in the liquid phase.
The second issue to face in three-phase batch reactors is light transfer.In fact, photons are required to travel from light source through reaction medium to the photocatalyst surface, where they are absorbed and activate the photocatalyst.At 20 • C and with 361 nm light wavelength, the water refractive index is 1.34795, whereas the refractive index value is 1.000464 for CO 2 and 1.000256 for water vapour [126,139].This means that light transfer in aqueous solutions is much more difficult in liquid media compared to gas phase systems.
For light harvesting, a variety of photoreactors geometries aimed at maximising catalyst photoactivation, were reported.In the liquid phase, the most commonly-used systems feature a quartz window that allows light to enter [64,140,141].More recently, some papers in the literature have reported the use of an annular design that allows a homogeneous light transfer to the catalyst, mainly in the radial direction [142].For example, Matějová and co-workers used a batch reactor with this geometry: the UV lamp is cast at the centre of the reactor to maximise light harvesting efficiency and it irradiates the aqueous catalyst suspension and CO 2 is bubbled through it to reach saturation, as depicted in Figure 3 [107].
ChemEngineering 2018, 2, x FOR PEER REVIEW 9 of 32 thermodynamics hindered photocatalytic performances for values higher than 10 bars [137].Even tuning temperature represents a challenge, since its increase positively influences kinetics and mass transfer, but it reduces CO2 solubility and favours electron-hole recombination.Kaneco and coworkers performed photoreduction in liquid CO2, overcoming solubility problems [138] at 20 °C, but, compared to common photocatalytic reactions, pressure is extremely high (6.5 MPa) in order to maintain CO2 in the liquid phase.
The second issue to face in three-phase batch reactors is light transfer.In fact, photons are required to travel from light source through reaction medium to the photocatalyst surface, where they are absorbed and activate the photocatalyst.At 20 °C and with 361 nm light wavelength, the water refractive index is 1.34795, whereas the refractive index value is 1.000464 for CO2 and 1.000256 for water vapour [126,139].This means that light transfer in aqueous solutions is much more difficult in liquid media compared to gas phase systems.
For light harvesting, a variety of photoreactors geometries aimed at maximising catalyst photoactivation, were reported.In the liquid phase, the most commonly-used systems feature a quartz window that allows light to enter [64,140,141].More recently, some papers in the literature have reported the use of an annular design that allows a homogeneous light transfer to the catalyst, mainly in the radial direction [142].For example, Matějová and co-workers used a batch reactor with this geometry: the UV lamp is cast at the centre of the reactor to maximise light harvesting efficiency and it irradiates the aqueous catalyst suspension and CO2 is bubbled through it to reach saturation, as depicted in Figure 3 [107].Finally, the use of a fine suspended catalyst might lead to fouling of radiation source, lower active surface contact area and higher separation cost for product collection.Despite technological advances, some of these issues such as high refractive index and poor solubility cannot be completely eliminated, reducing the potentialities of these systems.
Over the years, gas-solid photoreactors have become popular in the literature.This solution finally provides a solution for issues related to CO2 solubility in water.In fact, as any gas, water vapour and CO2 are perfectly mixable and water vapour diffusion in CO2 constant is relatively high (0.138 cm 2 •s −1 ); thus, it is possible to tune reactants' ratio and perform photoreduction in CO2 excess while avoiding water splitting.Moreover, separation is easier since reagents and products are in gas phase while the catalyst is solid.
Mixing the reagents can be performed using different physical phenomena and technological solutions, which differ in their control of the CO2/H2O ratio.Bazzo and Urawaka reported the use of moist quartz wool to generate water vapour in situ [143]; vapour feed control was not precise leading Finally, the use of a fine suspended catalyst might lead to fouling of radiation source, lower active surface contact area and higher separation cost for product collection.Despite technological advances, some of these issues such as high refractive index and poor solubility cannot be completely eliminated, reducing the potentialities of these systems.
Over the years, gas-solid photoreactors have become popular in the literature.This solution finally provides a solution for issues related to CO 2 solubility in water.In fact, as any gas, water vapour and CO 2 are perfectly mixable and water vapour diffusion in CO 2 constant is relatively high (0.138 cm 2 •s −1 ); thus, it is possible to tune reactants' ratio and perform photoreduction in CO 2 excess while avoiding water splitting.Moreover, separation is easier since reagents and products are in gas phase while the catalyst is solid.
Mixing the reagents can be performed using different physical phenomena and technological solutions, which differ in their control of the CO 2 /H 2 O ratio.Bazzo and Urawaka reported the use of moist quartz wool to generate water vapour in situ [143]; vapour feed control was not precise leading to errors in CO 2 /H 2 O ratio calculations.Collado and co-workers used a controlled evaporator mixture to maintain reactants ratio [144] whilst other researchers, such as Tahir and Amin [145] or Cybula et al. [146] employed a water bubbler to saturate flowing CO 2 with water vapour, as reported in Figure 4.In this way, it is possible to tune the water amount controlling temperature, pressure and carbon dioxide flow.However, in the reported papers temperature is not controlled, though it affects water vapour pressure and thus CO 2 /H 2 O ratio, decreasing the tests reproducibility.
ChemEngineering 2018, 2, x FOR PEER REVIEW 10 of 32 4. In this way, it is possible to tune the water amount controlling temperature, pressure and carbon dioxide flow.However, in the reported papers temperature is not controlled, though it affects water vapour pressure and thus CO2/H2O ratio, decreasing the tests reproducibility.Gas-solid systems also allow a great choice of catalyst introduction techniques.Packed bed reactor design is the easiest technological solution, since low pressure drops can be achieved, controlling catalyst particle size and providing promising results [147].However, it suffers from a low irradiated surface area to volume ratio that negatively affects photon harvesting, and thus, light absorption and scattering [148].According to Kočí and co-workers [149], an annular reactor, where the catalyst is embedded within two concentric cylinders and the radiation source is in the centre, improved irradiation homogeneity, despite issues with gases mixing due to small cross-section.
To overcome inhomogeneous irradiance, several solutions have been developed.The use of optical fibres instead of a single light source provides high transmission and irradiation uniformity [150].This kind of irradiation is generally coupled with the use of honeycomb monoliths, which also minimise pressure drops.The catalyst is layered within monolith inner walls by wash-coating or, as reported by Ola and Maroto-Valer [151], it is synthesised in situ by a sol-gel method.However, in these systems, mass transfer of reagents and products to and from the catalytic sites might be too critical and channel opacity might decrease light harvesting efficiency [152].
Finally, another possibility for gas-solid reactors is thin film reactors.In this configuration, the photocatalyst is not immobilised onto beads, fibres or monoliths, but it is deposited on a plate, or even better, on the surface of the photoreactor.In this case, irradiation and light distribution are influenced by geometry and light source with significant effects on global photoactivity [153].Several geometries were reported in the literature, for example, Pathak et al. used Nafion membranes to support photocatalyst film [154] whereas Tan and co-workers used a quartz rod within the photoreactor [69].
Ideally, a solar light driven photoreactor must have [129]:  high coverage area and homogeneous catalyst distribution with good exposure to light;  high CO2 velocity and high mass transfer;  intimate contact between reagents, catalyst and photons;  efficient light harvesting.
Comparing both photoreactors (whose features are summarised in Table 3), gas-solid systems are the most promising option for the design of a photoreactor due to their flexibility in process development and fewer limitations due to mass-related physical phenomena (i.e., CO2 adsorption and light scattering through reaction medium).Gas-solid systems also allow a great choice of catalyst introduction techniques.Packed bed reactor design is the easiest technological solution, since low pressure drops can be achieved, controlling catalyst particle size and providing promising results [147].However, it suffers from a low irradiated surface area to volume ratio that negatively affects photon harvesting, and thus, light absorption and scattering [148].According to Kočí and co-workers [149], an annular reactor, where the catalyst is embedded within two concentric cylinders and the radiation source is in the centre, improved irradiation homogeneity, despite issues with gases mixing due to small cross-section.
To overcome inhomogeneous irradiance, several solutions have been developed.The use of optical fibres instead of a single light source provides high transmission and irradiation uniformity [150].This kind of irradiation is generally coupled with the use of honeycomb monoliths, which also minimise pressure drops.The catalyst is layered within monolith inner walls by wash-coating or, as reported by Ola and Maroto-Valer [151], it is synthesised in situ by a sol-gel method.However, in these systems, mass transfer of reagents and products to and from the catalytic sites might be too critical and channel opacity might decrease light harvesting efficiency [152].
Finally, another possibility for gas-solid reactors is thin film reactors.In this configuration, the photocatalyst is not immobilised onto beads, fibres or monoliths, but it is deposited on a plate, or even better, on the surface of the photoreactor.In this case, irradiation and light distribution are influenced by geometry and light source with significant effects on global photoactivity [153].Several geometries were reported in the literature, for example, Pathak et al. used Nafion membranes to support photocatalyst film [154] whereas Tan and co-workers used a quartz rod within the photoreactor [69].
Ideally, a solar light driven photoreactor must have [129]: • high coverage area and homogeneous catalyst distribution with good exposure to light; • high CO 2 velocity and high mass transfer; • intimate contact between reagents, catalyst and photons; • efficient light harvesting.
Comparing both photoreactors (whose features are summarised in Table 3), gas-solid systems are the most promising option for the design of a photoreactor due to their flexibility in process development and fewer limitations due to mass-related physical phenomena (i.e., CO 2 adsorption and light scattering through reaction medium).Considering reactants feed, bubbling gaseous CO 2 seems the easiest most efficient way to obtain CO 2 -rich reactant mixtures, avoiding the use of sacrificial agents, although it is possible to obtain a stable and constant CO 2 /H 2 O ratio with careful control of CO 2 flow and bubbler temperature.

