Chinese Journal of Chemical Physics  2017, Vol. 30 Issue (3): 312-318

#### The article information

Liu Li-ming, Jin Zhao

Effects of Oxygen Vacancy on the Adsorption of Formaldehyde on Rutile TiO2(110) Surface

Chinese Journal of Chemical Physics, 2017, 30(3): 312-318

http://dx.doi.org/10.1063/1674-0068/30/cjcp1703049

### Article history

Accepted on: March 31, 2017
Effects of Oxygen Vacancy on the Adsorption of Formaldehyde on Rutile TiO2(110) Surface
Liu Li-ming, Jin Zhao
Dated: Received on March 22, 2017; Accepted on March 31, 2017
Hefei National Laboratory for Physical Sciences at Microscale, and Key Laboratory of Strongly-Coupled Quantum Matter Physics, Chinese Academy of Sciences, and Department of Physics, and Synergetic Innovation Center of Quantum Information & Quantum Physics, University of Science and Technology of China, Hefei 230026, China
*Author to whom correspondence should be addressed. Jin Zhao, E-mail:zhaojin@ustc.edu.cn
Abstract: Oxygen vacancy (Ov) has significant influence on physical and chemical properties of TiO2 systems, especially on surface catalytic processes.In this work, we investigate the effects of O v on the adsorption of formaldehyde (HCHO) on TiO2(110) surfaces through firstprinciples calculations.With the existence of Ov, we find the spatial distribution of surface excess charge can change the relative stability of various adsorption configurations.In this case, the bidentate adsorption at five-coordinated Ti (Ti5c) can be less stable than the monodentate adsorption.And HCHO adsorbed in Ov becomes the most stable structure.These results are in good agreement with experimental observations, which reconcile the long-standing deviation between the theoretical prediction and experimental results.This work brings insights into how the excess charge affects the molecule adsorption on metal oxide surface.
Key words: TiO2    Formaldehyde    Oxygen vacancy    Excess electrons
Ⅰ. INTRODUCTION

Titanium dioxide (TiO$_2$) is a versatile material for a wide range of applications, to name a few, as UV blocking pigments and sunscreen in industry, mixed conductor and synthetic single crystals semiconductor in electronic devices, most importantly, as abundant and toxic free photocatalyst which have been focused since the water splitting work done by Fujishima and Honda in 1972 [1].

Formaldehyde (HCHO) on TiO$_2$ is of particular interest because in numerous organic catalytic reactions HCHO is a key species (reactant, intermediate, or product) such as resins synthesis [2], CO$_2$ reduction [3], and hydrogen production [4]. Besides, HCHO is a common interior pollutant which causes severe health issues, TiO$_2$-based HCHO decomposing devices play a crucial role in the control of air pollution.

The adsorption of HCHO on TiO$_2$(110) surface has attracted attentions from both theoretical and experimental researchers for the last two decades [5-9]. Despite numerous innovational materials based on TiO$_2$ were synthesized to catalyze decomposing HCHO, the theoretical predictions and experimental observation of adsorption configurations of HCHO on TiO$_2$ were not in accordance, even on the most widely investigated rutile TiO$_2$ (110) surface. In 2001 for the first time, Idriss $et\; al$. theoretically addressed the adsorption configurations and energies of HCHO on TiO$_2$(110) surface using a cluster model based on density functional theory (DFT) calculations [10]. Although only one monodentate adsorption configuration was taken into consideration and the cluster model may not be an accurate description of extended surface, nevertheless, this work pioneered the ambition of using accurate DFT calculations to reveal the micro picture of aldehydes such as HCHO adsorption on metal oxides. In 2011, Haubrich $et\; al$. reported a systematic investigation of the effects of different surface and subsurface point defects on the adsorption of HCHO on TiO$_2$(110) surfaces through DFT calculations [11]. They declared a bidentate adsorption configuration which has almost twice the adsorption energy of other monodentate adsorption configurations on five-coordinated Ti (Ti$_\mathrm{5c}$). From then on, theoretical studies continued confirming that the bidentate adsorption configuration is the most stable structure. However, the bidentate adsorption configuration was hardly confirmed from experimental observations. For instance, in 2013, Yuan $et\; al$. reported the photocatalytic oxidation of methanol on TiO$_2$(110) surface by means of thermal desorption spectroscopy (TPD) and X-ray photoelectron spectroscopy (XPS) [7]. They found the formation of HCHO and the shift of C 1s level indicated that no bidentate configuration was detected. In early 2016, Yu $et\; al$. reported, at low coverage (sub-monolayer) and low temperatures (45 K to 65 K), the most abundant species of HCHO on TiO$_2$(110) surface is a chemisorbed monomer bounded to Ti$_\mathrm{5c}$ sites (Lewis acid) in a tilted monodentate configuration [12]. To understand the deviation between theoretical predictions and experimental observations is urgent.

