Insight into the Roles of Metal Loading on CO₂ Photocatalytic Reduction Behaviors of TiO₂

Abstract

The photocatalytic reduction of carbon dioxide (CO₂) into value-added chemicals is considered to be a green and sustainable technology, and has recently gained considerable research interest. In this work, titanium dioxide (TiO₂) supported Pt, Pd, Ni, and Cu catalysts were synthesized by photodeposition. The formation of various metal species on an anatase TiO₂ surface, after ultraviolet (UV) light irradiation, was investigated insightfully by the X-ray absorption near edge structure (XANES) technique. CO₂ reduction under UV-light irradiation at an ambient pressure was demonstrated. To gain an insight into the charge recombination rate during reduction, the catalysts were carefully investigated by the intensity modulated photocurrent spectroscopy (IMPS) and photoluminescence spectroscopy (PL). The catalytic behaviors of the catalysts were investigated by density functional theory using the self-consistent Hubbard U-correction (DFT+U) approach. In addition, Mott–Schottky measurement was employed to study the effect of energy band alignment of metal-semiconductor on CO₂ photoreduction. Heterojunction formed at Pt-, Pd-, Ni-, and Cu-TiO₂ interface has crucial roles on the charge recombination and the catalytic behaviors. Furthermore, it was found that Pt-TiO₂ provides the highest methanol yield of 17.85 µmol/g_cat/h, and CO as a minor product. According to the IMPS data, Pt-TiO₂ has the best charge transfer ability, with the mean electron transit time of 4.513 µs. We believe that this extensive study on the junction between TiO₂ could provide a profound understanding of catalytic behaviors, which will pave the way for rational designs of novel catalysts with improved photocatalytic performance for CO₂ reduction.

1. Introduction

The severe adverse effects of global warming resulting from excessive carbon dioxide (CO₂) emission arouses the need for urgent research into CO₂ reduction. CO₂ conversion to valued-added chemicals or fuels has gained enormous research interest as a game-changing technology for sustainable development [1,2,3,4,5]. Artificial photosynthesis, which mimics natural photosynthesis using renewable solar energy and water to convert CO₂ to manageable chemicals while leaving oxygen as a by-product, has been considered as one of the most green and sustainable technologies [6,7,8]. This method is also particularly attractive due to its ability to convert CO₂ to value-added hydrocarbons using ambient temperature and pressure [2,9]. Several semiconductors, including TiO₂ [10,11,12], CuO [13], g-C₃N₄ [14], Bi₂WO₆ [15], ZnO [16], SrTiO₃ [17], and CeO₂ [18], have been applied for photocatalytic reduction of CO₂. Among these semiconductors, anatase TiO₂ appears to be one of the most utilized catalysts due to its high performance, nontoxicity, high stability, and low cost [19,20,21]. However, rapid charge recombination is one of the important drawbacks, limiting the performance of TiO₂, and its large band gap also results in low CO₂ reduction efficiency [22].

Various strategies, such as surface modification, forming heterojunction and band alignment, and doping with metals and non-metals, have been reported as effective strategies to overcome these limitations and promote photocatalytic CO₂ reduction performance [3,23,24,25,26]. In particular, metal-TiO₂ composites have been shown to promote overall photocatalytic activity by reducing the recombination rate of the photogenerated charges and increasing light harvesting efficiency [27,28,29,30,31]. Generally, after irradiating the photocatalyst by incident light, the photogenerated electrons can transfer from the conduction band (CB) of photocatalysts across potential barriers to the contacting metal [32]. Therefore, metal acts as an electron sink for retarding the charge recombination rate. This can align the energy band between the metal and the semiconductor by shifting the Fermi level of the semiconductor to the metal located below the CB states of the semiconductor, and generating the semiconductor-metal heterojunction; namely the Schottky barriers [33]. The Schottky barrier effectively traps electrons, reducing the flow of electrons back to the semiconductor [23]. For example, Su et al. [3] studied the effect of Pd-loaded TiO₂ on CO₂ photoreduction. They found that the presence of Pd could enhance the CO₂-to-methane conversion by around two orders of magnitude compared to the bare TiO₂. It is well known that metal loading on semiconductors can enhance the photocatalytic CO₂ reduction performance [3,22,34,35,36,37], however the roles and underlying mechanisms of metals remain unclear. Some intrinsic challenges and critical factors, including surface molecular structures, charge transfer behaviors, and charge recombination rate during reduction, are also debatable. Moreover, insights into the interaction of adsorbed CO₂ with the semiconductor-modified surface as the catalytic sites are still expected to be further explored as the structure and the cation sites on the modified Ti surface composition are also involved in the catalytic pathways and selectivity of products. They can lower the reaction barrier to activate CO₂, and stabilize CO₂ intermediates to enhance CO₂ photoreduction.