Proposed Reaction Pathway
Reforming processes and in particular PR, as shown in the introduction, could be a very promising answer for both low-cost hydrogen production and storage.Nonetheless, hydrogen can also be produced from pure water through so-called photocatalytic water splitting [155].As pointed out by Ma et al. [156], a lot of papers make claims about "water splitting" while using a so-called hole scavenger [157]; but since no oxygen is evolved in this case, it is more proper to term these reactions as photoreforming or partial water splitting [156].Despite its attractiveness, this reaction is more difficult than PR, both from a thermodynamic and kinetic point of view, leading to lower hydrogen productivity [158].
The effect of these hole scavengers, or sacrificial reagents, as boosting agents in photocatalytic hydrogen production from water was first observed by Kawai and Sakata in 1980 [159].Today several compounds are known to be effective in PR with methanol [160], ethanol [161] and glycerol [162] being the most used.
The general mechanism of photocatalysis has been widely described by several authors [46,156].Briefly, upon absorption of a sufficiently energetic photon, charge carriers, unless they recombine, can be transferred to a suitable electron acceptor (for electrons) or donor (for holes) [163].CB energy should be higher than the acceptor while VB energy should be lower than the donor, so that reaction can occur [46].In a PR reaction, the acceptor is a hydrogen ion, yielding molecular hydrogen (H 2 ), while the donor, previously labelled as a hole-scavenger, is the reducing agent, either inorganic [164] or organic [165].For the oxidative reaction pathway that consumes the hole scavenger, two main pathways have been proposed, depending on the water/substrate ratio [166]:

•
Direct path, consuming the substrate directly by holes; • Indirect path, a hydroxyl radical-mediated mechanism, where these radicals are produced by interaction of holes with adsorbed water or surface hydroxyl moiety.
Beside these two main pathways, a deeper understanding of degradation mechanism is useful.A general mechanism scheme is reported in Figure 5, as proposed by Puga [46].
General proposed mechanisms of oxygenates PR (oxidative pathway).
It has been proposed that alcoholic moiety, one of the most abundant oxygen-containing functional groups in biomass, undergoes consecutive oxidation to carboxylic acid, eventually decomposing to an alkyl fragment or radical and carbon dioxide (decarboxylation) [167].Nonetheless, this mechanism seems to only be logical for polyols; ethanol is known to yield acetaldehyde as its main co-product [168][169][170], suggesting a key role played by the substrate's structure in enhancing C-C bond cleavage.Besides this mechanism, an alternative pathway was proposed by Bahruji et al. on Pd/TiO2 where alcohol was adsorbed as alkoxide, then decarbonylation yielded an alkyl fragment and CO, which further oxidized to CO2 by holes [171].Concerning the asformed alkyl fragments, several pathways have been proposed, for example, recombination with either hydrogen atoms or other alkyl fragments yields hydrocarbons [172], while coupling with hydroxyl radicals gives alcohols [173], that can be further oxidised until complete mineralization of the carbon skeleton to CO2.This suggests a higher availability of these radicals improves substrate degradation.Finally, unexpected by-products suggest other side-pathways such as etherification and esterification [174], ketonisation [175], radical coupling [176] and hemiacetalisation [168].
Thus, the reaction mechanism is a complex phenomenon, strongly substrate and catalystdependent.Nonetheless, the key point of reaction pathway knowledge is to understand how byproducts are produced and how these side processes can be correlated to catalyst and reaction conditions, as will be explained in the following sections.