Recently, the disagreement between the computational and experimental results started decreasing. In 2015, Zhang $et\; al$. reported that the spatial distribution of the extra charge near a O$_\mathrm{v}$ site strongly affected the binding of the Ti-bound formaldehyde, especially decreased the stability of bidentate adsorption configurations [9]. In late 2016, Feng $et\; al$. reported confirmation of the bidentate configurations of HCHO on Ti$_\mathrm{5c}$ through STM observations [13].

From the view of experimental endeavor, stoichiometric TiO$_2$ surfaces are very challenging to grow. They are easily being reduced or reconstructed with various defects and reconstruction. Among various point defects which make TiO$_2$ a typical $n$-type semiconductor, oxygen vacancy (O$_\mathrm{v}$) is very common. Removing one oxygen atom from the surface would break two Ti-O bonds and release two electrons back to surface, those unpaired electrons are defined as excess charge. It is believed to have significant influence on molecular adsorption and surface catalytic reactions. Therefore, a systematical study on how O$_\mathrm{v}$ affect the adsorption of HCHO is essential.

In this work, we reported the adsorption of HCHO on both stoichiometric and reduced TiO$_2$(110) surfaces. The interaction between molecular adsorption and excess charge was discussed in full details. The relative stability of various adsorption configurations in the presence of excess electrons induced by O$_\mathrm{v}$ was investigated through the comparison of geometries, energies and electronic structures. We found that the excess charge induced by O$_\mathrm{v}$ affects the adsorption of HCHO significantly. With the existence of O$_\mathrm{v}$, the bidentate adsorption at Ti$_\mathrm{5c}$, can be less stable than the monodentate adsorption. And HCHO adsorbed in O$_\mathrm{v}$ becomes the most stable structure. These results are in good agreement with experimental observations, which interpret the long-standing deviation between the theoretical prediction and experimental results.

Ⅱ. CALCULATION METHODS

Periodic DFT calculations were performed by using the vienna $ab$ $initio$ simulation package (VASP) [14, 15]. The generalized gradient approximation (GGA) functional was adopted with the Perdew-Burke-Ernzerhof (PBE) exchange-correlation description [16, 17]. The energy cutoff of 400 eV for plane-wave basis sets was used to expand the valence electronic wave function with the configurations of C(2s$^2$2p$^2$), H(1s$^1$), O(2s$^2$2p$^4$) and Ti(4s$^2$3d$^2$). The projector augmented wave (PAW) method was used to describe the electron-ion interaction. Dipole corrections were adopted to cancel the interactions between the excess charge and its periodic images for all calculations. A 5$\times$2 slab model containing 3 Ti-O-Ti tri-layer was chosen to simulate the TiO$_2$(110) surfaces. The atoms in the bottom Ti-O-Ti tri-layer were fixed to the positions within bulk TiO$_2$ during the structure optimization. To screen the effects of un-paired electrons from bottom Ti and O, the pseudo hydrogen saturation method was adopted. Since a Ti atom in TiO$_2$ bulk offers four electrons to bind to six O atoms and each O atom in bulk bonded to three Ti, pseudo hydrogens with the valence of 2/3 (H$_{0.66}$) and 4/3 (H$_{1.33}$) electron charge were added to bottom O and Ti correspondingly as shown in FIG. 1. The H$_{0.66}$-O and H$_{1.33}$-Ti bond lengths were determined to be 1.02 and 1.88 Å by the geometry optimization procedure which kept all Ti and O atoms fixed in bulk position then allowed the pseudo hydrogen atoms to relax. Then by fixing the bottom layer, the slab was optimized until the force on each atom is smaller than 0.02 eV/Å. To avoid the interlayer interaction, a vacuum layer of 15 Å was added between slabs. To correct the self-interaction error in DFT, we apply $U$=4.5 eV on Ti 3d orbitals.

