Insight into the photoexcitation effect on the catalytic activation of H₂ and C-H bonds on TiO₂(110) surface

The activation of H–H and C–H bonds (classified as R-H, R = H, C) is of vital importance in the hydrogenation and dehydrogenation processes for the catalytic conversion and functionalization of saturated or unsaturated hydrocarbon compounds. Ways to efficiently break these chemical bonds under mild condition are always at the heart of chemical industry [1-3], which constitutes one of the foundations for the controllable formation of value-added commodity [4-6]. Generally, the catalytic cleavage and functionalization of R-H bond have been extensively explored on kinds of catalysts, such as the noble metals (Pt [7-9], Rh [10-12], Au [13, 14] and et al. [15-18]), transition metal oxides (TMOs) and zeolites [19-21]. For example, the TMO-catalysts modified by single-atom metal doping or metal loading [22-26] were specifically manufactured to tailor the catalytic activity toward R-H bond activation. Despite these great progresses, these reactions are often performed at high temperatures (573–873 K) [27-29], which could lead to severe coke deposition and low product selectivity. Therefore, it is of paramount importance to develop more efficient catalysts or technologies for R-H bond activation under mild condition [30-33].

In this regard, the photocatalytic technology is believed to be capable of altering the reaction kinetics under mild condition owing to the presence of the excited carrier (h⁺/e⁻), and increasing experimental studies have demonstrated the feasibility for facilitating the R-H bond activation [34-36]. Taking the photocatalytic CH₄ activation as an example, it was proposed that CH₄ can be selectively converted to CH₃OH or HCHO under UV irradiation at room temperature over semiconductor catalysts, like TiO₂ [37], ZnO [38] and et al. [39-41]. However, the atomic-level insight into the photocatalytic effect is not unambiguously disclosed, to the best of our knowledge. Theoretically, owing to the difficulty in simulating photo-excited radicals properly on the semiconductor, the photocatalytic surface reaction has been relatively less studied with the detailed kinetic information provided [42-44]. It seems that most researches mainly focus on thermodynamic properties to evaluate the photocatalytic activity by considering the band structure of semiconductor and the potential of target reaction [45-48]. Kinetically, a comprehensive understanding of the nature of photocatalytic effect on R-H bond activation at the atomic-scale level is very limited. Specifically, one may ask, how does the photoexcitation affect the H₂ and C-H activation, and what is the quantitative difference between thermo- and photo-catalytic mechanisms? Moreover, there are different types of C-H bonds, e.g., C(sp³)-H, C(sp²)-H and C(sp¹)-H, and what are the general rule or difference of the photocatalytic effects on them?

Herein, we computationally explored the activation of H₂ molecule and diverse C–H bonds in thermo- and photo-catalysis on rutile TiO₂(110), one of the most studied semiconductors in photocatalysis. The theoretical results explicitly revealed the reaction mechanisms of H₂ activation; moreover, the activation of C-H bonds in diverse light-hydrocarbons, including acetylene (HC≡C–H), benzene (C₆H₅–H), ethylene (CH₂=CH–H), methane (CH₃–H), ethane (CH₃CH₂–H), propyne (CH≡CCH₂–H), toluene (C₆H₅CH₂–H) and propylene (CH₂=CHCH₂–H), were calculated and compared. The C–H bond energy (E_bond) and deprotonation energy (E_dep, corresponding to the acidity) were identified to be two main factors leading to the distinct behaviors of R–H bonds activation in thermo- and photo-excited conditions.

All spin-polarized DFT calculations were performed via Vienna ad initio Software Package (VASP) code [49, 50], using the Perdew-Burke-Ernzerhof (PBE) functional within the generalized gradient approximation (GGA) [51, 52]. The project-augmented wave (PAW) [53] method was used to represent the core-valence electron interaction, and the valence electronic states were expanded in plane wave basis sets with a cutoff energy of 450 eV. To ensure proper description of the valence electrons on Ti atom, a valence electron configurations 3p⁶4s²3d² (small core) was applied.

