Visible-light photocatalysts: Prospects and challenges

In this research update, we discuss the visible-light photocatalysis. Due to the potential of utilizing freely available solar energy for environmental remediation and fuel generation, this topic has been of increasing interest. Huge amount of work has been done in developing a large variety of photocatalyst materials, and advances have been made in understanding the process. Nevertheless, substantial challenges remain. Some of those challenges could possibly be solved by developing better materials, but in many cases, the biggest problem is whether photocatalysis could be scaled up to an industrial process that would be cost-competitive to existing technologies. Here, we discuss different types of visible-light photocatalysts and their applications and outline various challenges that need to be addressed in the development of practically relevant materials and systems.


I. INTRODUCTION
Due to its potential for green energy and environmental remediation, photocatalysis has been attracting increasing attention in recent years. Consequently, there are many reviews on this topic, focusing on different types of photocatalysts and different photocatalyst applications. [1][2][3][4][5][6][7][8][9] Considering the abundance of existing review papers on photocatalysis (over 2700 reviews since 2015, out of which over 1500 discuss visible or solar photocatalysis), in this article, we aim to provide a novel perspective on the issue by concise description of the problems and key strategies to tackle these problems, with selected examples of those strategies. Due to a huge number of published papers, providing a comprehensive overview of everything that has been reported even within a short time frame would be a daunting and not a particularly useful task since the abundance of data would cloud the key issues that need to be addressed. Furthermore, in the last five years, there have been limited fundamental advances in concepts and no significant breakthroughs in photocatalyst design. The principles are known, incremental progress has been made, and there is still lots of work to be done both in making these materials practically relevant (which is questionable for some applications) and in improving our understanding of the complex processes, in particular, in some complicated ternary or quaternary photocatalysts proposed.
The interest in photocatalysis has been rising since seminal works by Honda and Fujishima in 1972 10 and Reiche and Bard in 1979. 11 Various photocatalyst materials and photocatalytic applications have been investigated since then, but the commercial applications of photocatalysis have been scarce. The scarcity of commercial applications is typically attributed to the low photocatalytic activity, in particular, under visible or solar illumination. 1 Consequently, considerable efforts have been devoted into the development of photocatalyst materials with improved performance. Similarly, significant efforts have been devoted to the design of photocatalytic reactors, but this topic is beyond the scope of this review. When it comes to material design and different types of photocatalysts, the material development strategies are focused on maximizing the efficiency by targeting one or more steps in the photocatalytic reaction. The first step is the light absorption and the formation of the photogenerated electron-hole pairs. 1 The photogenerated carriers, then, migrate, and they can either recombine or reach the surface where they can participate in surface redox reactions. 1 Therefore, the strategies aiming at the improvement of photocatalytic performance target maximizing the absorption of light, maximizing charge transfer at the surface, and minimizing recombination by achieving efficient charge separation and maximizing surface redox reactions. 1 To achieve these goals, preferably more than one at a time, common methods investigated for performance improvement are morphology optimization (which can affect the surface active sites, as well as charge separation), doping (which can reduce the bandgap but sometimes can have negative effects on recombination losses), using sensitizers and/or co-catalysts (to increase visible absorption as well as provide more active sites and affect carrier dynamics), and using different types of heterojunctions. 1 We will discuss different types of photocatalysts in Sec. II.
Then, we will discuss the challenges of different types of applications, and we will mainly focus on common environmentally relevant applications, namely, pollutant degradation and solar fuel applications. Other applications of photocatalysis, such as applications in synthetic chemistry, will be only briefly discussed. Before we proceed with that, it is worthwhile to examine some basic premises. The first question is whether photocatalysis is commercially viable. The idea itself is immensely attractive-the use of the low cost environmentally friendly materials and solar energy which is essentially free for water treatment or the generation of hydrogen or hydrocarbon fuels. The answer to this question depends on whether there are less expensive alternatives which are good enough to do the job. The second question is whether it is truly necessary to have efficient visible photocatalysis since there are inexpensive and efficient UV light sources. The answer to this question will depend on the actual numbers, i.e., efficiencies under different excitations, reactor designs, and energy costs and efficiencies [for example, parity violating (PV) system for powering up UV illumination]. The third and final question is whether the photocatalysts are truly environmentally friendly. The answer to this question would be that it depends on the synthesis method, the reactor design, and the possibility of uncontrolled discharge of photocatalysts into the environment. We will discuss these questions in detail in Sec. III on the applications of photocatalysts. Finally, we will provide a brief description of recent advances in characterization techniques, which can provide answers to fundamental questions concerning photocatalysts and photocatalytic reactions, and discuss future outlook of visible photocatalysts.

II. TYPES OF PHOTOCATALYSTS
Popular photocatalyst materials, either as single material photocatalysts or as parts of heterojunctions (including Z-scheme systems), as illustrated in Fig. 1, are TiO 2 , [12][13][14][15][16] ZnO, 3,12,14 carbon nitride g-C 3 N 4 , 2,4,17 metal-organic framework (MOF) compounds, 8 graphene-based photocatalysts, 7 BiOCl, 18,19 black phosphorus, 20,21 ZnFe 2 O 4 , 22 etc. Among these materials, TiO 2 is still the one most commonly evaluated in different reactor designs, while g-C 3 N 4 is an increasingly popular component of photocatalysts for fuel generation due to its lower bandgap (∼2.8 eV), favorable conduction band (CB) edge position, and the abundance of surface groups that enable coupling to other photocatalysts. 23,24 However, graphitic carbon nitride has inferior chemical stability compared to TiO 2 , 15 and it can be susceptible to degradation by hydroxyl radicals. 25 It can also exhibit high recombination rates, low conductivity, insufficient visible-light absorption, and low solvent-accessible surface area. 24 Consequently, great efforts have been made in modifying the structure of polymeric carbon nitride to improve its performance. Some recently proposed materials, for example, metal halide perovskites such as CsPbBr 3 , 26 are also of interest, in particular, from the point of view of understanding the fundamental mechanisms in these remarkable materials. However, while there are many studies on designing novel photocatalysts, it should be noted that many materials investigated in research papers are unlikely to have practical application despite being promising in bench-scale evaluations. 15 Any material that contains toxic elements (CdS and CsPbBr 3 , for example), or rare and expensive components, or is fragile or chemically unstable is unlikely to be practically relevant, which likely contributes to the popularity of TiO 2 photocatalysis. 15 Single material photocatalysts (n-type and p-type semiconductors) are shown in Fig. 1. In addition to conventional semiconductor photocatalyst materials, MOFs have been attracting increasing attention recently. For a detailed review of MOFs for various photocatalytic applications, including hydrogen generation and CO 2 reduction, see Ref. 8. Similar to other photocatalysts, these materials still need significant improvements in achieved efficiencies, but they have been studied for considerably shorter time compared to conventional inorganic semiconductors. 8 In addition, organic polymers have also been attracting interest for photocatalysis applications. 27 Still, the majority of literature reports concern inorganic semiconductors, and hence, we will discuss those in more detail. However, the investigations of single material photocatalysts have been declining in popularity in recent years, mainly due to difficulties in simultaneously achieving narrow bandgap, favorable energy band alignments, and low recombination losses in a single material. Nevertheless, it is still worth mentioning common strategies for improving the performance of single material photocatalysts for visible-light photocatalysis since these strategies can still be employed to improve the performance of individual components of more complex photocatalysts. These strategies typically involve the optimization of crystal structure, surface area, and morphology; the manipulation of native defects; or the introduction of dopants to adjust the optical and electronic properties. In addition, sensitizers can be used to enhance the light absorption in the visible spectral range.