In the present study, the roles of loading metals, including Pt, Pd, Ni, and Cu, prepared by photodeposition on TiO₂ towards the photocatalytic reduction of CO₂ are investigated in many aspects simultaneously (i.e., band alignment, plasmonic effects, charge recombination, charge transfer, and surface chemistry), in order to gain an insight into true catalytic behavior. The plasmonic metal-TiO₂ nanostructures and their compositions are extensively characterized by various techniques, including X-ray absorption near edge structure (XANES), X-ray diffraction (XRD), transmission electron microscopy (TEM), UV-visible diffuse reflectance spectra (UV-Vis), and inductively coupled plasma–optical emission spectroscopy (ICP-OES). The influences of energy band alignment of different heterojunctions, charge recombination behaviors, and photonic efficiency of metal-TiO₂ are insightfully studied by intensity-modulated photocurrent spectroscopy (IMPS) and photoluminescence spectroscopy (PL), compared to the pristine anatase TiO₂. The interactions of CO₂ on TiO₂ and Pt-TiO₂ photocatalysts were studied by CO₂-TPD, combined with theoretical simulation by density function theory (DFT). Interestingly, Pt-TiO₂ showed the best photocatalytic CO₂-to-methanol performance among the metals studied (Pt, Pd, Ni, and Cu) with a methanol production rate of 17.85 µmol/g_cat/h, which is among the top unassisted photocatalysts that have been reported for CO₂-to-methanol conversion. The impressive performance is attributed to suppressed charge recombination, suitable band alignment, and appropriate surface chemistry.

2. Materials and Methods

2.1. Metal Deposited-Semiconductor Preparation

Metal-deposited TiO₂ semiconductors were prepared by the photodeposition method at room temperature. Next, 0.2 g anatase TiO₂ (98% TiO₂, Loba Chemie PVT. Ltd., Mumbai, India) was suspended in 50 mL of aqueous 2-propanol solution (99.8% V.S. Chem House, Bangkok, Thailand) (50 vol%). The mixture was purged with N₂. Various metal salts; namely, H₂Cl₆Pt.6H₂O (37.50 wt%. Sigma-Aldrich, St. Louis, MO, USA), PdCl₂ (99.999 wt% Sigma-Aldrich, Steinheim, Germany), Ni (NO₃)₂.6H₂O (97 wt%. Sigma-Aldrich, Steinheim, Germany) and CuN₂O_6.3H₂O (99 wt%. Sigma-Aldrich, Steinheim, Germany) were used as a metal source for Pt, Pd, Ni, and Cu, respectively. 1 mL of 0.1 μM metal solutions was gradually added into the catalyst suspension. The photodeposition of metals on the semiconductor was carried out under UV illumination. The suspension was irradiated with a mercury lamp (125 W), with a main emission in the UV range at 365 nm and a light intensity of 3.42 mW/cm² under continuous stirring for 2 h. The obtained samples were precipitated by centrifugation (4500 rpm for 10 min), and washed with DI water for two cycles. Then, the samples were dried at 103 °C for 24 h. The total metal content of each sample was determined by inductively coupled plasma–optical emission spectroscopy (ICP-OES) (Perkin Elmer, AVIO 200, Waltham, MA, USA) (listed in Table S1 of supplement).

2.2. Characterization of Photocatalysts

Transmission electron microscopy (TEM) (HF-3300, Hitachi, Japan) was used to observe the morphologies of the as-synthesized catalysts. UV-visible diffuse reflectance spectra (UV-DRs) (UV-3101PC, Shimadzu, Japan) was used to analyze the band gap energy of the samples. The oxidation states and species of the metals deposited on the surface of anatase TiO₂ were investigated by X-ray absorption near edge structure (XANES). XANES measurements were carried out with the fluorescent mode at the beamline 1.1 W and beamline 8, Synchrotron Light Research Institute (SLRI), Nakhon Ratchasima, Thailand. The data reduction of XANES spectra was performed using ATHENA program. CO₂ temperature programmed desorption (CO₂-TPD, Chemisorption analyzer; ChemStar TPX, Quantachrome Instruments, Boynton Beach, FL, USA) was carried out to investigate the interaction of CO₂ and the catalyst. Photoluminescence (PL) (Avaspec-2048TEC-USB2-2, Apeldoorn, The Netherlands) and Intensity-modulated photocurrent spectroscopy (IMPS) (Metrohm Autolab, Utrecht, The Netherlands) were used to determine the charge dynamics and recombination. IMPS was obtained using a Metrohm Autolab PGSTAT12. The modulation frequency ranged from 120 to 500 kHz. The Mott–Schottky technique was used to classify the semiconductor types, and also used to estimate the flat band potential V_fb and band alignment of the composites. The Mott–Schottky was obtained using the frequency response analyzer (Metrohm Autolab PGSTAT204, Utrecht, The Netherlands) with an applied bias ranging from 1.5 to −1.0 V (vs Ag/AgCl), and the frequency of impedance was fixed at 1 kHz with the RMS amplitude of 10 mV.

