Liquid vs . Gas Phase CO 2 Photoreduction Process : Which Is the Effect of the Reaction Medium ?

The use of carbon dioxide, the most concerning environmental issue of the 21st century, as a feedstock for fuels productions still represents an innovative, yet challenging, task for the scientific community. CO2 photoreduction processes have the potential to transform this hazardous pollutant into important products for the energy industry (e.g., methane and methanol) employing a photocatalyst and light as the only energy input. In order to design an effective process, the high sustainability of this reaction should be matched with the perfect reaction conditions to allow the reactant, photocatalyst, and light source to come together: therefore, the choice of reaction conditions, and in particular its medium, is a crucial issue that needs to be investigated. Throughout this paper, a careful study of carbon dioxide photoreduction in liquid and vapour phases are reported, focusing on their effect on catalyst performances in terms of light harvesting, productivity, and selectivity. Different from most papers in the literature, catalytic tests were performed under extremely low light irradiance, in order to minimise the primary energy input, highlighting that this experimental variable has a great effect on the reaction pathway and, thus, product distribution.


Introduction
Nowadays carbon dioxide emissions represent one of the most threatening issues to the environment due to its considerable contribution to several phenomena, such as the increase in global temperatures in recent centuries [1].The correlation between increasing CO 2 emissions and the use of fossil fuels has been undoubtedly established, but none of the actions put in place seems effective to harness CO 2 emissions [2,3].
Several strategies can be pursued to mitigate carbon dioxide emissions into the atmosphere and their effects [4][5][6]: though geologic or oceanic storage are already applied to this purpose [7][8][9], the transformation into valuable or highly-requested products is conceptually more favourable [10].A very attractive strategy is the conversion of carbon dioxide into fuels, which might lead to a real circular economy, avoiding the use of net CO 2 -producing energy sources.In this way, the product of hydrocarbon combustion, i.e., water and carbon dioxide, are converted into regenerated fuels in a process that was labelled by some authors as artificial photosynthesis [11][12][13][14].
Over the years, CO 2 photocatalytic reduction attracted attention, since it is able to reach this goal using a light source as the primary energy input which, in the foreseeable future, might be possibly obtained from the sun [15,16].The presence of a photocatalyst and its physicochemical properties are essential to the development of a robust and reliable process.At the moment, several semiconductors have been employed (such as ZnO, ZnS, WO 3 , Fe 2 O 3 , ZrO 2 [17][18][19][20]), but titanium dioxide-based materials are the most promising due to large availability, low cost, and the appropriate band-gap for this application [21,22].
Despite great effort on the development of an efficient photocatalyst, the choice of the optimal reaction conditions is not a secondary issue.Due to the novelty of this technology and the flourishing of publications on this topic, different rig configurations are reported in the literature.Since the first study by Inoue and co-workers [23], liquid phase systems have been widely developed.Water use, both as reagent and solvent, has been deeply investigated due to its great availability, inexpensiveness, and environmental friendliness.However, water as a solvent for this reaction suffers from a low carbon dioxide solubility [24,25], which hinders the possibility for CO 2 to interact with the catalyst dispersed in the reaction medium.To overcome this drawback, several strategies have been pursued over the years.
The substitution of water with organic solvents, such as methanol [26], isopropanol [27], or acetonitrile, proved to be effective to increase carbon dioxide solubility and, thus, photoactivity, despite a negative effect on the economic and environmental sustainability of the process.Another possibility is the use of alkali solutions as reaction media [28][29][30].In this case, carbon dioxide is dissolved as bicarbonate ions which are more stable in the aqueous medium but, unfortunately, are characterised by a higher reduction potential, further limiting process efficiency [31].
An alternative possible and most promising strategy is to boost the reaction conditions, which has been previously studied by some of the authors [32].Indeed, the increase of pressure allows enhancing carbon dioxide uptake in the liquid reaction medium, so as to increase reduction products and overall process productivity.
In the last decade, vapour phase reactors became popular over time [33][34][35][36], thanks to an easier reactant mixing, which leads to the possibility to tune the CO 2 /H 2 O ratio and also to perform this reaction under CO 2 excess [37], enhancing in this way carbon dioxide adsorption on the catalytic surface [38].As a matter of fact, Tahir and Amin showed that reagent ratio productivity and product selectivity to methane is maximised for a 1.5 CO 2 /H 2 O ratio, which is above the stoichiometric 0.5 value [39].
Carbon dioxide adsorption aside, light harvesting should be carefully considered because of the lower refractive index of gases when compared to water (1.000464 for CO 2 , 1.34795 and 1.000256 for liquid and vapour water, respectively, at 293 K, 361 nm light wavelength [40,41]).A variety of geometries has been brought up for photoreactors, aimed at maximising catalyst photoactivation.In the liquid phase, most commonly-used systems are featured by a quartz window that allows light to enter [42][43][44].More recently, some papers in the literature reported the use of an annular design that allows a homogeneous light transfer to the catalyst, mainly in the radial direction [32,45].
Generally, energy input is represented by UV light and irradiance, a metric for energy input, is usually in the range of 1000-3000 W•m −2 , which is extremely higher than UV light fraction in sunlight [15,[46][47][48].Very few papers report results from tests performed at lower irrandiances: for example, Woolerton and co-workers performed CO 2 photoreduction tests in the aqueous phase using UVA 450 W•m −2 irradiance [49], while, more recently Tahir et al. reported results at 200 W•m −2 in the vapour phase conditions [50].However, to the best of our knowledge, papers reporting CO 2 photoreduction tests using an irradiance below 100 W•m −2 have not been published yet, except for previous papers from the authors [32,39].The promising performances even at low irradiance conditions is an important step towards the use of the most sustainable and inexpensive light source, i.e., sunlight.
From this overview, it is clear that reaction conditions, and in particular the reaction medium, affect catalyst effectiveness in light harvesting and reactant interaction with it, thus playing a crucial role.Moreover, in literature a wide variety of photocatalysts and photoreactors have been reviewed Energies 2017, 10, 1394 3 of 14 but the reaction medium effect on catalytic performance is still not clear and, to the best of the author's knowledge, this topic has not been experimentally investigated yet.Therefore, the goal of the work is the assessment of experimental conditions for CO 2 photoreduction.In particular, titanium dioxide-based materials were tested both in the liquid and vapour phases for a direct comparison under extremely-low irradiance conditions to understand the reaction medium effect on process efficiency.