Catalyst Formulation
The ideal photocatalyst should be selective to complete organics mineralisation, stable and as cheap as possible in order to be applied on an industrial scale.The greatest challenge for a photocatalyst is strong visible light absorption and high carrier mobility (i.e., reduced recombination) [177] in order to harvest solar light efficiently.Moreover, visible light absorption allows lower energy losses since the energy required to excite the photocatalyst exceeds the thermodynamic requirements of reforming [178].Concerning photocatalytic material properties, little attention has been focused on important parameters such as band edges, optical absorbance and carrier mobility [177].Knowledge of these properties is important for efficient light harvesting and utilisation.
TiO2 is certainly the most studied and well-known material in photocatalysis due to its inexpensiveness, safety, chemical stability and photostability [156,179].TiO2 aside, different materials have been reported such as zinc oxide (ZnO) [180], copper oxides (CuO and Cu2O) [181], strontium titanate (SrTiO3) [182], sodium tantalate (NaTaO3) [183], indium tantalate (InTaO4) [184], cadmium sulphide (CdS) [185], tantalum nitride (Ta3N5) [186], tantalum oxynitride (TaON) [187], graphene It has been proposed that alcoholic moiety, one of the most abundant oxygen-containing functional groups in biomass, undergoes consecutive oxidation to carboxylic acid, eventually decomposing to an alkyl fragment or radical and carbon dioxide (decarboxylation) [167].Nonetheless, this mechanism seems to only be logical for polyols; ethanol is known to yield acetaldehyde as its main co-product [168][169][170], suggesting a key role played by the substrate's structure in enhancing C-C bond cleavage.Besides this mechanism, an alternative pathway was proposed by Bahruji et al. on Pd/TiO 2 where alcohol was adsorbed as alkoxide, then decarbonylation yielded an alkyl fragment and CO, which further oxidized to CO 2 by holes [171].Concerning the as-formed alkyl fragments, several pathways have been proposed, for example, recombination with either hydrogen atoms or other alkyl fragments yields hydrocarbons [172], while coupling with hydroxyl radicals gives alcohols [173], that can be further oxidised until complete mineralization of the carbon skeleton to CO 2 .This suggests a higher availability of these radicals improves substrate degradation.Finally, unexpected by-products suggest other side-pathways such as etherification and esterification [174], ketonisation [175], radical coupling [176] and hemiacetalisation [168].
Thus, the reaction mechanism is a complex phenomenon, strongly substrate and catalyst-dependent.Nonetheless, the key point of reaction pathway knowledge is to understand how by-products are produced and how these side processes can be correlated to catalyst and reaction conditions, as will be explained in the following sections.

