Ambient Air Puriﬁcation by Nanotechnologies: From Theory to Application

: Air pollution has been a recurring problem in northern Chinese cities, and high concentrations of PM 2.5 in winter have been a particular cause for concern. Secondary aerosols converted from precursor gases (i.e., nitrogen oxides and volatile organic compounds) evidently account for a large fraction of the PM 2.5 . Conventional control methods, such as dust removal, desulfurization, and denitriﬁcation, help reduce emissions from stationary combustion sources, but these measures have not led to decreases in haze events. Recent advances in nanomaterials and nanotechnology provide new opportunities for removing ﬁne particles and gaseous pollutants from ambient air and reducing the impacts on human health. This review begins with overviews of air pollution and traditional abatement technologies, and then advances in ambient air puriﬁcation by nanotechnologies, including ﬁltration, adsorption, photocatalysis, and ambient-temperature catalysis are presented—from fundamental principles to applications. Current state-of-the-art developments in the use of nanomaterials for particle removal, gas adsorption, and catalysis are summarized, and practical applications of catalysis-based techniques for air puriﬁcation by nanomaterials in indoor, semi-enclosed, and open spaces are highlighted. Finally, we propose future directions for the development of novel disinfectant nanomaterials and the construction of advanced air puriﬁcation devices. pollutants, including reaction pathways and key intermediates; and (4) a summary of developments in nanomaterial immobilization and purifying modules as applied in ﬁeld-tested techniques. Finally, we show how control systems based on APN technology can offer a path forward for protecting the general public from severe air pollution.


Introduction
China is facing daunting challenges in local and regional air quality due to its continued economic growth and industrialization. During the winters of 2013 and 2014, many of the cities in northern and eastern China (~15%) suffered from severe haze caused by particulate matter (PM) <2.5 µm in aerodynamic diameter (PM 2.5 ) [1]. High levels of PM 2.5 in the range of 500-800 µg/m 3 were recorded in one pollution episode in Beijing, and that resulted in strong impairments of visibility (<100 m) [2]. Other impacts of PM 2.5 pollution include adverse health effects, especially increased incidences of cardiorespiratory disease, as revealed in numerous toxicological studies [2]. In China, 1.1 million deaths in 2015 were ascribed to long-term exposure to high levels of PM 2.5 [3]. Thus, the following questions concerning the problem should be addressed: (1) what is the composition of PM 2.5 ? (2) what are the causes of severe haze formation? and (3) how can haze formation be controlled? An et al. published a comprehensive review of the current knowledge on the formation mechanisms of severe haze in northern China [4], and in that review, the synergetic effects of anthropogenic emissions and atmospheric processes over large scales were emphasized [4]. PM 2.5 is composed of fine particles produced from various physical processes (primary emission) and gas-to-particle conversion (secondary pollution) [4]. In 2014, Huang et al. interception, diffusion, electrostatic precipitation, etc. The removal efficiencies for these approaches vary as complex functions of particle size [19]. To prevent fine particles from being emitted into the atmosphere, conventional technologies are continuously being improved, and hybrid systems with combinations of two or more methods are being developed [2].
The methods for NO x abatement can be generally divided into in-combustion and post-combustion procedures [20]. For in-combustion methods, parameters such as reactant residence times, temperature peaks, and O 2 concentrations are adjusted to reduce NO x formation. The use of LNBs is an approach that belongs to the in-combustion methods for stationary and mobile sources. Before 2010, only medium-(≥100 MW and <300 MW) and large (≥300 MW)-size coal-fired power plants in China adopted the LNB technology [14]. During the 2015-2016 time period, the supercritical circulating fluidized bed (CFB) combustion has become the most promising technology and series of CFB boilers have been established in 350 MW plants in China. The advantages of fluidized bed combustor, such as compact furnace design and relatively uniform temperature, guarantee the effective burning of wide ranges of fuels and significant reduction of NO x and SO x emissions [21,22]. The recent works have been focused on evaluation of NO x emission from the chemical looping and calcium looping combustion processes, as they are emerging promising post-combustion technology for CO 2 capture for coal-fired burners. The fuel types and properties, flue gas oxygen feed and NO feed to the fluidizing gas are all great contributors to influence NO x emission [23,24].
Since 2010, selective catalytic reduction and selective noncatalytic reduction (SCR and SNCR, respectively) technologies have been gradually introduced-these methods belong to the post-combustion procedures [25]. Table 1 summarizes the primary post-deNO x methods, their principles, and performance in various applied scenarios. Although the SCR and three-way-catalysis (TWC) methods have become widely commercialized, the following problems need to be addressed: (1) high cost of the SCR technology for small-scale power plants and industrial boilers; (2) ammonia leaks-these have become a problem because NH 3 is a target of regulations; and (3) TWCs are not effective for diesel or lean-burn gasoline engines [11,12]. Under the current circumstances, large amounts of NO x are still being discharged into the environment and continue to cause deteriorations in air quality.
The abatement of anthropogenic VOCs is especially challenging because their composition, physical properties, reactivity, and emission sources are diverse. Typical end-of-pipe VOC abatement approaches are also summarized in Table 1. Generally, VOCs can be recovered by absorption, adsorption, condensation, membrane separation, and bio-filtration, or they can be destroyed by combustion, catalytic or biological oxidation processes [26]. To address the drawbacks with adsorption and combustion, researchers have developed catalytic VOC oxidation technologies in which VOCs are oxidized using a suitable catalyst at relatively low temperatures (250 to 500 • C). The catalysts are designed for dealing with effluent streams containing low VOC concentrations. The key challenge lies in choosing the best and most robust catalyst for a variety of pollutants because of the complexity of the VOCs mixtures [26]. Biological treatments offer a cost-effective and environmental-friendly approach to decompose VOCs and the odors emitted from agriculture, farming and sewage wastewater plants [27]. Hence, no single technology can be universally used to treat gaseous pollutant mixtures that differ in composition and concentrations. Combinations of technologies are especially favorable for situations requiring extreme purification.
The current air pollution mitigation methods have not been sufficient to prevent the frequent occurrence of haze episodes in China. Emission inventory studies in China have shown that the lowest uncertainties are mainly associated with the emissions from large point sources, such as power plants, cement plants, on-road vehicles, shipping, and residential combustion. On one hand, the lack of definite emission profiles for other industries or processes increases the difficulty of source pollution control [14]. On the other hand, meeting the stringent air quality regulations imposed by the WHO and local governments (e.g., Hong Kong) will require further technological innovations. According to WHO guidelines, the annual average PM 2.5 value should be <10 µg/m 3 , and the NO x level established by EU directive for the protection of human health is 105 ppbv. Therefore, an abundance of research opportunities exists for improving current technologies and for developing novel technologies to protect the environment and human health.

Ambient Air Purification by Nanotechnologies
Recent developments in nanotechnology have led to exceptional opportunities in the use of nanomaterials for environmental remediation. Compared with bulk materials, nanomaterials have larger surface areas and show remarkably different electronic, magnetic, and catalytic properties for PM < 100 nm, which make them superior for filtration, adsorption and photocatalysis [20,[31][32][33]. We discuss below the basic principles for ambient air purification with nanomaterials (APN), especially in relation to air pollutants at low concentrations (sub-ppb to ppm range), low emission intensities, and from dispersed sources. These applications take advantage of the special properties of the nanomaterials, including quantum confinement and surface/interface effects. Nanomaterials can be used for particle filtration, gas adsorption, bacteria disinfection and catalytic processes; and they have been components of various types of purification devices used to reduce particulate matter (PM), gaseous precursors (i.e., NO x , VOCs, etc.), and toxic pollutants (e.g., HCHO), as shown in Figure 1. Advantages of nanomaterials include moderate working conditions, no burdensome instrumental set-up, and low operating costs. Indeed, the APN technology often can be an ideal solution for improving the air quality in residential and urban areas, and it is especially well suited for applications in confined environments. As a result of these advantages, a substantial budget was allocated by Chinese government for research on nanomaterials and APN technology for pollution abatement, and APN will undoubtedly receive increased attention throughout the world in the future. Here, our objective is to present a comprehensive overview of APN technology from fundamental theory to real world air purifying applications. The specific topics to be discussed include (1) the development of novel nanofibrous filters for fine particle collection in low pressure-drop systems; (2) improvements in aerosol precursor gas adsorption and degradation kinetics from the perspective of nanomaterial design and parameter optimization; (3) a review of the fundamental mechanisms of photocatalytic and ambient temperature catalytic degradation of gas pollutants, including reaction pathways and key intermediates; and (4) a summary of developments in nanomaterial immobilization and purifying modules as applied in field-tested techniques. Finally, we show how control systems based on APN technology can offer a path forward for protecting the general public from severe air pollution.

