Towards scaling-up photocatalytic process for multiphase envi- ronmental applications: a review

Recently, we have witnessed a booming development of composites and multi-dopant metal oxides to be employed as novel photocatalysts. Yet the practical application of photocatalysis for environmental purposes is still elusive. Concerns about the unknown fate and toxicity of nanoparticles, unsatisfactory performance in real conditions, mass transfer limitations and durability issues have so far discouraged investments in full-scale applications of photocatalysis. Herein, we provide a critical overview of the main challenges that are limiting large-scale application of photocatalysis in air and water/wastewater purification. We then discuss the main approaches reported in the literature to tackle these shortcomings, such as the design of photocatalytic reactors that retain the photocatalyst, the study of degradation of micropollutants in different water matrices, and the development of gas-phase reactors with optimized contact time and irradiation. Furthermore, we provide a critical analysis of research-practice gaps such as treatment of real water and air samples, degradation of pollutants with actual environmental concentrations, photocatalyst deactivation, and cost and environmental life-cycle assessment.


Introduction
Air and water pollution represents a serious concern for the human health and the livelihood of entire ecosystems. A vast range of priority pollutants (most important regulated pollutants) and pollutants of emerging concerns (pollutants with severe health and environmental concerns but not yet regulated) can be found in industrial wastewaters and exhausts, or are released to the environment during the end-use of chemical products [1][2][3]. Air pollution, both outdoor and indoor, represents a major threat to the human health, killing an estimated seven million people worldwide in 2016 [4], due to pollutants such as particulate matter, polyaromatic hydrocarbons, carbon monoxide, nitrogen oxides, and volatile organic compounds (VOC). Widespread water pollutants include perfluoroalkyl substances (PFAS), pharmaceuticals and personal care products (PPCPs), pesticides, and endocrine-disrupting compounds (EDCs) [5].
Although most of these pollutants are present in the aquatic environment in very low concentrations (usually in the range of µg.L -1 and ng.L -1 ), they have severe direct and indirect repercussions on the environment and human health [6]. For instance, very low concentrations of antibiotics may favor the development of antibiotic resistance in certain microorganisms [7].
Conventional pollutant removal technologies present numerous shortcomings. For instance, conventional biological wastewater treatment methods are able to break C-C bonds in organic pollutants, but they have limited ability to destroy other strong bonds in source. This simple approach has been in fact employed in the majority of studies reporting the photocatalytic degradation of micropollutants or disinfection [31][32][33][34]. Notwithstanding the high photon absorption and efficient ROS interaction with pollutants in the case of dispersed catalysts, the photocatalyst reuse is hindered by the difficulty of recovering the nanostructured powder at the end of the process. To prove the stability and reusability of the dispersed photocatalysts, some researchers added drops of concentrated pollutants after the first photocatalytic cycle [7]; however, this approach is obviously not applicable in real-life. Furthermore, the potential release of photocatalyst powder represents another major drawback of setups using dispersed catalysts, because the toxicity of the nanoparticles and their repercussions on the environment are not yet fully understood.
Reactors employing photocatalysts in retained forms can facilitate reusability eliminating the need for powder recovery, which generally requires ultrafiltration, and, at the same time, they reduce the risks associated with nanoparticle release to the environment. We advocate for this strategy as a practical translation of bench-scale studies to full-scale water/wastewater treatment.
The fluidized-bed photocatalytic reactor is an example of photocatalytic systems that keep the photocatalyst within its boundary. In this system, the water flow is circulated with a proper velocity to keep the fluidized photocatalyst in a certain position inside the reactor [12]. In this sense, the catalyst remains in the reactor and thus there is no need for a collection step. However, the fluidized bed system was rarely used due to the low weight and high dispersibility of most photocatalysts [26]. Development of photocatalysts with bigger particle size to overcome the high dispersibility would be at the expense of the active surface area leading to lower light absorption. To attain appropriate photocatalytic suspensions, anchoring the photocatalyst on a porous substrate can also provide the additional benefit of promoting pollutant adsorption. For instance, Fang et al. prepared a composite macrostructure of TiO2 nanotubes and graphene for the adsorption and degradation of bisphenol A in a fluidized bed reactor [35]. This system, using a flow rate of 1 mL/min, showed 86% and 50% pollutant removals for, respectively, 0.05 mg/L and 0.5 mg/L bisphenol A solutions; the removal performance remained constant during 1000 min tests.