Catalyst Formulation
The ideal photocatalyst should be selective to complete organics mineralisation, stable and as cheap as possible in order to be applied on an industrial scale.The greatest challenge for a photocatalyst is strong visible light absorption and high carrier mobility (i.e., reduced recombination) [177] in order to harvest solar light efficiently.Moreover, visible light absorption allows lower energy losses since the energy required to excite the photocatalyst exceeds the thermodynamic requirements of reforming [178].Concerning photocatalytic material properties, little attention has been focused on important parameters such as band edges, optical absorbance and carrier mobility [177].Knowledge of these properties is important for efficient light harvesting and utilisation.TiO 2 is certainly the most studied and well-known material in photocatalysis due to its inexpensiveness, safety, chemical stability and photostability [156,179].TiO 2 aside, different materials have been reported such as zinc oxide (ZnO) [180], copper oxides (CuO and Cu 2 O) [181], strontium titanate (SrTiO 3 ) [182], sodium tantalate (NaTaO 3 ) [183], indium tantalate (InTaO 4 ) [184], cadmium sulphide (CdS) [185], tantalum nitride (Ta 3 N 5 ) [186], tantalum oxynitride (TaON) [187], graphene [188] and carbon nitride (C 3 N 4 ) [189].Besides pure materials, solid solutions have also been used to tune the band structure and improve visible light absorption.Examples include bismuth-yttrium vanadate (Bi x Y 1−x VO 4 ) [190] and zinc-cadmium sulphide (Zn 0.5 Cd 0.5 S) [191].In order to improve the catalytic activity of pristine material, several strategies have been widely reviewed in the literature [57,156], targeting visible light harvesting, reducing electron-hole recombination and improved reaction kinetics.Such approaches include nanostructure architecture [180,[192][193][194], doping [193,195,196], heterojunction [197][198][199][200][201][202][203], use of plasmonic metals [186,204,205], as well as co-catalysts [164, 206,207].Concerning nano-composites (joining of two or more materials at nanoscale), an interesting concept is the so-called Z-scheme.Bard suggested the first mechanism proposal of this system in 1979 [208] and the first all-solid-state Z-scheme system for overall water splitting, was developed by Sasaki et al. [209].As pointed out by Zhou et al., it involves the transfer of an electron from the lowest-lying VB to the highest-lying CB of the composite, thus gaining in reactivity for both electron and holes compared to heterojunctions [210].Few Z-scheme systems have been stated for PR [211,212], but very appealing systems are reported in the literature, such as CdS/Au/WO 3 [194] and g-C 3 N 4 /WO 3 [213].These nano-composites are able to fully harvest visible light, while exploiting high-reactive charge carriers for photocatalytic reactions, as pictured in Figure 6.Due to great potentiality of Z-scheme systems and the wide variety of available semiconductors [177,210], a lot of work can be done on assessing diverse materials on their nano-composites' photocatalytic properties.
ChemEngineering 2018, 2, x FOR PEER REVIEW 13 of 32 tune the band structure and improve visible light absorption.Examples include bismuth-yttrium vanadate (BixY1−xVO4) [190] and zinc-cadmium sulphide (Zn0.5Cd0.5S)[191].In order to improve the catalytic activity of pristine material, several strategies have been widely reviewed in the literature [57,156], targeting visible light harvesting, reducing electron-hole recombination and improved reaction kinetics.Such approaches include nanostructure architecture [180,[192][193][194], doping [193,195,196], heterojunction [197][198][199][200][201][202][203], use of plasmonic metals [186,204,205], as well as co-catalysts [164, 206,207].Concerning nano-composites (joining of two or more materials at nanoscale), an interesting concept is the so-called Z-scheme.Bard suggested the first mechanism proposal of this system in 1979 [208] and the first all-solid-state Z-scheme system for overall water splitting, was developed by Sasaki et al. [209].As pointed out by Zhou et al., it involves the transfer of an electron from the lowest-lying VB to the highest-lying CB of the composite, thus gaining in reactivity for both electron and holes compared to heterojunctions [210].Few Z-scheme systems have been stated for PR [211,212], but very appealing systems are reported in the literature, such as CdS/Au/WO3 [194] and g-C3N4/WO3 [213].These nano-composites are able to fully harvest visible light, while exploiting high-reactive charge carriers for photocatalytic reactions, as pictured in Figure 6.Due to great potentiality of Z-scheme systems and the wide variety of available semiconductors [177,210], a lot of work can be done on assessing diverse materials on their nano-composites' photocatalytic properties.Among the other approaches mentioned above for activity enhancement, the use of co-catalysts is an appealing method, in particular, by improving reaction kinetics [214].An increase in hydrogen evolution up to 2-3 magnitude order compared to pristine material has been reported [215,216].The role of co-catalysts is to trap excited electrons [217] or, less commonly, holes [218], and act as a reaction site [214].Reduction co-catalysts are the most used to increase catalytic activity and usually rely on noble metals such as gold (Au) [168,219], platinum (Pt) [193,220] and palladium (Pd) [171].Chiarello et al. observed increased reactivity upon co-catalysts loading, in order, Ag < Au < Pt, explained by their increasing work function, and thus electron-trap capability [174].Al-Azri et al. found an increase in activity, in order, Au < Pt < Pd, suggesting the higher reactivity of Pd is due to higher Fermi level and density of states despite it having a lower work function than Pt [178].Besides expensive noble metals, non-noble ones have been evaluated too.Copper (Cu) [170] and nickel (Ni) [221,222] have proved to be as active as noble metals in improving photocatalysts' activity.Metal Among the other approaches mentioned above for activity enhancement, the use of co-catalysts is an appealing method, in particular, by improving reaction kinetics [214].An increase in hydrogen evolution up to 2-3 magnitude order compared to pristine material has been reported [215,216].The role of co-catalysts is to trap excited electrons [217] or, less commonly, holes [218], and act as a reaction site [214].Reduction co-catalysts are the most used to increase catalytic activity and usually rely on noble metals such as gold (Au) [168,219], platinum (Pt) [193,220] and palladium (Pd) [171].Chiarello et al. observed increased reactivity upon co-catalysts loading, in order, Ag < Au < Pt, explained by their increasing work function, and thus electron-trap capability [174].Al-Azri et al. found an increase in activity, in order, Au < Pt < Pd, suggesting the higher reactivity of Pd is due to higher Fermi level and density of states despite it having a lower work function than Pt [178].Besides expensive noble metals, non-noble ones have been evaluated too.Copper (Cu) [170] and nickel (Ni) [221,222] have proved to be as active as noble metals in improving photocatalysts' activity.Metal oxides such as copper(II) oxide (CuO) [216,223], nickel oxide (NiO) [224] and cobalt oxide (CoO x ) [164] have also been reported to be suitable co-catalysts, although NiO has been proved to be converted in situ in metallic Ni, the truly active phase [222].Ni and Cu-based co-catalyst are surely the most reported [46], but cobalt-based materials also show interesting activity in hydrogen evolution [225] even though they are scarcely reported in literature.Moreover, few comparisons among these non-noble metal co-catalysts exist.Fujita et al. reported an increased reactivity, in order, CoO x < NiO < CuO on TiO 2 [226], while Sun et al. observed an improved hydrogen evolution, in order, RuO x < NiO < CuO < CoO x < Pt on a barium titanate [158], suggesting that interactions between the co-catalyst and the main semiconductor play a crucial role too.Alloying has been proved to boost the catalytic activity [227][228][229], probably due to synergistic effects among the two components as suggested by Jung et al. [230]; unfortunately, these reported alloys contain at least one noble metal, increasing the cost of the catalyst.Further work on purely non-noble metal alloys is necessary to gain a comprehensive knowledge of alloyed co-catalysts.Interesting alternatives to metal and metal oxides are metal sulphides, such as molybdenum disulphide (MoS 2 ) [231,232] and tungsten disulphide (WS 2 ) [233,234]; Zong et al. demonstrated a higher activity of MoS 2 compared to noble metals on CdS photocatalyst [235].Oxidation co-catalysts are widely studied for overall water splitting [155], and compared to oxide co-catalyst (RuO x , IrO x , CoO x ), cobalt phosphate (Co-Pi) has been reported to obtain higher activity in oxygen evolution [236] and to also exhibit self-healing properties in the presence of phosphate ions in solution, thus improving its stability [237].Recently, Di et al. reported an enhancing of activity in PR with Co-Pi co-catalyst, thus boosting the oxidative side of the reaction mechanism [238].Beyond the choice of the material, the loading is also important.Usually a very low amount of co-catalyst is able to boost the activity, for instance Kondarides et al. assessed 0.1-0.5% as the optimal Pt loading on TiO 2 [162] while Chen et al. estimated that 1.25% of CuO [223] and 0.5% of Ni [222] gave the best result on TiO 2 .Increasing the loading certainly increases the number of available active sites, but above a certain upper value the activity decreases due to enhanced charge carrier recombination, light-shielding effect or the decrease of active sites at the metal-semiconductor interface [162].Besides activity enhancement, the co-catalyst plays a significant role in selectivity [174,239], moreover, its size has been proved to have a remarkable role in enhancing both activity and selectivity of PR reactions [240].Finally, as reported by Kawai and Sakata in 1980 [159], the use of two co-catalysts (for both reduction and oxidation half-reactions) could remarkably improve photocatalytic performance, thus, suggesting further development on this side.
In Table 4 some of the most interesting used photocatalysts are summarized, reporting reaction conditions and product formation.
The key point of this section is that coupling two different materials is necessary in order to fulfil the requirement discussed at the beginning of this section.In this context, co-catalysts play a crucial role in catalytic activity and selectivity enhancement.