Surface engineering
The crystal structure of a semiconductor is known to affect its photocatalytic efficiency. 28 Since photocatalytic reactions occur at the active sites on the photocatalyst surface, the photocatalytic efficiency can be increased by nanostructure size and morphology optimization. 16 The number of active sites increases with the increase in the surface area, as well as the increase in the proportion of crystal facets with higher photocatalytic activity. 16,29 In general, morphological effects (for example, nanocrystals vs nanowires of the same material) on the photocatalytic activity can be interpreted as crystal facet-dependent reactivity or structural sensitivity. 28 Higher surface energy facets are thermodynamically unfavorable but can be achieved via careful optimization of the synthesis conditions. 16,29 In TiO 2 , increased photocatalytic activity can be achieved by optimizing {101} (reducing sites) and {001} (oxidizing sites) facets, while in ZnO increased activity is achieved on zinc terminated polar {0001} facets. 29

Defect engineering
Both cation [30][31][32][33] and anion 31,34 defects can affect photocatalytic activity. An example of the strategy of the use of native defects is using anion vacancies (oxygen vacancies 19 or nitrogen deficiency 35 ) to control photocatalytic activity. In addition, it was shown that individual defects could have either positive or negative effects on the photocatalytic activity, 36 for example, the introduction of oxygen and nitrogen defects in g-C 3 N 4 produced defects that are beneficial (O= =C, − −NHx species and N defects at the N-2C site) or detrimental (hydroxyl groups) for hydrogen evolution. 36 In addition to these possible defects, cyano group (C≡≡N) is another relevant defect in g-C 3 N 4 . 37 Native defects, such as oxygen vacancy, could not only introduce gap states to extend the photocatalytic response to longer wavelengths (depending on the semiconductor), but they can also enhance exciton dissociation, which contributes to enhanced charge carrier transfer and, consequently, photocatalytic activity. 34 Furthermore, defects can also serve as active sites for molecule chemisorption and alter the reaction pathways. 38 In addition to point defects (anion vacancies, cation vacancies, multivacancies, and interstitials), larger structural defects (edges in 2D materials and disorder) can contribute to photocatalytic activity. For example, surface disorder and native defects induced by reducing reactions in metal oxides (heating with a reducing agent) can result in a significant enhancement of visible-light absorption. 39 Furthermore, photocatalytic activity can also be affected by morphology optimization. [40][41][42] For example, in porous g-C 3 N 4 , enhanced hydrogen evolution and pollutant degradation were observed not only due to surface area increase but also due to improved charge carrier separation. 40 Increased photocatalytic activity for hydrogen generation was also reported for multishell g-C 3 N 4 . 41 For a detailed review of defect engineering in photocatalysis (for water splitting applications), see Ref. 43.

Doping
Doping is another relatively common method to extend the absorption response to longer wavelengths. It can also be used to modulate the growth of the photocatalyst to stabilize more reactive facets to enhance photocatalytic activity. 44 Common dopants for wide bandgap oxide photocatalysts, such as TiO 2 and ZnO, are nitrogen, sulfur, carbon, phosphorus, and fluorine as non-metal dopants and Ag, Mg, Cu, Rb, Co, Mo, Au, Pt, Fe, Cr, V, etc., as metal dopants. 12,15,45 Hydrogenation can also be used to extend the optical response of titania into the visible spectral range (i.e., black titania). 46,47 Nitrogen doping has been reported in porous carbon, 48 while dopants in g-C 3 N 4 include oxygen, sulfur, nitrogen, iodine, bromide, and phosphorus. 37,49,50 In addition, alkali metal (Li + , Na + , and K + ) 51 and chlorine 52 doping has also been reported for g-C 3 N 4 . In addition to single dopants, co-doping can be used. Alternatively, doping and defect engineering can be combined for synergistic improvements in performance. 48 It should be noted that while doping can indeed extend the absorption into the visible range by introducing midgap states, those states can also act as recombination centers and increase recombination losses, while the shift in the redox potential can negatively affect the surface redox reactions. 15 Another approach to extend the absorption range into the visible is the formation of solid solutions, such as oxynitrides, GaN:ZnO, or InN:ZnO. 53,54

Sensitizing
Increased visible-light absorption can be achieved using sensitizers. Sensitizers such as adsorbed organic dyes or metal complexes can be used for this purpose. 29 As an example, organic metal complexes, such as Ru(bpy) 2+ PERSPECTIVE scitation.org/journal/apm in decline in recent years, and the majority of research efforts have focused on the various composite materials.

B. Semiconductor heterojunction photocatalysts
To extend the light absorption into the visible range, but without the drawback of increased recombination, metal-semiconductor or semiconductor-semiconductor heterojunctions can be used. 1,15 For a comprehensive review of heterojunction photocatalysts, see Ref. 1. In the following, we will discuss different types of heterojunction photocatalysts.

Common heterojunction photocatalysts
Heterojunction photocatalysts can be classified on the basis of energy band alignments (type I, type II, and type III), as illustrated in Fig. 2, or they can be classified based on material morphology, such as partial contact photocatalysts and core-shell photocatalysts. An important factor to consider in the heterojunction design for efficient photocatalytic activity is the band edge alignment. 56 The band edge alignment at the interface is affected by band bending when two materials come in contact, and it can be significantly affected by interface properties (surface dipoles, the presence of interfacial defects that result in trapped charges, etc.). In general, this is a common phenomenon affecting all heterojunctions and it has been extensively studied in various electronic and optoelectronic semiconductor devices. When it comes to photocatalysts, however, there are some distinctive features less commonly observed in layered structures in electronic devices. For example, the band alignment and band bending can be strongly dependent on the particle sizes and their arrangement on each other. 57 Facet dependence has also been observed in band bending, which affects the charge transport across interfaces of different contacting planes and, consequently, photocatalytic activity. 58 Furthermore, band bending not only occurs in junctions between different materials, or different crystal phases of the same material (heterophase junctions), but can also occur between large and small particles of the same material with the same crystal structure. 59 In this case, the difference in the surface band bending results in an electric field that drives the charge carrier separation. 59 Depending on the morphology of two semiconductors, partial contact photocatalysts can have point contact where one of the photocatalysts is a nanoparticle, line contact for nanowire

PERSPECTIVE
scitation.org/journal/apm or nanoribbon/nanosheet catalysts, and face contact when both photocatalysts are two dimensional. 1 In addition to energy levels and morphology, another consideration for the choice of materials to form a heterojunction is the lattice mismatch, with materials having low lattice mismatch being preferred. 60 Higher lattice mismatch will result in increased strain, which in turn can lead to the increased concentration of interfacial defects that can increase recombination losses and affect energy level alignment at the interface. Various heterojunction photocatalysts have been reported, such as CeO 2 − −AgI, 61 56 In addition to more conventional heterojunctions involving common inorganic semiconductors, the combinations of semiconductor photocatalysts with various polymers and/or carbon-based nanostructures, such as chitosan, 12 reduced graphene oxide (rGO), 26 polyaniline and nitrogen-doped rGO, 71 carbon nanotubes (CNTs) and graphene, 72 polypyrrole, 73 and polyaniline, 74 have also been investigated. Furthermore, MOF-based heterojunctions, such as zeolitic imidazolate framework-8 (ZIF-8) nanoclusters on g-C 3 N 4 nanotubes, 75 have been investigated to combine photocatalytic action of g-C 3 N 4 with enhanced CO 2 adsorption of a MOF. In addition, ZIF-8 has also been used in MoO 3 /ZIF-8 core-shell nanorods for photocatalytic water purification [Cr(VI) reduction]. 76 Finally, we should also note that Janus nanostructures, i.e., nanostructures with surfaces with two distinct properties, can be used in heterogeneous photocatalysis, as described in a recent review. 77 Heterojunctions generally exhibit better performance compared to single material photocatalysts. Nevertheless, further improvements are still necessary to enhance visible-light absorption and improve charge carrier separation. To achieve this, specific types of heterojunctions, namely, Z-scheme heterojunctions and plasmonic photocatalysts, have been investigated to reduce recombination losses and utilize plasmonic effects, respectively.