The light source in fluidized-bed photoreactors can be placed at the center of the reactor or at the reactor walls, adopting either tubular UV lamps or LED lights [36,37] (Figure 1).

Figure 1.
Fluidized bed type LED reactor with immobilized catalyst on polymer supports; reproduced with permission from [37].
Alternatively, reactors using photocatalysts immobilized on a solid, fix support have been proposed as a solution allowing long operation periods with minimum catalyst loss [38]. Among them, reactors based on photocatalytic membranes have attracted increasing interest because immobilizing the photocatalysts at the membrane surface improves the mass transfer of pollutants to the photocatalytic sites [11]. The choice of the membrane materials is limited by the oxidative stress caused by the photocatalyst material: Suitable materials include polyethersulfone (PES) [39] or polyvinylidene fluoride (PVDF) membranes. Numerous types of photocatalyst immobilization procedures have been reported, such as dip coating [40], spray coating [41], layer-by-layer self-assembly [42] and blending with the polymer matrix [39]. Alternatively, porous photocatalyst morphologies can be used to obtain free-standing photocatalytic membranes [43,44]. During the operation of a photocatalytic membrane reactor, the contaminated water flows through the illuminated membranes for several cycles until the desired degradation is achieved, so the photocatalyst is kept in the reactor without the need for collection ( Figure 2). Consequently, flow rate represents a crucial parameter for the operation of membrane reactor, as it affects mass transport of the reagents to the catalytic sites as well as contact time between the pollutant and the photocatalyst [45]. Liu et al. immobilized ZnIn2S4 on a PVDF membrane by phase inversion method, for the photocatalytic degradation of fluvastatin [46]. Almost complete degradation of 10 mg/L fluvastatin was attained in 180 min of irradiation and the reuse for six cycles retained about 91.5% of the initial efficiency. In another study, Wei et al. developed a reduced graphene oxide membrane loaded with graphitic carbon nitride (g-C3N4) for the degradation of 5 mg/L rhodamine B under visible light [47]. The recycle reactor removed 98.7% of the dye in 90 min and a high permeability rate was retained for 300 min, but reuse was not studied. Fischer et al. loaded TiO2 on a PES membrane by dip coating and ultrasound modifications [48]. Interestingly, the produced membrane could achieve complete degradation of 13 mg/L methylene blue within 40 min for consecutive nine cycles. Other reactor configurations involving immobilized, fixed photocatalyst rely on the coating of internal reactor surfaces with the photocatalyst. Some examples of this strategy are reported in Figure 3. For instance, Behnajady et al. [49] reported a reactor design based on TiO2-coated longitudinal glass plates (15 x 290 mm), illuminated by 30 W UV-C lamps ( Figure 3a). The reactor was operated in a continuous-flow mode under different volumetric flow-rates and it could completely degrade 30 mg/L acid red 27 at a flow rate of 15 mL.min -1 . However, this type of reactor design is prone to photocatalyst deactivation due to the limited irradiated surface area [45]. In this respect, immobilization of the photocatalyst on substrates such as Rashig rings [50] can promote the overall photocatalyst surface area and mass transfer due to turbulent flow conditions within the reactor. Other concerns regard filming problems at the glazing/lamp surface due to accumulation of dirt from wastewater, and the need of air or oxygen purging to maintain aeration in the closed reactor.
