Photocatalysis represents an economical process for the conversion of solar energy for use in a broad range of applications, including environmental remediation, energy and fuel production. Intensive efforts have been undertaken to develop novel photocatalysts with improved performance in terms of broadband absorption, charge migration and reaction pathway efficiency. Strategies frequently applied in the design of photocatalysts include engineering of their nanostructures (nanoarchitectures and exposure of high energy facets [1, 2]), their interfaces and surfaces (nanocomposites with matched energy band positions and Ohmic contacts [3, 4]) and their band structures (doping and solid solutions [5]). Among these strategies, tailoring the vacancies in the lattices of traditional photocatalysts, such as graphitic-C₃N₄ (g-C₃N₄), titanium, tungsten or bismuth oxide-based compounds, has enormous potential for achieving high photocatalytic performance in diverse applications.

Vacancies, or point defects, are prevalent in all semiconductor materials because the conductivity of these materials originates from their inherently nonstoichiometric compositions. All chemical functions of materials, such as catalysis, adsorption or sensing behaviors, are dominated by their electronic structures and surface properties. The vacancies in a material play a critical role in the manipulation of the electronic structure, surface acidic and basic properties, dissociative adsorption, charge transportation, reactivity and other chemical properties of the material, and this has been known for a long time through both theoretical calculations and experimental observations [6-9]. However, the earliest study that directly discussed the influence of vacancies on photocatalytic performance was carried out about two decades ago by Yanagida et al. [10, 11]. In their reports, sulfur vacancies on the surface of CdS were shown to promote the photocatalytic reduction of CO₂ and the selective production of CO [10, 11]. Studies on oxygen vacancies (V_O) in photocatalysis increased dramatically in the 2000s and, accompanied by the development of visible light-driven TiO₂ catalysts [12], accelerated further after promising results were obtained from black TiO₂ and subsequent research [13-15]. Recently, vacancy-rich photocatalysts containing other vacancies besides V_O have been developed. These vacancies include anion vacancies such as sulfur [16], halogen [17-19], and nitrogen vacancies [20-23]; cation vacancies such as bismuth [24, 25], titanium [26, 27], and carbon vacancies [28, 29]; and combinations of both anion and cation vacancies [24, 30].

Vacancy engineering provides a number of notable advantages for high performance photocatalysts: (1) an approach to manipulating the compositions of catalysts without the introduction of impurities; (2) simultaneous tuning of the concentrations and types of vacancies in the bulk or surface lattice to meet requirements closely related to photocatalytic performance (e.g., narrow band gaps, enhanced optical absorption and improved charge separation); (3) a promising low-cost method using simple synthetic techniques to modify traditional photocatalysts composed of earth abundant elements, in contrast to modification using complicated procedures and rare elements; (4) a possible method to simultaneously influence thermocatalysis and photocatalysis by enabling surface chemisorption.

The critical functions of V_O on the photocatalytic performance of TiO₂ have been frequently reviewed [31-34]. In this minireview, we will introduce and discuss the latest progress regarding a broader range of anion and cation vacancies, and focus on their characterization, formation chemistry and fundamental benefits for photocatalysis.

It is essential to characterize the existence, concentration and type of vacancies present in a material. The existence of vacancies can be confirmed by measuring the stoichiometric ratios of elements using elemental analysis, X-ray fluorescence (XRF), X-ray photoelectron spectroscopy (XPS) or Rietveld refinement methods. Vacancies can also be directly observed by scanning tunneling microscopy (STM) [8], or indirectly characterized by cataluminescence [35], photoluminescence [36], electron spin resonance (ESR) [12], or other techniques. ESR is a typical measurement used to confirm the existence of V_O, which exhibits a signal at g = 2.004. A signal at g = 1.998 was observed in p-type TiO₂ and attributed to titanium vacancies (V_Ti) [26]. However, these methods cannot distinguish among the different types of vacancies or determine their concentrations.