Z-scheme photocatalysts
While common heterojunction photocatalysts improve the carrier separation and reduce charge recombination losses, it is still difficult to achieve efficient charge separation and strong redox capability simultaneously. 17 In a Z-scheme photocatalyst, as illustrated in Fig. 1, the transfer of electrons from the CB of one material to the valence band (VB) of the other material through an electron mediator ensures a larger difference between oxidation and reduction energy levels compared to a heterojunction. 17,78 Consequently, this type of photocatalyst is particularly promising for water splitting reactions, with a theoretical maximum solar energy conversion efficiency of 40%, which is higher than what can be achieved with single photocatalysts (∼30%). 79 The Z-scheme concept is actually not a particularly new idea since it was first proposed in 1979. 9 However, this concept has been attracting a lot more attention recently, with increased interest in photocatalytic fuel generation. For example, one of the important attractions of Z-scheme photocatalysts for water splitting is the possibility of combining the hydrogen evolution photocatalyst and oxygen evolution photocatalyst to achieve higher water splitting efficiency. Liquid Z-scheme systems were demonstrated for water splitting in 2001, followed by the demonstration of a solid-state Z-scheme system for water splitting in 2009. 9 In liquid Z-scheme systems, reversible redox mediators in solution have been commonly used. 17 Common electron mediators in liquid Zscheme systems are Fe 3+ /Fe 2+ , IO − 3 /I − , and NO − 3 /NO − 2 . 78,79 In addition to these, various cobalt-based organic compounds have been reported, 78,79 as well as transition metal polyoxometalates and vanadate ion shuttle. 79 However, due to backward reactions, low pH tolerance, and light harvesting issues in liquid systems, solid-state Z-scheme catalysts are of significant interest. 17 In a solid-state Zscheme photocatalyst, conductors (metal or carbon based) can serve as electron mediators. 17 The common conductor mediators include Au nanoparticles, Ag nanoparticles, rGO, and carbon nanotubes (CNTs). In addition, defective metal oxides, for example, Mn vacancies in MnO 2 , can provide redox couples, in this case Mn 3+ /Mn 4+ . 80 Furthermore, direct Z-scheme photocatalysts where electrons are transferred across the interface between two semiconductors have been reported. 17 Energy band alignment in a direct Z-scheme photocatalyst resembles that of a type-II heterojunction, but the charge carrier transfer pathways are distinctly different in these two structures. 81 In a type-II heterojunction, electrons move from the material with a higher conduction band minimum (CBM) to the one with a lower CBM position, while holes move in the opposite direction, minimizing recombination. However, this results in lower redox ability since the resulting redox levels correspond to a higher position of the valence band maximum (VBM) and lower position of the CBM between the two materials. 81 In a direct Z-scheme heterojunction, the electrons in the material with lower CBM recombine with the holes in the material with higher VBM, resulting in a larger difference between oxidation (lower VBM position) and reduction (higher position of the CBM) energy levels and, consequently, improved redox capability. 81 Compared to conventional Z-scheme photocatalysts, direct Z-scheme photocatalysts have an advantage of simpler structure, as well as lower absorption losses due to the light absorbed in the mediator. 78 It should be noted, however, that there is a possibility for the metallic mediator to contribute to the plasmonic enhancement of photocatalytic activity for a suitably constructed system. Different material combinations have been reported as Z-scheme photocatalysts to date, such as graphene bridged Ag 3 PO 4 /Ag/BiVO 4 , 82 Ag 2 Mo 2 O7/MoS 2 , 83 WO 3 /g-C 3 N 4 , 84 g-C 3 N 4 / SnS 2 , 23 TiO 2 /CuInS 2 , 85 TiO 2 /CdS, 86,87 α-Fe 2 O 3 /g-C 3 N 4 , 88,89 WO 3 / g-C 3 N 4 , 90 SrTiO 3 :La, Rh/C/BiVO 4 :Mo, 91 CdS/RGO/g-C 3 N 4 , 92 black phosphorus/BiVO 4 , 93 g-C 3 N 4 /Ag/MoS 2 , 94 g-C 3 N 4 /nanocarbon/ ZnIn 2 S 4 , 95 and MnO 2 /monolayer g-C 3 N 4 with Mn vacancies. 80 g-C 3 N 4 is a particularly popular material for Z-scheme photocatalysts, and a number of different material combinations have been reported. 17 Z-scheme photocatalysts in general have been attracting increasing attention for a variety of applications, including degradation of various pollutants as well as fuel generation (water splitting as well as CO 2 reduction). 78,79 They are among the most promising types of photocatalysts, especially when combined with additional modifications, such as plasmonic materials.

Plasmonic photocatalysts
For any of the photocatalyst classes of materials discussed above, it is possible to enhance photocatalytic activity by incorporating metal nanoparticles to utilize plasmonic effects, namely, resonant oscillations of the electrons within a plasmonic nanoparticle.

PERSPECTIVE scitation.org/journal/apm
The plasmonic wavelength of a material is dependent on its dielectric function, as well as size and morphology. 96 Plasmonic effects can be combined with other functions for the metal, for example, in solid Z-scheme photocatalysts. 94 They can also enhance the photocatalytic activity of metal-organic frameworks. 97 The enhancement of the photocatalytic activity by plasmonic nanoparticles can occur by different mechanisms, namely, hot electron injection, localized electromagnetic field enhancement, and plasmon resonance energy transfer (PIRET) from the metal to semiconductor through dipole-dipole interaction. 98 The most common nanoparticles are Ag nanoparticles due to their strong localized surface plasmonic resonance (LSPR), 99 although other metals have also been used for the purpose of plasmonic enhancement. While the use of noble metals (Ag and Au) for this purpose is common, 96,98 there have also been reports on the use of non-noble metals, which has an advantage of lower cost, and the use of more abundant materials. For example, the use of bismuth metal has been reported. 18,100 In addition, Al 96 and Cu 96,101 have also been proposed as suitable candidates for the plasmonic enhancement of photocatalysis since their plasmonic wavelengths can be tuned from 150 nm to 600 nm and from 600 nm to 1000 nm, respectively. 96 Furthermore, non-metallic materials, such as MoS 2 , have also been reported. 102 To further tune the properties and achieve increased enhancement, it is possible to use combinations of materials, such as Au/Ag/Cu 2 O core-shell nanorods 103 or Au/CuSe/Pt (where CuSe nanosheets and Au provide plasmonic effects, while Pt is a co-catalyst). 104 In addition, the use of connected instead of isolated nanoparticles could lead to an increased electromagnetic field due to strongly coupled structures and, consequently, increased photocatalysis enhancements. 105 Similarly, plasmonic dimers have demonstrated greater enhancement of hydrogen generation and CO 2 reduction compared to non-dimer structures. 60

III. APPLICATIONS OF PHOTOCATALYSIS
The applications of photocatalysis can be broadly divided into following categories: pollutant degradation, fuel generation, and synthetic chemistry applications. In this review, we will mainly focus on the main environmentally relevant types of applications, namely, pollutant degradation and fuel generation. Pollutant degradation has been demonstrated for a variety of chemical compounds. Disinfection applications can also be broadly defined as pollutant degradation. Photocatalytic fuel generation reactions include hydrogen evolution and CO 2 reduction. In this section, we will discuss these categories of photocatalyst applications.