Another alternative is represented by thin-film fixed-bed reactors, which are falling film recycle reactors using an immobilized photocatalyst. In these reactors, water flows in a laminar flow over an inclined flat plate, coated with the photocatalyst, and it is recirculated in a batch mode. This kind of setup does not require air/oxygen purging to maintain aerated conditions and can work using non-concentrated solar radiation [51]. For instance, Fouad et al. reported the use of W-TiO2 coating on inclined stainless-steel plates illuminated by two metal halide lamps placed above the reactor (Figure 3d) [52]. By this configuration, complete degradation of 30 mg/L sulfamethazine was attained in 120 min and the reuse for five consecutive cycles showed that catalyst could retain 90% of its initial performance. A similar geometry was used for the disinfection of real surface waters collected from the intakes of water treatment plants by Ru-WO3/ZrO2 using visible light [53]. The photocatalytic system could disinfect different water samples of 300 to 5000 CFU.mL -1 (CFU: colony-forming units) that include gram-negative and gram-positive bacteria and the coated plates kept their activity after four cycles of consecutive reuse. This reactor type was further developed by Samy  However, laminar flow adopted in thin film fixed bed reactors limits mass transfer, resulting in a slow treatment kinetics, hence large areas are required to attain an acceptable degradation efficiency. To solve this problem, thin film cascade photoreactors have been proposed. This setup uses turbulence introduced by waterfall to promote mass transfer ( Figure 3b). For instance, Stephan et al. developed a falling film recycle reactor comprised of fiber steps coated by TiO2, whereby the contaminated water flowed in a closed loop [56]. The reactor degraded 5 mg/L chlortoluron in less than 160 min, but reusability was not investigated. In this respect, thin film and cascade photoreactors require the development of affordable techniques for a highly stable immobilization of the photocatalyst on large flat surfaces.
In this respect, Sun et al. recently reported an interesting alternative, based on a magnetic-bed photocatalytic recycle reactor. In this study, magnetic photocatalyst particles of CoFe2O4-Ag2O were anchored by an external magnet to the bottom of the reactor ( Figure  3c) [57]. The magnetization of Ag2O by CoFe2O4 was not only beneficial for holding the photocatalyst, but it also reduced the band gap so that the composite could be activated by visible light. This system degraded 92% of methyl orange in 60 min, but the degradation was reduced to 80% after three cycles of reuse. Despite the fact that the efficiency of this reactor is lower than those of many other photocatalytic systems in the literature, it is a very novel and promising that could be developed on larger scales. Table 1 lists some recent examples of reactors with retained photocatalysts and their application for the removal of water pollutants. Despite the large amount of literature on the topic, there are still several issues that hinder the full-scale application of reactors with retained photocatalyst. First, reported processing capacities are generally limited, owing to mass transfer limitations and the inherent kinetics of the photocatalytic process. In order to promote degradation efficiency, higher residence time are needed, which however limits process throughput. Few investigations have reported on pilot scale reactors [51,58] and most of these studies have employed slurry based photo-reactors [59][60][61]. Moreover, numerous studies on reactor configurations focus merely on the main pollutant disappearance, without providing information on the reaction intermediates, which oftentimes are still toxic and poorly biodegradable compounds [62]. The evaluation of the long-term performance of the photocatalytic systems is also another scantly investigated point, as most studies report, at best, a few photocatalytic consecutive tests. In this respect, decreasing performance in consecutive tests is often indicative of photocatalyst leaching due to mechanical and thermal stress or to accumulation of pollutants and their transformation products on the active sites, which can translate in unsatisfactory system lifetime. [61] *100% means that the residual concentration was below the detection limit of the analytical method. 0

Photocatalytic activity in complex water matrices and real samples
1 In addition to considerations related to reactor design, other challenges regard the 2 role of the water matrix. Most studies report photocatalytic experiments on ideal lab con-3 ditions such as pure water solutions of a single pollutant at an easily quantifiable concen-4 tration (from few mg.L -1 to even max solubility) [63]. While rigorously controlled condi-5 tions are adopted for the sake of the characterization of novel photocatalysts, they are far 6 from the real scenario. In fact, real industrial wastewater streams always contain a mix of 7 pollutants in complex matrices, which include numerous electrolytes and natural organic 8 matter (NOM). Moreover, the concentrations of micropollutants in drinking and surface 9 water is generally in the range of ng.L -1 to µg.L -1 , hence far lower than in most of the liter-10 ature studies about photocatalytic remediation. 