 FIG. 1 Adsorption configurations of HCHO on stoichiometric TiO$_2$(110) surface ($\eta^1$-O$_\mathrm{b}$, $\eta^1$-O$_\mathrm{p}$ and $\eta^2$) in side and top views (along [001], [1-10], and [110] directions for upper, middle and lower schemas correspondingly in each box). The axis notations are omitted in the following for simplicity.

The formation energy ($E_\mathrm{formation}$) of oxygen vacancy is defined as:

 $E_\mathrm{formation}=E_\mathrm{perfect}-E_{\mathrm{O}_\mathrm{v}}-\frac{1}{2} E_{\mathrm{O}_2}$ (1)

$E_\mathrm{perfect}$ is the total energy of stoichiometric surface, $E_{\mathrm{O}_\mathrm{v}}$ is the total energy of the same surface with one oxygen vacancy and $E_{\mathrm{O}_2}$ is the energy of single oxygen molecule. To compare the stability of various adsorption configurations, we defined the adsorption energy ($E_\mathrm{ad}$) as:

 $E_\mathrm{ad}=E_\mathrm{TiO_2+HCHO}-E_\mathrm{TiO_2}-E_\mathrm{HCHO}$ (2)

where $E_\mathrm{TiO_2+HCHO}$ is the total energy of HCHO adsorbed on TiO$_2$(110) surface, $E_\mathrm{HCHO}$ and $E_\mathrm{TiO_2}$ are total energies of single HCHO molecule and clean TiO$_2$(110) surface correspondingly.

Ⅲ. RESULTS AND DISCUSSION A. The adsorption of HCHO on stoichiometric TiO$_\mathbf{2}$(110) surface

In agreement with previous reports [9, 11, 13, 18, 19], the calculated most stable adsorption configuration of HCHO on pseudo hydrogen saturated stoichiometric TiO$_2$(110) surface is the bidentate adsorption configuration ($\eta^2$) at Ti$_\mathrm{5c}$. FIG. 1 shows the top and side views of two monodentate ($\eta^1$) and one bidentate $\eta^2$ adsorption structures on Ti$_\mathrm{5c}$. For $\eta^2$ (FIG. 1(c)), two C-O bonds form sp$^3$ hybridization other than sp$^2$ hybridization as HCHO molecule in vacuum or monodentate adsorption configurations ($\eta^1$-O$_\mathrm{b}$, $\eta^1$-O$_\mathrm{p}$). Since $\eta^1$-O$_\mathrm{b}$ and $\eta^1$-O$_\mathrm{p}$ are almost identical either in geometry (FIG. 1(a) and (b)), adsorption energy or electronic structure [20], the $\eta^1$-O$_\mathrm{b}$ configuration was chosen as the representative for the two monodentate structures of HCHO at Ti$_\mathrm{5c}$ and remarked as $\eta^1$. The adsorption energy difference between bidentate and monodentate adsorption configurations is 0.7 eV as shown in Table Ⅰ.

Table Ⅰ Characteristic bonds (as notated in FIG. 1) and adsorption energies of monodentate and bidentate adsorption configurations at Ti$_\mathrm{5c}$ on reduced TiO$_2$(110) surface, total distortion (TD) is the sum of absolute bond-length change of corresponding bonds comparing to same adsorption configuration on stoichiometric surface ($\eta^1$$_-stoi and \eta^2$$_-$stoi).
B. Adsorption of HCHO on reduced TiO$_\mathbf{2}$(110) surface with O$_\mathrm{v}$ 1. BBO$_\mathrm{v}$ as a common defect on TiO$_2$(110) surface

For TiO$_2$(110) surface, bridge bonded oxygen vacancy (BBO$_\mathrm{v}$) is known as the most stable O$_\mathrm{v}$ structure. It is more stable than in-plane oxygen vacancy (IPO$_\mathrm{v}$) and O$_\mathrm{v}$ in sub-surface by 0.36 eV in our calculations. These results are in accordance with previous calculations [21]. Each O$_\mathrm{v}$ induces two excess electrons to the system. The two excess electrons prefer occupying sub-surface Ti which is in agreement with previous calculations [22, 23]. For reduced TiO$_2$(110) surface with BBO$_\mathrm{v}$, there are two sites favorable for HCHO adsorption. One is at surface Ti$_\mathrm{5c}$, the other is in BBO$_\mathrm{v}$.