Filtration
"The past experiences in the cities in the USA and recent data obtained by the Institute of Earth Environment in Xi'an, China indicated that it might take a decade or longer for particle filtration to be reduced to an acceptable level" [2]. Although regulations on PM 2.5 emission from industries and vehicles recently have been strengthened in China, adverse health effects remain a significant concern. Membranes/filters, including those made from electrospun nanofibers, have gained wide acceptance for applications in PM 2.5 collection owing to their superior filtration efficiency and low pressure drop.
Conventional filtration is effective for capturing large pollutant particles, but in practice, that approach has some significant pitfalls including frequent fouling, high energy use, and high operating costs. Unlike conventional multilayer air filters, electrospun nanofiber mats are characterized by high porosity and micrometer-sized void spaces, resulting in low resistance to airflow and low pressure drops [32,34]. High-power pumps are not needed for these materials.
Filtration performance is not only affected by pressure drop but also by filtration efficiency. Increasing the polarity and surface electrostatic charges of the polymers used to make nanofibers are common strategies for improving their filtration efficiency [32,35]. Several special functionalities can also be added to nanofibers by incorporating active components into the polymers used to make them. For example, as iron oxide accounts for a large fraction in the airborne dust, magnetic metal oxide nanoparticles (Fe 3 O 4 ) can be added to electrospun nanofibers to attract dust through magnetic attraction (Figure 2a  showing that freestanding nanofiber film can be easily transferred onto a facemask. Scale bar is 3 cm. (e,f) SEM images of Nylon-6 nanofibers before and after filtration. Scale bars in (e,f) are 2 µm. (a,b) Reprinted with permission from the reference [36]. (c-f) Reprinted with permission from the reference [34].
The simultaneous capture of aerosols and formaldehyde can be achieved with high efficiency (95%) at a low pressure drop (112 Pa) with composite polyacrylonitrile (PAN)/βcyclodextrin nanofiber membranes [37]. Dan Lv et al. [38] developed a multipurpose air filter made up of ZnO nanoparticles, polyvinyl alcohol, and konjac glucomannanbased nanofiber membranes. This type of filter exhibited superior filtration performance (>99.99%) for ultrafine particles compared with commercial HEPA filters; moreover, the flow resistance of the filters was low, and they displayed excellent photocatalytic and antibacterial activity functions.
In another study, a triboelectric nanogenerator (TENG) was incorporated into a selfpowering air filter made of electrospun polymer-coated stainless-steel wires. This type of device, when combined with a photocatalytic active component, showed remarkable performance in terms of PM 2.5 capture and formaldehyde decomposition [33]. For reallife applications, face masks and window materials made of nanofiber-loaded nonwoven fabrics have been developed to protect public health and to prevent outdoor PM 2.5 from entering indoor spaces (Figure 2c,d) [32,34]. There is good evidence that polymer-derived nonwoven nanofiber mats are suitable for environmental use due to their lightweight and tolerance of high-temperatures. In particular, ceramic nanofibers, specifically γ-Al 2 O 3 nanofiber membranes, with excellent thermal and chemical stabilities, are suitable for treating vehicle exhaust and for industrial filtration applications [34].
To develop efficient and stable air filtration devices, the mechanisms by which electrospun nanofibers capture PM need to be understood. In situ optical microscopy is a common way to study these processes [32], and studies of that type have shown that small PM quickly aggregates to form large spherical particles once they are captured by nanofibers. This is especially true for the points where the nanofibers join because large PM aggregates tend to accumulate there. Figure 2e,f clearly show that smoke PM formed a coating layer strongly wrapped around each nanofiber. These observations provide insights into how the effects of particle size and shape and surface properties of the polymer nanofibers affect filtration performance.

Adsorption
Adsorption is a non-destructive process that can be used to remove gaseous pollutants from air under ambient conditions, and there are two different processes, physisorption (non-covalent) and chemisorption (covalent), that may be involved. Sorption capacity largely depends on the following: (1) the physical structure and chemical composition of the sorption materials, including their specific surface area; (2) pore size/distribution and surface functional groups; (3) molecular structure, polarity, acidity and wettability of the target pollutants; and (4) adsorption conditions, principally temperature and humidity [18,39]. For NO x removal, Gomez-Garcıa et al. [12] showed that metal oxides, spinels, perovskites, zeolites, carbonaceous material, and heteropolyacids are primary sorbing materials with high capacities for uptake. Transitional metal (Co, Cu, Ni, Fe and Mn) oxides were found to have the best NO uptake because covalent bonds formed between the metal ions and NO [40][41][42]. In contrast, rare earth and alkaline earth metals, such as Ce, La, Ba and Na, showed comparatively low NO adsorption [12]. Zeolite exchange with transitional metal ions, and carbonaceous materials modified with iron or copper oxides also exhibited excellent NO uptake performance.
For VOC removal, carbonaceous materials of various forms are the most commonly used adsorbents [18]. Activated carbon (AC) is the most popular of these, and AC is widely used for gas sorption, with applications for alkanes, alcohols, ethers, aldehydes, and aromatics. However, due to its microporous (pore size ≤2 nm) and disordered pore structure, the movements of large VOC molecules are inhibited, and that results in slow adsorption kinetics [43][44][45]. Compared with AC particles, AC fibers (ACFs) with essentially a 1D structure are a novel material that exhibits fast adsorption kinetics due to the presence of short and straight micropores [28,46]. The ACFs also show unique adsorption behavior for nonpolar or weakly polar VOCs, such as toluene and formaldehyde, due to their nonpolar nature and lack of surface O groups. Furthermore, thermal treatments, acid treatments, or metal oxide modifications can be used to add oxygen functional groups to the ACFs which enhances the removal of polar VOCs [47]. ACFs are a promising candidate for commercial gas adsorption applications because they can be made into fabrics or other desired self-supporting forms.
With the advancement of nanotechnology, 1D carbon nanotubes and 2D graphene (oxidized and reduced states) nanosheets have become used for electronics, catalysis, photonics, and environmental remediation [48][49][50]. The extremely large surface areas (up to 3502.2 m 2 /g) of these nanomaterials has led to increased demands for their use in high capacity VOC adsorption applications compared with traditional AC [50,51]. Another advantage of nanomaterials made of graphene oxide is that the abundance of the surface oxygen functional groups facilitates the adsorption of the hydrophilic VOC molecules, but a drawback is that the aggregation of graphene oxide has severely restricted their commercial utilization as adsorbents.
Nonetheless, graphene oxide can be processed into various self-supporting forms or used as a supporting material for dispersed catalysts designed for simultaneous VOC adsorption and degradation [52,53]. Other carbonaceous materials, such as biochar and ordered mesoporous carbon, also have been investigated for use as VOC adsorbents [18]. Previous studies and reviews have shown that significant progress has been made in the control of VOCs with these materials [39,45,49,51,52], but the development of low-cost, high-efficiency (under a range of humidity conditions) and easily recyclable adsorbents remains a challenge for their applications in pollution control.

Photocatalysis
Nanomaterial-driven photocatalysis provides a green solution for removing air pollutants in the atmosphere. Early in 1986, Takeuchi and Ibusuki found that metal oxides in atmospheric PM played critical roles in the photochemical catalytic conversions of hydrocarbons and NO x . These are reactions that strongly enhance the formation of CO 2 and HNO 3 and decrease O 3 [54,55], and several photoactive metal oxides, including TiO 2 , ZnO and Fe 2 O 3 can be relatively abundant in the atmosphere, and heterogeneous reactions catalyzed by these photoactive materials have been considered for removing pollutants from the air.
Although photocatalysis has been studied for more than 40 years, water splitting for H 2 and wastewater treatment is still far from commercially viable due to the low energy conversion efficiency and complicated procedures for recycling photocatalysts. By contrast, significant numbers of patents issued for photocatalytic air purification reflect a growing interest in the adaptation of this approach for air quality applications [56]. Japan and several countries in the EU have pioneered basic research in this area and conducted field studies on the photocatalytic removal of air pollutants, whereas China has only recently started to focus on this field. In this section, photocatalysis principles, strategies for optimizing nanomaterials, effects of environmental conditions, reaction mechanisms, and nanomaterial immobilization are discussed in detail.

Principles of Photocatalysis and Major Nanomaterials
Photocatalysis is an advanced oxidation process (AOP) that involves irradiated semiconductors and causes not only the rate, but also the direction of a chemical reaction to be altered. In Fujishima-Honda energy band theory, [57] one electron (e − ) is excited from the valence band maximum (VBM) to the conduction band minimum (CBM) by absorbing one photon with sufficient energy, thereby simultaneously leaving a positive hole (h + ) in the valence band (Equation (1)). Photogenerated e − -h + pairs achieve spatial separation from surface reaction sites by diffusion or migration.
Once they have arrived at the photocatalyst surface, excited charge carriers can react with the pre-adsorbed H 2 O and O 2 to produce reactive oxygen species (ROS), such as hydroxyl (Equation (2)) and superoxide radicals (Equation (3)). These radicals possess extremely strong oxidation potentials, barely showing any selectivity towards pollutants, thereby enabling the complete conversion of the gaseous pollutants into harmless products. It is important to note that the ability to produce ROS and their oxidation and reduction capacities are mainly determined by the respective VB and CB positions of a specific photocatalyst. Figure 3 presents the band positions of some semiconductors and their relative position with the redox level of •OH and •O 2 − formation. The semiconductor photocatalysis principles related to environmental pollutant purification are also illustrated in Figure 3. As shown, the overall photocatalytic efficiency of the functional nanomaterials is determined by light absorption efficiency (η abs. ), charge separation efficiency (η sep. ) and most importantly, the surface reaction efficiency (η rea. ), as follows (Equation (4)): Various semiconductor nanomaterials have been employed for use in photocatalytic reactions; these include TiO 2 , ZnO, WO 3 , Fe 2 O 3 , SrTiO 3 , BiVO 4 , Bi 2 WO 6 , BiOX (X = Cl, Br, I), g-C 3 N 4 , etc., which can have different phases, morphologies, and modifications [58]. To improve the efficiency of each elementary step, various strategies have been used to modify the physicochemical properties of nanomaterials. For example, element doping (metal or nonmetal) to induce impurity states in wide bandgap semiconductors has been used to extend the light absorption range, while plasmonic noble metal (Ag and Au) deposition is a popular way to enhance light absorption and inhibit photoinduced charge recombination [59,60].
The construction of binary or ternary composites to form heterostructured interfaces has been widely used to promote charge separation and enhance light absorption. Facet modulation and oxygen vacancy engineering have become widely used for interfacial charge transfer enhancement [61,62]. Porous metal-organic-frameworks were also adopted to enhance gas adsorption to facilitate the photoreaction [63]. In Table 2, we highlight some recent emerging photocatalysts that have been used for the decomposition of gaseous pollutants and summarize their synthesis strategies, reaction conditions and performance.