11 12 2.2.1 Presence of interfering species and real water matrices 13 Most literature studies adopt photocatalytic tests involving a single pollutant at the 14 time, in a synthetic water matrix to avoid the interfering effects particularly from un-15 known organic species. The photocatalytic degradation of pollutants in a background-free 16 and transparent water is ideal for lab studies to investigate the interaction between the 17 catalyst and pollutant and the contribution of different ROS in the oxidation. However, 18 this is not the case in real practice like the treatment of surface water, industrial 19 wastewater, or groundwater, whereby a wide variety of substances can be found [64]. The 20 occurrence of organic or inorganic species other than the main pollutant may inhibit the 21 photocatalytic degradation or requires additional amounts of catalyst [65]. Moreover, 22 photocatalysts can exhibit largely different response to complex water matrices [62], 23 which underscores the limited value of comparisons of photocatalytic activity in ultrapure 24 water. 25 Some inorganic electrolytes, such as bicarbonates and chlorides, can have a detri-26 mental effect on the photocatalytic activity by acting as radical scavengers [28]. However, 27 the overall effect depends on the pollutant and on the relative concentration of the species. 28 Repousi et al. investigated the photocatalytic degradation of bisphenol A by Rh-TiO2 29 in ultra-pure water, humic acid solution, solution of inorganic salts found in bottled water, 30 and secondary treated wastewater [66]. The presence of humic acid enhanced the degra-31 dation of bisphenol A in this study, which was imputed to the reduced aggregation of Rh-32 TiO2 particles by the repulsion forces induced by the accumulation of humic acid on their 33 surface. On the other hand, the degradation was 30 times slower in the treated wastewater 34 matrix despite the whole organic content (TOC= 6.2 mg.L -1 ) was much lower than humic 35 acid organic content (TOC= 9.2 mg.L -1 ). Although the organic content of treated 36 wastewater is unknown, it was expected to contain persistent organics that survived dur- 37 ing the biological treatment and can scavenge the ROS during the photocatalytic reaction. 38 Awfa et al. examined the degradation of carbamazepine by a composite of carbon 39 nanotubes (CNTs) and TiO2 in two-stages treated domestic wastewater, river water, and 40 three commercial synthetic natural organic matter (NOM) solutions, viz., Suwannee River 41 humic acid (SRHA), Suwannee River reverse-osmosis isolates (SRNOM), and Suwannee 42 River fulvic acid (SRFA) [29]. The degradation of carbamazepine was reduced from 93% 43 in ultra-pure water to 87% and 40% in river water and treated wastewater, respectively. 44 Furthermore, inhibition by commercial NOM surrogates was in the order of SRHA, SRFA, 45 SRNOM indicating that the inhibitory effect depends on the characteristics of NOM such 46 as aromaticity (ratio of aromatic content to total organic content), molecular weight, and 47 light absorption. In another study, Ren et al. investigated the inhibitory effects of a lake 48 water and a synthetic solution of alginic acid sodium and citric acid on the degradation of 49 clofibric acid [64]. The results revealed that the inhibition of photocatalytic activity was 50 influenced by the molecular weight of NOM more than its concentration. It was also noted 51 that increasing the dissolved oxygen could decrease the inhibitory effect due to the en-52 hanced production of superoxide radicals, whereas changing the pH did not affect the 53 degradation rates. 54 NOM can inhibit the photocatalytic degradation of micropollutants by different con-55 current mechanisms according to the nature and concentration of existing organic sub-56 stances. Firstly, the photocatalytic activity could be reduced by the inner filter effect, i.e., 57 a lower effective irradiation of the photocatalyst, due to light absorption by organic com-58 pounds [29]. This effect is greatly dependent on the molecular structure of NOM, as com- 59 plex organic structures absorb light on a broader range than simple organics and second-60 ary by-products [67]. Specific UV absorbance (SUVA), which is the ratio of UV absorbance 61 at 254 nm (UV254) to dissolved organic carbon (DOC), describes aromaticity of NOM based 62 on the light absorption at 254 nm of the benzene rings. Higher SUVA values refer to ex-63 istence higher content in aromatic organics regardless of the NOM concentrations [68]. 64 Accordingly, water matrices with higher SUVA are expected to require higher light inten-65 sity to activate the surface of photocatalyst as a compensation for what will be absorbed 66 by NOM. 67 The photocatalytic activity can be also inhibited by the accumulation of NOM on the 68 surface of photocatalyst particles, which blocks the active sites and reduces the illumi-69 nated area [53]. This accumulation decreases the production of ROS even if the catalyst is 70 well illuminated. NOM with high aromatic content is more likely to be hydrophobic, and 71 hence it is more adsorbed at the photocatalyst surface [69]. Furthermore, aromatic NOM 72 quenches the holes in the valence band of the photocatalyst, which not only reduces the 73 degradation caused by the holes but also inhibits the production of hydroxyl radicals [70]. 