The Mott–Schottky equation (Equation (1)) involves the relationship between the capacitance and the biased voltage across the semiconductor/electrolyte interface. The derived Mott–Schottky plots were fitted by using the simple linear regression method.

$$\frac{1}{C_{SC}^2}=\frac{2}{{\epsilon }_0{\epsilon }_{r}e{A}^2{N}_D}\left(V-V_{fb}-\frac{k_{B}T}{e}\right)$$

(1)

where $C_{SC}^2$ is the space charge capacitance density (F). $V$ is the applied potential (V). ${\epsilon }_0$ (F·m⁻¹) and ${\epsilon }_r$ are the vacuum permittivity and relative permittivity of TiO₂, respectively [38]. $e$ and $k_B$ are electron charge (C) and Boltzmann’s constant (m²·kg·s⁻²·K⁻¹), respectively. A and T are the active surface area (cm²) and the absolute temperature (K), respectively. $N_D$ is the electron carrier density or donor concentration (cm⁻¹). Corresponding to this equation, the $V_{fb}$ can be extracted from the intercept between the extrapolated linear line and x-axis. In addition, $N_D$ was also evaluated from the slope of the equation [39]. The position of the Fermi level (E_F) relative to the conduction energy (E_C) can be calculated by using Equations (2) and (3):

$$E_C-E_F=\frac{kT}{e}ln\left(\frac{N_C}{N_D}\right)$$

(2)

$$N_C=2{\left(\frac{2\pi m_e^{*}kT}{h^2}\right)}^{3/2}$$

(3)

where $N_c$ is the effective density of state in the conduction band (cm⁻³). $h$ is Plank’s constant (m²·kg·s⁻¹). The $N_c$ was calculated by setting $m_e^{*}$ as $10m_0$ [40]. Where the $m_e^{*}$ and $m_0$ are the density of state effective mass for the electrons of anatase TiO₂ and the mass of the free electron (kg), respectively.

2.3. CO₂ Photoreduction

The CO₂ photoreduction was carried out in a closed system under UV- light (Hg-125 W). In a typical procedure, 0.1 g of the photocatalyst was dispersed in 50 mL of DI water. Prior to starting the reaction, N₂ (Linde, UHP 99.999%) gas was first purged for 30 min to remove air, then CO₂ was subsequently flowed into the system for 30 min to ensure that all oxygen and N₂ were removed. The pressure in the reactor was kept at 1 atm, and the UV-light was irradiated to start the reaction. The resulting products from the photocatalytic CO₂ reduction were measured by a gas chromatograph (GC-14 Shimadzu, Kyoto, Japan) equipped with a flame ionization detector (FID, Porapak Q mesh 50/80 Column) and a thermal conductivity detector (TCD, GC-SCI 310C) to identify and quantify the products. The product selectivity is calculated as Equation (4):

$$\%Selectivity=\frac{Xi\times 100}{{\displaystyle \sum }Xi}$$

(4)

where Xi is product yield, including CH₃OH and CO.

2.4. Density Functional Theory (DFT) Calculations

The reduced TiO₂ surface was modeled by creating an oxygen vacancy on the surface. The simulations were carried out by an efficient density functional theory using the self-consistent Hubbard U-correction (DFT+U) approach [41,42] implemented in the Vienna Ab initio Simulation Package (VASP) [43,44,45,46]. The DFT+U methodology has been known as an ad hoc method that improves the description of d-states of the transition metals (3d-orbital in case of Ti) by implementing U-correction, solving the underestimated electronic interactions problems and providing a more accurate estimation than the standard DFT method [47,48]. The applied U value of 3.5 eV for Ti atoms was selected based on other works, which have performed the calculations of CO₂ adsorbed on reduced TiO₂ surface with different values of U and showed comparable calculated results to the experiments [47,49]. The effective Projector Augmented Wave (PAW) pseudopotentials [50] were constructed to describe the electron exchange and correlation effects. The calculations were performed within the generalized gradient approximation (GGA) and the Perdew–Burke–Ernzerhof (PBE) functional [51]. The self-consistent (SCF) field tolerance and the ionic force convergence threshold were 1.0 × 10⁻⁵ eV and −0.01 eV/Å, respectively. The kinetic energy cut-off for the plane wave basis set was set to 500 eV. The Monkhorst–Pack mesh sampling [52] k-points of 2 × 2 × 1 was used. The Methfessel–Paxton scheme of order two with a value of the smearing parameter σ of 0.03 eV was employed and the spin-polarized calculations were carried out. Bader charge analysis was performed using VASP—VTST [53,54,55]. For geometry optimization, the coordinates of the atoms in the two bottom layers were kept fixed while the rest of the atoms were allowed to relax. A vacuum space between slabs of 15 Å was set. To construct a reduced surface, an O atom at the bridge site (2c-O) was removed [47,56,57]. The optimized metal clusters of tetramers Pt, Pd, Ni and Cu were located on the optimized reduced TiO₂ surface, following the configurations proposed by the literature [46,58,59,60,61]. The 3 × 1 supercell of the anatase TiO₂ (101) was constructed with six layers of the (101) surface (Ti₃₆O₇₂). To study the CO₂ adsorption on M₄-TiO₂ surfaces, both bent and linear CO₂ molecules were considered.