Light Absorption
Light-harvesting ability of a photocatalyst governs the wavelength and number of photons it can utilize to trigger chemical reactions. Traditional metal oxides and oxyhalides, such as TiO 2 , ZnO, SrTiO 3 , and BiOCl are wide bandgap (Eg > 3.0 eV) semiconductors that only respond to ultraviolet (UV) light with wavelengths <400 nm. However, UV light only accounts for a small fraction, 3-5%, of the entire solar spectrum [75][76][77]. To expand the light absorption range, researchers have developed techniques for element doping (metal or non-metal) into bulk semiconductor lattice structures to induce the formation of the interband hybrid energy levels that enable charge excitation by photons with lower energies [78][79][80]. For example, Wang et al. [64] prepared B-doped nanotubular g-C 3 N 4 which had an enhanced NO removal rate compared with its undoped counterpart (Figure 4a). On the contrary, Pichat et al. [81] observed that Cr-doped TiO 2 expanded light absorption region into visible light and considerably decreased photocatalyst oxidation activity, and that was attributed to Cr 3+ impurity states facilitating electron-hole recombination. Conflicting statements about the effectiveness of element doping abound in the literature, and those discrepancies often have been due to inaccurate controls on the doping process or other problems with the investigations.
Noble metal nanoparticle (i.e., Ag and Au) deposition with unique surface plasmon resonance (SPR) properties has become a popular method for broadening the visible light utilization of wide bandgap semiconductors [82,83]. Unfortunately, the high price and scarcity of the noble metals restrict real world applications in environmental management; indeed, more cost-effective and stable photosensitizers need to be developed. Carbon nanodots (CDs) are biocompatible and inexpensive photosensitizing materials notable for their up-conversion photoluminescent properties; that is, they can convert long wavelength light to shorter wavelengths [66,84,85]. For instance, Liu et al. observed that CDs harvested near-infrared light (≥600 nm) and then fluoresced at shorter wavelengths to excite TiO 2 nanosheets for surface reactions [86].
The intrinsic photosensitizing mechanisms for CDs are still not completely understood because the electronic structures of CDs are complex, and the CDs can act as either electron donors or acceptors [87]. Recently, non-noble metal Bi nanoparticles have been reported to induce surface plasmon resonance [88]. Gao et al. also fabricated Bi nanoparticleanchored ZnWO 4 microspheres, which showed robust photocatalytic ability for ppb-level NO removal under visible light irradiation ( Figure 4b) [67]. Further improvements in the utilization of visible solar energy will be needed for practical applications of CDs to environmental cleaning-up. Ideally, the catalysts ultimately chosen also should be green, inexpensive, and durable.

Charge Separation
Upon illumination, excited e − /h + pairs in semiconductors can be generated on time scales of femtoseconds (fs). Then, the time needed to initiate surface reactions between surface charges and adsorbed substrates takes several nanoseconds to milliseconds. However, photoexcited e − /h + recombination occurs within picoseconds and that is more rapid than the charge migration and transfer kinetics [89]. Therefore, charge carrier recombination is highly undesirable because the solar light energy is wasted, resulting in poor efficiency. Hence, remarkable efforts have been devoted to the engineering and synthesis of unique nano-structures that can facilitate e − /h + pair separation.
For bulk semiconductors in a space-charge depleted state, the internal electric fields drive electrons and holes apart. As a representative example, layered g-C 3 N 4 shows improved photocatalytic activity due to a stacked interlayer electric field which facilitates photogenerated carrier separation [90]. Li et al. [91] increased the internal electric field magnitude of Bi 3 O 4 Cl nanosheets by regulating the percentage of (001) facet exposure. In forming electric fields, the construction of p-n or n-n heterojunctions is a popular strategy for enhancing interfacial charge separation [92][93][94][95]. As shown in Figure 4c, the p-n heterojunction interface is constructed between ZnFe 2 O 4 (p-type) and Bi 2 O 2 CO 3 (n-type). To achieve Fermi level equilibrium, the band positions of ZnFe 2 O 4 tend to move upward while those of Bi 2 O 2 CO 3 tend to move downward, so that the electric field is formed at the heterojunction interface, which promotes photogenerated e − /h + separation under illumination [69].
Some heterojunctions with intimate contact can be craftily constructed in a one-step procedure. For example, Wang et al. reported the synthesis of Bi 2 O 2 CO 3 /g-C 3 N 4 and g-C 3 N 4 /LaCO 3 OH layered heterojunction nanomaterials through a one-step hydrothermal method using g-C 3 N 4 itself as precursor to provide CO 3 2− [93,94]. The intimate contact between heterostructure components ensures excellent charge separation and strong photocatalytic activities for the degradation of gaseous NO under visible light.

Surface Adsorption Properties
Given the typically low pollutant concentrations (sub-ppb to ppm level) in the atmosphere and the requirement for water vapor and O 2 adsorption for the production of ROS, it is instructive to consider surface adsorption properties at the molecular level. As more pollutant molecules become adsorbed onto a photocatalyst surface, the reaction probability between ROS and adsorbed pollutants increases, and this is favorable for photoinduced charge separation. Several approaches have been shown to effectively improve substrate adsorption; the first of these is to choose a nanomaterial surface with specific sorption sites for targeted molecules or to create those sites by surface modification. For example, LaFeO 3 perovskite oxide exhibits specific chemisorption towards NO due to the empty states in the Fe 3d orbital [96].
A second approach is to tune the exposed facets of the nanomaterials that come into contact with pollutant molecules. This is effective because the atomic arrangements and coordination can vary among different facets, thereby leading to variations in adsorption behavior [61,69]. Third, the oxygen vacancies (OVs) of the metal oxide nanomaterials possess abundant localized electrons-this endows the material surface with enhanced ability to activate adsorbed H 2 O and O 2 , thereby promoting ROS generation and degradation efficiency [97,98]. An example of this is the work of Zhang et al. [86] who introduced OVs in a SrFe x Ti 1−x O 3−δ photocatalyst and showed that the modification led to enhanced O 2 adsorption and activation for ROS generation (Figure 4d). In addition to the direct modification of surface adsorption properties, combining porous and inert materials featuring large specific surface areas is another feasible approach for promoting gas adsorption. Materials such as silica [99], alumina [100], zeolites [101], grapheme [102], and AC [103,104] all offer practical support matrices. Of all supporting materials, AC is the most commonly used due to its low price, extremely porous structure, and large surface area. An increased adsorption of air pollutants onto the AC phase followed closely by mass transfer to the catalyst sites provides for efficient photodegradation and adsorbent regeneration [105].