Positron annihilation spectroscopy is a powerful technique for investigating defects in materials, and can provide information on vacancy types and their relative concentrations [24, 37, 38]. Three positron lifetimes (τ₁, τ₂ and τ₃) and their relative intensities were obtained from the positron lifetime spectra of TiO₂ and CeO₂ nanoparticles. The positron lifetimes, from shortest to longest, were attributed to small V_O in the bulk, larger clusters of V_O on the surface, and large voids in the material, respectively. The relative concentration of bulk defects to surface defects can be estimated from the ratio of the intensities of the two corresponding peaks (τ₁ and τ₂). It was thus shown that a lower relative concentration of bulk vacancies in a TiO₂ crystal corresponds to a higher photocatalytic performance because of improved charge separation [37]. In contrast, there is a positive relationship between the presence of large V_O clusters on the surface of CeO₂ and the reducibility and reactivity of the photocatalyst [38]. Moreover, four positron lifetime components were observed in the spectra of ultrathin BiOCl nanosheets and nanoplates [24]. The shortest of these (~250 ps) was attributed to single isolated bismuth vacancies (V_Bi), and the second shortest (~325 ps) to Bi³⁺-oxygen vacancy associates. According to their relative intensities, the associated vacancies were dominant in ultrathin BiOCl nanosheets [24]. However, despite the success of positron annihilation in defect characterization, further knowledge of the various vacancies is required, and more advanced techniques are also desirable for a better understanding of the fundamental relationships between vacancies in the lattices of photocatalysts and their photocatalytic properties.

The anion vacancies in photocatalytic materials are most often oxygen, nitrogen, sulfur, and halogen vacancies. Owing to the instability of sulfide photocatalysts, there are very few reports on photocatalysts containing sulfur vacancies [16]. In contrast, V_O in metal oxides have frequently been investigated because most photocatalysts are n-type metal oxide nanoparticles, in which the conductivity originates from V_O. Fig. 1 shows the oxygen and titanium vacancies in a defect-containing TiO₂ lattice without dopants.

Although V_O occur naturally in materials, higher density V_O can be generated by aliovalent ion doping (e.g., nitrogen doping) [39], formation of two-dimensional ultrathin materials [24], solid-state chemical reduction by NaBH₄ [40-42], plasma treatment [12], addition of chelation precursors with carboxylic acids [43], hydrothermal reduction with graphene oxide [44], or annealing in an inert or reducing gas atmosphere (e.g., hydrogen) [13, 45]. Solid-state chemical reduction with NaBH₄ is a facile method, and the resulting colored products feature an amorphous shell with V_O [40]. However, hydrogen treatment (HT) has also attracted considerable attention after the pioneering work by Mao's group [13]. In fact, HT has been used historically to tailor the optical and electric properties of single crystal oxide materials [46]. This method was applied to TiO₂ nanoparticles and found to be effective in improving their photooxidation performance in as early as 2003 [47]. However, it is only since the exciting results with black TiO₂ that HT has become an important area of research and been explored extensively for the treatment of TiO₂, α-Fe₂O₃, WO₃, BiVO₄, and other photocatalysts with various nanoarchitectures [48-51]. After HT, both high density V_O and the resulting defect disorders are generated on the surface layer of the nanomaterials. Some results have indicated that the change in the properties of the materials was mainly caused by the high density of surface V_O trap sites [36]. However, others have suggested that both V_O and the surface structural disorder play a synergistic role, in which V_O mainly exists in the bulk of the lattice, but the surface structural disorder is almost stoichiometric [52].

The improved photocatalytic performance after HT has been ascribed to two main effects. First, after HT, the optical absorption of the photocatalysts was greatly extended to the near infrared (NIR) range (~1200 nm). The visible light absorption was attributed to the localized mid-bandgap V_O occupied by an unpaired electron in an antibonding orbital at a surface site [36]. These V_O states are located at 0.7-1.0 eV below the conduction band (CB) (Fig. 2(A) [12, 46, 52]). Therefore, the electrons can be excited from the valence band (VB) and localized V_O states to the CB or from the VB to localized V_O states by lower energy photons, and this results in a considerably broadened absorption. Second, a high V_O density causes an increase in the donor density in the photocatalysts because the V_O function as electron donors. After HT, the donor density in TiO₂ nanowires was three orders of magnitudes higher than the density before HT [14]. This increased donor density not only improves the charge transport in the lattice, but also promotes electron transfer at the interface because the upward shift of the Fermi level toward the CB as a result of the increased donor density augments the band bending. However, the considerable improvement in the photoactivity was mainly ascribed to the enhanced performance in the UV region, but not in the visible light region [14, 15]. This was confirmed by time-resolved fluorescence (TRF) and femtosecond transient absorption (TA) spectroscopy, which indicated that the charge carriers slowly decayed on UV excitation, but showed considerable recombination on visible light irradiation [34]. It remains a challenge to effectively utilize the absorption in the vis-NIR range, despite the fact that the integration of photo- and thermocatalysis has great potential for maximizing the use of low energy photons [15, 51, 53-55]. Besides the two benefits mentioned above, HT has also been found to improve the photostability of WO₃ by V_O production and formation of substoichiometric WO_3-x, which was highly resistant to reoxidation and dissolution induced by peroxo species [49].