2. HCHO adsorption on Ti$_\mathrm{5c}$

For the adsorption of HCHO at Ti$_\mathrm{5c}$ on reduced TiO$_2$(110) surface with BBO$_\mathrm{v}$, although adsorption configurations are preserved from those on stoichiometric surface, the adsorption energies and relative stability among these adsorption configurations are significantly affected by the presence of BBO$_\mathrm{v}$.

The first distinguishing feature is that the adsorption energies decrease about 0.4 eV for monodentate adsorption configuration ($\eta^1$) on reduced TiO$_2$(110) surface than on stoichiometric surface. We labeled different Ti$_\mathrm{5c}$ sites according to their relative position with BBO$_\mathrm{v}$ as online$_{-}$n, online$_{-}$nn, and offline as shown in FIG. 2. As shown in Table Ⅰ, for monodentate adsorption configurations, the adsorption energies for $\eta^1$$_{-}offline, \eta^1$$_{-}$online$_-$nn, $\eta^1$$_{-}online_{-}n are -1.24, -1.19, and \mbox{-1.17 eV}. Comparing to -0.83 eV of HCHO on stoichiometric surface in monodentate configurations, molecular adsorption on reduced surface is stabilized. The reason for the energy shift is the electrostatic interaction between excess charge induced by BBO_\mathrm{v} and dipole moment from adsorbed HCHO molecules.  FIG. 2 The top view (along [110] direction) of adsorption configurations of HCHO on TiO_2(110) surface with O_\mathrm{v}. Different configurations are labeled as "offline, online_-nn and online_-n" by the relative distance between the adsorption sites and O_\mathrm{v}. Another significant effect is the site sensitivity for bidentate adsorption configuration (\eta^2) of HCHO on reduced TiO_2(110) surface. It is clearly presented that, for \eta^2 configurations, the adsorption energy could vary from -1.99 eV to -1.12 eV at different Ti_\mathrm{5c} sites (FIG. 2(a)-(c)). The adsorption energy dependents on the distance between HCHO and BBO_\mathrm{v}. At offline site, HCHO binds to bridge bonded oxygen which has no O_\mathrm{v} in the same row. The local environment is similar to that on stoichiometric surface except there is electrostatic interaction between sub-surface excess electron and adsorbed HCHO molecule. Thus, the adsorption energy decreased from -1.57 eV on stoichiometric surface to -1.99 eV on reduced surface. The 0.4 eV energy shift is similar to the adsorption energies change of monodentate adsorption configuration (\eta^1) between stoichiometric and reduced surfaces. For HCHO at Ti_\mathrm{5c} closer to BBO_\mathrm{v} (\eta^2$$_-$online$_-$nn and $\eta^2$$_-online_-n), lattice distortion becomes larger after the molecular adsorption in Table Ⅰ. The aggravated distortion would dramatically affect the spatial distribution of excess electrons. The changed distribution of excess electrons alters the electrostatic interaction strength and the adsorption energies for \eta^2$$_-$online$_-$nn and $\eta^2$$_-online_-n become -1.76 and -1.12 eV. The significant site sensitivity of bidentate adsorption configuration is due to the change of the localization of excess charge induced by BBO_\mathrm{v}. For \eta^2$$_-$offline, similar to the monodentate adsorption configurations, the excess charge all localized in sub-surface separately as shown in FIG. 4(a)-(d) correspondingly. In this case, the overall electrostatic interaction is attractive leading to stabilize molecular adsorption. For bidentate adsorption structures ($\eta^2$$_-online_-nn and \eta^2$$_-$online$_-$n), the aggravated distortion induced by molecular adsorption would drive excess charge to surface localizing closely around BBO$_\mathrm{v}$ site then the overall electrostatic interaction is repulsive and the molecular adsorption is destabilized as shown in FIG. 4(e) and (f).

 FIG. 4 Distribution of O$_\mathrm{v}$ induced excess electrons after adsorption of HCHO on reduced TiO$_2$(110) in side (along [001] direction) and top (along [110] direction) views. The dashed red circles filled with white represent the position of BBO$_\mathrm{v}$ while the empty dashed red circles depict healed BBO$_\mathrm{v}$ by HCHO adsorption. Spatial distribution (orbitals) of two excess electrons were depicted in yellow.