Effects of Environmental Conditions
The overall reaction performance of photocatalysts can be influenced by environmental factors except for the physicochemical properties of the materials themselves. Here, light conditions, initial contaminant concentrations, relative humidity (RH) and coexisting pollutants are discussed in relation to their effects on reaction performance.
(1) Light intensity The effects of external light intensity on photocatalysis are closely correlated to charge carrier generation, which plays a key role in photocatalytic efficiency and quantum yield. Behnajady [106] found a positive correlation between light intensity and final pollutant removal rate in a study with a well-tested model supported with experimental data. In contrast, a study by Stafford et al. [107] suggested that surface reactions occurred even at low light intensities, because the slow rate of the photoinduced hole generation caused intermediate species to reach adsorption equilibria on the catalyst surface, thereby ensuring the presence of active sites to be occupied by pollutant species. Similarly, Ohko et al. [108] found that the quantum yield increased gradually with decreasing light intensity in the photocatalytic decomposition of gaseous 2-propanol. The quantum yield in that study reached a maximum value even when the light intensity was extremely low.
More generally, the adsorbed concentration and light intensity in photocatalytic reactions are governed by the first-order kinetics (Equation (5)) as follows [109]: Here, r is the reaction rate, k is the first-order rate constant, Γ is the concentration per unit real surface area and I is the light intensity. Under light-sufficient conditions, the α value is between 0 and 1 [110].
(2) Pollutant concentration Due to the kinetic constraints, photocatalysis alone is usually not effective for treating high pollutant concentrations, and even in the ppb and ppm ranges, pollutant removal rates can be quite variable. At low concentrations (<1 ppm), adsorption rates are slower than photo-oxidation rates, and in those cases, mass transfer is the rate determining step. When the pollutant concentration increases to a high level (ppm range), photooxidation becomes the more important rate determinant. Moulis and Krysa [111] observed that removal efficiency was high at low initial concentration of contaminants for all VOCs studied (including n-hexane, n-butyl acetate and toluene) because the surface adsorption capacity of the photocatalysts was not limited by the number of active sites [112].
Along these lines, Mo et al. [113] found that the photodegradation efficiency for toluene decreased by~30% when its initial concentration increased from 1 to 4 ppm. Thus, after the maximum capacity of the photocatalyst has been reached, any further increments in the pollutant concentrations do not improve the performance. Nonetheless, high contaminant concentrations can result in enhanced reaction kinetics when the ratedetermining step is controlled by mass transfer. Photocatalysis is especially well-suited for the abatement of indoor and ambient air pollutants that are in the ppb to sub-ppm concentration range.
(3) Relative humidity Compared to other parameters, the RH effect (or water vapor level) is significant and complex because the water vapor concentrations in the air are in the thousand ppm range, which is almost always much higher than that of gaseous pollutants, and as a result, competitive adsorption exists. Moreover, the presence of water vapor is extremely important because hydroxyl radicals, one of the dominant ROS species, can be directly generated by H 2 O interactions with photoinduced h + . On the other hand, in the absence of water vapor, the photocatalytic mineralization of the pollutants evidently can be inhibited. The photocatalysis of toluene oxidation has been evaluated for dry and wet conditions in a study by Debono et al. [114], and it was shown that toluene conversion to CO 2 was favored by wet conditions.
The optimal RH value can vary among pollutant types, that is, hydrophilic or hydrophobic VOC molecules may be affected differently. In gas-phase chlorobenzene removal, for example, RH is positively correlated with the adsorption and reaction processes, and the photocatalytic oxidation rate can remain stable even at high RH (over 80%) [115]. On the other hand, it has been shown that a relatively high water vapor level inhibited acetone photocatalytic oxidation [116], and more generally, the effect of RH can vary for different photocatalyst surfaces. For 1-butene photooxidation, Cao et al. [117] found that SnO 2 was extremely sensitive to RH effects and showed a significant negative relationship between RH and the photocatalytic oxidation rate. Meanwhile, TiO 2 films maintained stable performance under identical test conditions. The RH level also affects the generation of intermediates as illustrated in a study by Sleiman et al. [118] who showed that RH is a key factor in the photocatalytic oxidation of toluene. At low RH or in the absence of water vapor, a high level of mineralization (~90%) can be achieved, but several toxic by-products are produced and accumulate simultaneously.

(4) Coexisting pollutants
In practice, air purification inevitably involves multiple pollutants with different polarities, molecular sizes, hydrophilicities, and chemical reactivities; and one substance can either promote or inhibit the removal rate of others in such complex systems [119]. For example, in the photocatalytic degradation of formaldehyde, the presence of NO promotes formaldehyde conversion, whereas SO 2 inhibits it because SO 4 2− ions can bind to the formaldehyde adsorption sites on the catalyst surface [120]. Indeed, competitive adsorption kinetics and interactions among individual components are thorny issues that are encountered when dealing with coexisting pollutants.
For instance, Galli et al. [121] investigated the photocatalytic oxidation of 17 types of molecules in a mixed system and found that the gas molecules competed for adsorption sites on the catalyst surface according to their polarity and initial concentration. Similarly, Debono et al. [122] studied the photocatalytic oxidation of a mixture of toluene, decane, and trichloroethylene, and they found that several reactive chlorinated species were produced when trichloroethylene was undergoing degradation, thereby affecting reaction kinetics. These studies indicate that a single photocatalyst probably will not be able to treat all types of gaseous pollutants, and therefore, it would be advantageous to develop composite nanomaterials that are effective against an array of them.

Generation of Intermediates, Catalyst Deactivation, and Regeneration
Ideally, the final products of photocatalytic pollutant decomposition should be nontoxic, innocuous, and not cause secondary pollution. Given the slow kinetics of the charge transport and interfacial charge transfer, however, the formation of intermediates and by-products is almost inevitable. Moreover, toxic intermediates can accumulate on the surface of functional nanomaterials, occupy the active reaction sites, and consequently lead to a degradation in performance. For instance, Cao et al. [123] reported acute deactivation of TiO 2 during toluene photocatalytic degradation, and that was attributed to intermediates from incomplete oxidation, such as benzoic acid and benzaldehyde, that accumulated on the catalyst surface.
Gas chromatography-mass spectrometry (GC-MS) and in situ DRIFTS characterizations are techniques commonly used to study intermediate species formed over catalyst surfaces used for the degradation of VOCs and NO x . However, the mechanisms that govern selectivity and that cause the generation of toxic intermediates and by-products have not been fully established. To avoid the generation of toxic intermediates, researchers have investigated the deposition of metal oxides on catalyst surfaces to change the reaction pathways [123][124][125][126]. For example, Cui et al. applied in situ DRIFTS to probe the NO photoconversion route to NO 3 − by depositing SrO clusters on a g-C 3 N 4 surface, and it was shown that SrO facilitated NO activation via Sr-NOδ(+) formation [125]. Similarly, Li et al. employed GC-MS to show that dibenzothiophene over H 2 O 2 -assisted hydrothermal synthesis of TiO 2 -SiO 2 was initially photocatalytically oxidized to sulfoxides, then further oxidized to sulfones. That sequence optimized selectivity in a diesel-based photocatalyticadsorptive desulfurization process [127].
In terms of preferred reaction pathways for photocatalysis, the generation of sufficient ROS in photocatalysis is generally considered an effective method for suppressing unfavorable intermediates, and this is normally realized by the strategic design of the photocatalysts. For example, Li  − . That approach was effective because the co-existence of OVs and Bi metal had synergistic effects on ROS generation [126]. Besides, the geometric structures of ROS may also result in selectivity for photocatalytic reactions. In fact, a recent study showed that well-designed BiOCl(001) with surface OVs favored the formation of side-on superoxide radicals that could completely oxidize NO to nitrate (O 2 − B+ NO(g) → NO 3 − ) without any barriers, as shown in Figure 5a [128].
Air pollutants exhibiting different adsorption activations on varied facets can also induce differences in photocatalytic conversion routes. For instance, different NO adsorption activation patterns on the (110) and (001) facets of Bi 2 O 2 CO 3 essentially predetermine the facet-dependent photocatalytic conversion pathways of NO oxidation (Figure 5b) [129]. With sufficient localized electrons, NO and corresponding toxic intermediates can be depleted or easily activated as has been verified in the work of Dong et al. [130] For open structures that facilitate O 2 diffusion, such as TiO 2 nanotubes, the formation of intermediates during the photocatalytic degradation of VOCs can be inhibited by supplying sufficient O 2 . This method is also considered as a practical strategy for avoiding photocatalyst deactivation [131].
Photocatalyst regeneration is another extremely important consideration for practical applications in air purification because selectivity and stability need to be optimized to achieve safe and long-term photocatalytic air purification. In most cases, catalysts partially recover their activity after thermal treatment at high temperatures [132,133]. Sonication and UV irradiation, combined with H 2 O vapor or H 2 O 2 can also help the catalysts regain their original activity [133,134].