Materials Characterisation
For the purposes of this work, two commercial titanium dioxide-based materials were chosen to compare their photocatalytic performances with a lab-made titanium dioxide sample.Thus, a preliminary characterisation was performed to assure material suitability for photocatalytic application.
As shown in the N 2 physisorption isotherms (Figure 1), all three samples show a type-IV isotherm typical of mesoporous materials.
Energies 2017, 10, 1394 3 of 14 based materials were tested both in the liquid and vapour phases for a direct comparison under extremely-low irradiance conditions to understand the reaction medium effect on process efficiency.

Materials Characterisation
For the purposes of this work, two commercial titanium dioxide-based materials were chosen to compare their photocatalytic performances with a lab-made titanium dioxide sample.Thus, a preliminary characterisation was performed to assure material suitability for photocatalytic application.
As shown in the N2 physisorption isotherms (Figure 1), all three samples show a type-IV isotherm typical of mesoporous materials.P25, which is the most commonly used benchmark material [51], provides the lowest surface area (50 m 2 •g −1 ) and the isotherm shows a narrow hysteresis loop at high relative pressures (between 0.8 and 1 p/p0), indicating a narrow distribution of pores with an average pore size of 15 nm.The labmade sample (labelled as TiO2) shows a similar isotherm in shape, but a greater nitrogen adsorption, indicating a higher surface area.As a matter of fact, TiO2 has a 110 m 2 •g −1 , which is double that of P25, and an average pore size of 25 nm, whereas MIRKAT 211 is characterised by a much higher surface area (217 m 2 •g −1 ) and is characterized by a higher nitrogen adsorption at low relative pressures and a wider hysteresis loop, corresponding to a wider and non-homogeneous pore size distribution.
Considering the materials' crystal structures, there are some differences among the three samples, as shown by X-ray diffraction (XRD) diffractograms reported in Figure 2.  P25, which is the most commonly used benchmark material [51], provides the lowest surface area (50 m 2 •g −1 ) and the isotherm shows a narrow hysteresis loop at high relative pressures (between 0.8 and 1 p/p 0 ), indicating a narrow distribution of pores with an average pore size of 15 nm.The lab-made sample (labelled as TiO 2 ) shows a similar isotherm in shape, but a greater nitrogen adsorption, indicating a higher surface area.As a matter of fact, TiO 2 has a 110 m 2 •g −1 , which is double that of P25, and an average pore size of 25 nm, whereas MIRKAT 211 is characterised by a much higher surface area (217 m 2 •g −1 ) and is characterized by a higher nitrogen adsorption at low relative pressures and a wider hysteresis loop, corresponding to a wider and non-homogeneous pore size distribution.
Considering the materials' crystal structures, there are some differences among the three samples, as shown by X-ray diffraction (XRD) diffractograms reported in Figure 2.
In MIRKAT 211 and the TiO 2 sample, the only observed crystal phase is anatase, in accordance with JCPDS card No. 00-002-0387; this structure is the most suitable crystal phase for photocatalytic applications, due to the highest electron-hole life among titanium dioxide polymorphs [52].Moreover, due to the desired application for CO 2 photoreduction, the anatase conduction band is more reductive than the rutile band gap, thus, in principle, more suitable to reduce CO 2 , which is a highly energy-demanding reaction [53,54].
indicating a higher surface area.As a matter of fact, TiO2 has a 110 m 2 •g −1 , which is double that of P25, and an average pore size of 25 nm, whereas MIRKAT 211 is characterised by a much higher surface area (217 m 2 •g −1 ) and is characterized by a higher nitrogen adsorption at low relative pressures and a wider hysteresis loop, corresponding to a wider and non-homogeneous pore size distribution.