74 NOM and its transformation products further scavenge the generated ROS in the polluted 75 water inhibiting the oxidation of the target pollutants. 76 Although several studies investigated the influence of NOM on the photocatalytic 77 degradation of micropollutants, the effect of the photocatalytic process on the NOM com-78 position still needs to be clarified. This may be attributed to the challenges associated with 79 the characterization of molecular structure variations of NOM, which makes many studies 80 limited to aromaticity, fluorescence, etc. [71]. This research gap can be filled by advanced 81  Investigating the removal of micropollutants in the range of ng.L -1 to µg.L -1 to simu-89 late natural water streams and groundwater is hindered by the accuracy and detection 90 limit of the analytical methods. In the case of target analysis, photometric and chromato-91 graphic methods are often used to evaluate the photocatalytic degradation. In most cases, 92 these methods are appropriate for the mg.L -1 scale because neither the accuracy nor the 93 detection limit are adequate to work under 0.1 mg.L -1 . 94 The environmental concentrations of emerging micropollutants usually fall below 95 this range and also in the case of several priority pollutants, threshold limits are getting 96 more and more rigorous. For instance, the maximum allowed concentration of pesticides 97 in drinking water in the European Union is 0.1 µg.L -1 , while the maximum annual average 98 concentration of perfluorosulfonic acid and its derivatives is 0.65 ng.L -1 [72]. 99 Recently, gas or liquid chromatography coupled with tandem mass spectrometry 100 (MS/MS) allowed the analysis of micropollutants with an accuracy of a few ng.L -1 and the 101 identification of transformation products by suspect screening [2]. However, these tech-102 nologies are not widely available due to their high cost. Researchers who have no contin-103 uous access to these technologies tend to choose model pollutants with high concentra-104 tions to get clear peaks in conventional liquid or gas chromatography. purposes [72]. This issue may be solved by nontarget analysis, whereby thousands of or-108 ganic compounds could be detected with relative dominancy without the need for refer-109 ence standards [73]. For instance, high-resolution mass spectrometry (MS) can detect the 110 mass-to-charge ratios (m/z) with a precision of 10 -5 Da allowing accurate identifications of 111 organic compounds by computer-assisted data-processing tools [2]. With the spreading 112 of these technologies, we expect to witness photocatalysis studies aiming at investigating 113 the treatment of real water/wastewater based on non-target analysis and very low con-114 centrations in the near future. 115 Another approach evaluates the photocatalytic degradation by the removal of dis-116 solved organic carbon (DOC), chemical oxygen demand (COD), or detoxification. For in-117 stance, Arcanjo et al. used TiO2 modified by iron oxide and hydrotalcite for the treatment 118 of a textile industry wastewater [74]. Although the color removal reached 96%, the COD 119 and DOC removals were not as highly (less than 20% and 10% for the COD and DOC in 120 360 min respectively). The modest removal is imputed to the formation of recalcitrant by-121 products that may require a longer time for degradation. The toxicity was assessed by the 122 130 Many researchers claim they developed the state-of-the-art photocatalyst but provide 131 no information about its cost, toxicity, or life-cycle [11]. These issues are of crucial im-132 portance in terms of the actual full-scale applicability of any photocatalytic technology. 133 The economic sustainability of photocatalytic wastewater treatment on large scale is 134 generally discussed in terms of electrical energy per order (EÉO) values [76,77]. While sev-135 eral advance oxidation processes, including ozonation, UV/H2O2, UV/persulfate, and 136 UV/chlorine present median EEO values < 1 kWh/m 3 , hence are considered competitive for 137 drinking water applications, the EEO for photocatalysis with UV activation are generally 138 > 10-100 kWh/m 3 [78,79]. To make photocatalytic water remediation a competitive ap-139 proach for municipal wastewater treatment, this value needs to be reduced. 140 Although evaluating the energy efficiency of the process via EEO is encouraged, com-141 parisons among different technologies using this single parameter can be misleading. 