3. Results and Discussion

3.1. Characterization of Photocatalysts

The particle sizes and morphologies of Pt, Pd, Ni, and Cu-loaded TiO₂ prepared by photodeposition method and measured by Transmission Electron Microscopy (TEM) are shown in Figure 1a–e. As seen in the TEM images, both TiO₂ and metal nanoparticles are in a spherical shape. The metal nanoparticles are in good distribution (Supplement Information, Figure S1). The average particle size of Pt, Pd, and Ni was approximately 4–5 nm, while Cu-TiO₂ sample showed a larger particle size of (~12 nm). Moreover, there are also metal signals distributing throughout the whole samples, suggesting that there could be metals in other forms such as ions, clusters, or single atoms existing in the samples. The XRD patterns of the samples with different types of metal loaded on anatase TiO₂ are displayed in Figure 1f. We found that the characteristic peaks of TiO₂ (anatase, PDF 71-1167) were clearly observed, while the characteristic peaks belonging to Pt, Pd, Ni, and Cu species were invisible, due to the low concentration of the metals. The optical properties of the as-synthesized photocatalysts characterized by UV-Vis spectrophotometer are shown in Figure 1g. Their band gap energies were calculated using Kubelka–Munk equation derived from UV-visible diffuse reflectance (UV-vis DRs) spectra. According to the diffuse reflectance spectra, it was found that all samples have quite similar light absorption edges of around 400 nm, which corresponds to the similar band gap of 3.2 eV based on the Kubelka–Munk equation (inset of Figure 1g).

The species and oxidation states of the fresh and spent metal-decorated TiO₂ photocatalysts were further investigated by XANES. The XANES spectra accompanying the first order derivative of Pt-, Pd-, Ni-, and Cu-loaded TiO₂ samples were compared to the standards illustrated in Figure 2. Figure 2a,b demonstrate the XANES spectra and the first order derivative of platinum samples (Pt L₃-edge), compared to the spectra of H₂Cl₆Pt precursor and the reference standard materials, including Pt foil and PtO₂ (representing the oxidation states of 0 and +4, respectively). Normally, Pt⁰ exhibits absorption edges at 11,567.9, while Pt⁴⁺ provides the edge energy at 11,567.4 eV (Table S2). We observed that the absorption edges of the fresh and spent Pt-TiO₂ revealed the edge energy was close to that of the Pt⁰, indicating that both the fresh and spent Pt-TiO₂ catalysts were mainly metallic [62]. As shown in

Table 1. The linear combination analysis results of XANES spectra with standards.

Standards	Fresh (%)	Spent (%)
0.1%Pt/TiO2
Pt foil	0.724	1.000
PtO2	0.121	0.000
H2Cl6Pt	0.155	0.000
0.1%Pd/TiO2
Pd foil	0.822	0.905
PdO	0.178	0.095
0.1%Ni/TiO2
Ni foil	0.970	0.828
NiO	0.000	0.000
Ni(OH)2	0.030	0.172
0.1%Cu/TiO2
Cu foil	0.000	0.000
CuO	0.399	0.068
Cu2O	0.601	0.932

, which tabulates the linear combination fit of the XANES spectra, the metallic form of Pt in the fresh sample was 72.4%, while that of the spent Pt-TiO₂ was 100%. This evidence suggests that some of Pt⁴⁺ could be further reduced to form metallic Pt during the CO₂ reduction.