More importantly, vacancies function as active sites and enhance the adsorption of reactants (e.g., oxygen, water, or organic molecules), promote electron migration from materials to adsorbates, and can even directly participate in specific photocatalytic or photochemical reactions. One fundamental feature of V_O-rich oxides is their ability to chemisorb oxygen species, which also contributes to the stabilization of V_O. Engineering of nondopant V_O in CeO₂ is an important method for promoting the catalytic oxidation of CO by providing coordinatively unsaturated sites [56]. Interestingly, the inverse reaction, selective photochemical conversion of CO₂ to CO, was also found to be mediated by V_O, which facilitated electron migration from CeO₂ to the chemically adsorbed and preactivated CO₂ molecules. Simultaneously, owing to the consumption of V_O in the reaction, CO₂ photoreduction slowed down and was eventually terminated [57].

Based on their behavior in both photo- and thermocatalysis, vacancies could bridge them and enable a synergistic effect for the utilization of solar energy, even in the vis-NIR range. V_O were observed as electron-transfer mediators in the thermocatalytic decomposition of CO₂ on mesoporous WO₃, in which V_O were consumed to produce C/CO [51]. A synergistic photo/thermocatalytic process with a high selectivity for the conversion of CO₂ into CH₄ was accordingly developed using this V_O-rich WO₃ [51], which represents an important strategy for the combination of photo- and thermocatalysis to realize CO₂ reduction. However, it would be more desirable to achieve combined photo/thermocatalysis using only solar energy, thus avoiding external energy inputs. Recently, highly selective aerobic coupling of amines to the corresponding imines was achieved under NIR illumination (> 700 nm) on V_O-rich WO₃ nanosheets (Fig. 2(B) [45]. The low energy photons were used to generate superoxide radicals (O₂^•-) in a chemisorbed state at V_O sites. This was achieved by electron transfer from V_O to the chemisorbed oxygen as a result of the strong electronic coupling between them. Band excitation at the reduction level for free O₂/O₂^•- (-0.33 V vs. NHE) is not required in this process. Surprising stability was obtained owing to the recovery of surface V_O in the reaction between O₂^•- and neighboring activated benzylamine. The utilization of low energy photons is thus a promising strategy for the selective conversion of substrates using V_O.

The presence of nitrogen vacancies (V_N) is an intrinsic property of nonstoichiometric nitride photocatalysts, such as Ta₃N₅ [58, 59] and metal-free g-C₃N₄ [20, 22]. V_N-containing g-C₃N₄ was first produced by controlling the polycondensation temperature of dicyandiamide in an open container, and promoted the photodecomposition of RhB [22]. In this material, the number of short- and long-lived charge carriers was decreased and increased, respectively. This suppressed charge-carrier recombination was attributed to the trapping of charge carriers in the localized electronic states of V_N, and the accordingly reduced spatial overlap. Thermal treatment in a hydrogen or nitrogen atmosphere, sulfur doping, and microwave synthesis have also been reported to produce g-C₃N₄ with V_N [20, 21, 23, 60-62]. Compared with traditional modification of only the surface layer of photocatalysts, homogeneous modification features a complete change in the electronic structure of the photocatalyst, and can be realized by using layered nanostructured precursors [63]. Hydrogenation of the layered structural melon was used for the homogeneous self-modification of g-C₃N₄ with V_N, and this material had a greatly reduced band gap (2.03 eV) [20]. However, a reduced redox ability was expected owing to the band shifts in this modified g-C₃N₄. In addition to its enhanced performance in photocatalytic oxidation and water splitting, V_N-containing g-C₃N₄ also shows great potential in nitrogen fixation [23, 61, 62]. As well as promoting interfacial charge transfer from the catalyst to the adsorbed N₂, the V_N in g-C₃N₄ can also serve as selective chemical adsorption sites for the adsorption, activation and reduction of N₂ because of their similar properties to those of N₂-imprinted polymers [23]. The use of V_N-containing nitride photocatalysts with a negative conduction band may also be a promising strategy for the reduction of nitrogen-containing compounds.