We adopted a p(3 × 2) model of rutile TiO₂(110) with four TiO₂ atomic layers to accommodate the organic molecule and minimize the interaction between periodically repeated slabs. The vacuum between slabs was ~10 Å and a 2 × 3 × 1 k-point mesh was used during the optimizations. The transition states (TSs) were searched by the constrained optimization scheme and were verified when (i) all forces on the atoms vanish and (ii) the total energy is a maximum along the reaction coordinate but a minimum with respect to the rest of the degrees of freedom [54-56]. All the atoms were allowed to relax, until the force on each atom was less than 0.05 eV/Å. The van der Waals forces was taken into account throughout this work to consider the possible weak interaction between intermediates and TiO₂(110) surface [57, 58]. To model the photogenerated h⁺/e⁻, here we took the method of removing/adding electrons directly from the system with an additional neutralizing background charge applied in VASP due to the large dielectric constant of rutile TiO₂ [59]. Moreover, the photogenerated carriers usually tend to delocalize on TiO₂ system in GGA calculations due to the self-interaction error. To overcome this issue, DFT+U method with an on-site Coulomb interaction U added to the Ti-3d (U = 4.2 eV) and O-2p orbitals (U = 6.3 eV) was adopted for structure optimization and energy calculation [60], and the HSE06 hybrid functional was used to verify the electronic structure. Such an approach has been successfully applied to describe the photocatalytic reactions (e.g., methanol oxidation, oxygen evolution reaction) on TiO₂ [61, 62].

The adsorption energy of surface species (X) was defined as E_ad(X) = E(X/sur) – E(sur) – E(X), where E(sur), E(X) and E(X/sur) are the energy of catalyst surface, X species in the gas phase and X species adsorbed on the catalyst surface, respectively. Notably, at a specific temperature, the large entropy contributions of gaseous molecules were considered to estimate the adsorption free energy (G_ad) [63, 64].

As Fig. 1a shows, the rutile TiO₂(110) surface is terminated by two types of unsaturated sites: two-fold coordinated bridge O and five-fold coordinated Ti (denoted as O_br and Ti_5c, respectively), which constitute the possible active centers for H-H and C-H bonds activation. Firstly, the adsorption and dissociation of H₂ molecule under the thermocatalytic condition were briefly examined as a benchmark. It was found that H₂ can weakly adsorb on Ti_5c site with an adsorption energy, E_ad(H₂), of −0.20 eV, which corresponds to an adsorption free energy, G_ad(H₂), of 0.20 eV at room temperature. Subsequently, we tested two possible H₂ dissociation routes, including homolytic and heterolytic cleavage pathways, according to previous studies [30, 31, 65]. In the homolytic mode, H₂ activation is accompanied by the formation of a H^· and a H⁺ (i.e., H₂ + Ti_5c⁴⁺ + O_br²⁻ → H^· + Ti_5c³⁺ + O_brH⁻); the activation barrier is as high as 1.73 eV and the enthalpy change reaches 0.70 eV (see energy profiles in Fig. 1b). Alternatively, when the reaction proceeds via the heterolytic cleavage mode, which synergistically occurs at the O_br and Ti_5c dual sites (i.e., H₂ + Ti_5c⁴⁺ + O_br²⁻ → Ti_5c⁴⁺-H⁻ + O_brH⁻), the barrier decreases a lot to 0.58 eV. This indicates that the heterolytic mode is energetically more favorable for H-H bond scission than the homolytic mode. Moreover, the H atoms dissociated from the heterolytic cleavage would be eventually captured by O_br and Ti_5c sites yielding a H⁺ and a H⁻ anion, evidenced by their Bader charges of 1.00 |e| and −0.44 |e| (Fig. 1c), respectively. Notably, these barriers are slightly higher than those reported by Hu et al. (1.39 eV and 0.37 eV in homolytic and heterolytic modes, respectively), owing to the self-interaction error of TiO₂ system ignored in their study [65].

Figure 1

Figure 1.  (a) Ball-and-Stick model of rutile TiO₂(110) surface and H₂ adsorption configuration, in which H_a is the H atom approaching to O_br site and the other H atom is named as H_b. (b) Energy profiles of thermo-catalytic H₂ dissociation via homolytic and heterolytic modes, respectively, and (c) the optimized structures and spin density distribution of initial state (IS), transition state (TS) and final state (FS). The barrier (E_a) and enthalpy change (ΔH) are labeled (in eV). The distances of some key bonds and Bader charges of important intermediates are labeled in black (in Å) and blue (in |e|), respectively. Ti, gray; O, red; H, white.