142 Other factors should be considered, such as the embedded energy costs of consumable 143 chemicals [11]. In this respect, life-cycle assessment (LCA) provides a more complete pic-144 ture of the overall sustainability of a remediation process. LCA is an integrated procedure 145 to compile the environmental impacts associated with a process due to the materials, en-146 ergy consumption, and emissions from cradle-to-grave, based on a huge number of data-147 bases [80]. Despite the wealth of literature dedicated to new photocatalysts, the environ-148 mental impacts associated with the production and operation of these catalysts are usually 149 ignored. Only a few studies compared the impacts of photocatalysis by TiO2 for the re-150 moval of micropollutants with other treatment processes. Magdy et al. found that photo-151 catalytic degradation of phenol by TiO2 in a compound parabolic collectors (CPCs) reactor 152 was more eco-friendly than electro-Fenton and activated carbon adsorption processes, 153 whereas the photo-Fenton process in the same CPCs reactor had less environmental im-154 pacts [81]. The main contributors to the environmental impact were the production of 155 TiO2, electricity used for circulating the water in the reactor, and residual transformation 156 products in the same order. Muñoz  On the contrary, Pesqueira et al. found that the photo-Fenton process had higher environ-159 mental impacts than TiO2 photocatalysis for degradation of a group of pharmaceuticals 160 [82]. This dissimilarity may be attributed to the difference in the chemical dosages and 161 reaction times. It should be noted that LCA analyses of photocatalytic processes have been 162 scarcely reported in the literature. A complete LCA analysis is indeed very challenging 163 because most new complex photocatalysts are not included in LCA databases yet, so re-164 searchers have to build the LCA inventory to include the precursors and energy consump-165 tion used for the catalyst synthesis. Moreover, notable gaps still exist in our knowledge, 166 particularly concerning the toxicity and environmental impact of novel nanostructures. 167 168 For more than three decades, the scientific community has devoted a great deal of 169 effort to developing photocatalytic processes for the removal of a range of air pollutants 170 [83][84][85][86][87][88]. Photocatalysis has demonstrated to be effective for the removal of pollutants in 171 gas phase at relatively low concentrations [89]. Different photocatalytic materials and 172 photoreactors have been proposed [85, [89][90][91]. However, similar to the case of photocata-173 lytic water treatment [10], the process scale-up for the photocatalytic air purification is still 174 an open challenge [92]. 175 When it comes to the application of photocatalysis in real air system, different ap-176 proaches can be envisaged: (iv) Giant solar photoreactors for CO2 reduction have been suggested as possible 202 strategy against climate change [106]: this approach has the additional benefit of produc-203 ing value-added chemicals or fuels from the reduction process. However, research in this 204 field is still in the early stages. 205 New applications are also emerging. In this respect, Horváth et al. [107] have recently 206 reported a photocatalytic mask filter based on TiO2 nanowires, which could be used as a 207 protection against airborne viruses including covid-19, and could be photocatalytically 208 sanitized under UV irradiation ( Figure 5 Mass transfer and contact time are key parameters in photocatalytic air processing, 222 since in air systems, only adsorbed pollutants can undergo degradation by direct oxida-223 tion by positive holes or surface photoproduced ROSs. While this issue is generally less 224 felt in lab scale, batch reactors, where contact times range from seconds to minutes [113], 225 the design of photocatalytic reactors for treating large air volumes requires a careful opti-226 mization of mass transfer issues [114] in order to achieve a fast degradation of pollutants 227 without the accumulation of undesired by-products. In real applications, high flow rates 228 are needed to maximize mass transfer but this comes at the cost of decreasing the contact 229 time between the pollutant and the photocatalyst surface to less than 1 s, which often re-230 sults in an incomplete degradation of the pollutant. To solve this problem, different strat-231 egies have been reported including modifications in the reactor geometry to decrease face 232 velocity [105] or increase in the photocatalytic surface area and irradiance [109]. It should 233 be noted that the ideal contact time depends on the adsorption affinity between the reac-234 tants and the photocatalyst/support [115]. 235 Costa Filho et al. [109] proposed a micro-meso-structured photoreactor for the re-236 moval of gas phase n-decane by simulated solar light ( Figure 6). The adopted geometry 237 allowed a uniform irradiance of the TiO2 photocatalyst, yielding a degradation rate of 238 6.6 mmol m −3 s −1 and no loss of activity after 72 h of consecutive use. Fluidized bed photoreactors have also showed a great potential for the photocatalytic 244 purification of contaminated air. Light irradiation can be external (Figure 7e) [116,117] or 245 internal (Figure 7f) [118], and can also utilize LED lamps [115]. However, the leaching of 246 photocatalysts from the bed was observed in this type of reactors [119]. 247 Very recently, Saoud et al. reported a pilot scale reactor for gas phase VOC remedia-248 tion and E. coli inactivation based on a photocatalytic textile, based on optical fiber, irra-249 diated by UV LED [115]. A 66% degradation of 5 mg.m −3 butane-2,3-dione was measured 250 when the system operated at 2 m 3 .h −1 . 251 More conventional geometries adopting honeycomb monoliths have been already 252 commercialized for the photocatalytic purification of air (Figure 7a,b). Dijk et al.