Considering the materials' crystal structures, there are some differences among the three samples, as shown by X-ray diffraction (XRD) diffractograms reported in Figure 2.  Comparing the two profiles, the TiO 2 sample provides sharper and more defined peaks than MIRKAT 211, indicating a higher crystallinity in the former sample.As a matter of fact, the MIRKAT 211 manufacturer states that this product is made up of 40 wt % anatase, while the rest is amorphous TiO 2 [55,56].Instead, according to Rietveld analysis, it is composed of 98 wt % anatase and the remaining part is amorphous.Conversely, P25 shows not only anatase-related peaks, but also diffraction peaks due to rutile (JCPDS card No. 76-1940).In fact, it is known that this material is made up of a mixture of the two crystal structures in a ca.75/25 anatase/rutile ratio [57].Despite rutile's fast electron-hole recombination, the co-presence of the anatase and rutile crystallites creates an electronic circulation from the photoexcited rutile lattice trapping sites, the most easily excitable, yet unstable, to anatase ones which are the most resistant to electron-hole recombination [58].This feature is also observable in the diffuse reflectance spectroscopy (DRS) spectra reported in Figure 3 and is often correlated to the enhanced lifetime of the photogenerated charges.
Energies 2017, 10, 1394 4 of 14 In MIRKAT 211 and the TiO2 sample, the only observed crystal phase is anatase, in accordance with JCPDS card No. 00-002-0387; this structure is the most suitable crystal phase for photocatalytic applications, due to the highest electron-hole life among titanium dioxide polymorphs [52].Moreover, due to the desired application for CO2 photoreduction, the anatase conduction band is more reductive than the rutile band gap, thus, in principle, more suitable to reduce CO2, which is a highly energy-demanding reaction [53,54].
Comparing the two profiles, the TiO2 sample provides sharper and more defined peaks than MIRKAT 211, indicating a higher crystallinity in the former sample.As a matter of fact, the MIRKAT 211 manufacturer states that this product is made up of 40 wt % anatase, while the rest is amorphous TiO2 [55,56].Instead, according to Rietveld analysis, it is composed of 98 wt % anatase and the remaining part is amorphous.Conversely, P25 shows not only anatase-related peaks, but also diffraction peaks due to rutile (JCPDS card No. 76-1940).In fact, it is known that this material is made up of a mixture of the two crystal structures in a ca.75/25 anatase/rutile ratio [57].Despite rutile's fast electron-hole recombination, the co-presence of the anatase and rutile crystallites creates an electronic circulation from the photoexcited rutile lattice trapping sites, the most easily excitable, yet unstable, to anatase ones which are the most resistant to electron-hole recombination [58].This feature is also observable in the diffuse reflectance spectroscopy (DRS) spectra reported in Figure 3 and is often correlated to the enhanced lifetime of the photogenerated charges.MIRKAT 211 and the TiO2 sample provide a typical sharp absorption below 390 nm, corresponding to a 3.2 eV band gap (typical of anatase), in accordance to the XRD patterns and data reported in the literature [59].Differently, P25 provides a small absorption between 390 and 410 nm due to the small fraction of rutile and below 390 nm absorption due to anatase.Considering the polymorph structure of this material, the P25 overall band gap is 3.1 eV, which is lower than the pure anatase titanium dioxide.
In conclusion, all materials are suitable, in principle, for CO2 photoreduction despite their different morphological and structural properties.P25 differs in surface, crystal, and electronic properties from the other two samples, which are indeed characterised by a much higher surface area and anatase as the only crystal phase.MIRKAT 211 and the TiO 2 sample provide a typical sharp absorption below 390 nm, corresponding to a 3.2 eV band gap (typical of anatase), in accordance to the XRD patterns and data reported in the literature [59].Differently, P25 provides a small absorption between 390 and 410 nm due to the small fraction of rutile and below 390 nm absorption due to anatase.Considering the polymorph structure of this material, the P25 overall band gap is 3.1 eV, which is lower than the pure anatase titanium dioxide.
In conclusion, all materials are suitable, in principle, for CO 2 photoreduction despite their different morphological and structural properties.P25 differs in surface, crystal, and electronic properties from the other two samples, which are indeed characterised by a much higher surface area and anatase as the only crystal phase.

