**3. Titanium Dioxide: Synthesis and Surface Modification Strategies for Enhanced CO2 Photoreduction**

As shown in Figure 7, TiO2 exists as one of three mineral phases: anatase, brookite, and rutile. Anatase-phase TiO2 is the most common phase of TiO2, but has a band gap of 3.2 eV, and is therefore weakly active under visible light. Rutile-phase, with a band gap of 3.0 eV, has the strongest visible light absorption, whereas brookite (band gap = 3.3 eV) has the weakest [65]. The use of mixed-phase TiO2 enhances both the visible light harvesting ability and the electron–hole separation of TiO2 [66]. The crystalline phase of TiO2 is highly dependent on its preparation technique.

**Figure 7.** Crystal structures of TiO2. Reproduced from work in [67].

The sol–gel method is widely used among researches for the preparation of bare and doped TiO2-based photocatalysts both as thin films or powders [68]. This is attributed to the many benefits of this simple and cost effective method, which include the synthesis of nano-sized crystallized powder of high purity and homogeneity at relatively low temperature, possibility of tuning stoichiometry and morphology and easy preparation of composite materials [69,70]. Other photocatalysts such as ZrO2 [71], ZnO [72], and WO3 [73] have also been fabricated via the sol–gel process.

In general, the one-pot sol–gel procedure, shown in Figure 8, consists of two processes: hydrolysis and condensation. During these processes, the metallic atoms of the precursor molecules bind to form metal oxides and metal hydroxides. In the case of TiO2, titanium alkoxides, such as titanium isopropoxide or titanium n-butoxide, are used as the precursor. Alcohol and acid must also be introduced into the reaction system as reaction modifiers. A densely cross-linked 3D TiO2 gel with large specific surface area is formed as an end product after allowing the mixture of precursor, alcohol and acid to stir for several hours [70]. The choice of reaction parameters including reactants molar ratio, solution pH, reaction temperature, and reaction time, significantly affect the morphology of the final catalyst [74,75]. For instance, different reactant molar ratios would result in different hydrolysis rates which would in turn return structurally different TiO2 catalysts [76,77]. Another example is the choice of acid where using acetic acid in the sol–gel synthesis of TiO2 was shown to favor the formation of anatase-phase TiO2, whereas using hydrochloric acid favored the formation of brookite and rutile mixed-phase TiO2. Post-synthesis treatment techniques such as aging, drying and annealing are sometimes utilized to enhance the activity of the synthesized TiO2 photocatalyst [78]. To develop TiO2 films, the viscous sol may be deposited on a substrate via film coating techniques [79,80].

**Figure 8.** Sol–gel method for the preparation of bare and doped TiO2-based photocatalysts.

The hydro/solvo thermal method is another one-pot TiO2 synthesis technique that takes place in an aqueous solution above room temperature and atmospheric pressure [81]. More specifically, the TiO2 precursor (titanium alkoxides) is mixed with an aqueous solution of an acid or a base for 16 to 72 h at high reaction temperature (110–180 ◦C). Post-synthesis treatment, which includes washing

the obtained precipitate with DI water and then dispersing it in HCl solution before calcination, is usually performed to enhance the nanostructure of TiO2. Quenching, or rapid cooling, is also commonly applied post-synthesis to enhance the photocatalytic activity of TiO2 [82]. The hydro/solvo thermal process yields highly pure and well-defined TiO2 nanocrystals with narrow particle size distribution [70]. Depending on the synthesis process parameters, which include the choice of precursor, the concentration of acidic/alkaline solution and the reaction temperature and time, different TiO2 morphologies may be obtained. This ability to control the synthesis process to obtain TiO2 as nanoparticles, nanotubes, nanoribbons, or nanowires is a great advantage of the hydro/solvo thermal treatment technique [83–85].

A number of studies have reported the significant CO2 photocatalytic conversion improvement of TiO2-based catalysts when synthesized under supercritical conditions. Camarillo et al. [86] prepared TiO2 catalysts in supercritical CO2 via a hydrothermal method. The synthesized TiO2 catalyst was used for the photoreduction of CO2 to methane, where the catalyst exhibited better photoactivity than the standard reference catalyst Evonik P-25. When prepared in supercritical CO2, TiO2 displayed improved CO2 adsorption, enhanced charge separation and stronger visible light absorption. The highest methane production rate of 1.13 μmol/g-cat/h was obtained when diisopropoxititanium bis(acetylacetonate) precursor and isopropyl alcohol were used in the catalyst preparation.