Considering the Pd-decorated TiO₂ samples, the Pd L₃-edge XANES spectra and their first order derivative were compared with Pd foil, PdO, and PdCOCl₂ (Figure 2c,d). The edge energies of the fresh and spent Pd-TiO₂ were found to be 3175.4 and 3175.5 eV, which are significantly close to that of Pd foil, confirming the metallic Pd⁰ oxidation state [63]. This result is also consistent with the result from linear combination analysis, which shows 82.2% and 90.5% of Pd⁰ in the fresh and spent Pd-TiO₂ samples, respectively. Figure 2e,f display the XANES spectra and the first order derivative of the Ni-loaded TiO₂ samples, compared with the nickel standard references; namely, Ni foil, NiO, Ni(OH)₂, and Ni(NO₃)₂ which have the edge energies of 8340.9, 8350.4, 8349.9, and 8350.4 eV, respectively [64]. The pre-edge of both fresh and spent Ni-TiO₂ were found at 8341.2 eV and 8339.9 eV, respectively. The linear combination fitting of Ni-TiO₂ samples shows that the Ni species in the fresh and the spent Ni-loaded TiO₂ were close to Ni foil, confirming the existence of Ni⁰ in these samples.

Figure 2g shows the Cu K edge XANES spectra of the Cu-loaded TiO₂ samples, which are compared to Cu foil, CuO, and Cu₂O standards. Interestingly, the XANES features of both fresh and spent Cu-loaded TiO₂ show similar edge energy positions, which are 8995.4 and 8995.8 eV, respectively. The linear combination analysis, as indicated in Figure 2g shows unclear species of Cu. The presence of mixed oxidation states of the CuO and Cu₂O can be clarified by the first derivatives of the absorption edges, as shown in Figure 2h. We observed that Cu¹⁺ and Cu²⁺ components existed in both samples while the metallic copper disappeared. According to the investigation of the metal oxidation states by XANES from the linear combination analysis as tabulated in Table 1, it can be noted that the metal nanoparticles resulting from photodeposition in the fresh and spent Pt-, Pd-, and Ni-loaded TiO₂ were mostly in metallic form. However, Cu-TiO₂ favorably formed Cu¹⁺ species with a ratio of 60.1 and 93.2% for fresh and spent Cu-TiO₂, respectively.

Previous research has explored the metallic behaviors on the photodeposition and photo-oxidation of propanol using the photodeposition methodology [65]. The proposed mechanism for metal photodeposited TiO₂ is given in Equations (5)–(7). H⁺ is a proton produced by the photo-oxidation of propanol with holes, as shown in Equation (6). Metal ions were reduced over TiO₂ by reacting with photogenerated electrons, resulting in the formation of metallic particles, as validated by XANES:

$$Ti{O}_2+hv\to Ti{O}_2(e^{-}{}_{CB}+h^{+}{}_{VB})$$

(5)

$$C_3{H}_{7}OH+h^{+}{}_{VB}\to {}^{\ast }C{}_3{H}_{6}OH+H^{+}$$

(6)

$$M^{n+}+n{e}^{-}{}_{CB}\to M(0)$$

(7)

CO₂ temperature programmed desorption (CO₂-TPD) was carried out to investigate the interaction of CO₂ reactant on the catalyst as shown in Figure 3 for the adsorption temperature from 50 to 900 °C. The main spectra can be assigned to the molecularly adsorbed bidentate carbonates ${(\mathrm{b-}\mathrm{HCO}}_3^{2-})$ (380–550 °C) and monodentate carbonate ${(\mathrm{m-}\mathrm{HCO}}_3^{2-})$ (550–760 °C), corresponding to strong basic sites of catalysts [66]. The peak intensity of ${\mathrm{b-}\mathrm{HCO}}_3^{2-}$ and ${\mathrm{m-}\mathrm{HCO}}_3^{2-}$ over Pt-TiO₂ was lower than TiO₂, suggesting that more medium and strong basic sites were formed over TiO₂. However, as compared to pristine TiO₂, the chemical desorption peaks of Pt-deposited TiO₂ have shifted to a higher temperature, implying the stronger basicity of its adsorption sites [67]. The ${\mathrm{b-}\mathrm{HCO}}_3^{2-}$ and ${\mathrm{m-}\mathrm{HCO}}_3^{2-}$ species were generated from CO₂ molecules combined with oxygen atoms or metal atoms of the cocatalyst [68].