It should be noticed that the bidentate adsorption configuration, which is the most stable structure on stoichiometric surface, could become the least stable one with the interaction with O$_\mathrm{v}$. Especially when HCHO adsorbing at Ti$_\mathrm{5c}$ close to BBO$_\mathrm{v}$ in FIG. 3, the bidentate adsorption configuration ($\eta^2$$_-online_-n) is less stable than all other monodentate adsorption configurations.  FIG. 3 Comparison of adsorption energy vs. adsorption site between monodentate and bidentate adsorption configurations on stoichiometric and reduced TiO_2(110) surfaces. 3. HCHO adsorption in BBO_\mathrm{v} The adsorption of HCHO in BBO_\mathrm{v} heals the vacancy in a manner of inserting oxygen end of HCHO into the vacancy. The symmetric adsorption configuration (sym_-$$\eta^1$-O$_\mathrm{v}$) with molecular oxygen bonded to two five-folded Ti in BBO$_\mathrm{v}$ has the adsorption energy of -1.46 eV (FIG. 5(a)-(c)). Beside this symmetric monodentate adsorption configuration (FIG. 5(a)), another asymmetric monodentate adsorption structure was found with an extra hydrogen bond between molecular H and adjacent bridge oxygen (sym$_-$$\eta^1-O_\mathrm{v}). Despite an extra hydrogen bond formed (FIG. 5(b)), the asymmetric adsorption configuration has the adsorption energy of -1.45 eV which is similar to symmetric one because the Ti-O_\mathrm{m} bonds are stretched on one side and compressed on the other side. Notice that the adsorption energies of monodentate adsorption configurations at BBO_\mathrm{v} are, not only about 0.6 eV lower than monodentate adsorption configurations on stoichiometric surface, but also about 0.2 eV lower than monodentate adsorption configurations at Ti_\mathrm{5c} with O_\mathrm{v}.  FIG. 5 Adsorption configurations of HCHO in BBO_\mathrm{v} on TiO_2(110) surface in side and top views (along [001] [1-10], and [110] directions for upper, middle and lower schemas correspondingly in each box). A bidentate adsorption configuration (\eta^2-O_\mathrm{v}) can be formed with an extra chemical bond of C-O_\mathrm{b}. The adsorption energy of \eta^2-O_\mathrm{v} is -1.70 eV which is the most stable one among all presented adsorption configurations of HCHO close to BBO_\mathrm{v} (FIG. 5(c)). In other words, if there are BBO_\mathrm{v}s on TiO_2(110) surface, HCHO favors adsorbing at O_\mathrm{v} sites rather than Ti_\mathrm{5c} sites near BBO_\mathrm{v}. The adsorption of HCHO in BBO_\mathrm{v} has minor influence on the distribution of excess charge. For sym_-$$\eta^1$-O$_\mathrm{v}$ and asym$_-$$\eta^1-O_\mathrm{v}, a small amount of excess electron (less than 20%) involved in the Ti-O_\mathrm{m} bond formation, the majority of excess electrons (larger than 80%) still locate at sub-surface (FIG. 4(g) and (h)). For \eta^2-O_\mathrm{v} configuration, another C-O bond forming sp^3 hybridization then no need of excess charge for chemical bonding thus the excess electrons remain localized in sub-surface (FIG. 4(i)). 4. Effects of HCHO adsorption on the electronic structure Molecular adsorption affects the electronic structure of reduced TiO_2(110) surface in two ways. On one hand, molecular orbitals could emerge in band gap as HOMO or LUMO of the system; on the other hand, the defect states originated from BBO_\mathrm{v} could be affected by molecular adsorption. For monodentate adsorption configurations at Ti_\mathrm{5c} (\eta^1$$_-$offline, $\eta^1$$_-online_-nn, and \eta^1$$_-$online$_-$n), the highest occupied molecular orbitals are about 0.3 eV bellow the valence band maximum (VBM) of TiO$_2$ (FIG. 6). The gap states originated from BBO$_\mathrm{v}$ are stabilized by molecular adsorption with the downward energy shift of 0.5 eV compared to clear reduced surface. The stabilization of excess charge states is because of the electrostatic attraction between adsorbed HCHO and excess electrons.

 FIG. 6 Density of states (DOS) for various adsorption configurations of HCHO on reduced TiO$_2$(110) surface. Black line represents total DOS and red line represents the partial DOS of HCHO with the magnification of 10 times for clarity.