A. Pollutant degradation
The use of various photocatalysts for pollutant degradation has been extensively explored, with some recent examples summarized in Table I. The problems from material design point of view that need to be solved are the same as in any other photocatalytic application, i.e., efficient photocatalyst should exhibit high light absorption, low recombination of photogenerated charges, and efficient charge transfer to the reactants. While there are numerous examples of the use of photocatalysis for the degradation of various dyes, this type of application is of less interest in modern photocatalysis research, based on the declining number of publications on dye degradation work published in high impact factor journals. Contributing to the declining interest in photocatalytic dye degradation work is declining significance of insights gained into the photocatalytic mechanism from this type of studies. While the dye degradation can be (in theory) justified by the huge amount of dye pollutants discharged by various industries, including textile industry, and while experiments on dye degradation are very simple and straightforward, the lack of new insights into the process, complications in the interpretation of results, and the lack of practicality of application for industrial wastewater treatment have resulted in declining importance of this type of studies. Since the dyes can undergo irreversible transformation through different pathways during the photocatalytic degradation experiment, other characterizations in addition to the loss rate are needed to evaluate the photocatalytic activity of the material. 106 Consequently, other compounds such as phenol have been proposed to compare photocatalytic activities. 106 The investigation of degradation of phenol is sometimes combined with other applications. For example, the degradation of phenol 107 (as well as organic dyes 108 ) and hydrogen evolution were demonstrated in photocatalytic fuel cells. In addition to the shift in the interest in the types of pollutants to be degraded, there has also been a shift in the interest in photocatalytic pollutant degradation in general. The development trends of a research field can be estimated from the trends of publications in high impact factor journals (although the journal impact factor is not a measure of the quality of the work, generally this type of journals tends to publish more papers in areas currently considered to be "hot"). We can observe that while the number of total publications on photocatalysis is still growing, the number of papers on photocatalytic degradation in high impact factor journals has definitely declined, as illustrated in Fig. 3. Nevertheless, it is still relevant to discuss some specific types of pollutants, as well as limitations of photocatalysis in water purification and opportunities in air purification.

Emerging pollutants
Among different pollutants, the use of photocatalysis for the degradation of emerging pollutants has been attracting increasing attention. In the case of emerging persistent organic pollutants, conventional treatment measures are either insufficiently effective or they simply transfer the pollutant to another phase (from water to sludge, for example), which could cause secondary pollution. It should be noted, however, that while the lack of selectivity of photocatalytic degradation is often listed as an advantage of photocatalysis, this means that emerging pollutants that are often present at very low concentrations (ng/l to μg/l) would likely still persist, while pollutants with higher concentrations will be more significantly degraded. Therefore, photocatalytic degradation of emerging pollutants, if incorporated at all, likely could only be used as a last treatment stage to only remove low concentrations of pollutants, which could not be removed by other methods. Furthermore, it should be noted that the presence of natural organic matter and various anions and cations in the water can affect the overall effectiveness and reaction pathways of photocatalytic degradation of pollutants. 13,15,109 The presence of other species in the water in addition to target pollutants limits the effectiveness of the process by absorbing or scattering light and by scavenging generated reactive oxygen species. 15 Furthermore, there is a possibility of generation of toxic, assimilable, and sensory-unpleasant   transformation products if photocatalysis is used for water treatment. 110 a. Pharmaceuticals. There are different categories of emerging pollutants, including various personal care products, pharmaceuticals, and persistent organic pollutants. Antibiotics belong to the category of emerging pollutants since their extensive use in human, veterinary, and agricultural applications has resulted in contamination of surface, ground, and drinking water. 82,99,111,112 Pharmaceutical factories and hospital wastewater could also contribute to the presence of pharmaceuticals in water. 99,113 In the environment, antibiotics can form unknown complex mixtures with other organic pollutants or heavy metals, affect the biological organisms in the environment, and contribute to the increase in antibiotic resistant infections. 83 They are typically not completely removed during the standard wastewater treatment process but can be removed by photocatalysis. It has been shown that photocatalytically treated antibiotic contaminated water did not promote antibacterial resistance over a period of one year. 113 118 and ZnFe 2 O 4 /rGO. 119 In some cases, both antibiotic pollutant degradation and hydrogen evolution have been demonstrated. 116 In addition to antibiotics, the degradation of other pharmaceuticals and personal care products is also of significant interest. 120,121 For example, the degradation of naproxen, a common non-steroidal anti-inflammatory drug, was demonstrated using a photocatalyst consisting of g-C 3 N 4 , carbon quantum dots, and single-atom dispersed silver. 120 Another carbon nitride based heterojunction photocatalyst has also been demonstrated to effectively degrade ibuprofen. 62 In addition, TiO 2 was shown to be effective for the degradation of carbamazepine, diclofenac, and ibuprofen in a reactor using visible LEDs as the light source. 121 b. Plastics industry pollutants-endocrine disruptors and microplastics. Another class of emerging pollutants of significant interest are endocrine disruptors, such as bisphenol A, which is commonly used in epoxy resins and polycarbonate plastics. 122 Different materials have been proposed for bisphenol A degradation, such as CeO 2 − −AgI, 61 cyclodextrin decorated TiO 2 spheres, 123 Ag 2 O/Bi5O7I, 63 and reduced graphene oxide (rGO) supported on silica and zirconia. 122 Microplastics represent another class of emerging pollutants that are a significant concern due to their persistence in the environment. They originate from primary sources, i.e., micrometer-sized plastics discharged into the environment such as high density polyethylene particles from personal care products, and secondary sources, i.e., from breakdown of larger plastics pieces. 124 Some promising results on microplastic degradation using photocatalysis have been reported, but overall the number of publications is low. The materials that have been reported to be effective for microplastic degradation include nitrogen-doped TiO 2 124 and ZnO. 125 It should be noted that the demonstration of photocatalytic degradation of microplastics is relatively recent compared to the research on other emerging pollutants, although it has been known that the incorporation of photocatalytic particles (ZnO or TiO 2 ) into plastics can affect their weathering under UV illumination.