Vapour Phase Photocatalytic Results
Before performing the activity runs, blank tests were carried out in order to avoid any bias in data collection and interpretation.In fact, traces of organic species can definitely lead to misleading results [60].In the case of our freshly-prepared sample, only inorganic salts were used in the preparation of the catalyst and residual sulphate ions were eliminated by washing.However, in order to check the presence of carbonaceous species possibly deriving from manipulation, the following tests were performed: without light, catalyst, or reactants.In none of the three cases we observed any hydrocarbon formation.Moreover, a test with catalyst, light, and water (so, without CO 2 only) was performed, and no C-based product was detected, indicating the absence of carbonaceous species on the surface.Thus, it is possible to state, first of all, that this reaction is not a photochemical reaction, but a photocatalytic one.Moreover, catalysts are photostable and do not contain any trace of carbon from their manipulation, and collected data are not affected by carbonaceous species on the photocatalytic surface.
Photocatalytic tests with these three samples in the vapour phase are reported in Figure 4.In both testing rigs, it was chosen to use low-irradiance conditions (i.e., 50 W•m −2 ) in order to minimise energy input and unequivocally determine and appreciate the differences in catalytic behaviours due to material morphology and experimental conditions.Such a low irradiance with respect to the literature reports is suggested by the useful energy fraction that can be reasonably exploited from solar light irradiation.In the case of bare titanium dioxide, radiation energy for a material excitation must be greater than, or at least equal to, its band gap (BG), i.e., 3.2 eV for anatase and 3.0 for rutile, which limit its application to UV irradiation.To increase VIS light absorption, TiO 2 's BG is usually narrowed in the case of environmentally-benign photooxidations, but this usually leads to a decrease in conduction band's reductant properties.In fact, reduction potential for CO 2 is quite close to TiO 2 's conduction band potential, strongly limiting band gap reduction strategies.
Energies 2017, 10, 1394 5 of 14 tests were performed: without light, catalyst, or reactants.In none of the three cases we observed any hydrocarbon formation.Moreover, a test with catalyst, light, and water (so, without CO2 only) was performed, and no C-based product was detected, indicating the absence of carbonaceous species on the surface.Thus, it is possible to state, first of all, that this reaction is not a photochemical reaction, but a photocatalytic one.Moreover, catalysts are photostable and do not contain any trace of carbon from their manipulation, and collected data are not affected by carbonaceous species on the photocatalytic surface.Photocatalytic tests with these three samples in the vapour phase are reported in Figure 4.In both testing rigs, it was chosen to use low-irradiance conditions (i.e., 50 W•m −2 ) in order to minimise energy input and unequivocally determine and appreciate the differences in catalytic behaviours due to material morphology and experimental conditions.Such a low irradiance with respect to the literature reports is suggested by the useful energy fraction that can be reasonably exploited from solar light irradiation.In the case of bare titanium dioxide, radiation energy for a material excitation must be greater than, or at least equal to, its band gap (BG), i.e., 3.2 eV for anatase and 3.0 for rutile, which limit its application to UV irradiation.To increase VIS light absorption, TiO2's BG is usually narrowed in the case of environmentally-benign photooxidations, but this usually leads to a decrease in conduction band's reductant properties.In fact, reduction potential for CO2 is quite close to TiO2's conduction band potential, strongly limiting band gap reduction strategies.In vapour phase tests, the only detected products were methane and hydrogen.The former was the most abundant product with all three catalysts.Methane derives from complete carbon dioxide photoreduction and no trace of other carbonaceous species, like CO or formaldehyde, were detected.Karamian and co-workers [61], reported that, in most cases, in gaseous systems CO is the first intermediate product of CO2 photoreduction by water vapour.However, reaction conditions, and in particular temperature, irradiance, and reaction time, can modify the reaction pathway and, thus, product distribution.In particular, when CO2 deoxygenation is faster than dehydrogenation, methane production is favoured with respect to oxygenated compounds [62].This is the case of vapour phase reaction, characterised by CO2 excess.This mechanism involves the formation of C In vapour phase tests, the only detected products were methane and hydrogen.The former was the most abundant product with all three catalysts.Methane derives from complete carbon dioxide photoreduction and no trace of other carbonaceous species, like CO or formaldehyde, were detected.
Energies 2017, 10, 1394 6 of 14 Karamian and co-workers [61], reported that, in most cases, in gaseous systems CO is the first intermediate product of CO 2 photoreduction by water vapour.However, reaction conditions, and in particular temperature, irradiance, and reaction time, can modify the reaction pathway and, thus, product distribution.In particular, when CO 2 deoxygenation is faster than dehydrogenation, methane production is favoured with respect to oxygenated compounds [62].This is the case of vapour phase reaction, characterised by CO 2 excess.This mechanism involves the formation of C radicals that recombine with H originated from water [63].
At the same time, hydrogen can be produced from water splitting into hydrogen and oxygen, which, in this case, can be potentially responsible for reductant consumption.In all cases, here, selectivity to methane was higher than 98%, indicating that CO 2 photoreduction is favoured with respect to water splitting.
Among the three samples, P25 is the least active material in methane production and this experimental evidence is ascribable to its lower surface area, which limits the number of available photocatalytic sites and adsorbed species.Conversely, MIRKAT 211, despite the highest surface area, provides lower methane production than the TiO 2 sample, which is the best performing material.This photocatalytic behaviour is ascribable to the co-presence of two favourable features, i.e., relatively high surface area and high crystallinity of the anatase phase, which seems to be more effective on photoactivity.Indeed, small crystals expose boundary defective sites that may act as electron-hole recombination centres, thus decreasing the photocatalytic activity.This means that, to increase TiO 2 efficiency in CO 2 photoreduction in the gas phase under low-irradiance conditions it is necessary to formulate materials characterised by high surface area and high crystallinity in the anatase phase at the same time.