The choice of precursor significantly affects the photocatalytic property of the synthesized catalyst. Bellardita et al. [87] prepared TiO2 photocatalysts via the hydrothermal method using two different precursors, titanium tetrachloride and titanium butoxide. The TiO2 photocatalysts were tested for the reduction of CO2 and results showed that the photocatalysts synthesized from titanium tetrachloride favored the formation of formaldehyde, whereas those synthesized from titanium butoxide favored the formation of methane. The same study also examined the use of copper-doped TiO2 for the photoreduction of CO2. Results showed that doping TiO2, synthesized with titanium tetrachloride precursor, with 1 wt.% copper enhanced the formaldehyde production rate. However, doping TiO2, synthesized with titanium butoxide precursor, with 1 wt.% copper decreased the methane production rate.

The catalyst morphology plays a crucial role in determining its photocatalytic efficiency. TiO2 nanostructures (nanosheets, nanorods, nanowires, nanotubes, nanobelts, etc.) have been shown to exhibit superior performance for photocatalytic reduction of CO2 [26]. These nanostructures provide large surface area, reduced grain boundaries, and facile charge transport paths. Specifically, reduced grain boundaries and defects usually lead to a direct pathway for electron transport to catalytic sites, inhibiting the electron–hole recombination rate.

The use of unmodified TiO2 photocatalysts for the reduction of CO2 under UV light has been reported by many to be inefficient [37]. This is mainly attributed to the reduction potential (−0.5 V) of electrons in the TiO2 CB which is much lower (i.e., more positive) than the theoretical thermodynamic requirement for the single electron reduction of CO2 (−1.90 V). To drive any reduction process, the potential of the CB must be more negative than that of the reduction reaction [39]. This is true in the case of the multiple electron reduction of CO2 to methane (−0.24 V) or to methanol (−0.38 V). On the other hand, the co-presence of strong reducing electrons and free protons is detrimental as they may interact to produce molecular hydrogen. This may be avoided by introducing spatially separated centers within the TiO2 lattice in order to trap the charges and reduce their recombination [37].

Several modification strategies have been suggested to help overcome the drawbacks of TiO2. Many of these strategies are shown in Figure 9 and they include metal deposition, metal/non-metal doping, loading of carbon-based material, formation of semiconductor heterostructures, and dispersion on high surface area supports. All of the aforementioned techniques serve to enhance the separation of electrons and holes and to extend the absorption of light into the visible range. Loading TiO2 with carbon-based materials and dispersing TiO2 on inert supports offers the added advantages of increasing the concentration of surface electrons, improving the surface adsorption of reactants (specifically CO2), and reducing the agglomeration of TiO2 nanoparticles. Nonetheless, the utilization

of carbon-based materials hinders the absorption of light while high surface area supports lower the light utilization efficiency since they absorb and scatter part of the radiation resulting in a waste of photons. Although TiO2-based heterostructures are complex to synthesize and are considered to be relatively unstable, they effectively separate oxidation and reduction sites improving the photocatalytic performance of TiO2 semiconductors. To overcome limitations and drawbacks, many researches have suggested to employ hybrid systems which combine two or more of these different modification techniques [40]. The advantages and disadvantages of the main modification strategies along with the performance results from the most important CO2 photoreduction studies are summarized in Tables 1 and 2 below [19,26,37,88–90]. Further details on the pros and cons of the mentioned TiO2 surface modification techniques are discussed in the following individual Sections 3.1–3.5

**Figure 9.** Photocatalyst modification strategies. Reproduced from work in [40]. Copyright 2016 Elsevier B.V.


**Table 1.** Advantages and disadvantages of TiO2 modification techniques.


**Table 2.** Photocatalytic performance of various TiO2-based catalysts in the reduction of CO2\*.

\* Abbreviations: NP: nanoparticle; NW: nanowire; NR: nanorod; NT: nanotube; CVD: chemical vapor deposition; HT: hydrothermal; ST: solvothermal; IMM: immersion; IMP: impregnation; MWCNT: multiwall carbon nanotubes; UV: ultraviolet.

#### *3.1. Metal Deposition*

Although considered an expensive modification technique, metal deposition enhances the electron–hole separation in a semiconductor, improving the photoreduction efficiency of CO2 [40,107]. As the work function of the metal increases, the metal's ability to accept the photogenerated electrons also increases [36,40]. This in turn enhances the electron–hole separation and the overall photocatalytic activity of TiO2. Some of the most commonly deposited metals arranged in order of highest work function include platinum (5.93 eV), palladium (5.60 eV), gold (5.47 eV), and silver (4.74 eV) [25,40].

Wang et al. [91] demonstrated high CO2 photoreduction efficiency of one-dimensional (1D) TiO2 single crystals coated with ultrafine Pt nanoparticles. This efficient Pt–TiO2 nanofilm, shown in Figure 10, was synthesized via chemical vapor deposition and exhibited selective formation of methane with a maximum yield of 1361 μmol/g-cat/h. Deposition of Pt nanoparticles enhanced the electron–hole separation leading to a more efficient photocatalytic performance. Li et al. [36] also proposed the use of Pt-deposited TiO2 for the photoreduction of CO2 to methane. Results showed that depositing 0.2 wt.% Pt on mesoporous TiO2 yields methane with a production rate of 2.85 μmol/g-cat/h. Tostón et al. [108]

examined the photocatalytic activity of 1 wt.% Pt–TiO2 photocatalysts prepared in supercritical CO2. The synthesized photocatalyst was used to reduce CO2 into methane. The study indicated that the use of supercritical CO2 results in a photocatalyst with higher surface area, crystallization degree, pore volume, visible light absorbance, and methane production rate (0.245 μmol/g-cat/h).