3.2. CO₂ Photocatalytic Reduction

The photocatalytic reduction of CO₂ was performed with liquid H₂O under UV-light irradiation using various metal-loaded TiO₂ samples as a photocatalyst. For the control experiment, the photocatalytic CO₂ reduction was performed without photocatalysts and light irradiation. There was no reduced CO₂ detected during the reaction, indicating that the reduced CO₂ products were generated by photocatalytic reactions. The influence of the metals (Pt, Pd, Ni, and Cu) on CO₂ photoreduction was carefully investigated, and the results are shown in Figure 4. The amount of metal deposited on TiO₂-based photocatalyst was fixed at around 0.1 wt%, and this was confirmed by ICP-OES (Supplement Information, Table S1). Methanol and CO were the major and minor products, respectively (Figure S2). After 2 h photoreaction, methanol was produced approximately 3.12 µmol/g_cat/h over the pristine TiO₂. We found that metal loading can significantly promote the generation of the reduced CO₂ products, compared to the pristine TiO₂. Different dopants (Pt, Pd, Ni, and Cu) on TiO₂ resulted in different degrees of enhancement on the photocatalytic CO₂ reduction performance. The highest amount of methanol yield was found over the Pt-TiO₂ catalyst with the production rate of 17.85 µmol/g_cat/h with selectivity of 96.41%, following by Cu, Pd, and Ni-TiO₂, which can produce methanol of 9.98, 9.35, and 6.09 µmol/g_cat/h, respectively (Figure 4). The CO production over all catalysts seems negligible, as no higher than 2.5 µmol/g_cat/h of CO was detected in any samples. The possible products in the liquid phase, such as formaldehyde, were also undetectable, as analyzed by a UV-Vis spectrophotometer. Moreover, 0.1 wt% Pt-TiO₂ was carried out on the photoreaction without the involvement of CO₂ to verify methanol and CO produced by CO₂ photoreduction over metal loaded TiO₂. The result showed that no product was detected (Supplement Information, Figure S3). Furthermore, CHN analysis of 0.1 wt% Pt-TiO₂ catalyst showed that the amount of carbon on the catalyst was negligible (Table S3). This evidence indicated that methanol and CO were produced by the photoreduction of CO₂. Pt-TiO₂ also showed excellent stability, as the catalyst can maintain >90% of its original reactivity after running for three cycles (Supplement Information, Figure S4 and Table S8).

To attain a greater understanding of CO₂ adsorption and reaction on TiO₂ supported catalysts loading with Pt, Pd, Ni, and Cu, we further computationally evaluated the structural and electronic characteristics of CO₂ adsorption on the supported tetramer metal clusters, including Pt, Pd, Ni, and Cu on the reduced surface of anatase TiO₂ (101). The anatase TiO₂ (101) surface consists of five-fold (5c-Ti) and six-fold (6c-Ti) coordinated Ti atoms and two-fold (2c-O) and three-fold (3c-O) coordinated O atoms at the surface (see Figure S6). These tetrameric metal clusters well represent both two- and three-dimensional deposited clusters. Three possible CO₂ adsorption sites found in this study could be categorized as: (i) CO₂ binding on the metal cluster; (ii) CO₂ binding at the interface between the metal cluster and the TiO₂; and (iii) linear CO₂ adsorbed on the TiO₂ surface (Supplement Information, Figures S5–S9 and Tables S4–S7).

To analyze the CO₂ adsorption systems, four properties of CO₂ were analyzed, including the adsorption energy of CO₂, the angle of O-C-O of adsorbed CO₂, the charge accumulation on the adsorbed CO₂ molecule, and the vibrational frequencies of the adsorbed CO₂. The adsorption energy ($E_{ads}$) of CO₂ was calculated according to Equation (8):

$$E_{ads}=E_{C{o}_2/M_4-Ti{O}_2}-E_{M_4-Ti{O}_2}-E_{C{O}_{2 \left(g\right)}}$$

(8)

where $E_{C{o}_2/M_4-Ti{O}_2}$ is the total energy of CO₂ adsorbed system, $E_{M_4-Ti{O}_2}$ and $E_{C{O}_{2 \left(g\right)}}$ are the total energy of the metal clusters located on the reduced TiO₂ surfaces and the energy of isolated CO₂, respectively. The results are illustrated in Figure 5. We found that the more negative $E_{ads}$ indicated the more stable CO₂ adsorption configuration. The difference in Bader charge (∆e) of CO₂ were the changes of CO₂ atomic charges upon adsorption. The negative ∆e implied electron accumulation. The more negative ∆e indicated the adsorbed CO₂ molecule gains more electrons. To confirm the key bond characteristics of adsorbed CO₂ anion, the vibration frequency calculations were obtained. Three key vibrational modes of CO₂ including symmetric (ν1), bending (ν2), and asymmetric (ν3) stretching modes were agreed to for other calculations for CO₂ anion adsorption on M₄-TiO₂ surfaces [47,59,69].