Liquid Phase Photocatalytic Results
In this case, blank tests were performed preliminarily.No evidence of activity under dark conditions or in the absence of catalyst or CO 2 was observed, indicating that carbon reactants, catalyst, and light are necessary for the reaction to proceed and that there are not any residual carbonaceous species from sample manipulation.Water splitting was not significant in this case, mainly based on the selection of the hole scavenger and operating conditions.
Considering the liquid phase activity results, no methane was detected for all three samples.However, differently from the vapour phase systems, higher amounts of hydrogen and C-containing products in the liquid phase (methanol, formic acid, formaldehyde) were detected.This was expected, since the mechanism of photoreduction is completely different in liquid or gas-phase system [64].A recent investigation on the reaction mechanism and product distribution vs. time, showed that liquid phase products accumulate as primary species in the liquid phase, followed by parallel and consecutive reactions.Among these, hydrogen and CO are produced by photoreforming of the newly-reduced organic compounds.Furthermore, in similar photocatalytic systems [61,65,66], it is reported that the reaction is triggered by the formation of peroxocarbonate species, which are reduced to formic acid, formaldehyde, and methanol afterwards.Due to the higher H 2 O/CO 2 ratio, it is plausible that •CO 2 − undergoes hydrogenation faster than deoxygenation, leading to all those products found in the liquid phase.However, as reported by Liu and Li [62], hydroperoxo species in water are characterised by a high redox potential, which makes them very unstable and, thus, unreliably detectable.Therefore, we can speculate that in our conditions the unstable peroxocarbonates are the first intermediates, then directly converted into formate ions, formic acid, and further reduction products.If CO 2 is absorbed in water at neutral pH (molecular form), it is consecutively reduced to formic acid, formaldehyde, and methanol.Formaldehyde is, instead, the first forming product when basic aqueous solutions are adopted, and further reduction to methanol or oxidation to formic acid may occur.Hydrogen and CO/CO 2 form in the gas phase, but their formation reaction is consecutive to the accumulation of organic compounds in the liquid phase and the depletion of Na 2 SO 3 , used as a hole scavenger [67].Indeed, after sulphite consumption, the formed organic compounds start acting as hole scavengers in a photoreforming reaction scheme.
When testing the present samples, we observed hydrogen as a predominant product in the gas phase, with increasing concentration passing from P25 (negligible), to MIRKAT 211, and to TiO 2 .Carbon monoxide was detected only for MIRKAT and TiO 2 in very small amounts compared to hydrogen (1% of the total gas phase products for both the materials).This indicates high selectivity to the full reoxidation to CO 2 of the organic compounds in the photoreforming step.
Additionally, the best outcomes, in terms of C-containing products, were obtained using sample TiO 2 , essentially for the same reasons outlined in the previous paragraph.Figure 5 shows the results in terms of TOF.Furthermore, P25 revealed the worst activity, due to a lower surface area, in spite of the slightly lower band gap.
Energies 2017, 10, 1394 7 of 14 Similar productivities of the liquid phase products were obtained for the TiO2 and the MIRKAT 211 samples, which is ca.double with respect to the P25 sample, and much higher than the CH4 amount achieved in the gas phase testing rig.This is explained, in part, considering the higher reduction potential needed for the reduction from CO2 to CH4 than to obtain partially-reduced compounds.On the other hand, this demonstrates the better performance achievable with liquidphase CO2 photoreduction, provided that improved solubility is reached using relatively high pressure testing.The product distribution in the liquid phase, considering P25 and MIRKAT 211, are in accordance with the consecutive reaction steps occurring at neutral pH [64].Indeed, the higher concentration of methanol is detected, being the final product formed during the CO2 photoreduction process.However, the product distribution in the case of sample TiO2 was different, since a maximum concentration was obtained for formic acid, with total C formed in the liquid phase being comparable with MIRKAT 211 and ca.double that of P25.The highest H2 productivity, and the lowest of methanol, are interpreted on the basis of a higher turnover rate, which depletes the sulphite hole scavenger faster, with consequently faster start-up of the oxidation of the organic products through photoreforming.