DownLoad: Full-Size Img PowerPoint

Secondly, we calculated the photocatalytic H₂ dissociation on TiO₂(110) surface. Generally, the generated photo-hole/electron pairs under light illumination thermodynamically tend to be trapped by O_br and Ti_5c sites and form O_br^·− and Ti_5c³⁺, respectively, corresponding to O_br²⁻ + h⁺ → O_br^·− and Ti_5c⁴⁺ + e^– → Ti_5c³⁺ [66-68]. As Figs. 2a and b show, our calculations verify that these two excited active sites (O_br^·− and Ti_5c³⁺) have insignificant effect on H₂ adsorption with E_ad(H₂) of −0.22 eV and −0.21 eV, respectively. With the assistance of Ti_5c³⁺, the barrier of H₂ dissociation is nearly the same as that in thermocatalysis (0.56 eV vs. 0.58 eV) with a very similar TS structure (see TS1 in Fig. 2c), implying little contribution of photo-electron on H₂ activation. Intriguingly, on hole-trapped TiO₂(110) surface, the TS of H-H bond cleavage changes from the dihapto configuration in the thermocatalytic condition to the monohapto one (Figs. 2c and 1c), and the barrier is reduced to 0.37 eV with an enthalpy change of −0.04 eV (i.e., H₂ + O_br^·− → HO_br⁻ + H^·). Notably, in comparison with the pristine homolytic cleavage mode in thermocatalysis, the barrier has largely declined (1.73 eV vs. 0.37 eV). With the generation of H^· radical, it would further re-adsorb on O_br and form H⁺ with a large energy release (−2.90 eV). More specifically, in this photocatalytic TS, one H in H₂ molecule is captured by O_br^·− and the other H suspends above the TiO₂ surface with the H···H and H_a···O_br bonds being 0.857 Å and 1.307 Å, respectively. In other word, it exhibits a radical-like TS, and the radical nature of H atom is also confirmed by Bader spin charge of 0.97 |e|. Therefore, the H-H bond activation mediated by photo-hole essentially follows the homolytic cleavage mechanism.

Figure 2

Figure 2.  (a, b) H₂ adsorption on TiO₂(110) in the presence of Ti_5c³⁺ and O_br^·−, respectively. (c) Energy profiles of H₂ dissociation in thermocatalysis via heterolytic mode (black line) or under light irradiation, including hole- and electron-assisted circumstances (red and blue lines, respectively).

DownLoad: Full-Size Img PowerPoint

To shed light on the promotion effect of photo-hole on H₂ dissociation, we quantitatively inspected the related electronic structure information. Firstly, the periodic natural bond orbital (NBO) [69] of two-site assisted TS complex in thermocatalysis was analyzed to uncover the bonding/antibonding nature of H_a-O_br and H_b-Ti_5c (see Fig. 3a), in which H_a/H_b denotes the H atom in H₂ close to the O_br/Ti_5c site, respectively (Fig. 1a). Quantitatively, the charge of H_a is calculated to be 0.65 |e| in the H₂ activation TS, indicating that the electron in H_a is partially shared by the forming H_a-O_br bond; however, it owns a charge of 1.06 |e| for the H_b atom in H_b-Ti_5c bond, implying the negligible electron transfer (versus that (0.93 |e|) of H_b atom in pristine H₂ molecule). Moreover, the occupancy of H_a-O_br bonding/antibonding is 0.91/0.20 |e|, but nearly zero for H_b-Ti_5c bond. Therefore, we can speculate that the activation of H₂ molecule depends mainly on the O_br site, whereas the Ti_5c site contributes to a small extent, which could well explain the greater promotion effect of photo-hole than photo-electron. Similar result can be drawn in the C-H bond activation of CH₄ molecule on rutile TiO₂(110) surface [70].

Figure 3

Figure 3.  (a) The periodic natural bond orbital (NBOs) of TS for H₂ dissociation in thermocatalysis via the heterolytic mode. (b, c) The projected densities of states (PDOSs) of pristine TiO₂(110) and hole-trapping TiO₂(110), respectively, using HSE06. The vertical dotted lines in the PDOSs denote the fermi levels.

DownLoad: Full-Size Img PowerPoint

Furthermore, the projected densities of states (PDOSs) of TiO₂(110) in the absence or presence of O_br^·− were calculated (see Figs. 3b and c). It can be seen that the presence of localized photo-hole does not lead to significant change in the whole band gap, but induces a new unoccupied state within the forbidden band. In principle, this hole-state is conductive to accepting electron and exhibits strong oxidizability, evidenced by the higher energy level of O_br^·− (~1.5 eV above the fermi level). In this sense, the superior performance of O_br^·− than O_br²⁻ on H-H bond rupture can be rationalized as follows. O_br^·− holds less electron (−0.69 |e|) than O_br²⁻ (−1.08 |e|), and this reactive O_br^·− site, as a stronger oxidative center, can more easily accept electron from the dissociated hydrogen atom after H₂ homolytic cleavage in photocatalysis, thereby largely decreasing the activation barrier.