Limitations for practical applications in water purifications
It should be noted that even for emerging pollutants, it is likely that the practical applications of photocatalysis will face the same problems as the applications in conventional wastewater treatment. 126 In simple terms, the biggest issues with practical applications of photocatalysis are problems with scaling up to an industrial process and being cost-effective and time-effective. 126 Despite the fact that photocatalysis has been around for almost half a century, so far there is only one commercial photocatalytic water treatment system, the Purifics Photo-Cat. 126 The system is not solar, but rather it uses UV lamps, and its efficiency can actually be higher if the photocatalyst is replaced with hydrogen peroxide dosing. 126 In addition, while a suspended photocatalyst reactor, i.e., the same type as used in the Purifics Photo-Cat, is the most viable and practical option, the treatment cost is further increased by photocatalyst recovery in this design. 14,126 Furthermore, while visible-light photocatalysis is a very popular topic among photocatalysis research papers, it should PERSPECTIVE scitation.org/journal/apm be noted that using solar energy instead of UV lamps would require much higher land area footprint, with basically two order of magnitude larger area needed to provide the same light energy (and that would apply only for a small part of the day and optimal weather). 126 Generally, photocatalytic water treatment systems are considered impractical in the water treatment industry since conventional technologies exhibit significantly higher energy efficiency. 15 Conventional advanced oxidation processes involving the use of ozone, chlorine, and/or hydrogen peroxide with or without UV irradiation typically have electrical energy per order values EEO (energy in kW h required to degrade the contaminant by one order of magnitude in 1 cubic meter of contaminated water) below 1 kW h/m 3 , while for UV photocatalysis this value exceeds 100 kW h/m 3 . 127 Thus, it is unrealistic to expect significant changes in the application of photocatalysts in industrial water treatment. Niche applications have been proposed as a possible way forward, and these include applications to specific pollutants, such as heavy metal reduction to a less toxic or less soluble form [Cr(VI) to Cr(III) 15,71,74,76,118,128,129 and U(VI) to U(IV) 130 ] and smaller scale applications. 15

Pathogen elimination
Photocatalysis can also be applied for water disinfection, and it is effective not only against bacteria 131 but also against viruses. 132 For a detailed review of antimicrobial activity of photocatalysts, see Ref. 133. Consequently, solar photocatalysis is often cited as a possible solution for ensuring safe drinking water supply in the developing countries. While the main challenge in the developed countries is the maintenance of existing infrastructure and the treatment of wastewater, in developing countries, the priority is obtaining safe drinking water by eliminating pathogenic microorganisms. 134 Similar to water treatment for pollutant removal, despite successes in lab-scale experiments, industrial scale photocatalytic disinfection is impractical. For example, various pilot photocatalysis projects, such as the Solwater TiO 2 /solar concentrator project, have confirmed unrealistic treatment times and volumes (4 h for disinfection in a 20 l reactor). 126 There are considerably more cost-effective solutions, which use solar energy and are applicable to the developing countries, such as solar pasteurization and solar disinfection (SODIS). 134,135 SODIS, in particular, provides a solution that is difficult to beat in terms of simplicity and low cost-it simply involves filling water into transparent containers such as plastic bottles with one side painted black (or alternatively they can be placed on a black surface) and leaving them out in the sun for a period of time. 134 The method can be improved by using parabolic concentrators and glass containers, which can shorten treatment time and eliminate the need for periodic replacement of plastic bottles. 134 While on the lab-scale improvements in SODIS by photocatalyst incorporation have been demonstrated, there have been no practical field designs due to difficulties and drawbacks for both suspended and immobilized photocatalyst reactors. 136 Therefore, the usefulness of photocatalysis for practical disinfection applications is limited. However, the investigation of the toxicity mechanisms of photocatalysts from the basic science point of view could still be potentially interesting since the mechanisms of action are still not well understood. 131 This includes the fact that antibacterial activity and photocatalytic pollutant degradation can proceed based on different pathways 61,137 and can exhibit different trends, i.e., efficient antibacterial materials are not necessarily good for pollutant degradation and vice versa.

Air purification
Finally, in addition to water treatment, photocatalysis could be applied to air treatment. [138][139][140] The pollutants of interest include NOx, SOx, CO, H 2 S, and volatile organic compounds (VOCs). 138 It has been proposed that different from water treatment and fuel generation, photocatalysis can be both economically and technologically competitive for air treatment. 138 For a detailed review of this application, see Ref. 138.