**Figure 10.** Enhanced CO2 photoreduction efficiency by ultrafine Pt nanoparticles deposited on TiO2 crystals. Reproduced from work in [91]. Copyright 2012 American Chemical Society.

Owing to the effect of localized surface plasmon resonance (LSPR) and to that of Schottky barrier formation, Au and Ag nanoparticles have been shown to enhance the visible-light activity of several semiconducting materials. In LSPR, an electromagnetic field is created and the photoreaction is improved via photon scattering, plasmon resonance energy transfer and hot electron excitation. On the other hand, the formation of a Schottky barrier enhances the photoactivity by trapping and prolonging the electron life [109]. Tahir et al. [92] described the photoreduction of CO2 using Au-decorated TiO2 nanowires, prepared by hydrothermal method. Deposition of 0.5 wt.% Au on the TiO2 nanowires yielded 1237 μmol carbon monoxide/g-cat/h and 12.65 μmol methanol/g-cat/h. Surface deposition of Au nanoparticles enhanced charge separation and improved photocatalytic activity under visible light through plasmon excitation. More specifically, the LSPR effect of Au nanoparticles promotes electrons to the CB of TiO2. The positively charged plasmas of Au nanoparticles then trap photogenerated CB electrons enhancing the separation of charges. Next, the LSPR effect of Au improves the energy of these trapped electrons and, as a result, the efficiency of CO2 photoreduction increases [109]. Kong et al. [93] proposed the use of Ag-electrodeposited TiO2 nanorods, prepared by hydrothermal method, for the photocatalytic reduction of CO2 to methane. The photocatalyst exhibited enhanced photoactivity with a methane yield of 2.64 μmol/g-cat/h. This improved performance was attributed to the plasmonic characteristics of the Ag nanoparticles.

Murakami et al. [110] photocatalytically reduced CO2 using Ag/Au-TiO2. The deposited metal nanoparticles act as reductive sites increasing the production of methanol compared to that of bare TiO2. Zhai et al. [94] photodeposited Cu and Pt nanoparticles on TiO2 (Evonik P-25) using copper sulfate and chloroplatinic acid, respectively, as the metal precursors. Results showed the photo-depositing TiO2 with 1.7 wt.% Cu and 0.9 wt.% Pt allowed to produce methane from CO2 with a yield of 33 μmol/g-cat/h. Yan et al. [95] used a sol–gel method to synthesize a novel Cu/C–TiO2 catalyst for the photoreduction of CO2 in water under UV irradiation. Co-deposition of Cu and C onto the TiO2 surface extended the light absorption into the visible range, reduced the electron–hole recombination rate and provided an increased number of reaction sites on the photocatalytic surface. The enhanced Cu/C–TiO2 photocatalyst reduced CO2 to methane with a production rate of 2.53 μmol/g-cat/h.

#### *3.2. Doping*

One of the many strategies used to enhance the activity of TiO2 photocatalysts is doping with metals such as copper, nickel, silver and cerium. This technique is the most widely applied method used to extend the light absorption range and suppress the recombination rate of TiO2. Although introducing metal nanoparticles into the TiO2 matrix leads to structural defects, their presence reduces the TiO2 bandgap, shifting the absorption threshold to visible [40]. The reduction in bandgap occurs as a consequence of the new energy level produced by the dispersion of dopants in the TiO2 lattice [111]. Not only that, but the doped metal may also act as an electron trap, enhancing the separation of the photogenerated electrons and holes during irradiation as illustrated in Figure 11. One major drawback of metal doping is the possibility of metal leaching and, consequently, catalyst deactivation. This might occur as a result of the photocorrosion of doped TiO2, especially when water is used as a reductant [37].

**Figure 11.** Electron–hole separation of metal-doped semiconductors. Reproduced from work in [112]. Copyright 1995 American Chemical Society.

To prepare metal-doped TiO2, the sol–gel method is commonly followed due to its versatility and simplicity. In a typical process, a TiO2 precursor is mixed with a metal-dopant precursor which has been dissolved in alcohol. After hydrolysis for several hours and at elevated temperatures, the reaction mixture is dried and a metal-doped catalyst in powder form is obtained [111]. Chemical vapor deposition and hydro/solvo-thermal techniques may also be used for the synthesis of metal-doped TiO2.