Room Temperature Catalysis
Room temperature (or ambient temperature) catalysis is considered as one of the most energy-efficient methods because there is barely any energy requirement for heating, and this makes it another appealing technology for eliminating low levels of NO x in semi-closed spaces, such as underground tunnels and for ppm levels of HCHO in new buildings [135]. Ambient temperature catalysis is a complete destruction/oxidation process that uses ROS, rather than liquid or AC adsorption, which need frequent replacement or regeneration. Two main advantages of room temperature catalysis are the energy savings and the absence of secondary pollution [136][137][138].
Indeed, considerable effort has been dedicated to the development of highly efficient and selective catalysts to achieve the complete degradation of HCHO and NO x at room temperature. For HCHO removal, transition metal oxides, including cobalt oxides [139], manganese oxides [140][141][142][143], nickel oxides [144] and titanium oxides [145], and noble metal nanoparticle catalysts (i.e., Ag, Au, Pt, Rh and Pd) [146][147][148][149][150] are commonly used. For NO x removal, transition metal oxides, especially cobalt and manganese oxides are often used [151]. Table 3 summarizes the types of catalysts commonly used for the oxidation of HCHO and NO x under ambient conditions.  The catalytic performance of nanomaterials is largely dependent on their structural properties (i.e., particle size, surface area, pore structure and phase structure), electronic properties, and surface/interface state (i.e., O species and surface composition). The designs of the catalysts are determined by the reaction conditions for which they will be used and the methods used for their synthesis. For example, Zhang et al. compared the activities of α-, β-, γand δ-MnO 2 for HCHO removal at 80 • C and found that tunnel structure and active lattice oxygen species were mainly responsible for the excellent catalytic performance of δ-MnO 2 [153]. The tunnel structure of δ-MnO 2 is built by edge-sharing octahedral MnO 6 with water molecules and selected cations (e.g., Li + , Na + , K + and Ca 2+ ) located between layers to balance excess charge [141,169,170]. Chen et al. also reported that the tunnel structure of manganese oxides is the major influencing factor for catalytic HCHO removal [171].
Cryptomelane with 1D tunnels exhibit excellent catalytic activity that can be explained by tunnel cross sections of ca. 0.46 × 0.46 nm that were similar to the dynamic diameter of the HCHO molecule [171]. Besides, OVs produced as a result of the large fractions of Mn 3+ ions in the MnO 2 framework served as active sites that promoted catalytic activity [172,173]. Bai et al. also reported that an increase in the quantities of Mn 3+ ions over Ag/MnO 2 can promote the formation of OVs and in so doing facilitate catalytic activity [154]. Thus, regulation of surface OVs is an effective means of promoting catalytic HCHO removal.
Recently, our group synthesized tetragonal-monoclinic nano-zirconia (ZrO 2 ) phase junction nanomaterials (TMZ) (Figure 6a), which exhibited higher conversion rates for HCHO into CO 2 compared with pure ZrO 2 . The enhanced catalytic property was attributed to the adsorption enhancement of Zr-rich sites and single electron activation at TMZ interfaces. Those effects cause C-H dissociation of HCHO and hydrogenation of O 2 , thus lowering the activation barrier for HCHO oxidation [174]. Our group also reported that nitrogen-doped, carbon-encapsulated small Co nanoparticles (Co@NC) (Figure 6b,d) for HCHO oxidation at room temperature achieved >80% HCHO conversion efficiency [168]. Metallic Co nanoparticles were identified as the active sites for O 2 dissociation through density-functional theory (DFT) calculations [175]. Panel (d) is enlarged image of (c). Reprinted with permission from reference [175]. (e) TEM and HRTEM images of MCO catalysts with different exposed facets (i and ii correspond to TEM and HRTEM images of MCO-111 facet, respectively; iii and iv correspond to TEM and HRTEM images of MCO-110 facet, respectively; v and vi correspond to TEM and HRTEM images of MCO-001 facet, respectively). Reprinted with permission from reference [176].
In related work, Li et al. systematically synthesized MnO x -CeO 2 to tailor the exposure of three major facets (Figure 6e), among which the MnO x -CeO 2 with exposed (111) facets (MCO-111) showed higher activity for formaldehyde oxidation and CO 2 selectivity compared with two low-index (110) and (001) facets [176]. Furthermore, the oxygen vacancies were highly stable on the (111) facet, and the high migration rates for bulk lattice oxygen were able to replenish the surface lattice oxygen consumed, resulting in good activity and stability. The highly active regions exposed at the (111)-dominated surfaces were evident in IR thermography images, which also showed that Lewis acid−base properties of (111) surface facilitated oxygen mobility and activation.
For NO elimination at room temperature, several types of catalytic materials, such as metal oxide catalysts (Co 3 O 4 ) [177], CrO x -based [166][167][168] and MnO x -based [159][160][161][162][163] catalysts), molecular sieves (Na-ZSM-5 [165]), and carbon nanofibers [164,178], have been studied thus far. The amorphous structures of these catalysts are critical for maintaining high activity and durability. For instance, Wang et al. [167,168] demonstrated that chromicbased amorphous oxide catalysts were active in ambient-temperature NO oxidation. These authors prepared a series of CrO x -ZrO 2 mixed oxide materials with high surface areas and homogeneous amorphous phases, and those mixed oxides exhibited 100% conversion rates for low NO concentration at 25 • C over extremely long times (nearly 30 h). A similar conclusion was reached by Shi et al. who showed that mesoporous, amorphous Mn-based oxides with low crystallinity and defects (OVs or Mn vacancies) were beneficial for high catalytic performance [159][160][161][162][163].
Nanostructured, Mg-doped manganese oxides (Mg-SC-OMS-2) showed low crystallinity with large surface areas and extensive defects, and they retained 99% NO removal efficiencies for as long as 8 h under an extremely high gas space velocity of 120,000 mL g −1 h −1 . The OVs in Mg-SC-OMS-2 improved the activity of the material for the additional activated O species generated via the Mars-van Krevelen (MvK) pathway [161]. In another study, the excellent performance of a nanoflower-like, weakly-crystallised manganese oxide was associated with Mn vacancies and hydrated surfaces, which can reduce the reaction barriers of the rate-determining steps [162]. Mesoporous X-Mn (X = Co, Ni, Zn) bimetallic oxides demonstrated substantially enhanced efficiencies in NO removal compared with their single-metal oxide (MnO x , Co 3 O 4 , NiO and ZnO) counterparts [159].
As discussed above, the addition of ZrO 2 to stabilise the key Cr 6+ redox sites can improve catalytic performance. In studies of Cr-Co, Cr-Fe and Cr-Ni oxide catalysts, the best performing catalyst achieved 100% NO conversion over~30 h of operation at a high space velocity of 45,000 mL g −1 h −1 [168]. Finally, the presence of the surface hydroxyl groups on chromic-based oxide catalysts is equally important for high catalytic activity and long-term stability [166]. In summary, amorphous structures, OVs, and surface hydroxyl groups can enhance the catalytic oxidation of NO x at ambient temperatures.

Noble Metal Loading
The types of noble metals and the amounts that are loaded on the catalysts can significantly influence performance. Zhang et al. investigated the effects of noble metals (i.e., Au, Rh, Pd and Pt) on catalysis of HCHO [147,155], and they reported the following order of activity for the supported noble metal catalysts under the same reaction conditions (20 • C, 100 ppm HCHO, 20 vol% O 2 , He balance, GHSV = 50,000 h −1 ): Pt/TiO 2 >> Rh/TiO 2 > Pd/TiO 2 = Au/TiO 2 . The enhancement of catalytic activity was attributed to a 'strong metal-support interaction' (SMSI) effect. Amongst a variety of influencing factors, including size, morphology, and electronic state of the supported catalysts, the SMSI effect is an intrinsic factor that can increase catalytic activity [179]. That is, SMSI can stabilise the morphology of metals, lead to strong reactant adsorption, and facilitate activation [180]. The SMSI effects also include enhanced O 2 chemisorption, which promotes the oxidation of pollutant molecules [180]. More specifically, the SMSI between Au and CeO 2 can weaken Ce-O bonds, thereby generating OV, and as a consequence, gas-phase O 2 molecules are readily activated to promote HCHO oxidation [181].
In terms of theory, DFT calculations indicate that the SMSI can lead to perturbations of electronic properties of the catalysts, and in doing so, enhance electronic interactions between components [182]. Moreover, SMSI can also modulate the chemical states of Ir species and affect reaction selectivity in that way [183]. Given the promising results obtained thus far, the SMSIs between noble metals and metal oxides and their associated effects on gas adsorption and O 2 dissociation for HCHO and NO x control should continue to be investigated.

Gas Hourly Space Velocity
In dynamic test systems of HCHO or NO x catalytic oxidation, GHSV is a critical influence on catalytic performance because it directly affects the residence time of pollutants in the reactor or the contact time between pollutants and catalysts. Generally, the removal efficiency of HCHO or NO x will decrease with increasing GHSV if other reaction conditions are held constant [156,[184][185][186]. For instance, Huang et al. studied the HCHO removal efficiency over 0.1% Pt-TiO 2 catalyst under different GHSV conditions [184]. The results showed that complete removal (100%) of HCHO was achieved at the GHSV of 40,000 h −1 (at room temperature, 10 ppm HCHO concentration, 50% humidity), while the 99.1%, 97.8% and 90.5% removal efficiencies of HCHO were obtained at GHSVs of 80,000, 160,000 and 240,000 h −1 , respectively.
Liu et al. studied trends in HCHO conversion over Pt-TiO 2 for increases in GHSV from 5,000 h −1 to 250,000 h −1 (30 • C, 10 ppm HCHO inlet concentration), and that study showed that the HCHO conversion decreased at first, and gradually reached a plateau when GHSV increased to 80,000 −1 [185]. That is because high GHSV reduces the residence time of HCHO molecules over catalysts. Li et al. found that the HCHO conversion efficiency significantly decreased from 98% to 37% when GHSV increased from 12,000 to 360,000 h −1 , especially at 20 • C over a Pd/CeO 2 catalyst [156]. For higher reaction temperatures, the decrease in HCHO conversion became less significant.