Effect of Reaction Medium
In both experimental conditions, despite surface area being important, crystallinity and suitable crystal phase (i.e., anatase) represent the most important physicochemical properties for an efficient TiO2 photocatalyst.The choice of performing under equal and extremely low irradiance allows the even greater appreciation of these differences in the effectiveness of the different materials in light harvesting and, thus, in carbon dioxide conversion.
Results in vapour and liquid phase tests are directly compared here, which has never been done before from an experimental point of view.The tests have been carried out under specificallyoptimised conditions for both testing modes.It is clear that process activity and selectivity is driven by the reaction medium.From all experimental evidence, it is possible to suppose that a reduction in the titanium dioxide valence band undergoes two different pathways, as shown in Figure 6.In one case, fast deoxygenation leads to C species that are reduced afterwards whereas, in the liquid phase, hydroperoxo species undergo hydrogenation preferentially [68], yielding to possible intermediate Similar productivities of the liquid phase products were obtained for the TiO 2 and the MIRKAT 211 samples, which is ca.double with respect to the P25 sample, and much higher than the CH 4 amount achieved in the gas phase testing rig.This is explained, in part, considering the higher reduction potential needed for the reduction from CO 2 to CH 4 than to obtain partially-reduced compounds.On the other hand, this demonstrates the better performance achievable with liquid-phase CO 2 photoreduction, provided that improved solubility is reached using relatively high pressure testing.The product distribution in the liquid phase, considering P25 and MIRKAT 211, are in accordance with the consecutive reaction steps occurring at neutral pH [64].Indeed, the higher concentration of methanol is detected, being the final product formed during the CO 2 photoreduction process.However, the product distribution in the case of sample TiO 2 was different, since a maximum concentration was obtained for formic acid, with total C formed in the liquid phase being comparable with MIRKAT 211 and ca.double that of P25.The highest H 2 productivity, and the lowest of methanol, are interpreted on the basis of a higher turnover rate, which depletes the sulphite hole scavenger faster, with consequently faster start-up of the oxidation of the organic products through photoreforming.

Effect of Reaction Medium
In both experimental conditions, despite surface area being important, crystallinity and suitable crystal phase (i.e., anatase) represent the most important physicochemical properties for an efficient TiO 2 photocatalyst.The choice of performing under equal and extremely low irradiance allows the even greater appreciation of these differences in the effectiveness of the different materials in light harvesting and, thus, in carbon dioxide conversion.
Results in vapour and liquid phase tests are directly compared here, which has never been done before from an experimental point of view.The tests have been carried out under specifically-optimised Energies 2017, 10, 1394 8 of 14 conditions for both testing modes.It is clear that process activity and selectivity is driven by the reaction medium.From all experimental evidence, it is possible to suppose that a reduction in the titanium dioxide valence band undergoes two different pathways, as shown in Figure 6.In one case, fast deoxygenation leads to C species that are reduced afterwards whereas, in the liquid phase, hydroperoxo species undergo hydrogenation preferentially [68], yielding to possible intermediate products.The liquid phase product distribution of sample TiO 2 confirms the proposed mechanism (74% formic acid, 18% formaldehyde, 8% methanol).Therefore, different reaction pathways are active when operating in the two different media, which led to direct production of the most reduced product (methane) in the gas phase system whereas, in the liquid phase, intermediate reduction products are formed together with H2.
As for testing in the liquid phase, it should be remarked that the key for significant productivity is related to the possibility to achieve sufficiently high CO2 solubility in water by applying a sufficiently high pressure.

Materials
The following reagents were used as received: TiOSO4•xH2O•yH2SO4 (Ti assay > 29% Sigma Aldrich, Milan, Italy), sodium hydroxide (assay > 97% Carlo Erba, Milan, Italy), and 2-propanol (assay 99.8% Fluka, Milan, Italy).Two standard TiO2 reference materials were used.The first one is the most commonly used titanium dioxide, i.e., P25 from Degussa (Essen, Germany), used as a reference for comparison with the literature.The second one is MIRKAT 211, which has been purchased by Euro Support s.r.o (Amersfoort, The Netherlands).This commercial titania has been chosen since it is characterised by a very large surface area (217 m 2 •g −1 ) and it is in the anatase form, i.e., typically the most suitable titania crystalline phase for this photocatalytic application.

Synthesis of TiO2
The precipitation method has been chosen to synthesise titania samples.In a typical synthesis, a 1.2 M titanyl sulphate solution and a 9.0 M NaOH solution are added dropwise and simultaneously to 200 mL of distilled water under vigorous stirring, in order to keep pH neutral.The Ti(OH)4 suspension has been aged at 60 °C for 20 h.Afterwards, the precipitate has been filtered and washed with distilled water to remove sulphate ions.The absence of sulphates has been verified by means of a barium chloride test [69].Wet Ti(OH)4 has been dried overnight at 110 °C and calcined at 400 °C for 4 h in air flow to obtain TiO2.