For bidentate adsorption configurations at Ti$_\mathrm{5c}$ ($\eta^2$$_-offline, \eta^2$$_-$online$_-$nn, and $\eta^2$$_-online_-n), molecular orbital contributes a gap state about 0.2 eV above VBM. As the adsorption of HCHO approaching to BBO_\mathrm{v} (from \eta^2$$_-$offline to $\eta^2$$_-online_-nn then \eta^2$$_-$online$_-$n), the molecular gap state shifts upward a little bit due to the larger local structural distortion in Table Ⅰ. Gap states induced by BBO$_\mathrm{v}$ are stabilized for $\eta^2$$_-offline and \eta^2$$_-$online$_-$nn configurations also due to the electrostatic attraction between adsorbed HCHO and excess electrons. For $\eta^2$$_-online_-n configuration which is close to BBO_\mathrm{v}, these excess charge gap states are destabilized by shifting up about 0.1 eV because the two excess electrons are closely distributed around BBO_\mathrm{v} site due to molecular adsorption. As for HCHO adsorbing in BBO_\mathrm{v} in sym_-$$\eta^1$-O$_\mathrm{v}$ and asym$_-$$\eta^1-O_\mathrm{v} configurations, an empty state from the hybridization of O 2p and Ti 3d emerged about 0.7 eV bellow conduction band minimum (CBM). The excess charge states have about 0.5 eV energy downward shift similar to the monodentate adsorption at Ti_\mathrm{5c} (\eta^1$$_-$offline, $\eta^1$$_-online_-nn, and \eta^1$$_-$online$_-$n). Molecular adsorption is stabilized by excess charge located mainly (larger than 80%) in sub-layer. For HCHO in BBO$_\mathrm{v}$ in $\eta^2$-O$_\mathrm{v}$ configuration, molecular O 2p orbitals contribute a gap state about 0.1 eV above VBM which is similar to HCHO at Ti$_\mathrm{5c}$ in $\eta^2$ configuration. But the excess charge is still located in sub-surface which is similar to HCHO at Ti$_\mathrm{5c}$ in $\eta^1$ configuration.

From above analysis, we argued that the bidentate adsorption configurations of HCHO at Ti$_\mathrm{5c}$ have great potential to be the hole scavenger, while HCHO in BBO$_\mathrm{v}$ in the sym$_-$$\eta^1-O_\mathrm{v} and asym_-$$\eta^1$-O$_\mathrm{v}$ configurations have the great potential to be the electron scavenger in photocatalytic process on TiO$_2$(110) surfaces. And molecular adsorption could either stabilize or destabilize the excess charge states by altering the electrostatic interaction between molecular adsorption and BBO$_\mathrm{v}$.

Ⅳ. Conclusion

In this work, we investigated the adsorption of HCHO on stoichiometric and reduced TiO$_2$(110) surfaces through first-principles calculations. Compared to stoichiometric surface, oxygen vacancy in reduced TiO$_2$ would induce excess electrons to the system and affect the relative stability of molecular adsorption configurations. For HCHO adsorbing at Ti$_\mathrm{5c}$, on one hand, excess electrons would stabilize molecular adsorption for monodentate adsorption. On the other hand, for bidentate adsorption, the adsorption stability depends strongly on the adsorption site. The adsorption is stabilized when the adsorption is far away from BBO$_\mathrm{v}$. Yet it can be significantly destabilized when the adsorption is close to BBO$_\mathrm{v}$ and becomes less stable than the monodentate adsorption. The adsorption in BBO$_\mathrm{v}$ is the most stable configuration near BBO$_\mathrm{v}$ among all the configurations we investigated. Our calculations reconcile the long-standing discrepancies between theoretical predictions and experimental observations of the HCHO/TiO$_2$ system and provide valuable insights into the effects of O$_\mathrm{v}$ on the molecular adsorption on TiO$_2$.

Ⅴ. Acknowledgments

This work was supported by the National Natural Science Foundation of China (No.21373190, No.21421063, No.11620101003) and the National Key Foundation of China, Department of Science & Technology (No.2016YFA0200600 and No.2016YFA0200604), the Fundamental Research Funds for the Central Universities of China (No.WK3510000005), the support of National Science Foundation (No.CHE-1213189 and No.CHE-1565704). Calculations were performed at Environmental Molecular Sciences Laboratory at the PNNL, a user facility sponsored by the DOE Office of Biological and Environmental Research.