Photocatalysis for air purification
[120] 253 have designed an internally illuminated honeycomb-based photoreactor for air treatment 254 (Figure 7d) with high surface-to-volume ratio for air purification. This reactor is made for 255 multiphase use [110], wherein the photocatalyst is coated via sol-gel in the inner walls of 256 monolith channels. However, the drawback of this system is the need of separated illumi-257 nation of each monolith channel. Transport cellulosic monoliths allow the penetration of 258 artificial or natural solar light through all monolith channels (Figure 7c) [121][122][123][124][125]. 259 Another reactor geometry for air treatment is tubular flow photoreactors enclosing 260 lamps (Figure 7g,h) [126,127]; for enhanced mass transfer and surface-to-volume ratio, 261 multi-walls based reactors have been constructed as shown in (Figure 7i)  Hybrid techniques combining photocatalysis and other AOPs hold promise for the 277 treatment of large volumes of air due to synergistic effects. Filho et al. [131] have combined 278 O3/UV/TiO2 for the oxidation of VOC in air streams in NETmix micro-photoreactor system 279 (Figure 8a). While single ozonation exhibit fast oxidation towards n-decane but very slow 280 mineralization, the hybrid system O3/UV/TiO2 showed a synergistic removal and miner-281 alization. Zadi et al. [102] combined non-thermal plasma and photocatalysis for the oxi-282 dation of propionic acid and benzene in refrigerated food chambers (Figure 8b). By the 283 use of photocatalysis as a single process, the authors reported serious gas toxicity. The 284 combination of photocatalysis with non-thermal plasma overcomes such a toxicity issue, 285 and on top of that, an enhanced photocatalytic efficiency was observed together with an 286 excellent regeneration of the photocatalytic surface.

294
The efficiency of air purification by photocatalysis depends on a series of operational 295 parameters and environmental conditions. However, comparatively few studies consid-296 ered these all important parameters in photocatalytic tests [105,109] and even less report 297 actual field tests of pilot scale reactors [133]. 298 First of all, the role of relative humidity should be considered. Water vapor is a ubiq-299 uitous component of air and industrial exhaust. The relative humidity of the treated efflu-300 ent can vary greatly depending on the environmental conditions and process parameters, 301 but it is generally present in concentrations far higher than those of the pollutants are. 302 Water vapor is also known to greatly affect the photocatalytic process [113]. However, 303 there is no consensus on the role of air humidity on the photocatalytic performance [115]. 304 Water vapor seems to play different, conflicting roles on gas phase photocatalysis and the 305 overall beneficial or detrimental effect depends on parameters such as the type of pollu-306 tant and its concentration, the photocatalyst adsorption capacity and air relative humidity. 307 It is worth to mention that water competes for adsorption at the photocatalyst active sites, 308 which can lead to a detrimental effect in terms of pollutant adsorption and hence photo-309 catalytic activity [133]. Competition for adsorption between the reaction intermediates 310 and water can also result in a detrimental effect in terms of mineralization [133]. On the 311 other hand, water can react with photogenerated charges to generate reactive radicals, 312 promoting the photocatalytic degradation. The injection of water vapor can enhance the 313 formation of HO • species on the surface of the photocatalyst [134]. Conversely, in dry air 314 system, the yield of photoproduced ROS can be reduced and it is mostly related to O2 -• 315 species. 316 Air samples also contain a mixture of pollutants, generally with individual concen-317 trations in the order of the ppbv. Most literature studies involve photocatalytic tests with 318 a single gas pollutant, often with concentrations in the ppm range. Higher pollutant con-319 centrations are generally associated with improved reaction rates (until rate reaches its 320 plateau), poorer efficiency of removal and lower mineralization [ The deactivation of photocatalyst systems is a serious issue that should be solved for 328 a continuous processing [34]. It takes place during the photocatalytic reaction due to irre-329 versible adsorption of recalcitrant by-products or due site blocking by carbonaceous resi-330 dues or dust particles on the surface [120]. In gas phase photocatalysis, deactivation is a 331 more pressing concern than in liquid phase, as there is no water solvent to help to remove 332 products and intermediates from the surface. 