With the understanding of photoexcitation effect on modulating H₂ dissociation, we are now at the position to explore the C-H bond activation in a series of different organic molecules on TiO₂(110), aiming to uncover the general rule of the photo-hole effect, which includes (i) sp¹–C group: HC≡C–H; (ii) sp²–C group: C₆H₅–H and CH₂=CH–H; (iii) sp³–C group: CH₃–H, CH₃CH₂–H, CH≡CCH₂–H, C₆H₅CH₂–H and CH₂=CHCH₂–H. At first, the thermally-driven C–H bond activation processes were studied. By comparing the barriers in homolytic and heterolytic cleavage modes (Table 1), we can see that heterolytic route is not always the most favorable one for all C-H bonds, which differs from H–H bond activation. In detail, the HC≡C-H, C₆H₅–H, CH₂=CH–H, CH₃–H and CH₃CH₂–H bonds prefer to be activated via the heterolytic cleavage mode, whereas the homolytic mode for the others (i.e., CH≡CCH₂–H, C₆H₅CH₂–H and CH₂=CHCH₂–H) (Fig. 4a and Table 1). Additionally, except for HC≡C–H (0.13 eV), almost all the reaction barriers of the most favored pathway are above 0.90 eV, which indicates the difficulty of C–H bond activation at room temperature. In the presence of O_br^·−, we identified that the homolytic mode becomes more favored for C–H bonds in all but HC≡C–H (Table 1). Particularly, the activation barriers for the CH≡CCH₂–H, C₆H₅CH₂–H and CH₂=CHCH₂–H bonds are extremely small (< 0.10 eV) as compared to those in the thermocatalytic condition; for the usually-known inert C–H bonds in benzene, ethylene and methane, the barriers can also be reduced to below 0.52 eV. Overall, these results imply that for all these C–H bond in light hydrocarbons, the photo-hole can prompt their activation evidently, except for HC≡CH molecule.

Table 1

Table 1.  Reaction barriers (E_a) of various R-H bonds cleavage on TiO₂(110) in thermo- and photo-catalysis (heterolysis/homolysis, abbreviated as hetero-/homo-, respectively). E_bond is the R-H bond energy defined with the reaction energy of homolytic process (R-H → R^·+ H^·) and E_dep is the deprotonation energy from the heterolytic process (R-H → R⁻+ H⁺), taken from the NIST WebBook [71]. Note that the TSs of photocatalysis in heterolytic mode are hardly accessible in geometry, and thus their barriers are not shown here. The unit is eV.

DownLoad: CSV

Figure 4

Figure 4.  (a) Comparisons for the most favored activation barrier (E_a) of C-H bonds in the thermal and photo-hole assisted conditions, as well as the corresponding TSs for group-Ⅰ, -Ⅱ and -Ⅲ C-H bonds (b-d), respectively, in which the distances of H_a···O_br bonds are labeled in black (in Å). (e) Correlations between E_a and E_bond in the thermal and photo-hole assisted conditions, respectively.

DownLoad: Full-Size Img PowerPoint

In addition, it is remarkable that the activation of these C–H bonds exhibits different behaviors, and is modulated by O_br^·− to different degrees in comparison to the thermocatalysis (Fig. 4a). According to the results revealed above, these C–H bonds may be, therefore, classified as three groups: (group-Ⅰ) the ones (CH₃–H, CH₃CH₂–H, CH₂=CH–H and C₆H₅–H, Fig. 4b) that follow the heterolysis mode in thermocatalysis and change to the homolysis mode in photocatalysis, to which the photo-effect is beneficial (see ΔE_I in Fig. 4e); (group-Ⅱ) the ones (CH≡CCH₂–H, C₆H₅CH₂–H and CH₂=CHCH₂–H, Fig. 4c) that follow the homolysis mode in thermocatalysis and keep it in photocatalysis, where the photo-effect is more efficient than group-Ⅰ (see ΔE_Ⅱ in Fig. 4e); (group-Ⅲ) HC≡C–H (Fig. 4d), which prefers the heterolysis mode in both thermo- and photo-catalysis, and the presence of O_br^·− will inhibit the HC≡C–H bond activation (see ΔE_Ⅲ in Fig. 4e). One may naturally ask, what are the origins for these three different behaviors? Specifically, why do the group-Ⅱ C–H bonds tend to follow the homolytic cleavage mode, differing from the others? For HC≡C–H, why does it always follow the heterolysis mode, and why is the presence of O_br^·− adverse to its C–H bond scission?

Toward these questions, firstly, we quantitatively tested the dependence of the E_a values via homolytic/heterolytic routes in the thermocatalytic condition on the corresponding C–H bond energies (E_bond). Two overall linear relations can be obtained with R² of 0.87/0.96 (Fig. 4e), respectively. More intriguingly, as E_bond increases, E_a in the homolytic route increases, but decreases in the heterolytic one, and the crossing point of the two curves is at E_bond = 4.2 eV. These correlations clearly suggest that the choice of homolysis versus heterolysis in the thermocatalytic condition strongly depends on the C–H bond energy: the heterolysis is favored in the strong bond-energy region, while the homolysis is preferred in weak bond-energy region. For the group-Ⅱ C–H bonds, they own the relatively weaker E_bond among all the studied ones (Table 1), due to the stability of dehydrogenated molecule-fragment (i.e., CH≡C-CH₂^·, C₆H₅–CH₂^· and CH₂=CH-CH₂^·) with extended π-electron conjugation, which rationalizes the preference of homolytic cleavage in thermocatalysis.