Reaction Conditions
Photocatalyst improvement is just part of the story; reaction conditions play an important role in efficiency enhancement of the overall process.Herein, we focus attention on liquid vs gas phase reaction conditions, since, even though some remarkable differences exist, they have scarcely been highlighted in the literature for PR reaction.
Similarly to CO 2 photoreduction, PR has been extensively carried out in the liquid phase [161,162,168,191,216,241]. Gas phase conditions have been barely reported and only volatile compounds such as methane [242,243], methanol [244,245] and ethanol [170,175,246] have been used.Nevertheless, as reported by Ampelli et al., gas phase conditions have several advantages over liquid conditions [170], such as: Several other reaction parameters can affect both the activity and selectivity of PR reactions; these will be briefly summarised here, along with the underlying issues and key points.
Since photocatalysis is a light-driven process, the obvious parameters are light wavelength and intensity in both the liquid and gas phase.Concerning the former, it is known that the lower the incoming photon wavelength, the higher the hydrogen evolution rate [247] and the apparent quantum yield (AQY) [243].This experimental behaviour is unexpected since the as-formed charge carrier experienced a fast thermalisation (a few hundred femtoseconds) [248], losing their extra-energy at the bottom of CB.However, Xu et al. proposed two explanations for faster methanol dissociation on TiO 2 (110) surface: direct injection of electrons in methanol's lowest unoccupied molecular orbital (LUMO), or a phonon-assisted methanol dissociation using energy from thermalisation [249].Speaking of the latter (light intensity or irradiance), as previously explained in Section 2.1, lower irradiances improve photon utilization and boost AQY.Unfortunately, few reports exists about the effect of light intensity on PR activity [46]; despite this, a wide variety of light intensities have been used from 2 W•m −2 [175] to 250 W•m −2 [250].It is worth highlighting that light wavelength and intensity affect the photon utilization efficiency, rather phenomena involving reagents, products or reactions intermediates such as adsorption, desorption or change in activation energy.
Although the energy source in photocatalysis is light, thermal energy can play a significant role within a narrow temperature range (20-90 • C) [46].It is widely known that a higher temperature increases hydrogen yield for both the gas [175] and liquid phase [162,251] as reaction medium.Concerning the former, Taboada et al. suggested such evidence is due to improved desorption of reaction intermediates [175].Nonetheless, Caravaca et al. observed that on Pt/TiO 2 , methanol reforming occurs without irradiation at temperatures higher than 160 • C, thus suggesting a shift to a pure thermocatalytic process [245].
Substrates and their concentrations strongly influence both activity and selectivity.Several compounds can be used as hole scavengers in PR, but the most interesting are certainly oxygenates organic compounds because they have higher reactivity than hydrocarbons and potentially lower environmental impact (biomass or biomass-derived compounds) [42].Hydrocarbons and fossil fuels have been used too [173].Yoshida et al. reported photocatalytic methane conversion that, despite requiring noble metal co-catalyst and high methane-water ratio, successfully yielded hydrogen and CO 2 [242,243].We have cited this process since it could be an appealing way to move from traditional thermocatalytic reforming processes to milder and less energy-demanding PR processes on well-established fossil fuel-based technologies.
Concerning oxygenates, methanol, ethanol and glycerol are certainly the most used as mentioned in Section 3.1.Focusing on the relationship between substrate structure and hydrogen evolution activity, Chen et al. observed, using a Pd/TiO 2 catalyst, increased reactivity with an increasing number of hydroxyl moiety and α-hydrogen on the substrate, and reported the following order of reactivity: glycerol > 1,2-ethanediol > 1,2-propanediol > methanol > ethanol > 2-propanol > tert-butanol [178].Bahruji et al. also reported a similar effect on alcoholic substrates (methanol, ethanol, 2-propanol, 1-propanol, 1-butanol and tert-butanol) [171].Regarding selectivity, it is known that methanol [252] and glycerol [162] afford a complete conversion to CO 2 (mineralisation), while ethanol gives acetaldehyde as the main co-product [168], suggesting a key role of hydroxyl moiety in activation of C-C bond cleavage.It is also known that side pathways yield alkanes from the alkyl chain of alcohols [171], thus lowering hydrogen yield.It is worth noting that these alcohols can be obtained from biomass through chemical [253][254][255] or biochemical [256] processes, while glycerol is an abundant by-product of the biodiesel industry [257].Other organic compounds have been also used, such as carboxylic acids and aldehydes, since they constitute bio-oil [258] and wastewater [259].Wu et al. compared formaldehyde and formic acid, recognised as intermediates in methanol PR, with methanol itself, and found an increase in reactivity as follows: methanol < formaldehyde < formic acid [240].Likewise, acetic acid was found to be more active than acetaldehyde, while formic acid has showed higher hydrogen yield [252], thus confirming higher reactivity of carboxylic acids and lower reactivity for compounds with longer alkyl chains.Carbohydrates have been extensively used in PR reactions too and are of particular interest as the main components of biomass [260].Kawai and Sakata firstly reported the use of these compounds in photocatalytic hydrogen evolution, observing a H 2 /CO 2 ratio close to stoichiometric value, meaning complete mineralisation of sugar occurred, and a decrease in reactivity moving from simple sugars to polysaccharides [159].Kondarides et al. studied the reactivity for simple sugars (lactose, cellobiose, maltose), and like Kawai and Sakata, they observed a decrease in hydrogen evolution for starch and cellulose due to their more complex structure [261].Further study by Caravaca et al. assessed that the size of cellulose also plays a significant role, and moving toward a real lignocellulosic substrate, observed lower yield in hydrogen, although it was higher than pure water [221].Finally, Speltini et al. have reported some interesting results on waste PR, in particular on olive mill wastewater (OMW) [220] and swine sewage [262], although no complete conversion to CO 2 was observed.PR reaction has also been carried out in a pilot plant using wastewater as substrate and yielding hydrogen, although no complete mineralisation to CO 2 was observed in this case [241].
Another important parameter is reagent concentration.Langmuir's equation has often been observed to fit experimental data on PR reactions in the liquid phase with different substrate and photocatalysts [167,191,263]: where r H 2 is the reaction rate, C 0 is the initial substrate concentration, k H 2 is the rate constant for hydrogen evolution and K is the adsorption constant.This means that the phenomenon is controlled by substrate adsorption on the catalyst's surface [167].It has been observed that the addition of little amounts of a hole scavenger (substrate) can increase hydrogen evolution compared to pure water [162], making this process suitable for simultaneous pollutant removal and hydrogen production from very diluted aqueous medium.As the concentration of hole scavenger increased, the hydrogen production was improved too, in both liquid [264] and gas phase [168] systems.Nevertheless, besides a certain upper water-substrate ratio, a decrease in activity or a constant value has been observed in both gas [174] and liquid phase [191,262] conditions, meaning that water becomes the limiting reagent.Moreover, Chiarello et al. suggested that water plays a key role in gas phase conditions, not only as oxidant but also as a proton conductor, allowing a facile transfer from oxidation to reduction sites [166].
Effect on the activity aside, the water-substrate ratio also affects selectivity improving, for higher ratios, complete conversion to CO 2 and reduction of side-products [174].
In the liquid phase, medium acidity and catalyst concentration also have a significant role.The effect of pH is strongly substrate-dependent: methanol [240], acetic acid [172] and cellulose [265] PR on TiO 2 was favoured at neutral pH; OMW, a phenolic-rich mixture, gave better results in acidic media [220] while ethanol [264], glycerol [163], glucose [167] and sewage [262] showed best performances in alkaline solutions.The acidity also plays a crucial role in selectivity.Sakata et al. observed a remarkable decrease in methane formation due to side reaction, in acetic acid PR at higher pH, despite the lower hydrogen yield [172].As observed by Simon et al., using photoluminescence and transient absorption spectroscopy, alkali-enhanced activity on ethanol PR is due to improved formation of hydroxyl radicals from hydroxyl ion-rich medium, that boost organics decomposition [266].Finally, one should keep in mind that pH also affects substrate adsorption on catalyst's surface [191,220], catalyst's particle dispersion [240] and pH-dependence on CB and VB [172,266].
Concerning the concentration of catalyst in the suspension reaction medium, it is known that too low an amount of catalyst hinders the activity while too much powder in suspension reduces light penetration [162,262].