Halogen vacancies (V_H) are easily generated in bismuth oxyhalides, which have unique layered structures consisting of positively charged[Bi₂O₂]²⁺ slabs and interleaved negative halogen slabs with nonbonding interactions along the c-axis [17, 19]. Because of the weak van der Waals interactions, halogen ions can diffuse out of the layered structure by calcination, or be replaced by other halogen ions through ion exchange in the solution or solid state. Generally, bismuth oxyhalides possess diverse halogen-deficient crystal phases, which can be obtained simply by tuning the precursor ratios or/and calcining at relatively high temperatures. However, we recently observed that Bi₇O₉I₃ containing iodine vacancies could be obtained by calcination below the phase transition temperature [17, 18]. The iodine deficiency was indicated by the decrease in the Bi/I atomic ratio from 2.36 to 2.80 after calcination at 350 ℃. It was found that shallow V_H improve the carrier transport properties, whereas deep V_H have the opposite effect [64]. Homogeneous modification of the bismuth oxyiodides with iodine vacancies is expected because of their layered structure, and the whole band gap increased from 2.04 to 2.31 eV after calcination at 350 ℃. Moreover, the enlargement of the band gap was confirmed by theoretical calculations, which also indicated the downward shift of the band position and the transition of the semiconductor behavior from p to n type. These changes are expected to increase the oxidation ability of the photogenerated holes, improve the electron mobility, and accordingly contribute to an enhanced photooxidation performance. However, although V_H can easily be created in layered structural bismuth oxyhalides, there are concerns about the stability of these compounds, especially in the aqueous phase, which greatly restricts their application in water purification.

There are very few reports on cation vacancies in the field of photocatalysis. In contrast to anion vacancies and cation interstitials, which are shallow donors, cation vacancies are shallow acceptors and always induce p-type conductivity. The benefits of V_Ti in TiO₂ are expected to include: (1) increase in the electrical conductivity of the ionic components; (2) increase in the mobility of electrons and interfacial transfer of charge carriers, which suppresses recombination and enhances the photocatalytic activity; (3) generation of room-temperature ferromagnetism; and (4) provision of surface active sites for water adsorption and formation of active complexes, and the corresponding enhanced photoactivity in water splitting [65-67]. Some experimental results have indicated that the band gap of TiO₂ is independent of the V_Ti concentration [26, 65]. Theoretical calculations have also suggested that V_Ti would not generate new states in the band gap. However, the presence of V_Ti in TiO₂ nanosheets caused a slight downward shift in the VB maximum and a considerable downward shift in the CB minimum; thus resulting in a net narrowed band gap [67].

It is difficult to generate stable cation vacancies in traditional n-type oxides, for example, V_Ti in TiO₂. Nowotny et al. [68] found that the transport kinetics of V_Ti in TiO₂ were much slower than those of V_O and titanium interstitials, and accordingly resulted in quasi-quenches of the V_Ti species. Therefore, V_Ti are expected to be restricted to the surface or near-surface regions, which could be one of the reasons for the difficulty in synthesizing V_Ti-containing TiO₂. However, p-type TiO₂ containing V_Ti can be engineered in the following ways: (1) generation of defect disorder by calcination above 1573 K [69]; (2) prolonged oxidation (p(O₂) = 75 kPa) at temperatures above 1273 K [27, 65, 70]; (3) incorporation of aliovalent dopant ions [71, 72]; and (4) controlled calcination of lamellated titanium glycerolate [26]. According to Rietveld refinement results, high concentrations of titanium defects were present in iron-doped TiO₂ [71] or TiO₂ with a particle size of several nanometers [73]. Moreover, the concentration of V_Ti increased with increasing iron content and decreased with increasing particle size, but remained stable when the vacancies were balanced by hydroxyl ions or protons in the lattice. Recently, Zou's group [26, 74] developed stable p-type ZnO and TiO₂ with cation vacancies by calcination of the glycerolate precursors. These p-type oxides are formed through the generation of oxygen-rich structures during the decomposition of interlayer organic groups at high temperatures, which creates distortion and thus generates cation vacancies in the lattice. This method represents a promising way of producing photocatalysts with cation vacancies. However, although the generation of V_Ti has been greatly improved and the benefits have been theoretically discussed, the characterizatio n of these vacancies, and their stability and effects in photocatalytic interfacial processes are far from clear.