Relative Humidity
Similar to what happens in photocatalytic reactions, the presence of water vapor can also affect other types of catalysts at room temperature. For example, Chen et al. demonstrated that water vapor is beneficial for HCHO removal over supported gold catalysts [187,188]. Under a range of RH from 50%-75%, HCHO was completely oxidised to CO 2 and H 2 O at room temperature (80 ppm HCHO, 34,000 h −1 GHSV). However, the conversion efficiency of HCHO decreased by more than 10% in the absence of water. An abundance of water vapor favours the production of surface hydroxyl groups which promote the conversion of HCHO to formate, an important intermediate that can be further oxidised to CO 2 and H 2 O.
An overabundance of water, however, can induce the competitive adsorption of HCHO and H 2 O on active sites, and that will impede HCHO removal. More specifically, Huang et al. compared the influence of RH on the removal efficiency of HCHO over 0.1% Pt/TiO 2 , and the results showed the following decreasing trend in conversion efficiency: 25% ≈ 50% > 75% > 97.5% >> 0% (10 ppm HCHO, 25 • C and 80,000 h −1 GHSV) [184]. For NO oxidation, Shi et al. reported that water vapor enhanced catalytic stability over mesostructured, amorphous manganese oxides [172]. The maximum NO removal efficiency of 90% lasted over 237 h under a RH range of 50-90%, and the removal efficiency barely showed any effects of RH The stability was due to the formation of volatile HNO 3 molecules (i.e., 3NO 2 + H 2 O = 2HNO 3 + NO), which are easily carried away by that air flow, rather than forming NO 3 − that could block the active sites. Similarly, Mg-SC-OMS-2 showed an increase in NO removal under high humidity at room temperature. On one hand, H 2 O can help replenish O-containing species, but on the other, water vapor can impede NO 3 − deposition [161]. In another study, water vapor was found to inhibit NO removal on a Fe-Mn oxide catalyst, and the removal efficiency significantly decreased under 88% RH due to the competitive inhibition of the NO x adsorption/absorption sites by water molecules [160]. The evidence summarized above indicates that there is an optimal RH range for each unique catalyst, and this should continue to be investigated.

Catalytic Conversion Mechanisms
Generally, three typical reaction mechanisms, namely, the (i) Langmuir-Hinshelwood (L-H), (ii) Eley-Rideal (E-R) and (iii) Mars-van-Krevelen (MvK) mechanisms are involved in the catalytic conversion of pollutants at room temperature. In the L-H mechanism, a two-stage redox process takes place as follows: (1) The catalyst surface is partially oxidised by O 2 from the gas phase to form surface O species-this is non-equilibrium dissociative adsorption of O and not the classical oxidation on a catalyst surface. (2) Then, surface O species react with the pollutants that are weakly pre-adsorbed on the surface, and in the meantime, the oxidised catalyst surface is reduced by pollutant molecules [189][190][191]. For the L-H mechanism, the controlling step is the surface reaction between two reactant molecules adsorbed on adjacent active sites. Different from the L-H mechanism, in the E-R mechanism, one reactant molecule from the gas phase reacts with another reactant molecule that was previously adsorbed on the catalyst surface. In the case of the MvK mechanism, the lattice oxygen on the catalyst surface participates in the reaction, and it can be supplemented by the oxygen obtained from the gas phase [149,192].
Catalytic oxidation is a highly complex process, and reactions of this type often are affected, if not controlled, by the behaviour of reaction intermediates [193]. Theoretical calculations can be used to predict reaction pathways and mechanisms, and that information can be used to better understand the process. For example, on the surface of Mn-doped, oxygen-defective CeO 2 , HCHO can adsorb onto Ce-O bridge sites and O 2 onto adjacent O defect sites. Following that, HCHO is dehydrogenated to HCO-, and the dissociated H combines with O 2 to form -OOH. The HCO-continues to dehydrogenate to CO which then reacts with O from the dissociation of -OOH to form CO 2 and H 2 O [194].
The process just described is an example of the L-H mechanism. Metal oxides with alterable cation oxidation states can serve as active sites that trigger HCHO oxidation, which is under control of the MvK mechanism [149,181,195]. Specifically, on Au/CeO 2 surfaces, HCHO captures a lattice oxygen and forms HCOOH; this is followed by O 2 occupying the oxygen defect site. The O 2 later dissociates and reacts with HCOOH to form H 2 CO 3 , which in turn further dissociates into CO 2 and H 2 O [181]. For Pt-Bi 2 WO 6 nanosheet catalysts, HCHO is first adsorbed through H bonding with a surface hydroxide radical. The HCHO captures one lattice oxygen and forms CH 2 O 2 (dioxymethylene, DOM) with one remaining oxygen defect. Subsequently, the DOM undergoes isomerization and transforms into HCOOH. The O 2 adsorption and steps that follow are similar to what occurs on Au/CeO 2 surfaces [195]. For HCHO oxidation on K-Ag/Co 3 O 4 surfaces, HCHO is first dehydrogenated to HCO-, and that is followed by the acquisition of one lattice oxygen to form CHOO and a OV site which can be repaired by O 2 adsorption. Finally, CO 2 and H 2 O are formed through metal cation centred re-oxidation [149].
Surface OVs can promote the activation of oxygen molecules and the replenishment of ROS. As shown in Figure 7a, on Co 3 O 4 catalysts with surface OVs, the transfer of active oxygen can be significantly facilitated by a synergistic effect between Co 3+ /Co 2+ and O/Vo redox. Once HCHO molecules adsorb onto the surface, they can be readily oxidized into formate species and eventually converted into CO 2 and H 2 O by reacting with ROS [196]. For ferrihydrite (Fh) supported Pt catalysts (Pt/Fh) (Figure 7b), HCHO and O 2 were first simultaneously adsorbed onto the Fh and Pt surfaces, respectively; wherein the abundant surface hydroxyl groups on Fh were beneficial to HCHO adsorption via hydrogen bonding between an oxygen atom of HCHO and a hydrogen atom of the surface -OH. Then, the active oxygen radicals generated through O 2 splitting attacked the carbon atom of HCHO to from DOM intermediates, which further transformed into formate by the addition of an active surface oxygen atom. The adsorbed formate was subsequently oxidized into adsorbed CO 2 (CO 2ads ) and H 2 O (H 2 O ads ) by an active hydroxyl radical and finally desorbed from the Pt surface [197]. Reprinted with permission from reference [196]. (b) Reaction pathway for complete oxidation of HCHO over the Pt/Fh. Reprinted with permission from reference [197]. (c) Schematic illustration and reaction pathway of catalytic-oxidation of HCHO on the Co@NC catalyst at room temperature. Reprinted with permission from reference [175].
By combining the results of in situ DRIFTS and DFT calculations, our group proposed a MvK reaction pathway for HCHO oxidation over cobalt nanoparticles encapsulated in porous nitrogen-doped carbon (Co@NC) (Figure 7c). As proposed, HCHO and O 2 molecules initially adsorb onto Co@NC. Then, the adsorbed O 2 on the Co 0 surface dissociates into active oxygen radicals, oxidizing the adjacent HCHO to formate species, which are further oxidized to adsorbed CO 2 and H 2 O that ultimately desorb from the Co surface [175].
Various mechanisms for NO catalytic oxidation over other catalysts also have been uncovered; for instance, Cr 2 O 3 with exposed (001) surfaces tends to follow the E-R mechanism, in which NO in the gas phase directly reacts with adsorbed O 2 to form NO 2 . This accounts for the strong adsorption of O 2 onto unsaturated Cr3c sites [142,198]. For Mgdoped manganese oxide surfaces, the NO catalytic oxidation process follows the typical MvK mechanism. That is, the lattice oxygen is consumed by reaction with NO, and this is accompanied by the reduction of Mn 4+ to Mn 3+ . Subsequently, Mn 3+ is reoxidized to Mn 4+ through oxygen replenishment to close the redox cycle [161]. A similar mechanism has been observed for amorphous CrO x -ZrO 2 catalysts [167].
Ideal catalysts for industrial applications would be efficient, selective, and durable. In practice, however, the accumulation of intermediate hetero atoms that can cause poisoning by halogens or sulfur leads to incomplete oxidation. Moreover, the growth of catalyst particles inevitably leads to catalyst deactivation [139,142,190,199,200]. In essence, because the MvK mechanism is a redox reaction, [201] strong interactions between the catalysts and surface-active species can hinder continuous redox cycling, and this can also cause catalyst deactivation [167,201,202].

Catalyst Immobilization
Functional nanomaterials are essentially solid powders, and therefore, immobilization on support materials or films is necessary for practical applications, otherwise it would cause severe pressure drops or mass loss [136]. The choice of an appropriate support for immobilizing photocatalysts is based on several considerations as follows: (I) photocatalytic activity should not be inhibited after the immobilization process. For example, a structured ceramic foam was used to immobilize a g-C 3 N 4 photocatalyst, and its photocatalytic performance was not degraded but enhanced [203]. (II) catalyst powders must possess good adhesion properties for the support material because any released nanomaterials could pose threats to human health [204]. (III) the support material should have a large specific surface area and be chemically inert; and (IV) the support material should be mainly composed of raw materials and inexpensive. In addition to ceramic foams, other types of substrates such as optical fibres [205], porous lava [206], fiberglass [207], quartz [208], glass [209], and stainless steel [210] are often used for immobilization. Glass substrates are extensively utilised because of their good light transmittance and ease of cleaning [211].
Sol-gel synthesis [212], dip coating [213], chemical or physical vapor deposition [214], and electrophoretic methods [215] have been used to immobilise photocatalysts on various supports. Of these, the sol-gel method, which exhibits high versatility, has been widely adopted for depositing different precursors, such as titanium alkoxide, titanium tetrachloride or titanium halogenide, on substrates [216]. Recently, Huang et al. fabricated protonated g-C 3 N 4 /Ti 3+ self-doped TiO 2 nanocomposite films (p-CN/TiO 2 ) on glass substrates, and that appears to be an effective new approach for making photocatalysts adhere to a substrate at room temperature [217]. For room temperature catalyst immobilization without the need for light penetration, honeycomb ceramics [218], cellulose fibres [219], and AC [220,221], are common solid supports. Of these, AC exhibits superior mass transfer due to its high surface area and high porosity [222].
Liu et al. investigated the mechanism for NO oxidation over Na-ZSM-5 molecular sieves with high SiO 2 /Al 2 O 3 ratios [165]. The results showed that the NO catalytic oxidation process was accompanied by a significant adsorption of NO 2 , thereby causing a dynamic equilibrium between NO 2 adsorption and NO oxidation. In another study, Wang et al. found > 60% NO removal at room temperature on porous carbon nanofibers (PCNFs) for an inlet NO concentration of 20 ppm [178]. The removal mechanisms included NO adsorption, catalytic oxidation to NO 2 , and partial reduction to N 2 by PCNFs due to the large surface area and abundance of N species. The development of air purification technologies based on functional nanomaterials does present significant practical challenges, but there is a good potential for achieving the final objectives.