Characterization of the Photocatalysts
XRD patterns have been collected on a Bruker D8 Advance powder diffractometer with a sealed X-ray tube (copper anode; operating conditions, 40 kV and 40 mA) and a Si(Li) solid state detector Therefore, different reaction pathways are active when operating in the two different media, which led to direct production of the most reduced product (methane) in the gas phase system whereas, in the liquid phase, intermediate reduction products are formed together with H 2 .
As for testing in the liquid phase, it should be remarked that the key for significant productivity is related to the possibility to achieve sufficiently high CO 2 solubility in water by applying a sufficiently high pressure.

Materials
The following reagents were used as received: TiOSO 4 •xH 2 O•yH 2 SO 4 (Ti assay > 29% Sigma Aldrich, Milan, Italy), sodium hydroxide (assay > 97% Carlo Erba, Milan, Italy), and 2-propanol (assay 99.8% Fluka, Milan, Italy).Two standard TiO 2 reference materials were used.The first one is the most commonly used titanium dioxide, i.e., P25 from Degussa (Essen, Germany), used as a reference for comparison with the literature.The second one is MIRKAT 211, which has been purchased by Euro Support s.r.o (Amersfoort, The Netherlands).This commercial titania has been chosen since it is characterised by a very large surface area (217 m 2 •g −1 ) and it is in the anatase form, i.e., typically the most suitable titania crystalline phase for this photocatalytic application.

Synthesis of TiO 2
The precipitation method has been chosen to synthesise titania samples.In a typical synthesis, a 1.2 M titanyl sulphate solution and a 9.0 M NaOH solution are added dropwise and simultaneously to 200 mL of distilled water under vigorous stirring, in order to keep pH neutral.The Ti(OH) 4 suspension has been aged at 60 • C for 20 h.Afterwards, the precipitate has been filtered and washed with distilled water to remove sulphate ions.The absence of sulphates has been verified by means of a barium Energies 2017, 10, 1394 9 of 14 chloride test [69].Wet Ti(OH) 4 has been dried overnight at 110 • C and calcined at 400 • C for 4 h in air flow to obtain TiO 2 .

Characterization of the Photocatalysts
XRD patterns have been collected on a Bruker D8 Advance powder diffractometer with a sealed X-ray tube (copper anode; operating conditions, 40 kV and 40 mA) and a Si(Li) solid state detector (Sol-X) set to discriminate the Cu Kα radiation.Apertures of divergence, receiving, and detector slits were 2.0 mm, 2.0 mm, and 0.2 mm, respectively.Data scans have been performed in the 2θ range of 5-75 • with a 0.02 • step size and counting times of 3 s/step.Quantitative phase analysis and crystallite size determination have been performed using the Rietveld method as implemented in the TOPAS v.4 program (Bruker AXS) using the fundamental parameters approach for line-profile fitting.The determination of the crystallite size was accomplished by the Double-Voigt approach and calculated as volume-weighted mean column heights based on integral breadths of peaks.
N 2 adsorption-desorption isotherms at 196 • C were performed using a MICROMERITICS ASAP 2000 analyser (Micromeritics Instruments Corporation, Norcross, GA, USA) to obtain information on the surface properties.All samples were previously outgassed at 200 • C for 2 h.The mesopore volume was measured as the adsorbed amount of N 2 after capillary condensation.The surface area was evaluated using the standard BET [70] equation and the pore size distribution was obtained using the BJH method applied to the isotherm desorption branch [71].
Diffuse reflectance spectroscopy (DRS) of carefully ground powders was performed with a Thermo Scientific Evolution 600 spectrophotometer, equipped with a diffuse reflectance accessory praying-mantis sampling kit (Harrick Scientific Products, NY, USA).A Spectralon1 disk was used as the reference material.The experimental absorption versus lambda plot was elaborated using the Kubelka-Munk function [72].The band gap energy (Eg) of the catalysts are determined by the intercept of a linear fit to the absorption edge and they can be estimated using the standard equation, which is based on the relationship between frequency (c/λ) and photon energy (Eg = 1240/λ).