With the participation of O_br^·− in the C–H bond activation, a similar linear correlation exists between the E_a following the homolysis mode in the photocatalytic condition and E_bond as well (R² = 0.95, Fig. 4e). One can see from Fig. 4e that this line is overall downward shift in parallel as compared to that of the homolysis mode in the thermocatalytic condition, which again illustrates the promotion effect originating from the high oxidizability of O_br^·−. Accordingly, we can explain the different promotion effect of O_br^·− on the C–H bonds of group-Ⅰ versus group-Ⅱ, which gives ΔE_Ⅱ > ΔE_I. For group-Ⅱ C–H bonds, the homolytic mode is favored in both thermo- and photo-catalytic conditions, and the promotion extent (ΔE_Ⅱ) of O_br^·− is the largest as seen from Fig. 4e. Differently, for group-Ⅰ C–H bonds, the heterolysis mode is more favored in the thermocatalytic condition, and it will change to the homolysis mode in photocatalysis; thus, the photo-promotion effect (ΔE_I) is relatively weakened relative to ΔE_Ⅱ. Moreover, it is worth noting that ΔE_I becomes smaller and smaller as the E_bond increases.

Secondly, as illustrated in Fig. 4e, with the increase of E_bond, we can see that the thermodynamically driven heterolytic mode becomes more and more feasible. When above ~5.3 eV, it will be even more favored than the photo-driven process, essentially ascribed to that the localized hole at O_br^·− cannot provide enough energy to compensate the C–H bond energy. Specifically for HC≡C–H activation, the large E_bond of HC≡C–H is unfavorable for the occurring of homolytic mode. Moreover, as we know, HC≡CH together with the group-Ⅱ molecules (CH≡CCH₃, C₆H₅-CH₃ and CH₂=CHCH₃) displays weak acidity [72, 73], consistent with their relatively small deprotonation energies (E_dep) in Table 1. These two factors imply that the HC≡C–H is more susceptible to releasing hydrogen in the form of proton (i.e., HC≡C–H → HC≡C⁻ + H⁺). In this circumstance, the hydrogen in HC≡C–H would more easily combine with the O_br²⁻ site via heterolysis rather than homolysis, which accordingly corresponds to a lower barrier of 0.13 eV (2.10 eV in homolysis). When the O_br^·− is present, owing to the additional Coulomb repulsion between positively charged H⁺ and O_br^·− (versus that in H⁺···O_br²⁻), the activation process is kinetically less favored than that in the pristine O_br²⁻ (0.78 eV vs. 0.13 eV in photo- and thermo-catalysis, respectively). Thus, the HC≡C–H bond prefers to be activated through the heterolytic mode and photo-hole would suppress this activation process, as a result of its strong bond energy and acidity. Notably, the similar result was also observed in the heterolytic O–H bond cleavage in the photocatalytic CH₃OH oxidation revealed in our previous study (Table 1) [61].

In summary, we have quantitatively studied the activation mechanism of H₂ and a series of C–H bonds on TiO₂(110) in the thermo- and photo-catalytic conditions, aiming to reveal the general photoexcitation effect on modulating the chemical bonds breakage. The main results can be summarized as follows:

(i) The thermal-driven H₂ activation on TiO₂(110) tends to obey the heterolytic cleavage mode, and the O_br^·− radical formed from the photo-hole localization would change the mode to be homolytic one and evidently facilitate the activation process.

(ii) We identified that the examined C–H bonds can be classified as three groups in terms of the bond energy and acidity: group-Ⅰ (CH₃–H, CH₃CH₂–H, CH₂=CH–H and C₆H₅–H), group-Ⅱ (CH≡CCH₂–H, C₆H₅CH₂–H and CH₂=CHCH₂–H) and group-Ⅲ (HC≡C–H). The bond energy order is group-Ⅱ < group-Ⅰ < group-Ⅲ; meanwhile, group-Ⅱ/-Ⅲ C–H bonds own weak acidity. The O_br^·− species can facilitate the activation of group-Ⅰ/Ⅱ C–H bonds to different degrees, whereas inhibits the one in HC≡C–H.

(iii) The different C–H activation behaviors in group-Ⅰ to group-Ⅲ could be largely attributed to the C–H bond energies and acidities. For the group-Ⅰ/-Ⅱ C–H bonds with relatively moderate/low bond energies, they obey the heterolytic/homolytic cleavage mechanism in thermocatalysis, and will uniformly obey the homolysis mechanism driven and simultaneously promoted by the highly oxidative O_br^·−, where the promotion effect on group-Ⅱ C–H bonds is more evident with nearly negligible barriers.