One of the most commonly used metal dopants incorporated into TiO2 semiconductors is copper. Different metal loading levels of Cu-doped TiO2 were tested for the photocatalytic reduction of CO2 under UV light. For instance, Wu et al. [96] reduced CO2 to methanol using a Cu-doped TiO2 catalyst under UV irradiation. The catalyst was synthesized via a hydrothermal method and exhibited a band gap of 3.3 eV. Doping TiO2 with 1.2 wt.% copper produced methanol with a yield of 0.45 μmol/g-cat/h. Tseng et al. [58] reported the photocatalytic reduction of CO2 into methanol under UV irradiation using copper-doped TiO2, prepared using the sol–gel method. The results of this study showed that doping the TiO2 catalyst with 2 wt.% copper gave the highest methanol yield of 12.5 μmol/g-cat/h. The copper doping lowered the electron–hole recombination probability, which consequently boosted the photocatalytic efficiency. Nasution et al. [97] studied the use of Cu-doped TiO2 (Evonik P-25), prepared by an impregnation method, for the photocatalytic reduction of CO2 under UV irradiation. Results showed that doping TiO2 with 3 wt.% copper yields a maximum methanol production rate of 194 μmol/g-cat/h. All of the aforementioned studies suggest that the photocatalytic reduction of CO2 using a Cu-doped TiO2 catalyst under UV illumination yields methanol as the main reaction product.

Due to the LSPR effect of copper nanoparticles, Cu-doped TiO2 has been shown to exhibit enhanced photocatalytic activity under visible light. The strong local electron field from the LSPR effect improves the energy of the trapped electrons resulting in an enhanced photocatalytic CO2 reduction process. The concentration of copper dopant, however, must be carefully considered. Although excess copper nanoparticles improve the separation of charges, they also lead to the formation of larger copper particles. These larger particles have a weaker LSPR effect and as such considerably lower the photocatalytic efficiency [113].

The metal dopant concentration significantly affects the photocatalytic activity of doped TiO2. Depending on its concentration, the metal dopant can either hinder or promote the anatase-to-rutile transformation during calcination. The crystal structure transformation to rutile reduces the catalyst's surface area and removes any active species present on the anatase surface, thereby lowering the catalytic performance of doped TiO2 [114]. To maintain high anatase levels and therefore better photocatalytic activity, low concentrations of metals (0.1–0.5 mol %) are commonly doped into TiO2. In addition to the structural transformation, the dopant loading level also influences the recombination of photogenerated charges. By acting as recombination centers, metal dopants, can significantly increase the electron–hole recombination especially when present at high concentrations (>3 mol %), therefore lowering the photocatalytic efficiency [115]. Kwak et al. [98] investigated the photocatalytic reduction of CO2 to methane using a nickel-doped TiO2 catalyst, prepared via a solvothermal method. Doping with 0.1 mol % nickel gave the highest methane yield of 14 μmol/g-cat/h. Matˇejová et al. [99] examined the photoactivity of cerium-doped TiO2, synthesized by the sol–gel method, for the reduction of CO2. The doped cerium shifted the light absorption of the catalyst into the visible range and enhanced charge separation. Results showed that doping TiO2 with 0.28 mol % cerium gave the highest methane yield of 0.889 μmol/g-cat/h.

Nonmetal dopants, such as nitrogen, carbon, sulfur, and fluoride, are commonly used to help reduce the TiO2 band gap, consequently increasing the absorption of visible light [26]. Substitutional nonmetal doping typically introduces defect states localized at the impurity site which reduce the band gap and induce absorption in the visible region. Furthermore, these defect states, which form below the conduction band, have been shown to be good acceptors of electrons. Nonetheless, nonmetal doping also leads to the formation of oxygen vacancies. These oxygen vacancy defects generally act as charge recombination centers which are detrimental in photocatalytic reactions and must therefore be avoided [116]. As the quantity of nonmetal dopant increases, the amount of defects increases and the photocatalytic activity consequently decreases. Therefore, in nonmetal doping, extra care must be taken in optimizing the dopant concentration for enhanced visible light absorption and improved photocatalytic activity with an acceptable extent of defects [117]. One strategy commonly used to reduce the recombination of charges in nonmetal doped TiO2 is co-doping with an electron donor–acceptor pair [118]. Similar to metal-doping, non-metal doping is usually performed via the sol–gel process, although other techniques including hydro/solvo-thermal methods may also be applied.

Studies have shown that the main reaction product resulting from the photocatalytic conversion of CO2 using non-metal dopants is formic acid. For example, Zhao et al. [100] studied the photocatalytic activity of nitrogen-doped TiO2 nanotubes prepared by a hydrothermal method, where titanium (III) chloride and hexamethylene tetramine were used as precursors. As a doping agent, nitrogen extended the light absorption of TiO2 into the visible range. The formic acid yield was 1039 μmol/g-cat/h, the methanol yield was 94.4 μmol/g-cat/h, and the formaldehyde yield was 76.8 μmol/g-cat/h. Xue et al. [101] investigated the non-metal, substitutional doping of TiO2 with carbon using citric acid as a precursor. The doped carbon lowered the TiO2 band gap, shifting the light absorption into the visible range, and improved the charge separation efficiency. CO2 was photocatalytically reduced to yield 439 μmol formic acid/g-cat/h.