These experimental results reveal that the reaction mechanism of CO₂ photoreaction over various metal-deposited TiO₂ photocatalysts could proceed through the carbene pathway. This research fits well with the work reported by Habisreutinger and co-workers [70]. CO₂ photoreaction could be initiated through the chemisorbed CO₂ molecules on the heterogeneous catalyst and form the adsorbed ${\mathrm{CO}}_2^{•-}$ species on the surface [71], confirmed by computational DFT result. This reaction is more likely to occur at the metal-TiO₂ interfaces rather than on the pure TiO₂ or the metal surfaces, as indicated by the more negative E_ads values (Tables S4–S7). The simultaneous photogenerated holes react with adsorbed water or hydroxide ions ${\mathrm{OH}}_{ads}^{-}$ to generate oxygen and proton. Subsequently, the ${\mathrm{CO}}_2^{•-}$ reacts with the adsorbed ${\mathrm{H}}^{•}$, which is produced by the reduction of ${\mathrm{H}}^{+}$ before cleavage to form carbon monoxide and hydroxide ion (OH⁻) [67]. The adsorbed CO can desorb from the catalyst sites to produce CO as a product due to the weak CO adsorption of catalyst surface [72]. The adsorbed CO can also combine with an additional two electrons to form carbon residue on the surface, and then react with three ${\mathrm{H}}^{•}$ radicals to form ˙CH radical, carbene, and methyl radicals. Methyl radicals can further react with hydroxyl radical to form methanol [73]. As shown in Figure 5a, Pt and Cu provide moderately negative E_ads, which could facilitate both the adsorption and desorption processes of CO₂ molecules, while Ni provides very negative E_ads. In principle, Cu should also show good CO₂-to-methanol conversion performance, as it has suitable E_ads and good charge accumulation on CO₂ molecules. However, our experiment found that Cu can easily turn to copper oxide and, hence, Pt appears to be the most promising among the plasmonic metals studied in this work.

The enhancement of methanol yield from the modified TiO₂ with metal loading could be explained by the formation of a Schottky barrier that could reduce e⁻/h⁺ recombination [3,34,74,75,76]. To gain insight into the charge recombination kinetics, intensity-modulated photocurrent spectroscopy (IMPS) and photoluminescence spectra (PL) were implemented to study the behaviors of charges. The IMPS results clarify the photogenerated charge transfer, as shown in Figure 6a. The frequency minimum in the complex plane of the IMPS plot can be used to calculate the mean transit time of the photogenerated e⁻ according to Equation (9) [77]:

$${\tau }_c=\frac{1}{2\pi fc}$$

(9)

where f_c is the minimum point frequency (Hz) of the IMPS response. The smaller τ_c indicates the better charge transfer [77]. The results showed that the τ_c values of Pt-, Cu-, Pd-, and Ni-loaded TiO₂ were 4.513, 4.613, 4.665, and 4.668 µs, respectively. Obviously, Pt-decorated TiO₂ showed the lowest value of τ_c, indicating enhanced charge transfer ability compared to other samples. In contrast, Ni-TiO₂ has the highest values of τ_c, indicating the poor charge transfer kinetics. These results are in good agreement with the actual photocatalytic CO₂ reduction performance. Furthermore, PL was applied to further examine the charge transfer characteristics of Pt-TiO₂ compared to the pristine TiO₂, as shown in Figure 6b. The spectra of pure TiO₂ showed higher PL intensity of the peak emission, indicating the higher charge recombination rate [78]. It can be noted that the presence of metals decorated on the TiO₂ surface can enhance the photocatalytic reduction of CO₂ by inhibiting the recombination rate [76].

Figure 7a shows the Mott–Schottky plots of various synthesized photocatalysts deposited on FTO substrates. The linear regression of all metal-modified TiO₂ samples showed a positive slope, indicating that the catalysts are n-type semiconductors. The V_fb (vs Ag/AgCl) extracted from the Mott–Schottky results of TiO₂, Pt-TiO₂, Pd-TiO₂, Ni-TiO₂, and Cu-TiO₂ are −0.53, −0.23, −0.34, −0.53, and −0.23 V, respectively. These V_fb can be used to calculate the Fermi energy (E_F vs. vacuum) of the catalysts. As shown in

Table 2. The electronic properties derived from Mott–Schottky results.

Catalyst	Vfb (V vs. Ag/AgCl)	ΦM	EF	EC	EV	ND	ECB-EF
		(eV)	(eV vs. Vacuum)			(cm−3)	(mV)
TiO2	−0.53	-	−4.17	−4.12	−7.32	1.23 × 1020	47.83
Pt/TiO2	−0.23	−5.65 (Pt)	−4.47	−4.45	−7.65	3.27 × 1020	22.60
Pd/TiO2	−0.34	−5.22 (Pd)	−4.36	−4.32	−7.52	1.96 × 1020	35.72
Ni/TiO2	−0.53	−5.04 (Ni)	−4.17	−4.12	−7.32	1.32 × 1020	45.98
Cu/TiO2	−0.23	−4.65 (Cu)	−4.47	−4.44	−7.64	2.64 × 1020	28.14

, Pt-TiO₂ has the lowest E_F (−4.47 eV), which is more negative than the unmodified TiO₂ (−4.17 eV). Furthermore, it was found that the E_F of the catalysts decreased as the work function (Φ_M) of the deposited metals increased. It was believed that the higher work function of metal (low E_F) can cause the more downward shifting of E_F in TiO₂.