Roadside
Air pollution due to excessive NO x emissions is a major issue near roads where traffic is heavy and in tunnels and underground parking lots, and this is a special concern for people who spend a significant amount of time in those places. To address the problem, TiO 2 photocatalysts have been combined in novel ways with cement for pavements or coverings on clays and other construction materials to eliminate NO x in urban environments since the end of 1980s [223][224][225]. This process is regarded as a passive air purification strategy because natural sunlight is used to activate the photocatalytic material, and pollutant molecules freely diffuse onto the surface, with examples shown in Figure 8. Laboratory studies by Bianchi et al. [226] were conducted to evaluate the deNO x performance of full-sized industrial photocatalytic porcelain tiles. Under realistic gas flow rates and NO x concentrations (100 ppb) conditions, the NO x conversion in July was found to be higher than that in September, and that was explained by the difference in light irradiance.
Although reasonable photocatalytic performance of these engineered construction materials has been achieved in the laboratory, field studies have shown mixed results. For example, Maggos et al. found that a decrease in NO x concentration levels in an indoor car park where TiO 2 paint had been applied was significant, whereas Gallus found no photocatalytic beneficial effect on NO x , O 3 , or VOCs in an artificial street canyon in Italy compared with a reference site [227][228][229]. In another full-scale experiment supported by the EU, TiO 2 containing concrete pavement on a street in Hengelo, The Netherlands showed a maximum of 45% NO x removal efficiency under ideal weather conditions [230]. A positive result was also obtained in field studies in Copenhagen where NO abatement reached a peak higher than 45% at the summer solstice [231]. Our recent field study suggested even higher deNO x performance was achieved over g-C 3 N 4 /TiO 2 composite pavements during rush hour, indicating the diffusion and adsorption of low level NO x in open space are equally crucial [232]. The discrepancies in air cleaning effects of these engineered materials were likely caused by differences in the photocatalytically active surface area-to-volume ratios (S active /V) at the field sites, but other factors can also help explain the differences.
For instance, daytime and nighttime studies should be conducted to provide an integrated value for reduction efficiency in urban environments. Sampling height is another important factor that can influence the remediation results because sampling few centimetres away from the active photocatalyst surface results in a much higher NO x reduction value compared with sampling at a height of 3 m, which is typical of an urban network monitoring station. Variations in pollutant emissions and dispersion can affect transport to the active surface, and that can affect the remediation results [229]. Although the results have not always been positive, this innovative technology deserves continued research because it may be suitable for the treatment of low concentrations of pollutants (sub-ppm level), and it has low implementation and operating costs compared with other techniques.

Indoor Environment
There are many sources for formaldehyde, toluene, and various other VOC species, and NO x , in household environments; these include combustion sources, especially emissions from cooking; outgassing from interior furnishings and paints; and the infiltration of polluted outdoor air. Indoor air pollution has become a serous concern because pollutant concentrations can be high and exposures long-indeed, humans spend > 80% of their time indoors and that increase the risk of harmful effects. Early in 1995, the U.S. Environmental Protect Agency started investigations into the photocatalytic oxidation of indoor VOCs, including formaldehyde, benzene and polycyclic aromatic hydrocarbons (PAHs).
Indoor environments have conditions that are different from the outdoors, such as low air flows (<1 L min −1 ) and low visible light intensities, and only nontoxic photocatalytic materials are suitable for indoor use [233]. Taking these constraints into consideration, Lorencika et al. developed a water-based photocatalytic coating (C-TiO 2 ) on wallpaper that had made use of visible light, and that material showed 18% deNO x performance for an initial NO x concentration of 500 ppb. These researchers suggested that the pretreatment method they used to expose the active component was directly related to the deNO x effectiveness. Indoor lighting optimization also would be a way to enhance photocatalytic efficiency. Gandolfo et al. [234] studied the effectiveness of TiO 2 -embedded photocatalytic wall paints on NO 2 conversion in an indoor environment, and HONO, a potentially harmful intermediate, was found to be formed while NO 2 was being destroyed. This shows that thorough evaluations of potential problems with APN methods, especially the generation of secondary pollutants, need to be conducted in addition to the more straightforward assessments of the efficacy of NO removal.

Tunnels and Indoor Parking Lots
Conditions unfavourable for pollutant dispersion cause city tunnels and indoor parking lots to be among the most polluted places, often with high levels of NO x , VOCs, and PAHs generated by vehicles, especially during rush hours and when large numbers of vehicles are starting their engines [235]. In 2012, a city tunnel in Rome was renovated with photocatalytic, cement-based paint on interior and exterior walls and a new lighting system [236]. In two successive environmental monitoring campaigns of 30 working days each, the NO x abatement values were found to be an improvement of over 20% compared with a non-renovated tunnel. The best reduction value was observed in the centre of the tunnel, where the light intensity and wind speed were almost constant.
In 2015, another city tunnel renovation program was conducted in the Leopold II tunnel in Brussels, Belgium by researchers from five European countries. That study was funded as the Demonstration of Photocatalytic Remediation Processes on Air Quality Project, [228,230] and it showed a different effect compared with the study in Rome. In the heavily polluted Belgian tunnel, the photocatalytic surfaces were deactivated by dust and soot, there was low light irradiance, and wind speeds were high. Those conditions led to poor photocatalytic activity and unexpectedly low NO x removal. Although the final results were disappointing, that study showed that high light intensity, low wind speeds, large active surfaces, and superior photocatalytic activity are required for successful remediation. In 2007, the EU sponsored a projected called the Photocatalytic Innovative Coverings Application for Depollution Assessment in which a photocatalytic coating was used to treat air pollution in an indoor parking lot. Results of that study, conducted by Maggos et al. [237] showed that the deNO x ability of TiO 2 -containing paint was~20%, which was lower than that obtained in the laboratory, and that relatively low value was thought to be because organic compounds had interfered with NO x oxidation in the parking lot.
The literature summary presented above shows that passive APN systems can be simple, cost-effective solutions for combating air pollution in some applications. However, the performance of these systems can be compromised in heavily polluted areas and in places with low light intensity and variable weather conditions. The rapid deactivation of photocatalysts and the generation of secondary pollutants are primary concerns relative to this technology.

Active Systems for APN
As discussed above, passive APN systems have significant limitations for effective pollutant abatement and their durability is an issue, and therefore, active APN systems have been introduced. These active systems typically incorporate light sources and blowers, and they often are combined with other techniques, such as filtration or adsorption (examples shown in Figure 8). In the discussion that follows, these systems are categorised based on their applications.

Air Cleaners for Indoor Environment
The efficiencies of passive APN systems can be severely restricted by insufficient mass transport and residence time of pollutants on the photocatalyst surface that were not ideal. These problems can be countered by installing a blower or a fan in the system to force air to flow over the photocatalyst surface. A typical example of an active APN apparatus is an indoor air cleaner. Conventional air cleaners use filtration to remove particle matter from the air, but they are generally ineffective for gaseous pollutants. Incorporating filtration with nanomaterial-based catalysis is a way to improve the performance of air cleaning units and increase the durability of the catalysts.
For example, Ao et al. [238] reported that an air cleaner with active carbon and HEPA filters had an extremely low capacity for NO removal. The addition of a TiO 2 filter to that system, however, resulted in the removal of almost 90% of the initial NO in 120 min under irradiation with a UVC lamp. The use of a combined TiO 2 /active carbon filter showed a further enhancement in NO removal, and what is more, the release of NO 2 was inhibited due to its adsorption onto the active carbon. Indeed, the adsorption of the gaseous contaminants on active carbon has two benefits (1) to increase the residence time of the pollutants under high flow rates and (2) to regenerate the adsorbent [56]. These studies demonstrate the advantages of combining filtration, adsorption, and photocatalysis for indoor air purification systems.
In addition to efficient mass transport, strong and uniform irradiance for air purifying devices is critically important for ensuring a high level of performance and suppressing the generation of undesirable intermediates. For example, Denny et al. [239,240] found that during the photodegradation of ethanol, the use of two blacklight blue (BLB) lamps with high irradiance reduced the production of acetaldehyde, a toxic intermediate. After replacing the BLB lamp with a UV-LED lamp, the peak concentration of acetaldehyde decreased by more than 10-fold due to the more uniform irradiation of the LEDs on the photocatalyst surface.
Slimen and Fujishima developed an indoor air purifier that used a TiO 2 -impregnated Ti mesh filter as a cleaning unit, and it was found to be effective in decomposing cigarette smoke [241]. Weon et al. [69] developed a (001) facet-dominated TiO 2 nanotube photocatalyst filter, which showed improved activity for toluene, acetaldehyde, and formaldehyde removal compared with (101) facet exposed TiO 2 nanotubes (TNTs). These researchers further evaluated the removal of VOCs by a commercial air cleaner equipped with a (001)-TNT filter, and they found that it exhibited an average removal efficiency of 72% within 0.5 h; this level of performance met the Korean air cleaner standards protocol [69]. Feng et al. combined a triboelectric nanogenerator (TENG) with a photocatalysis technique to develop a self-powered electrostatic filter for formaldehyde removal [33].