Photoreactors Design
Hence, we want to consider reactor configurations, which is hardly reported and reviewed for PR.Liquid phase reactions are usually carried out in slurry-type reactors, in which the powdered photocatalyst is suspended in the aqueous medium, allowing a good irradiation pattern of the catalyst.The major drawback of this configuration is the catalyst recovery, as experienced in water photoremediation.Immobilisation on a solid support avoids complex separation procedures, though it lowers catalytic activity due to mass-transfer limitations [267].Concerning gas phase reactions, several different reactor configurations have been reported, such as packed bed [244], plate thin-film [170] and coated honeycomb [175], although no direct comparison among them has been done.It is worth observing that reactor design, as reported in Section 2.2, has a crucial role in photocatalysis, since as we recently report for CO 2 photoreduction, a proper irradiation pattern creates higher productivity with lower amounts of catalyst [74].Nonetheless, there is plenty of work that can be done on PR reactors' design.
Eventually, the purification of hydrogen in gaseous streams from CO 2 and other by-products should be addressed as issues, in particular, for scaling-up this technology.This problem is actually handled by traditional hydrogen production processes (e.g., methane steam reforming, MSR) [268].CO 2 removal, the sole H 2 co-product in stoichiometric PR reaction, can be gained by absorption in liquid alkaline solution, as in post-combustion CO 2 capture technologies [269].The resultant CO 2 -rich solution can be used for CO 2 photoreduction in the liquid phase as previously shown in Section 2.2; the combination of these two processes should fulfil both H 2 purification (from CO 2 ) and avoids CO 2 emission into the atmosphere, while maintaining photocatalysis advantages of mild conditions and solar light utilization.Nonetheless, up until now, none of these processes have been coupled to a PR reaction rig in both gas and liquid phase medium equipment.This additional knowledge can be useful for deeper study on reactor design, for scaling-up PR technology.
In conclusion, the choice of reaction conditions has a crucial impact (Table 5) on process energy efficiency and on enhancing both activity and selectivity.Currently a lack of knowledge about the effect of some parameters such as light irradiance and reactor configuration, as well as no reported works on H 2 purification directly from gaseous PR streams, are critical issues, particularly for scaling-up PR processes and light utilisation efficiency assessment.Moreover, the incomplete conversion to CO 2 , i.e., the selectivity of the process, is a crucial issue, since it reduces hydrogen yields, requiring further processing steps in order to "clean" the effluents.This additional knowledge could be useful for evaluating the industrial reliability of this technology as a cheap H 2 source, particularly if one wants to compare PR with other available H 2 production processes.