333 Deactivation is most severe in the presence of aromatic compounds [135,136], how-334 ever it has been reported for a wide range of species [137]. For instance, van Dijk et al. 335 reported photocatalyst deactivation within 80 min of photocatalytic operation for the ox-336 idation of cyclohexane [120]. Despite the importance of this topic for the commercial ap-337 plication of photocatalytic technologies, comparatively few studies have investigated the 338 mechanisms of deactivation, regeneration and the photocatalyst lifetime. 339 Several strategies to reactivate the photocatalyst have been proposed, including treat-340 ment with water vapor [120], high temperature treatment [120,137], oxidation with H2O2 341 [137] and UV irradiation [138]. The use of a more oxidizing atmosphere and better mass 342 transfer have been reported to slow down deactivation [136]. 343 Combining photocatalysis with another AOP generally promotes the lifetime of the 344 photocatalyst. Zadi et al. [102] reported that the introduction of non-thermal plasma to 345 the photocatalytic system can lead to continuous regeneration of the photocatalyst in air 346 system. It was found by Ribeiro et al. [139] that addition of ozone allows the use of TiO2 347 without deactivation up to 77 h.

350
Notwithstanding the myriad of studies on the application of photocatalysis to air and 351 water purification, this technology is still far from full-scale, commercial application. 352 To avoid concerns about leaching in the environment of the photocatalysts and its 353 toxicity, several attempts to design photocatalytic reactors with retained photocatalysts 354 have been presented. The designs involve the immobilization of the photocatalyst on a 355 suitable support or in monoliths. Numerous studies showed photocatalytic reactor de-356 signs suitable for reusing the same catalyst for a few cycles with minimal loss of perfor-357 mance; however, more extensive tests on the photocatalyst lifetime should be performed 358 and photocatalyst leaching should be more systematically determined. 359 Another critical issue is related to the adopted water matrix and the use of more re-360 alistic test conditions. Studies reporting the photocatalytic degradation of micropollutants 361 in different water matrices (including treated municipal wastewaters, river waters, and 362 commercial NOM) showed that the photocatalytic activity was inhibited due to the ab-363 sorption of light and scavenging of ROS by the organic backgrounds. Tests in ultrapure 364 water often provide indications that are not transferable to such complex systems. There 365 is the need for robust photocatalyst able to retain its activity in a complex water matrix 366 and in multi-pollutant conditions. 367 The commercial viability and environmental sustainability of the photocatalytic tech-368 nology require more attention to the determination of the costs and LCA analysis of the 369 developed materials and processes. The application of photocatalytic remediation to mu-370 nicipal wastewater treatment has been so far hindered by scale-up difficulties and costs 371 uncompetitive with conventional technologies and other AOPs. However, niche applica-372 tions, such as water treatment in remote locations or the treatment of waste streams from 373 aquaculture and hydroponics, are more promising for short-term applicability of the pho-374 tocatalytic technology. 375 Photocatalytic air purification offers a range of applications, some already at the com-376 mercial stage. However, issues remain in terms of reactor design, especially for the treat-377 ment of large volumes of air without the emission of undesired reaction intermediates. 378 Also, in this case, tests in more realistic conditions should be performed, including con-379 siderations about relative humidity, multi-pollutant interactions, and photocatalyst deac-380 tivation. 381 In general, there is the need of pilot scale studies and tests on real samples to revamp 382 industrial interest in the technology.