### *3.3. Carbon-Based Material Loading*

The incorporation of carbon-based materials into TiO2 catalysts is gaining wide interest in the field of photocatalysis. This is mainly due to the carbon-based materials' properties such as their abundancy, low cost, good corrosion resistance, high electron conductivity, large specific surface area, and tunable surface properties. Although in most cases they reduce the light absorbed by TiO2, loading the catalyst with carbon based-materials, such as carbon nanotubes or graphene, enhances the electron–hole separation, the electron concentration on the TiO2 surface, and the CO2 adsorption through π–π conjugations between CO2 molecules and the carbon-based material [40].

Carbon nanotubes are regarded to be among the most remarkable emerging materials mainly due to their unique electronic, adsorption, mechanical, chemical, and thermal characteristics [119]. Advances in the synthesis techniques of multiwalled carbon nanotubes (MWCNTs) have resulted in a significant reduction of the material's cost, consequently allowing for their use on a large industrial scale [103]. In addition to that, the mesoporosity of carbon nanotubes, which favors the diffusion of reacting species, is an added value of this widely investigated material [120]. Many recent studies have focused on developing synthesis methods to combine carbon nanotubes with semiconducting catalysts, such as TiO2. This new hybrid composite is believed to enhance the efficiency of many photocatalytic applications, especially with its improved charge transfer properties and the formation of new active sites [103,120]. The main disadvantage in synthesizing MWCNT/TiO2 hybrids is the need for treating the carbon nanotubes with strong acids in order to introduce the functional groups to the surface [120]. A range of different techniques has been employed for the fabrication of MWCNT/TiO2 composites: mechanical mixing [121], sol–gel [122], electrospinning [123,124], and chemical vapor deposition [125]. Depending on the choice of the synthetic strategy, composite materials with different coating uniformity and varying physical characteristics are prepared. More specifically, uniform TiO2 coatings on carbon nanotubes may be achieved through electrospinning and chemical vapor deposition, whereas other methods such as sol–gel typically yield nonuniform coatings with TiO2 aggregates being randomly dispersed on the carbon nanotube surface [126]. On the other hand, chemical vapor deposition and electrospinning are complex techniques which require the use of specialized equipment.

Using the process shown in Figure 12, Gui et al. [102] prepared a MWCNT/TiO2 nanocomposite for the photocatalytic reduction of CO2. The MWCNTs enhanced the TiO2 photoactivity under visible light, yielding 0.17 μmol methane/g-catalyst/h. Xia et al. [103] studied the effect of using two different methods of synthesis—sol–gel and hydrothermal—for the preparation of MWCNT-supported TiO2 composite. The sol–gel method formed anatase-phase TiO2 nanoparticles on the MWCNT, whereas rutile-phase TiO2 nanorods were formed when the hydrothermal method was used. In addition to that, the MWCNT/TiO2 composite prepared via sol–gel mainly reduced CO2 to ethanol (29.87 μmol/g-cat/h), whereas the composite prepared hydrothermally mainly reduced CO2 to formic acid (25.02 μmol/g-cat/h). The efficiency of the CO2 photoreduction reaction was significantly enhanced with the incorporation of MWCNTs. More specifically, the MWCNTs helped reduce the agglomeration of TiO2 nanoparticles and helped increase the separation of the photogenerated electron–hole pairs. Ong et al. [104] demonstrated the photocatalytic activity of CNT-Ni/TiO2 nanocomposites in reducing CO2 to methane under visible light. The nanocomposite, prepared by chemical vapor deposition, exhibited a reduced band gap of 2.22 eV and a methane yield of 0.145 μmol/g-cat/h.

**Figure 12.** Enhanced visible light responsive MWCNT/TiO2 core–shell nanocomposites for photocatalytic reduction of CO2. Reproduced from work in [102]. Copyright 2013 Elsevier B.V.

Graphene is a 2-D, single layer of graphite with outstanding physiochemical properties including low production costs, ease of scalability, exceptional catalytic performance, high mechanical strength, high porosity, high thermal and electrical conductivity, remarkably high CO2 adsorption capacity, high transparency, high flexibility, and high specific surface area. Furthermore, the exceptional photocatalytic properties of graphene, which include zero band gap, large BET area and high electron mobility, have encouraged its use in many photocatalytic applications [127]. Many different types of semiconducting photocatalysts have been coupled with graphene with the aim of enhancing the overall efficiency of photocatalysts [128,129]. The preparation process greatly affects the morphology, structure, size, properties, and activity of the composite photocatalyst [130]. The methods commonly used for the preparation of graphene-based TiO2 are generally divided into two main categories: in situ crystallization (including hydrothermal/solvothermal, sol–gel, microwave assisted and others) and ex situ hybridization (including solution mixing, self-assembly, electrospinning and others). It is important to note that molecular linkers between TiO2 and the graphene sheets are not required during the synthesis of the composite [131]. This is a great benefit since molecular linkers might act as electron traps decreasing the overall photocatalytic efficiency of the graphene-based TiO2 composite.