Tunnel
The largest tunnel air cleaning system in central Hong Kong presently in use consists of eight units with advanced air cleaning equipment to cleanse the air for a nearby green space and commercial centre. Each cleaning unit is composed of an electrostatic dust collector, which is used to remove suspended particulate matter, and a special unit for NO 2 emitted from motor vehicles. A total of 5,400,000 m 3 of air in the tunnel can be treated each hour, and the removal efficiency of PM 10 and NO 2 can be as high as 80% (http://www.3hk.cn/hk/2019/01/21/119325.html (21 January 2019), in Chinese).

Photocatalytic Solar Tower for an Industrial Zone
In Japan, small, unregulated industrial establishments emit low concentrations VOCs, which can lead to local air pollution. To control this problem, researchers devised a unique tower-like photoreactor that photodecomposes VOCs and is suitable for installation in confined spaces [242]. A cylindrical mirror with gratings was used to focus solar irradiation onto the interior of a reactor. A TiO 2 -coated ceramic tube in the reactor effectively eliminated toluene vapor at 5 ppm, and its performance was almost solely dependent on solar intensity. Long-term monitoring for six months showed almost no decrease in photocatalytic activity, which suggests that the solar tower is suitable for the long-term remediation of atmospheric VOCs.

Catalytic Street Lamps
Recently, our group developed an enhanced outdoor air cleaning device for the elimination of PM and NO x that combines filtration and photocatalysis with lighting. Practical considerations for structural design, key parameter optimization, and performance in real environments were taken into account during the development of these units. For these devices, pre-treatment with a filtration unit is used to remove particles, and a photocatalysis unit is used for further purification. The performance of the device was evaluated through mathematical modelling and monitoring in a field deployment. The average elimination rates of NO were over 55% and 40% on sunny and rainy days, respectively, and the average filtration efficiency of PM 2.5 reached 70-80%. The overall purification efficiencies for PM 2.5 , PM 10 and NO x for 200 devices installed in the Xi'an International Horticultural Expo Park were 4.3%, 9.0% and 16.1%, respectively, during a long-term deployment. Plans are being made to install enhanced outdoor air cleaning devices in large areas in the near future to improve air quality.

Hybrid Solar-Assisted Large-Scale Cleaning System (HSALSCS) for Large-Scale Air Cleaning
At present, active APN systems are most often used for indoor environments and relatively small spaces. Last year, Nature News reported on a prototype HSALSCS developed by Chinese and American scientists to combat urban smog in Xi'an, a provincial capital in central China [243]. The design and development of the HSALSCS have been discussed in detail by J. Cao elsewhere [244]. Briefly, a 60 m-high concrete chimney was built on top of a large glass solar collector, within which electric-powered blowers move the air through a series of filter banks. Particles are separated by the filter banks, and clean air billows from the top of the chimney.
Q. Cao et al. [245] from the aerosol group at the University of Minnesota ran a transient 3-D numerical model of fluid and heat flow for the system and validated the model with field measurements in January 2017. The analysis showed that the system's cleaning performance was largely dependent on physical parameters and factors, especially the amount of solar irradiation, solar collector width, and tower width and height. The pressure drop of the filters also was critically important. If the pressure drop becomes extremely high, additional fans (or blowers) must be used to strengthen the flow field [246].
During a one month field monitoring period, a 19% decrease in PM 2.5 concentration within a 10 km area was registered in air monitors surrounding the HSALSCS compared with monitors in other parts of the city. This study received strong positive feedbacks from world-renowned atmospheric scientists and chemical engineers who noted that this prototype provides an innovative approach for combating the public health hazard of urban smog. The HSALSCS was selected as one of "28 Incredible 'Made in China' Innovations that are Changing the World" in February 2019 (28 incredible 'Made in China' innovations that are changing the world (msn.com), 28 June 2019. Reprinted with permission from reference [238,242,243] and (c) passive system of APN with TiO 2 embedded concrete pavements applied on roads, bridges, and tunnels. Reprinted with permission from reference [230,236].
In summary, the studies on the photocatalytic removal of air pollutants with nanomaterials have shown promising results, but their performance has varied with the specifics of the applications. Indeed, enhancements to catalytic activity will require the design more effective nanomaterials, and that in turn, will depend on a better understanding of the underlying mechanisms involved in the interactions between the nanomaterial surface/interfaces and pollutant molecules.

Conclusions and Future Perspectives
APN is a novel technology for combating air pollution that is characterized by low energy consumption and high efficiency. This review provides a general background on air pollution in China and highlights the need for developing ambient air control technology as a supplement to source control methods. Next, the most pressing and critical issues are considered; these include filtration, adsorption, and mild catalysis (photocatalysis and ambient temperature catalysis), including nanomaterial design and optimization, key environmental conditions, nanomaterial deactivation and regeneration. Finally, applications of photocatalytic air cleaning, and cleaning devices integrated with adsorption, filtration, or catalytic degradation technologies are discussed.
Although promising achievements in air pollution control technologies already have been made, future research is urgently needed to directly address the following issues: 1. The continued development of novel nanomaterials that feature strong visible light (up to 700 nm) absorption, fast charge transport, and low e − /h + recombination is still desirable, especially for practical air pollution abatement applications.
2. Replacement of noble metals and rare metals in catalysts with more abundant materials, including non-metallic substances, such as metal-free g-C 3 N 4 photocatalysts, carbon, perovskite, or other natural clay nanomaterials with high adsorption properties. This will benefit large-scale production and broaden applications.
3. Understanding the mechanism involved in interfacial charge transfer and ROS formation; the key factors governing reaction pathways and material deactivation deserve special attention.
4. Design of new APN devices to target pollutants in a range of micro-environments, and in combination with other cleaning devices, to combat street and even urban-scale air pollution.
5. With the outbreak of novel coronavirus disease (COVID-19) and the serious threat to public health that it poses, research opportunities are now emerging for investigations into the disinfection of airborne pathogens with nanomaterials. The formation of byproducts by nanomaterials is minimal, and this adds to their attractiveness compared with traditional disinfectants such as O 3 , H 2 O 2 , ClO 2 , or UV irradiation [247]. Recently, some nanomaterials have been shown to provide broad-spectrum disinfectant properties; these include nanosilver (nAg), TiO 2 and carbon-based nanostructures [248][249][250][251].
The mechanisms that provide disinfection are another area of active investigation. For example, released silver ions (Ag + ) with high redox potentials can strongly bind to thiol groups (-SH) of viral or bacterial proteins, and this is thought to cause the denaturation of DNA molecules and lead to permanent disinfection and broad spectrum antimicrobial effects [247]. Disinfection from TiO 2 is due to the production of ROS through exposure to light-those substances disrupt the respiratory chains of bacteria and cause oxidative damage to cell membranes and nucleic acids. Nano-TiO 2 photocatalytic deactivation is effective for Gram-positive and Gram-negative bacteria and viruses [248,249]. The antimicrobial effects of CNT are different from nAg or TiO 2 because the damage caused by the tubes results from the physical puncturing of cell membranes [247]. Finally, the antimicrobial mechanism for the graphene family of materials has been reported to involve oxidative damage caused by ROS.
Optimization of the oxidation state, size, shape, and electronic structure of nanomaterials and surface functionality is a way to enhance antimicrobial activity. Overall, the deactivation mechanisms for viruses/bacteria can be quite variable among nanomaterials, and improvements in their effectiveness will require further research on material optimization and long-term stability, and verification of the mechanisms also will be needed. There are, however, some indications that the attachment of antimicrobial nanomaterials onto support polymers is a critical step toward practical air disinfection, and this is another area that deserves further investigation.
A literature survey has shown that Japan, Hong Kong, the United States and European Union countries have conducted a considerable number of trials in which nanomaterialbased catalysts were used for outdoor and indoor air purification. In comparison, China has made fewer attempts along these lines, and the results of only a few studies have been published. Therefore, efforts should be made in China to develop new, cost efficient, and functional air cleansing nanomaterials and to employ them in ways that benefit our society. Given the continued problems of air pollution and the growing numbers of pollutants expected in the future, APN technology can become more efficient and practical through focused research, and we believe it will pay large dividends in the mediation of air pollution throughout the world.