B. Fuel generation
Photocatalytic fuel generation includes photocatalytic water splitting to produce hydrogen and photocatalytic CO 2 reduction to produce hydrocarbon fuels. As all photocatalytic processes, photocatalytic fuel generation reactions involve light absorption, followed by separation and transport of charge carriers, and finally surface redox reaction, in this case the reduction of water or CO 2 by photogenerated electrons. 7 For efficient redox reactions, however, it is necessary for the energy band positions (valence band for oxidation reaction by holes and conduction band for reduction reaction by electrons) to match the oxidation and reduction potentials of water splitting and the reduction potential of CO 2 . 7 The number of potential photocatalysts that can produce hydrogen from water (conduction band more negative than the redox potential of H 2 /H 2 O and valence band more positive than the redox potential of O 2 /H 2 O) is limited, as illustrated in Fig. 4. 7 Similarly, for CO 2 reduction, the conduction band edge needs to be more negative than the corresponding redox potential. 7 Furthermore, reactants should be adsorbed onto the photocatalyst, while reaction products should desorb, and excess opposite charge carriers, such as holes in the carbon dioxide reduction process, should be scavenged before they recombine with photogenerated electrons. 6 While charge separation is typically addressed by photocatalyst design (heterojunctions and Z-scheme photocatalysts), attention should also be devoted to reactor design to separate reaction products to prevent back reactions. 6 For the CO 2 reduction, aqueous photoreactors and gas-phase photoreactors can be used. 6 Both water splitting and CO 2 reduction reactions are "uphill" chemical reactions, resulting in an increase in the Gibbs free energy. Therefore, for both types of reactions, the biggest problem is the low efficiency of the process. However, hydrogen production by water splitting is a simple reaction and a two electron process. In contrast, CO 2 reduction is complex and requires 6-8 electrons, with a range of possible reaction products. 6,7 Consequently, due to increased complexity of reactions, the efficiency of CO 2 reduction is even lower than that of hydrogen production. 7 Unfortunately, from the point of view of mitigating climate change, CO 2 reduction is more attractive since it can generate hydrocarbon or alcohol fuels from CO 2 . Consequently, the photocatalysis research into fuel generation has been rapidly developing. Several detailed reviews of CO 2 reduction 6 and hydrogen production 9 have been published recently. A particularly attractive concept in solar fuel generation is the artificial photosynthesis, 6 where useful fuels can be generated from carbon and water. Unlike natural photosynthesis, which produces carbohydrates, artificial photosynthesis produces a range of partially reduced compounds, such as CO, CH 4 , and methanol. 26 The most common products of CO 2 reduction are carbon monoxide and methane. Photocatalytic CO 2 reduction rarely results in high-grade carbon species (Cn, n ≥ 2), but, nevertheless, such complex processes are possible. 144 For example, the evolution of CH 4 and C 2 H 6 on Au@TiO 2 yolkshell hollow spheres is illustrated in Fig. 5. 144 It can be observed that while Au@TiO 2 results in the evolution of C 2 H 6 , TiO 2 alone (including standard photocatalyst P25) does not generate high-grade carbon species, and the generation of C 2 H 6 was attributed to the plasmonic enhancement of the electric field by Au nanoparticles within titania shells. 144 Due to significant interest in the generation of high-grade carbon species, the improvement of understanding of the processes involved in producing Cn, n ≥ 2, carbon species is crucial for the development of high performance photocatalysts for solar fuel generation. Both reactions also utilize similar photocatalyst materials since the values of redox potentials for CO 2 and H 2 O reduction are close 28 and tend to use co-catalysts. Nanoparticle co-catalysts (most commonly, noble metals) are used to provide active sites and reduce activation energy for the reaction. 145 In addition to increasing the activity, co-catalysts can also affect the charge separation, selectivity, stability, and back reactions. 69 Co-catalysts commonly promote either reduction or oxidation reactions, such as hydrogen evolution reaction (HER) or oxygen evolution reaction (OER) in water splitting. However, they can also increase the rates of backward reactions, for example, hydrogen oxidation reaction (HOR) and oxygen reduction reaction (ORR) in water splitting. Therefore, an important part of the design of the photocatalytic water splitting system is the selective deactivation of reverse reactions. For example, the use of oxides of Ni, Ir, and Ru instead of pure metals suppresses ORR while retaining the promotion of HER and OER. 9 Although there have been examples of co-catalyst-free fuel generation photocatalysts, 146 the majority of reported photocatalysts for both water splitting and CO 2 reduction include some form of co-catalysts. Co-catalysts can be classified as noble metal and rare earth co-catalysts, noble-metal-free co-catalysts, metal-free co-catalysts, and biological co-catalysts. 69 For a detailed review of co-catalysts for CO 2 reduction, see Ref. 69. Noble-metal co-catalysts are very commonly used for both water splitting and CO 2 reduction. Various co-catalysts used in recent works include Pt as the most common co-catalyst, 41,49,88,90,[147][148][149] as well as Cu I /Pd, 33 Ru, 91 Pd, 54 and Rh. 54 Noble metals combined with metal oxides could also be used as co-catalysts, such as Au− −TiO 2 . 150 However, due to high cost and scarcity of noble metals and rare earths, possible alternatives are the use of single-atom catalysts, which are generally promising for photocatalytic applications, [151][152][153][154] or the development of non-noble-metal co-catalysts. For a detailed review on single metal atom photocatalysis, see Ref. 153.
It should be noted that non-noble metals can also serve as singleatom co-catalysts, 155 although non-noble-metal materials are more commonly investigated in other forms, and 2D materials and clusters are of significant interest in addition to single-atom catalysts. 156 For example, 2D materials, such as MoS 2 , can serve as effective cocatalysts due to abundant active sites at edges. 65 Other metal sulfides have also been reported to be effective co-catalysts, such as CoSx polyhedrons. 4 Other non-noble-metal co-catalysts have also been reported, such as amorphous NiO, 66 Co 2 P, 157 Co(bpy) 2+ 3 (bpy = 2 ′ ,2 bipyridine), 70 and MnOx, 150 as well as metal-free co-catalysts, such as phosphorene, 69 Ti 3 C 2 MXene, 158 and graphene oxide. 159 In addition, dual co-catalysts, such as NiS/carbon dots, 160 CoP, and Pt 161 have been reported. Dual co-catalysts can enable the enhancement of both oxidation and reduction reactions, but co-loading of different types of co-catalysts has been less explored to date. 9,162 In addition to co-catalysts, some research works report the use of photosensitizers, such as Ru-based dye. 55 It should also be noted that the reactions are typically affected not only by the material choice for the co-catalyst but also by the co-catalyst size and morphology. 163,164 Both efficiency and selectivity can be affected by co-catalyst size and morphology since these factors (size and terminating facets) determine the concentrations and types of active sites present on the surfaces. 163,164 Another feature that both reactions have in common is that reports in the literature often involve half-reactions, with sacrificial agents and/or electron donors added. 9,165 However, it should be noted that sacrificial half-reactions are generally "downhill" chemical reactions, different from true water splitting and/or CO 2 reduction, which are uphill processes. 9 Consequently, photocatalytic activities for true water splitting and sacrificial reactions are not directly correlated. 9 Furthermore, since the common sacrificial donors, such as alcohols and triethanolamine, contain carbon and hydrogen, they can contribute to the measured yield of reaction products and result in erroneous estimations of yield. 165 Despite the calls for improved reporting on heterogeneous photocatalysis and summaries of best practices, 165,166 the majority of literature reports still use sacrificial agents and only investigate the reduction part of the reaction. For a more detailed discussion of overall water splitting, see Ref. 166.
It should be noted that when discussing progress made to date, due to the fact that reaction rates are strongly dependent not only on the photocatalyst used but also on the actual experimental conditions used for photocatalytic activity testing, it is not possible to provide meaningful comparisons of the numbers reported in different studies. We, therefore, simply list some examples of relevant materials and discuss important issues for the commonly investigated ones. We will also list some recent achievements in terms of apparent quantum yield (AQY) or apparent quantum efficiency (AQE) since the information on AQY or solar-tohydrogen (STH) efficiency provides some basis for comparison in water splitting applications. 168 Among few studies that report apparent quantum efficiency for water splitting, reported figures include the following AQEs at 420 nm (unless stated otherwise): 27.8% on mesoporous N-doped g-C 3 N 4 , 49 28.5% on CdS/MoS 2 core-shell nanowires, 57% (with NaCl) on polymeric CN with shorter interlayer distance, 169 8.57% on crystalline CN nanosheets, 147 44.35% on α-Fe 2 O 3 /g-C 3 N 4 Z-scheme photocatalysts, 88 <0.1% for NiO/g-C 3 N 4 , 66 7.05% for g-C 3 N 4 /ZnIn 2 S 4 nanoleaf, 68 19% on SrTiO 3 :La, Rh/C/BiVO 4 :Mo, 91 36.5% on CdS/RGO/g-C 3 N 4 , 92 0.89% on black phosphorus/BiVO 4 , 93 and 2.78% on g-C 3 N 4 co-modified with MnOx and Au− −TiO 2 . 150 The exceptionally high AQY of polymeric CN prepared by co-condensation of urea with oxamide in molten salt (KCl/LiCl) mixture was attributed to reduced interlayer stacking and extended visible-light absorption, which resulted in increased generation of dissociated surface charge carriers that could participate in photocatalytic reactions. 169 The XRD patterns of CN samples prepared under different conditions, high resolution TEM image of the optimized sample, and proposed scheme for interlayer charge transfer are shown in Fig. 6, where decreased interlayer spacing can be clearly observed. It can, thus, be observed that there is a huge variation in values reported for different materials, and despite the fact that AQE/AQY values provide at least some basis for performance comparisons, it would be highly beneficial to the progress of the field to standardize the experimental conditions (reactor geometry, illumination, and additives if any).

D. CO 2 reduction
A number of different photocatalysts for CO 2 reduction have also been reported, 6 and various synthetic methods for their preparation have been recently reviewed. 171 Various photocatalysts PERSPECTIVE scitation.org/journal/apm       are very commonly investigated, with g-C 3 N 4 attracting increasing attention in recent years. These materials can serve as photocatalyst supports or can be used directly as photocatalysts, and they have been extensively reviewed. 173 Literature reports are summarized in Table III. Similar to water splitting, for CO 2 reduction, AQE is defined as the number of reacted electrons divided by the number of incident photons and is the parameter of interest. 6 However, CO 2 reduction would typically generate different reaction products. Thus, efficiency is often described using the total consumed electron number (TCEN). 6 However, this number takes into account the reactor volume, catalyst mass, and reaction time, 6 which makes direct comparisons of different materials difficult. Compared to water splitting, AQE/AQY values have been reported considerably less frequently for CO 2 reduction, and they are typically in the range of several percent. 146,150 TCEN is more commonly reported, 6 but in this case as well, huge variation among different literature reports can be observed. Thus, the entire field of photocatalytic fuel generation would benefit from some standardization of experimental conditions, the development of standard test reactor configurations, and requirements for standardized reporting. This would enable direct comparisons between different materials, which would likely facilitate faster development of improved performance photocatalysts. While at present relative comparisons to either P25 or pure g-C 3 N 4 can provide some information, this is insufficient for designed development of efficient photocatalysts.