Photoactivity Tests in the Liquid Phase
Catalysts were suspended in water with a concentration of 0.3 g•L −1 at neutral pH.Na 2 SO 3 (Sigma Aldrich, 98%) was used as a hole scavenger, with a concentration of 1.7 g•L −1 .All these products were used without further purification.The gas phase was analysed by gas chromatography (Agilent, Santa Clara, CA, USA mod.7890) equipped with a thermal conductivity detector (TCD) and a proper set up for the quantification of H 2 , CH 4 , CO, and polar/non-polar light gases.The liquid mixture has been analysed by means of a HPLC (Agilent 1220 Infinity) using a column (Alltech OA-10308, 300 mm_7.8 mm) with UV and refractive index (Agilent 1260 Infinity) detectors.Aqueous H 3 PO 4 solution (0.1 wt %) was used as the eluent.The consumption of Na 2 SO 3 was evaluated by iodometric titration.This method is based on the oxidation of sulphites (SO 3 2− ) into sulphates (SO 4 2− ) by iodine produced from a solution with a known concentration of potassium iodate (KIO 3 , Sigma Aldrich, 98%) and potassium iodide (KI, Sigma Aldrich, 99%) in acid environment and the subsequent titration of the iodine excess with sodium thiosulphate (Na 2 S 2 O 3 , Sigma Aldrich, 98%).The equivalence point of the titration was detected using starch solution as an indicator.Photocatalytic reduction of CO 2 with H 2 O was carried out at constant pressure of 7 bar, at a constant temperature of 80 • C for 24 h.The scheme of the experimental setup and details are shown elsewhere [30].Briefly, it is an innovative pressurized batch photoreactor with a medium-pressure Hg lamp as the source of radiation.The emission range was between 254 nm and 364 nm.The lamp was cooled using an inner internal air flow of 2000 L h −1 .This air flow was selected to tune the average measured irradiance in the reactor to 50 W/m 2 , for comparison with testing in the gas phase (vide infra).The emitted power was periodically measured by means of a photoradiometer (Delta OHM HD2102.2).
Before starting the irradiation of the reaction mixture the system was outgassed at constant CO 2 flow of 15 mL•min −1 at a pressure of 13 bar with the aim to eliminate air from the reactor head space.Then, in order to saturate water with CO 2 , a static pressure of 7 bar of CO 2 was applied overnight.

Photoactivity Tests in the Vapour Phase
The catalytic apparatus was reported in a previous work [39].CO 2 photoreduction was performed using a borate glass thin film reactor (length: 33 mm, height: 18 mm, thickness: 2 mm).Here the catalyst (10 mg) was inserted by depositing the catalyst suspended in 2-propanol on the light-exposed side of the reactor.Before tests, the catalyst was left overnight under helium flow to desorb all residual carbonaceous species.
The samples were illuminated using a 125 W mercury UVA lamp (purchased from Helios Italquartz s.r.l. with emission range 315-400 nm shielded by a special tubular quartz, to block radiations whose wavelength is lower than 350 nm), with an average irradiance of 50 W•m −2 , considering all emission wavelengths.Afterwards, a gaseous mixture of carbon dioxide and water has flowed through the reactor.Compressed CO 2 (99.99%) regulated by a mass flow controller was carried through a water bubbler kept at 40 • C to generate a CO 2 and H 2 O vapour mixture (13.3 CO 2 /H 2 O molar ratio).The reactor was closed when the system reached the equilibrium state and this point was taken as the beginning of the reaction.Therefore, the reaction was not performed under a continuous gas flow, but it took place under static conditions.A total of 9.2 µmol of CO 2 and 0.7 µmol of H 2 O were present within the sealed reactor.In all catalytic tests, the reaction time was 6 h.
The reaction products were analysed by a gas chromatograph (HP G1540A, Hewlett-Packard Company, Wilmington, DE, USA) equipped with a Porapak Q column and a TCD detector, appropriately calibrated to separate and quantify H 2 , CH 4 , and CO.Activity results are expressed in turn-over frequencies (TOFs) in µmol•gcat −1 •h −1 , as commonly used in the literature [42,73].

Conclusions
Breakthroughs in CO 2 photoreduction relies not only on the development of efficient photocatalysts, but also on finding the most suitable reaction conditions, and in particular reaction medium, which has never been investigated before.Throughout this work, it was shown that, to appreciate differences and effectiveness of photocatalytic materials, reaction conditions are not a trivial matter, but dramatically affect photocatalytic performances.
The choice of low-irradiance conditions allowed, on one side, to appreciate the differences in photocatalytic behaviour and, on the other, to minimise the energetic input.
Whatever the reaction conditions are, it was established that the crystal phase (i.e., anatase) and crystallinity increase titanium dioxide photoactivity, which excelled in lab-synthesised samples in comparison with commercial benchmark materials.
However, the choice of reaction medium dramatically drives the reaction pathway with consequences on process selectivity: vapour phase deoxygenation drives selectivity towards methane whereas, in the liquid phase, water dehydrogenation allows obtaining intermediate oxidation state products, i.e., formic acid, formaldehyde, and methanol.

Figure 2 .
Figure 2. P25, MIRKAT 211, and TiO2 XRD patterns.A represents the diffraction peaks of the anatase phase, R, those of the rutile phase.

Figure 2 .
Figure 2. P25, MIRKAT 211, and TiO2 XRD patterns.A represents the diffraction peaks of the anatase phase, R, those of the rutile phase.

Figure 2 .
Figure 2. P25, MIRKAT 211, and TiO 2 XRD patterns.A represents the diffraction peaks of the anatase phase, R, those of the rutile phase.

Figure 6 .
Figure 6.Different reaction mechanisms in vapour and liquid phases.