In general, both the hydrothermal and the solvothermal processes are characterized by their high reactivity, low energy requirement, mild reaction conditions, relatively environmental set-up, and simple solvent control. Under optimal conditions, both techniques can yield large quantities of graphene-based photocatalysts at low cost [129]. Furthermore, TiO2 nanostructures with high crystallinity are typically produced through the one-pot hydrothermal/solvothermal approach without the need for post-synthesis calcination [132]. Experimental conditions, including precursor concentration, pH, temperature, and time, significantly affect the reaction pathway and the nanomaterial's crystallinity and, therefore, must be carefully considered [129,132].

In sol–gel techniques and through a series of hydrolysis and condensation steps, a liquid colloidal solution "sol" containing graphene/graphene oxide and TiO2 precursor is transitioned into a xerogel [129,132]. The hydroxyl groups present on the surface of the graphene oxide sheets provide nucleation sites for hydrolysis [129]. Consequently, the TiO2 nanoparticles will be strongly bonded to graphene, offering a great advantage over other synthesis methods [133].

Solution mixing is fairly simple and involves only the mixing of a TiO2 precursor in a suspension of graphene oxide under vigorous/ultrasonic agitation. Simple treatments, including drying and calcination, may be utilized after synthesis to improve the photocatalytic efficiency of the final composite material [130]. Furthermore, graphene oxide may be reduced to graphene sheets by the addition of agents such as amines, sodium borohydride, and ascorbic acid [129]. It is important to note that detrimental defects in the graphene sheets may arise as a result of long exposure time and high power ultrasonication. In addition to that, formation of chemical bonds between TiO2 and graphene may be difficult due to the mild operating conditions [130]. Nonetheless, this technique provides the benefit of uniformly distributing the TiO2 nanoparticles over the graphene sheets, ultimately improving the photocatalyst's activity [129].

Self-assembly is also another highly efficient, time-saving and cost effective synthesis technique [134,135]. This method offers structure and size control via component design and is applicable in various fields. The major disadvantages include the need for surfactants, sensitivity to environmental factors and production of graphene with low mechanical strength [136]. Surfactants are required in order to help disperse graphene oxide sheets and in order to improve TiO2 nanoparticle loading [132]. Self-assembly based on the electrostatic attraction between negatively charged graphene oxide sheets and positively charged TiO2 nanoparticles has been used to synthesize layered graphene-based TiO2 semiconductors in a simple and cost-effective manner [132,137].

Tu et al. [138] loaded TiO2 nanoparticles onto graphene nanosheets via an in situ method. The addition of graphene significantly increased the specific surface area of the catalyst, consequently generating more adsorption and reaction sites on the catalytic surface. The 2 wt.% graphene/TiO2 catalyst was used to reduce CO2 to ethane with a yield of 16.8 μmol/g-cat/h. Ong et al. [105] deposited nitrogen-doped TiO2 nanoparticles onto graphene sheets using a solvothermal method. The nitrogen-doped TiO2/graphene catalyst reduced CO2 to methane with a yield of 0.37 μmol/g-cat/h. The efficient charge separation of graphene and the enhanced absorption of visible light were attributed to the improved photocatalytic performance of TiO2.

#### *3.4. Heterostructures*

Sensitizing wide bandgap semiconductors with narrow bandgap semiconductors or dye molecules to form heterostructures is another method used to improve the photoactivity of many photocatalysts. This strategy provides means of increasing absorption toward the visible light region, at the same time enhancing the electron–hole pair separation, setting apart the reduction and oxidation sites for an improved performance [40,139]. Examples of narrow bandgap semiconductors that have been coupled with TiO2 for photocatalytic reduction of CO2 under visible light irradiation include CdS [140], Bi2S3 [141], CdSe quantum dots (QDs) [142], PbS QDs [143], and AgBr [144]. To enhance the adsorption of CO2 molecules, materials with high intrinsic basicity, such as cobalt aluminum hydroxide, can be applied as sensitizers of TiO2 [145]. Four different types of heterostructures exist depending on the charge carrier separation mechanism: conventional type-II, p-n, direct Z-scheme and surface heterojunction. However, these mechanisms will not be further discussed as they are considered to be outside the scope of this review. More details may be found elsewhere [40,146,147].