(iv) Owing to the strong bond energy and weak acidity, we found that the HC≡C–H bond consistently prefers to be activated via the heterolytic cleavage mode, and O_br^·− would inhibit this activation as a result of additional Coulomb repulsion between the dissociated proton and O_br^·− (versus the pristine O_br²⁻ on TiO₂(110)).

This work provided a general atomic-level description on the photo- versus thermo-catalytic activation of H₂ and various C–H bonds, which may explicitly deepen our understanding of the photocatalytic effect on modulating the chemical bonds breakage.

Declaration of competing interest

The authors report no declarations of interest.

Acknowledgments

This project was supported by National Nature Science Foundation of China (Nos. 21873028, 91945302), National Ten Thousand Talent Program for Young Top-notch Talents in China, Shanghai Shu-Guang project (No. 17SG30), and the Fundamental Research Funds for the Central Universities

References

1. [1]

   R. Khorasani, P.E. Fleming, Comput. Theor. Chem. 1096 (2016) 89-93.  doi: 10.1016/j.comptc.2016.09.033
   

2. [2]

   J. Berkowitz, G.B. Ellison, D. Gutman, J. Phys. Chem. 98 (1994) 2744-2765.  doi: 10.1021/j100062a009
   

3. [3]

   B.B. Wayland, Polyhedron 7 (1988) 1545-1555.  doi: 10.1016/S0277-5387(00)81781-7
   

4. [4]

   L. Meng, Z. Chen, Z. Ma, et al., Energy Environ. Sci. 11 (2018) 294-298.  doi: 10.1039/C7EE02951A
   

5. [5]

   X. Li, W. Wang, F. Dong, et al., ACS Catal. 11 (2021) 4739-4769.  doi: 10.1021/acscatal.0c05354
   

6. [6]

   M. Huš, D. Kopac, B. Likozar, J. Catal. 386 (2020) 126-138.  doi: 10.1016/j.jcat.2020.03.037
   

7. [7]

   J.L. Gland, G.B. Fisher, E.B. Kollin, J. Catal. 77 (1982) 263-278.  doi: 10.1016/0021-9517(82)90167-1
   

8. [8]

   G.E. Gdowski, R.J. Madix, Surf. Sci. 119 (1982) 184-206.  doi: 10.1016/0039-6028(82)90292-8
   

9. [9]

   R. Van Lent, S.V. Auras, K. Cao, et al., Science 363 (2019) 155-157.  doi: 10.1126/science.aau6716
   

10. [10]

    X. Qi, Y. Li, R. Bai, et al., Acc. Chem. Res. 50 (2017) 2799-2808.  doi: 10.1021/acs.accounts.7b00400
    

11. [11]

    S.R. Neufeldt, G. JiméNez-Osés, J.R. Huckins, et al., J. Am. Chem. Soc. 137 (2015) 9843-9854.  doi: 10.1021/jacs.5b03535
    

12. [12]

    J.C. Lewis, R.G. Bergman, J.A. Ellman, Acc. Chem. Res. 41 (2008) 1013-1025.  doi: 10.1021/ar800042p
    

13. [13]

    M. Wijzenbroek, D. Helstone, J. Meyer, et al., J. Chem. Phys. 145 (2016) 144701.  doi: 10.1063/1.4964486
    

14. [14]

    Q. Wu, L. Zhou, G.C. Schatz, et al., J. Am. Chem. Soc. 142 (2020) 13090-13101.  doi: 10.1021/jacs.0c04491
    

15. [15]

    S. Mukherjee, F. Libisch, N. Large, et al., Nano Lett. 13 (2013) 240-247.  doi: 10.1021/nl303940z
    

16. [16]

    T. Engel, H. Kuipers, Surf. Sci. 90 (1979) 181-196.  doi: 10.1016/0039-6028(79)90018-9
    

17. [17]

    D.F. Padowitz, S.J. Sibener, Surf. Sci. 254 (1991) 125-143.  doi: 10.1016/0039-6028(91)90645-9
    

18. [18]

    Y. Sun, S. Zhang, W. Zhang, et al., Chin. J. Chem. Phys. 31 (2018) 485-491.  doi: 10.1063/1674-0068/31/cjcp1805120
    

19. [19]

    F. Zaera, D. Chrysostomou, Surf. Sci. 457 (2000) 89-108.  doi: 10.1016/S0039-6028(00)00337-X
    

20. [20]

    M. L. Yang, Y.A. Zhu, C. Fan, et al., Phys. Chem. Chem. Phys. 13 (2011) 3257-3267.  doi: 10.1039/c0cp00341g
    

21. [21]