In contrast to V_N-containing g-C₃N₄, g-C₃N₄ containing cation carbon vacancies (V_C) can be developed by thermally treating bulk g-C₃N₄ in an NH₃ or Ar atmosphere [28, 29]. In addition, V_C-induced room-temperature ferromagnetism of g-C₃N₄ has been confirmed [29]. The dramatically improved photocatalytic hydrogen evolution was attributed to the higher donor density and the notably longer lifetime of the charge carriers. Moreover, V_C-containing g-C₃N₄ displayed considerably higher H₂O₂ production than both pure g-C₃N₄ and V_N-containing g-C₃N₄ [28]. This was not only attributed to the enhanced optical absorption and higher donor density of V_C-containing g-C₃N₄, but also to its improved O₂ adsorption, the promotion of electron transfer for O₂ reduction, and most importantly, the transformation of O₂ reduction into a two-electron process, in which the amino groups around the V_C had an indispensable role. This suggests that cation vacancies in g-C₃N₄ have great potential for certain photocatalytic applications.

Bismuth oxide-based photocatalysts have attracted considerable attention, owing to their narrow band gap and high activity in the visible light range, as well as the relative abundance of elemental bismuth [18, 75]. Bismuth vacancies (V_Bi) are also ubiquitous in nonstoichiometric bismuth-oxide compounds [24, 25, 76, 77]. Various bulk defects, including V_O and V_Bi, were observed in ball-milled BiPO₄ [76]. These defects degraded the photocatalytic performance of BiPO₄, but were easily removed by reflux or calcination. Recently, a stable V_Bi-containing Bi₆S₂O₁₅ was developed by increasing the ratio of the sulfate to bismuth sources in the hydrothermal synthesis [25]. The resulting catalyst was characterized as core-shell nanowires with a disordered shell layer containing a high concentration of V_Bi. These vacancies contributed to the efficient separation of charge carriers, and reduced the size of the band gap by elevating the whole band and producing shallow surface states above the VB.

The components and concentration of vacancies in the surface lattice are facet-dependent; therefore, two-dimensional materials with different facets and thicknesses have different vacancy compositions. In general, V_Bi dominate on the surface of BiOCl nanoplates. However, when the thickness of the nanoplates was reduced to 2.7 nm, Bi³⁺-oxygen vacancy associates were observed to dominate on the surface of these ultrathin BiOCl nanosheets. This resulted in a strongly negatively charged surface and a smaller band gap, and the separation of photogenerated charges was promoted [24]. Specific facets of ultrathin photocatalytic materials are exposed on light irradiation or chemical reaction; this could provide a delicate strategy for tuning the available vacancies for photocatalytic applications.

The latest progress in this field provides a greater fundamental understanding of the relationship between a diverse range of vacancies and their related photocatalytic properties, and the growing knowledge has been successfully applied to realize advanced photocatalytic processes for water splitting, N₂ fixation, H₂O₂ production, decontamination, organic synthesis, and CO₂ reduction, among others. Vacancy-rich photocatalysts display extended optical absorption, facilitate charge separation and enhance the selective chemisorption of reactants, which all result in improved photoreactivity. However, further efforts are required to deal with the following open problems or challenges. (1) More powerful characterization techniques are needed to reveal the existence, format and concentration of the vacancies. This is essential for the understanding the association between vacancies and photocatalytic performance. Currently, we cannot easily discern the different types of vacancies, let alone the format of the vacancies in the lattice surface, subsurface or bulk. (2) Current knowledge on the formation chemistry and fundamental functions of anion or cation vacancies in photocatalytic processes is insufficient and far from clear, especially for those photocatalysts containing both donor and acceptor vacancies [24, 30]. (3) It is desirable that vacancies in photocatalysts retain their function and remain unchanged for long periods of time. However, although we can obtain stable vacancy-rich compounds, and we understand the dependence of stability on the vacancy format according to theoretical calculations [78], we do not yet know the exact factors that control the stability of vacancies, or how to manipulate the raw materials and synthetic methods to tune the format of the vacancies. It is expected that increasing attention on vacancies in photocatalysts and the development of vacancy-engineering techniques will mean that their use in solar photocatalytic applications will be greatly accelerated in the near future.