Reference
 [1] A. Fujishima, and K. Honda, Nature 238 , 37 (1972). DOI:10.1038/238037a0 [2] Y. H. Lee, C. A. Kim, W. H. Jang, H. J. Choi, and M. S. Jhon, Polymer 42 , 8277 (2001). DOI:10.1016/S0032-3861(01)00342-1 [3] B. Kaesler, and P. Schönheit, Eur. J. Biochem. 186 , 309 (1989). DOI:10.1111/ejb.1989.186.issue-1-2 [4] C. Winner, and G. Gottschalk, FEMS Microbiol. Lett. 65 , 259 (1989). [5] G. Busca, J. Lamotte, J. C. Lavalley, and V. Lorenzelli, J. Am. Chem. Soc. 109 , 5197 (1987). DOI:10.1021/ja00251a025 [6] C. B. Xu, W. S. Yang, Q. Guo, D. X. Dai, T. K. Minton, and X. M. Yang, J. Phys. Chem. Lett. 4 , 2668 (2013). DOI:10.1021/jz401349q [7] Q. Yuan, Z. F. Wu, Y. K. Jin, L. S. Xu, F. Xiong, Y. S. Ma, and W. X. Huang, J. Am. Chem. Soc. 135 , 5212 (2013). DOI:10.1021/ja400978r [8] Q. Yuan, Z. F. Wu, Y. K. Jin, F. Xiong, and W. X. Huang, J. Phys. Chem. C 118 , 20420 (2014). DOI:10.1021/jp5061733 [9] Z. R. Zhang, M. R. Tang, Z. T. Wang, Z. Ke, Y. B. Xia, K. T. Park, I. Lyubinetsky, Z. Dohnálek, and Q. F. Ge, Top. Catal. 58 , 103 (2015). DOI:10.1007/s11244-014-0349-6 [10] L. Kieu, P. Boyd, and H. Idriss, J. Mol. Catal. A-Chem. 176 , 117 (2001). DOI:10.1016/S1381-1169(01)00231-X [11] J. Haubrich, E. Kaxiras, and C. M. Friend, Chemistry 17 , 4496 (2011). DOI:10.1002/chem.201002588 [12] X. J. Yu, Z. R. Zhang, C. W. Yang, F. Bebensee, S. Heissler, A. Nefedov, M. R. Tang, Q. F. Ge, L. Chen, B. D. Kay, Z. Dohnálek, and Y. M. Wang, C. Wöll Christof, J. Phys. Chem. C 120 , 12626 (2016). DOI:10.1021/acs.jpcc.6b03689 [13] H. Feng, L. M. Liu, S. H. Dong, X. F. Cui, J. Zhao, and B. Wang, J. Phys. Chem. C 120 , 24287 (2016). DOI:10.1021/acs.jpcc.6b08797 [14] G. Kresse, and J. Hafner, Phys. Rev. B 48 , 13115 (1993). DOI:10.1103/PhysRevB.48.13115 [15] G. Kresse, and J. Furthmuller, Phys. Rev. B 54 , 11169 (1996). DOI:10.1103/PhysRevB.54.11169 [16] J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. 77 , 3865 (1996). DOI:10.1103/PhysRevLett.77.3865 [17] J. P. Perdew, M. Emzerhof, and K. Burke, J. Chem. Phys. 105 , 9982 (1996). DOI:10.1063/1.472933 [18] H. Z. Liu, X. Wang, C. X. Pan, and K. M. Liew, J. Phys. Chem. C 116 , 8044 (2012). DOI:10.1021/jp210465u [19] M. R. Tang, Z. R. Zhang, and Q. F. Ge, Catal. Today 274 , 103 (2016). DOI:10.1016/j.cattod.2016.01.057 [20] L. M. Liu, and J. Zhao, Surf. Sci. 652 , 156 (2016). DOI:10.1016/j.susc.2015.12.036 [21] M. V. Ganduglia-Pirovano, A. Hofmann, and J. Sauer, Surf. Sci. Rep. 62 , 219 (2007). DOI:10.1016/j.surfrep.2007.03.002 [22] P. M. Kowalski, M. F. Camellone, N. N. Nair, B. Meyer, and D. Marx, Phys. Rev. Lett. 105 , 146405 (2010). DOI:10.1103/PhysRevLett.105.146405 [23] T. Shibuya, K. Yasuoka, S. Mirbt, and B. Sanyal, J. Phys. Condens. Matter 24 , 435504 (2012). DOI:10.1088/0953-8984/24/43/435504