To synthetize heterostructured photocatalysts, several fabrication methods, including chemical vapor deposition, atomic layer deposition, hydrothermal/solvothermal, and ion exchange reactions, have been developed [147]. Chemical vapor deposition allows for the sequential deposition of multiple materials on a substrate surface. This synthesis route follows a two-step growth procedure: (1) growth of inner core semiconductor on a suitable substrate, and (2) deposition of outer layer shell semiconductor on the core surface. Compared to chemical vapor deposition, atomic layer deposition is usually more favored due to its enhanced benefits which include precise control of film thickness at the atomic level and conformal growth of complex nanostructures [148,149]. Studies have also shown that atomic layer deposition enhances the light trapping and carrier separation properties of the photocatalytic heterostructure [150,151]. A more convenient preparation technique is the hydrothermal/solvothermal

method. Both the precursor dissolution and the reaction rate are enhanced by this process. Ion exchange is another approach used to synthesize photocatalytic heterostructures. In this novel process, cations at the interface of the two semiconducting materials are exchanged [152,153]. Without changing the shape of the ionic semiconductor, ion exchange reactions selectively yield a dimer structure with an epitaxial hetero-interface [147].

Li et al. [141] prepared a Bi2S3/TiO2 nanotube heterostructure for the photocatalytic reduction of CO2 into methanol. The incorporation of Bi2S3 enhanced the visible light absorption and the photocatalytic performance of the TiO2 nanotubes. The methanol yield of the Bi2S3/TiO2 nanotube heterostructure was reported to be 44.92 μmol/g-cat/h. Qin et al. [38] investigated the conversion of CO2 to methyl formate on a CuO-TiO2 heterostructured composite in methanol. The photocatalytic activity was greatly enhanced due to the heterojunction between CuO and TiO2. More specifically, the surface-junction improved the transfer of charges by facilitating the migration of electrons from the TiO2 CB to the CuO VB as depicted in Figure 13a. Consequently, the probability of charge recombination was lowered and the photocatalytic efficiency was enhanced. As can be seen in Figure 13b, the TiO2 catalyst loaded with 1 wt.% CuO exhibited the highest methyl formate yield of 1602 μmol/g-cat/h. One of the most significant approaches was carried out by Nguyen et al. [154], who employed a metal doped TiO2 catalyst sensitized with ruthenium dye (N3 dye) for the photoreduction of CO2 into methane. The methane yield of the N3-Dye-Cu (0.5 wt.%)-Fe (0.5 wt.%)/TiO2 was 0.617 μmol/g-cat/h. The improved photoreduction of the dye-sensitized TiO2 is attributed to the full absorption of visible light by the N3-dye along with the efficient charge transfer in the N3 dye-TiO2 system.

**Figure 13.** Photocatalytic conversion of CO2 to methyl formate (MF) over CuO–TiO2. (**a**) Band positions of CuO and TiO2. (**b**) Formation rate of MF for the different catalysts. Reproduced from work in [38]. Copyright 2010 Elsevier Inc.

#### *3.5. Dispersion on Supports*

The dispersion of TiO2 as a nanoparticle on various types of supports enhances the catalyst's product selectivity, pore structure, and electronic properties. In addition to that, this modification technique eliminates the need for post treatment separation and provides high surface area and mass transfer rate. An ideal supported-TiO2 photocatalyst must have effective light absorption properties, be resistant to degradation induced by the immobilization technique and provide firm adhesion between the support and the catalyst. Key challenges of this strategy include mass transfer limitations and low light utilization efficiency. TiO2 photocatalysts may be immobilized by dip or spin coating onto substrates such as fibers, membranes, glass, monolithic ceramics, silica, and clays. In dip coating, the thoroughly cleaned supports are immersed in a coating precursor solution as is demonstrated in Figure 14a. After being slowly pulled out of the solution, the coated support is then dried out to remove excess solution and moisture [155]. The withdrawal speed of the substrate, number of coating cycles and the TiO2 solution viscosity determine the catalyst film thickness [23]. In spin coating, the

precursor is deposited on the substrate by a dispenser while the substrate is rotated as demonstrated in Figure 14b. The rotation continues until uniform distribution of the precursor layer is achieved. Although the spinning process will result in the partial evaporation of the precursor solvent, the substrate will still require additional thermal treatment to stabilize the layer. The thickness of the films obtained by spin coating depends on the concentration of the precursor solution and the rotation speed [156]. Due to its scale-up applicability and higher controllability, dip coating is usually preferred over the spin coating technique [157]. In addition to that, superior structural and optical properties, such as enhanced crystallinity and higher average particle size, have been observed for dip-coated films in comparison to spin-coated ones [158,159].

**Figure 14.** Dip-coating (**a**) and spin-coating (**b**) of TiO2 photocatalyst on an inert support. Reproduced from work in [155,160]. Copyright 2017 Elsevier Ltd., 2016 Elsevier B.V.