    Y. Chen, D.G. Vlachos, J. Phys. Chem. C 114 (2010) 4973-4982.
    

22. [22]

    W. Zhang, M. Pu, M. Lei, Langmuir 36 (2020) 5891-5901.  doi: 10.1021/acs.langmuir.0c00644
    

23. [23]

    G. Righi, R. Magri, A. Selloni, J. Phys. Chem. C 123 (2019) 9875-9883.  doi: 10.1021/acs.jpcc.9b00609
    

24. [24]

    L. Brugnoli, A. Pedone, M.C. Menziani, et al., J. Phys. Chem. C 123 (2019) 25668-25679.  doi: 10.1021/acs.jpcc.9b06805
    

25. [25]

    Y. Chen, P. Hu, M.H. Lee, et al., Surf. Sci. 602 (2008) 1736-1741.  doi: 10.1016/j.susc.2008.02.036
    

26. [26]

    Y. Chen, J. Cheng, P. Hu, et al., Surf. Sci. 602 (2008) 2828-2834.  doi: 10.1016/j.susc.2008.06.033
    

27. [27]

    M.V. Bossche, H. Gronbeck, J. Am. Chem. Soc. 137 (2015) 12035-12044.  doi: 10.1021/jacs.5b06069
    

28. [28]

    F. Zasada, J. Janas, W. Piskorz, et al., ACS Catal. 7 (2017) 2853-2867.  doi: 10.1021/acscatal.6b03139
    

29. [29]

    Z. Zhu, W. Guo, Y. Zhang, et al., Carbon Energy 3 (2021) 519-540.  doi: 10.1002/cey2.127
    

30. [30]

    T. Whittaker, K.B.S. Kumar, C. Peterson, et al., J. Am. Chem. Soc. 140 (2018) 16469-16487.  doi: 10.1021/jacs.8b04991
    

31. [31]

    K. Sun, M. Kohyama, S. Tanaka, et al., J. Phys. Chem. C 118 (2014) 1611-1617.  doi: 10.1021/jp4099254
    

32. [32]

    S. Mukherjee, L. Zhou, A.M. Goodman, et al., J. Am. Chem. Soc. 136 (2014) 64-67.  doi: 10.1021/ja411017b
    

33. [33]

    N. Yodsin, C. Rungnim, S. Tungkamani, et al., J. Phys. Chem. C 124 (2019) 1941-1949.
    

34. [34]

    Y. Chen, S. Ji, W. Sun, et al., Angew. Chem. Int. Ed. 132 (2020) 1295-1301.  doi: 10.1002/ange.201912043
    

35. [35]

    H. She, H. Zhou, L. Li, et al., ACS Sustain. Chem. Eng. 6 (2018) 11939-11948.  doi: 10.1021/acssuschemeng.8b02217
    

36. [36]

    T. Yan, Y. Wang, Y. Cao, et al., Appl. Catal. A: Gen. 630 (2022) 118457.  doi: 10.1016/j.apcata.2021.118457
    

37. [37]

    C.F. Lien, M.T. Chen, Y.F. Lin, et al., J. Chin. Chem. Soc. 51 (2004) 37-42.  doi: 10.1002/jccs.200400007
    

38. [38]

    H. Song, X. Meng, S. Wang, et al., J. Am. Chem. Soc. 141 (2019) 20507-20515.  doi: 10.1021/jacs.9b11440
    

39. [39]

    X. Cao, T. Han, Q. Peng, et al., ChemComm 56 (2020) 13918-13932.
    

40. [40]

    P.V.L. Reddy, K.H. Kim, H. Song, Renew. Sustain. Energy Rev. 24 (2013) 578-585.  doi: 10.1016/j.rser.2013.03.035
    

41. [41]

    N. Feng, H. Lin, H. Song, et al., Nat. Commun. 12 (2021) 4652.  doi: 10.1038/s41467-021-24912-0
    

42. [42]

    R. Sun, C. He, L. Fu, et al., Chin. Chem. Lett. 33 (2022) 527-532.  doi: 10.1016/j.cclet.2021.05.072
    

43. [43]

    L. Fu, R. Wang, C. Zhao, et al., Chem. Eng. J. 414 (2021) 128857.  doi: 10.1016/j.cej.2021.128857
    

44. [44]

    J. Yu, C. He, C. Pu, et al., Chin. Chem. Lett. 32 (2021) 3149-3154.  doi: 10.1016/j.cclet.2021.02.046
    

45. [45]

    Y. Zhao, W. Gao, S. Li, et al., Joule 3 (2019) 920-937.  doi: 10.1016/j.joule.2019.03.003
    

46. [46]

    Y. Tian, L. Piao, X. Chen, Green Chem. 23 (2021) 3526-3541.  doi: 10.1039/D1GC00658D
    

47. [47]

    W.B. Jiang, J.X. Low, et al., J. Univ. Sci. Technol. China 50 (2020) 1361.
    