Conclusions
CO 2 photoreduction and biomass PR are very challenging processes that could withstand the transition from Anthropocene to Sustainocene, as cited in the introduction.The main advantage of photocatalysis is the direct use of sunlight, a very abundant and cheap primary energy source.Nonetheless, several issues hinder the application on an industrial scale; through this review the main concerns were described.It is clear that upgrading photocatalysis processes is a comprehensive challenge that requires a multidisciplinary approach.Reaction conditions are usually less investigated than photocatalyst formulation, but significantly affect the activity enhancement and further scale-up of these processes.Finally, reactor design is surely a key issue in scaling-up, since both mass-transfer and light-transfer phenomena heavily affect overall efficiency of the process.Though further work is needed, huge efforts have been made on improving photocatalysts, whilst process conditions and reactor design still need to be implemented to achieve promising results, particularly for PR.Thus, there is plenty of work that should be done to improve knowledge and applicability of photocatalysis to solar fuels production.

Figure 2 .
Figure 2. Different reaction mechanisms in the vapour and liquid phases.CB and VB energy levels have been taken from [109].

Figure 2 .
Figure 2. Different reaction mechanisms in the vapour and liquid phases.CB and VB energy levels have been taken from [109].

Figure 4 .
Figure 4. Example of a photoreactor for gas-solid CO2 photoreduction where a water saturator is employed.

Figure 4 .
Figure 4. Example of a photoreactor for gas-solid CO 2 photoreduction where a water saturator is employed.

Figure 6 .
Figure 6.Z-scheme of visible light harvesting g-C3N4/WO3 nano-composite.CB and VB energy levels have been taken from [213].

Figure 6 .
Figure 6.Z-scheme of visible light harvesting g-C 3 N 4 /WO 3 nano-composite.CB and VB energy levels have been taken from [213].
al. performed CO 2 photoreduction tests in the aqueous phase under 450 W•m −2 irradiance [84], whilst more recently Tahir et al. reported tests conducted at 200 W•m −2 in the vapour phase conditions

Table 2 .
Summary of proposed mechanisms found in the literature.

Table 2 .
Summary of proposed mechanisms found in the literature.

Table 3 .
Summary of features to consider in CO 2 photoreactor design.

Table 4 .
Summary of some used photocatalysts for biomass PR.

Table 5 .
Summary of features to consider on PR reaction improvement.