Bellardita et al. [87] investigated the photocatalytic activity of silica-supported TiO2 for the reduction of CO2. The primary product obtained from the photocatalytic reaction was acetaldehyde. Li et al. [106] prepared an ordered mesoporous silica-supported Cu/TiO2 nanocomposites via a sol–gel method for the photocatalytic reduction of CO2 in water under UV irradiation (Figure 15a). As previously discussed, one of the many benefits of sol–gel synthesis concerns the control of pore size and the pore size distribution of the final catalyst. The sol–gel method used to prepare the silica-supported Cu/TiO2 was carefully formulated to produce ordered mesoporous silica with a pore size of ~15 nm [161]. The ordered mesoporous structure of silica was clearly observed in transmission electron microscopy (TEM) images as shown in Figure 15c,d. Mesoporous silica, with its ordered pore networks, has a much higher surface area than agglomerates of silica nanoparticles with irregular mesopores and is therefore considered a better support for catalysts. The high surface area of the ordered mesoporous silica support (>300 m2/g) enhanced the dispersion of TiO2 nanoparticles and improved the adsorption of reactants on the catalytic surface. The main product of the CO2 photoreduction reaction was carbon monoxide when bare TiO2 was supported on silica. However, when Cu/TiO2 was supported on silica, the production rate of methane significantly increased. The deposition of Cu enhanced the separation of charges and improved the kinetics of the multi-electron reactions, consequently increasing methane selectivity. As can be deduced from Figure 15b, the 0.5 wt.% Cu/TiO2-Silica composite produced carbon monoxide and methane with a yield of 60 and 10 μmol/g-cat/h, respectively.

**Figure 15.** Photocatalytic reduction of CO2 on ordered mesoporous silica supported Cu/TiO2. (**a**) Schematic representation of experimental set-up; (**b**) production rates of CO and methane; (**c**) transmission electron microscopy (TEM) image of the 0.5 wt.% Cu/TiO2-Silica composite; (**d**) high resolution TEM image of a single TiO2 nanoparticle. Reproduced from work in [106]. Copyright 2010 Elsevier B.V.

#### **4. Conclusions and Future Perspective**

In this review, recent advances in the area of CO2 photoreduction have been discussed and the basic mechanism of CO2 photoreduction, particularly with water, has been described. Limited studies exist on the reaction mechanism and kinetics of CO2 photoreduction as a result of the complexity of the reaction which is due mainly to the possibility of forming various products and the presence of many reaction pathways. In situ techniques, such as FTIR and NMR, would help investigate crucial reaction kinetic parameters such as active sites, electronic states and reaction intermediates. Additionally, various modification strategies of TiO2 that may help overcome the limitations of current CO2 photoreduction technologies have been described. A comparative assessment between the different TiO2-based catalysts that have been reported in literature so far for CO2 photoreduction was made. This overview of the latest research trends in CO2 photoreduction may be used as a tool to help develop low-cost and highly-efficient TiO2-based semiconductors for an improved photocatalytic reduction of CO2.

Regardless of the significant theoretical advancements, the practical application of TiO2 photocatalysts for the photoreduction of CO2 is still far from being attained. This is mainly due to the lack of stable and visible light active photocatalysts. Researchers must focus on engineering a photocatalytic nanocomposite with an optimum band gap that simultaneously enhances charge separation and visible light absorption. A few suggested strategies include coupling wide band gap semiconductors with narrow ones, introduction of oxygen vacancies and defect sites and functionalization of the TiO2 surface. To enhance the adsorption and reduction of CO2 especially in the presence of water, a basic and hydrophobic TiO2 surface must be developed. The latest studies clearly demonstrate the considerable improvement in the yield and efficiency of the CO2 photoreduction process with modified TiO2 catalysts. More specifically, the major limitations of TiO2, including weak visible light absorption, increased charge recombination, and low CO2 adsorption capacity, have been greatly overcome by synthesizing nanocomposites with exposed reactive sites and improved morphology.

Surface modifications to TiO2-based photocatalysts cannot solely suffice the technical and economic feasibility requirements of CO2 photoreduction applications, especially if they are to be applied on a large industrial scale. One promising solution could be to combine photocatalysis with other CO2 conversion methods such as thermochemical and/or electrochemical catalysis. Another suggested solution could be to couple modified TiO2 with other visible-light active photocatalysts with the aim of improving the photocatalytic process under solar irradiation. Further enhancement might be attained if compressed CO2 (liquid or supercritical) is used for the photocatalytic reduction process. Although this increases the CO2 conversion efficiency, the overall cost will be considerably high due to intensive compression requirements. In this case, industries already utilizing compressed CO2 must be targeted.

**Funding:** We acknowledge financial support from Khalifa University of Science and Technology through the Research and Innovation Center on CO2 and H2 (RICH) (grant RC2-2019-007). We also acknowledge financial support from Abu Dhabi Education and Knowledge through the Award for Research Excellence (grant AARE17-8434000095).

**Conflicts of Interest:** The authors declare no conflict of interest.