48. [48]

    M. Harb, G. Jeantelot, J.M. Basset, J. Phys. Chem. C 123 (2019) 28210-28218.  doi: 10.1021/acs.jpcc.9b08145
    

49. [49]

    G. Kresse, J. Furthmüller, Phys. Rev. B 54 (1996) 11169.  doi: 10.1103/PhysRevB.54.11169
    

50. [50]

    G. Kresse, J. Furthmüller, Comput. Mater. Sci. 6 (1996) 15-50.  doi: 10.1016/0927-0256(96)00008-0
    

51. [51]

    J.P. Perdew, A. Ruzsinszky, G.I. Csonka, Phys. Rev. Lett. 100 (2008) 136406.  doi: 10.1103/PhysRevLett.100.136406
    

52. [52]

    J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865.  doi: 10.1103/PhysRevLett.77.3865
    

53. [53]

    G. Kresse, D. Joubert, Phys. Rev. B 59 (1999) 1758.
    

54. [54]

    H. Yuan, H. Yang, P. Hu, et al., ACS Catal. 11 (2021) 6835-6845.  doi: 10.1021/acscatal.1c01050
    

55. [55]

    H.F. Wang, D. Wang, X. Liu, et al., ACS Catal. 6 (2016) 5393-5398.  doi: 10.1021/acscatal.6b00764
    

56. [56]

    H.F. Wang, R. Kavanagh, Y.L. Guo, et al., Angew. Chem. Int. Ed. 51 (2012) 6657-6661.  doi: 10.1002/anie.201108981
    

57. [57]

    S. Grimme, J. Comput. Chem. 27 (2006) 1787-1799.  doi: 10.1002/jcc.20495
    

58. [58]

    S. Grimme, J. Antony, S. Ehrlich, et al., J. Chem. Phys. 132 (2010) 154104.  doi: 10.1063/1.3382344
    

59. [59]

    M. Setvin, C. Franchini, X. Hao, et al., Phys. Rev. Lett. 113 (2014) 086402.  doi: 10.1103/PhysRevLett.113.086402
    

60. [60]

    D. Wang, H.F. Wang, P. Hu, Phys. Chem. Chem. Phys. 17 (2015) 1549-1555.  doi: 10.1039/C4CP04159C
    

61. [61]

    J. Zhang, C. Peng, H. Wang, et al., ACS Catal. 7 (2017) 2374-2380.  doi: 10.1021/acscatal.6b03348
    

62. [62]

    D. Wang, T. Sheng, J. Chen, et al., Nat. Catal. 1 (2018) 291-299.  doi: 10.1038/s41929-018-0055-z
    

63. [63]

    H. Yuan, N. Sun, J. Chen, et al., ACS Catal. 8 (2018) 9269-9279.  doi: 10.1021/acscatal.8b02114
    

64. [64]

    H. Yuan, J. Chen, H. Wang, et al., ACS Catal. 8 (2018) 10864-10870.  doi: 10.1021/acscatal.8b03045
    

65. [65]

    G. Hu, Z. Wu, D.E. Jiang, J. Phys. Chem. C 122 (2018) 20323-20328.  doi: 10.1021/acs.jpcc.8b05251
    

66. [66]

    J. Zhang, P. Zhou, J. Liu, et al., Phys. Chem. Chem. Phys. 16 (2014) 20382-20386.  doi: 10.1039/C4CP02201G
    

67. [67]

    S. Wu, L. Wang, J. Zhang, J. Photochem. Photobiol. C Photochem. Rev. 46 (2021) 100400.  doi: 10.1016/j.jphotochemrev.2020.100400
    

68. [68]

    H. Sheng, Q. Li, W. Ma, et al., Appl. Catal. B 138 (2013) 212-218.
    

69. [69]

    B.D. Dunnington, J.R. Schmidt, J. Chem. Theory Comput. 8 (2012) 1902-1911.  doi: 10.1021/ct300002t
    

70. [70]

    M. Zhou, H.F. Wang, JACS Au. 2 (2022) 188-196.  doi: 10.1021/jacsau.1c00466
    

71. [71]

    P.J. Linstrom, W.G. Mallard, Chem. Eng. 46 (2001) 1059-1063.
    

72. [72]

    P. Pässler, W. Hefner, K. Buckl, et al., Ullmann's Encyclopedia of Industrial Chemistry, Wiley-VCH, Weinheim, 2000.

73. [73]

    K.M. Ervin, S. Gronert, S.E. Barlow, et al., J. Am. Chem. Soc. 112 (1990) 5750-5759.  doi: 10.1021/ja00171a013