E. Chemical synthesis reactions
Visible-light photocatalysis can also be used in chemical synthesis, for example, to achieve milder reaction conditions, such as lower synthesis temperature, or to replace toxic or expensive reagents depending on the type of chemical reaction. 174 The process has been recently reviewed, from the perspective of chemical reactions that can be achieved 174 or combinations of metallocatalysis and photocatalysis 5 or organocatalysis and photocatalysis. 175 The reactions where the use of photocatalysis was reported include C− −C bond formation, 174,176 carbon-heteroatom bond formation, αamine functionalization, decarboxylative coupling, cycloadditions, atom-transfer radical addition, and fluorination. 174 Some impressive improvements, such as the achievement of the Suzuki-Miyaura coupling reaction at room temperature 176 and progress in the synthesis of trifluoromethoxy and difluoromethoxy compounds, which are of interest in drug development, 177 have been reported. The expanded range of transformations and improvement in the reaction conditions have been achieved for a variety of chemical reactions, with the benefits ranging from marginal to transformative depending on the specific reaction. 174 A variety of photocatalyst materials can be used for this purpose, including those common in other applications, such as TiO 2 178 and carbon nitride. 179

IV. ADVANCES IN CHARACTERIZATION TECHNIQUES AND UNDERSTANDING OF THE PHYSICAL MECHANISMS
Understanding the mechanisms responsible for a photocatalytic activity of a material for a given reaction and evaluating practical applicability of that material requires comprehensive experimental characterization. The experimental characterization needs to include its optical properties (which for powder nanomaterial samples requires the use of diffuse reflectance UV-Vis spectroscopy), sample morphology [which typically requires scanning electron microscopy (SEM) or transmission electron microscopy (TEM)], sample composition [using energy dispersive x-ray spectroscopy, xray photoelectron spectroscopy (XPS), x-ray diffraction, and Raman spectroscopy], and surface properties [using XPS, Raman, and Fourier Transform Infrared spectroscopy (FTIR)]. 106 In addition, information about surface area can be obtained from Brunauer-Emmett-Teller (BET) surface area measurements, and relevant particle aggregation size and zeta potential in the medium of interest can be obtained from dynamic light scattering (DLS) measurements. 106 Additional characterization can include determination of energy levels either by cyclic voltammetry or by ultraviolet photoelectron spectroscopy (UPS). In-doped materials for low dopant concentrations, more sensitive composition analysis techniques may be needed, such as atomic absorption spectroscopy (AAS), inductively coupled plasma mass spectroscopy (ICP-MS), or ICP atomic PERSPECTIVE scitation.org/journal/apm emission spectroscopy (ICP-AES). 106 Further information on photogenerated charge carrier dynamics can be obtained from photoluminescence and time resolved-photoluminescence measurements. Additional experimental techniques may be employed to evaluate the nature of defects in a semiconductor, such as electron spin resonance (ESR)/electron paramagnetic resonance (EPR) for defects with unpaired electrons. 19 The evaluation of photocatalytic performance should, in addition to the determination of the experimental rate of the photocatalyzed reaction determined from experiments done on samples where equilibrium in the dark has been achieved, contain the determination of apparent quantum yield or apparent quantum efficiency. 106 It should be noted that the correct determination of these quantities can be complex, but it should be performed to enable reliable comparisons among different research groups. 106 Finally, it is also a good practice to demonstrate data reproducibility by performing the experiment multiple times and including error bars. 106 Stability is also of relevance for practical applications, and it can be evaluated by recovering the photocatalyst and performing repeated cycles on re-used photocatalysts, combined with the characterization of photocatalyst properties after re-use. 106 Furthermore, in the investigations of the reaction mechanisms, the use of isotopes ( 13 CO 2 and D 2 O) 150 is a highly valuable tool for unambiguous identification of reaction pathways. This is particularly important in CO 2 reduction, where a carbon containing residue (for example, from organic solvents and precursors employed in material synthesis) decomposing under irradiation could be an alternative source of detected reaction products. 6,171 In addition, while oxygen is expected to be generated from CO 2 reduction using water as an electron (hydrogen) source, very few studies report oxygen evolution. 171 In addition to these necessary characterization techniques, recent development of characterization techniques has provided more insights into the photocatalysis process. 180 In particular, in situ or operando techniques have facilitated narrowing the pressure gap and enabled a direct observation of the photocatalytic process as well as the surface structural changes in the photocatalysts. Traditional x-ray photoelectron spectroscopy (XPS) is a useful tool to study the elemental surface composition and the oxidation states of the catalysts, but it has to work under ultra-high vacuum environment. Recent advances of the in situ/operando near ambient pressure XPS (NAP-XPS) technique have allowed probing the underlying mechanism of the catalytic process under a more realistic condition. [181][182][183] Separating the reaction chamber and the energy analyzer with a specially designed nozzle and a differential pumping system, the NAP-XPS system could collect photoelectrons at a pressure of up to ∼20 mbar. 184 Currently, there are two designs for the reaction chamber: backfilling and reaction cell [Figs. 7(a) and 7(b)]. In both cases, there is a nozzle which acts as a slit or an aperture to separate the reaction environment and the high vacuum environment of differential pumping states. The nozzle head is placed close to the sample surface so that it is within the range of the inelastic scattering mean free path (λ) in order to minimize the kinetic energy loss of the photoelectrons. In the differential pumping unit, there are typically three pumping stages with electrostatic lenses installed to change the natural trajectory of the photoelectrons in order to maximize their collection [ Fig. 7(c)]. The advantage of the NAP system is that ex situ XPS studies of the catalyst before and after the reaction could not access the authentic chemistry of the catalyst surface during the reaction. In particular, inconsistent information could be generated as the used or unused catalysts are measured at room temperature in UHV for ex situ studies, while the catalyst at the catalysis state could be at a high temperature in the mixture of reactants. Usefulness of this technique in improving the understanding of fundamental mechanisms responsible for the behavior of specific photocatalysts has been demonstrated. For instance, the technique has revealed the surface chemistry changes of TiO 2 under water vapor and oxygen condition with and without UV irradiation. 185 It has been found that the presence of both water and oxygen is sufficient to sustain the catalytic capability of TiO 2 without photons for a period of time. In addition, the in situ irradiated XPS technique can be used to confirm the formation of a direct Z-scheme heterojunction. 87 Another in situ technique is the environmental transmission electron microscopy (E-TEM). Microstructural and chemical changes in the catalysts during an interaction with gas or liquid can be observed by this system up to several mbar with specially designed sample holders. 186 These types of experimental setups are very system-specific with customized design. The sample is normally tilted with respect to the holder tip plane to allow exposure to light and electrons at the same time. The light source for the reaction could be introduced either by lens modules or by fiber optics. 186 The lens-based system has a hollow holder shaft for the optical path (Fig. 8). A laser diode is used to provide illumination, and therefore, the choice of wavelength would be limited by the available laser diodes. As it is connected directly to the microscope vacuum system, the light source cannot be exchanged while the holder is in operation. Once it is evacuated, the heat produced by the laser diode during operation does not dissipate and, hence, the laser efficiency might not be constant over time. The fiber optic design could offer more feasibility for the connection of light sources, but the intensity of illumination will be limited. The two designs in combination have offered complementary options for the ex situ experiment. Since all the photocatalysis reactions happen on the surface, scanning tunneling microscopy (STM) is an ideal tool to provide information of the adsorption sites, active sites, as well as surface reconstruction of the catalysts. With the development of in situ reaction cells, high pressure STM (HP-STM), which allows STM observation under a reacting gas environment up to 100 mbar, has become more mature in the past decade. 187 The core unit of the HP-STM system is the in situ reaction cell (Fig. 9). It consists of the sample stage, heating facility, tip and scanning motors, and gas inlet system, and it is isolated from the main chamber. While the cell can be filled with reaction gas up to 100 mbar during operando reactions, the base pressure in the main chamber is maintained under UHV condition by using the differential pumping system. Therefore, the changes in morphologies, defects, or vacancies on the surface under reaction conditions can be recorded directly. For instance, metal and metal oxide catalysts were found to have different restructuring processes under different gas content exposure by HP-STM, 187 where this information could not be determined with the ex situ experiment. For example, nanoclusters were observed during the exposure of Pt to 1 Torr of O 2 at room temperature and were identified as Pt surface oxide. 187 The clusters on the surface disappeared after evacuating O 2 to UHV. Another example is the reconstruction of the ZnO non-polar surface under different gaseous environments. It was previously found that adsorption of water molecules on the surface can lead to certain reconstruction of the surface. With HP-STP, it was determined that hydrogen is also capable of producing similar reconstruction. 187 To introduce illumination in HP-STM, fiber  optics is a better choice as the lens module may affect the cooling efficiency of the STM.
The atomic force microscope based infrared spectroscopy (AFM-IR) is another recently developed tool to study the spatial chemical change in the surface in nanometer scale, which could be useful for in situ reaction investigation. 188 Instead of using the optical method to determine the absorption of infrared light, it makes use of the AFM tip to detect the thermal expansion during the infrared illumination (Fig. 10). The technique has realized up to a 5-10 nm spatial resolution of chemical mapping. It operates in ambient environment, and therefore, the chemical change in the solid surface during a photocatalytic reaction under atmospheric pressure can be determined on site. For instance, a nanoscale degradation study on oil paints that contain titania as inorganic pigments was performed with the AFM-IR instruments. 189 In addition, in vivo discrimination of gram-positive and gram-negative bacteria was also feasible with the technique, which could possibly extend to the studies of photocatalytic antibacterial effects. 190