The electron carrier density (N_D) of Pt-TiO₂ was calculated to be 3.27 × 10²⁰ cm⁻³, which was higher than that of pristine TiO₂. It is well known that a higher donor density implies a higher material conductivity, which facilitates the charge transport process [79]. Interestingly, Cu-TiO₂ has low E_F (−4.47 eV), which is comparable to Pt-TiO₂ even though the work function (Φ_M) of Cu is the lowest value (−4.65 eV) compared to the other interested metals. This might be due to the effect of Cu₂O, which was the main component of Cu-TiO₂ (see the linear combination analysis of XANES results). According to the report of Aguirre et al. [80], the E_F of TiO₂ can shift downward when it intimately contacted Cu₂O to equilibrate the E_F between two materials. Although the E_F of Cu-TiO₂ was equal to the E_F of Pt-TiO₂, the $N_D$ of Cu-TiO₂ was lower (2.64 × 10²⁰ cm⁻³). This could be the reason why Cu-TiO₂ has lower CO₂ reduction efficiency than that of Pt-TiO₂.

Therefore, the relationship between the metal’s work function obtained from the Mott–Schottky results and the CH₃OH yield is further analyzed as shown in Figure 7b. Normally, noble metals are known as efficient cocatalysts due to their large work functions. The metal’s work function (Φ_M) is the energy needed to bring the electron from the metal’s Fermi energy to a vacuum level [33,81]. The larger metal work function, the better electron trapping ability. Band bending is formed when noble metals and semiconductors make intimate contact [32]. Pt-TiO₂ showed the highest amount of methanol production, followed by Pd and Ni, respectively. This trend follows the order of the work function of the metals (see Figure 7b). The Schottky junction between Pt-decorated TiO₂ with energy band alignment is shown in Figure 7c. The work function of Pt is −5.65 eV versus E_vacuum, which is more positive than the conduction band of anatase TiO₂ (−4.12 eV vs. vacuum) [82]. Therefore, electrons can easily transfer from the CB of TiO₂ to Pt sites, which act as an electron sink [83]. On the other hand, metal with a smaller work function, such as Ni, causes a weaker driving force of electron migration [84]. As a result, the Pt-deposited on TiO₂ sample showed significantly higher photocatalytic activity than the unloaded TiO₂ due to the higher charge separation efficiency, which is also higher than the other metals [75,85]. However, the production of methanol was not in the trend over Cu-loaded TiO_2. According to the XANES result, Cu loading on TiO₂ was in a form of complex oxide. Therefore, the energy band alignment of Cu_xO-TiO₂ heterojunction was a semiconductor-semiconductor heterojunction, as revealed in Figure 7d. When Cu_xO with lower CB level contacts with TiO₂, which has higher level of CB, electrons in the CB of TiO₂ can be transferred to that of Cu_xO. The electrons and holes are transferred to the CB of Cu_xO and the VB of TiO₂, respectively [33]. As a result, the photoexcited electron-hole pairs can be separated by the electric field.

4. Conclusions

TiO₂ loaded with various metals (Pt, Pd, Ni, and Cu) was successfully synthesized by photodeposition method. We found that all samples are active for CO₂ photoreduction. On the other hand, various insightful characterizations reveal that Pt is the most promising metal, as it provides the largest work function when formed in heterojunction with TiO₂, and provides the most appropriate CO₂ adsorption and charge accumulation energies for methanol formation. which is also in good agreement with the experimental results. The large metal work function could enhance the charge transfer ability and suppress charge recombination. Electrons can transfer from the conduction band of a semiconductor to the metal surface and thus promote the photocatalytic reduction activity. When compared to Cu, Pd, and Ni-loaded TiO₂ photocatalysts, Pt-loaded TiO₂ photocatalyst also has the fastest charge transfer of 4.513 µs. Hence, Pt-TiO₂ can generate methanol (major product) with the rate of 17.85 µmol/g_cat/h, which is the highest photocatalytic CO₂ reduction activity among the M-TiO₂ that have been studied in this work, and among the top TiO₂-based photocatalysts that have been reported for photocatalytic CO₂-to-methanol conversion.