V. CONCLUSIONS AND FUTURE OUTLOOK
Among photocatalyst materials, the most promising approaches involve the use of complex multicomponent systems, such as heterojunction (including Z-scheme) photocatalysts, combined with co-catalysts and/or plasmonic particles. There is significant interest in the development of non-noble-metal plasmonic particles and co-catalysts to reduce the cost of the photocatalyst. However, what should also be considered is that even with the use of noble metals, efficiency targets have not been reached. In fact, in fuel generation applications, improvements of several orders of magnitude are needed when it comes to efficiency. Therefore, focusing on achieving as high efficiency as possible, combined with comprehensive material characterization and standardization of experimental conditions PERSPECTIVE scitation.org/journal/apm and test reactor design, and reporting AQE/AQY values instead of simply reporting reaction rates is expected to contribute to the advance of the field in a more significant manner. When it comes to practical applications, commercialization potential is often discussed taking into account the estimated cost of generated products or estimated photocatalyst costs, 171 but the challenges of scaling up the production are rarely considered. As already discussed, photocatalysis is impractical for large scale water treatment applications since it is not cost-effective or energy effective when scaled up to industrial level. 15,126 The median EEO value of the reported literature estimates for a photocatalytic process is 335 kW h/m 3 /order, which is significantly higher compared to ozonization (0.15 kW h/m 3 /order), O 3 /H 2 O 2 (0.2 kW h/m 3 /order), and UV/ozone and UV/hydrogen peroxide (0.7 kW h/m 3 /order and 0.75 kW h/m 3 /order, 0.15 kW h/m 3 /order, respectively). 127 Furthermore, the median EEO values for the UV/H 2 O 2 process decrease with increased capacity from lab-scale over pilot-scale to industrial scale. 127 In contrast, the energy consumption of a photocatalytic system increases on scaling up due to pumping and photocatalyst recovery requirements. 126 Consequently, in the only commercial photocatalytic water treatment system, Purifics Photo-cat intended for small-scale applications, the efficiency of the unit improved by replacing the titanium dioxide photocatalyst with hydrogen peroxide. 126 This would become even more pronounced when moving from pilot-scale to industrial scale water treatment due to the need to operate at higher pumping rates. Thus, a possible solution would be to focus on various niche applications, 15 such as small-scale pointof-use systems, or alternatively permanent pollutant degradation regardless of the cost (i.e., photocatalysis can degrade pollutants that are difficult to degrade by other methods), if this ever becomes a social necessity due to increasing pollution levels. Another possible way forward is to focus on optimizing the reactor designs, including the illumination sources such as UV-LEDs. 191,192 UV-LEDs would be particularly suitable for small-scale systems.
In terms of photocatalysis for fuel generation, no reported photocatalyst has come close to the required efficiency level for commercialization, and this is even without considering the technological challenges of scaling up to develop a system to be able to produce fuels at sufficient rates and quantities to be practically usable. It is estimated that the required solar-to-hydrogen efficiency for practical applications is 10% for overall water splitting. 9,166 For reference, current particulate photocatalysts can only achieve ≈1% under laboratory conditions, 166 which means improvement in the efficiency of more than one order of magnitude to obtain 10% efficiency in an industrial scale process. It should be noted that 10% efficiency under solar illumination could not be achieved with photocatalysts with absorption edges at or below 500 nm even for AQY of 100%. 9 In contrast, if the absorption edge could be extended to 600 nm, needed AQY would be 60%. 9 While the majority of review papers tend to discuss issues related to efficiency improvement challenges, it would be worthwhile to consider the challenges of scaling up. For example, a fuel cell of 1 kW consumes approximately 10 l of hydrogen per minute, which means either huge scale up of bench top reactors is required or some efficient method of storage is needed for photocatalytic hydrogen production to be practically relevant. The system lifetime also needs to be evaluated (and it should be noted that more narrow bandgap photocatalysts that are desirable from increasing STH point of view generally have inferior photostability), as well as safety issues, namely, the separation of oxygen and hydrogen when they are produced at large quantities. Estimates of economic realities for methanol generation from CO 2 reduction are similarly pessimistic. It was claimed that solar-to-fuel efficiency as high as 15% needs to be achieved to be economically competitive for methanol generation. 193 This involves huge challenges of 40% one-pass conversion of CO 2 and the achievement of methanol yield rate of ≈10 000 μmol/gcath. 193 It should be noted that these estimates are based on a number of assumptions made considering the cost estimates of a "solar refinery." 193 Furthermore, one of the reasons for the requirement of high solar-to-fuel efficiency is to reduce the sensitivity to solar irradiance and operational time. 193 Nevertheless, considering typical figures reported so far, 6 the achieved catalytic activity and fuel yields still need to be improved by several orders of magnitude in order to generate meaningful quantities of fuel.
Consequently, there are significant challenges in applications of photocatalysis at industrial scale. On the other hand, niche applications and small-scale systems may be viable and perhaps it would be more meaningful to focus on those. Increased collaboration between physicists, chemists, materials scientists, and engineers is needed to address the challenges of scaling up, and industry input is needed to reduce the gap between academic interest and practical application needs. Considering the efficiency challenges for visible-light photocatalysts to become relevant for practical applications, working on increasing the efficiency is considerably more significant than the development of alternative, lower cost materials. Furthermore, the robustness of the material should not be disregarded since it is highly relevant for the operation of slurry reactors and successful photocatalyst recovery, and some fragile multicomponent materials would simply not survive high pressures, high flow rates, and ultrafiltration recovery in a realistic industrial system.

ACKNOWLEDGMENTS
Financial support from the Seed Funding Grant for Applied and Translational Research, the University of Hong Kong, is acknowledged.

DATA AVAILABILITY
Data sharing is not applicable to this article as no new data were created or analyzed in this study.