Modified LaTiO2N Photocatalyst with Spatially

FULL PAPER Photocatalysts

www.afm-journal.de

La2O3-Modified LaTiO2N Photocatalyst with Spatially Separated Active Sites Achieving Enhanced CO2 Reduction Lei Lu, Bing Wang, Shaomang Wang, Zhan Shi, Shicheng Yan,* and Zhigang Zou And then, the CO2 is reduced by photo generated electrons with the protons to form hydrocarbon compounds such as CH4, a common product with eightelectron transfer.[6] This means that the CO2 reduction reaction is difficult in kinetics due to the multielectron process with high energy barriers in the two half reactions. Molecule activation was demonstrated to be an effective method to decrease the barriers. Solid base (e.g., alkali metal oxides) can activate CO2 to lower-energy carbonate (CO32−) species as a result of the basic sites from O2− ions of solid base chemically reacting with the acidic CO2 molecule.[10–15] Thus, a positive cor relation between CH4 yield and surface basicity of catalysts was observed with the modification of MgO or NaOH on catalyst surface.[14,15] Recently, alkaline lanthanum cation-containing semiconductors, such as LaPO4 and La2Sn2O7, were developed as photocatalyst for CO2 reduction, due to the fact that the alkaline lanthanum is more favored for CO2 adsorption.[16–19] In addition, it has been theo retically and experimentally demonstrated that crystal defects can also serve as active sites in photocatalytic reactions.[20–22] Scanning tunnel microscope observations have found that a strong interaction between H2O molecule and oxygen vacan cies (OVs) on (110) surface of TiO2 can lead to dissociation of H2O to surface hydroxyl.[20] And theoretical calculations also confirmed that CO2 molecule adsorbed at surface OVs on the same crystal face of TiO2 was in favor of the CO2 reduction into CO.[21] These results indicated that introducing solid base or OV defects on the surface of catalysts can lower the energy barrier of the CO2 reduction reaction by activating CO2 or H2O. How ever, a challenge to simultaneously activate the H2O and CO2 for achieving rate-matching two half reactions still remains due to the lack of advanced material design strategy. Here, we put forward a material design concept to simultane ously activate the CO2 and H2O at different spatial sites using a solid base to modify the photocatalyst with vacancy defects. And La2O3 modified LaTiO2N (La2O3/LaTiO2N) with OV defects were synthesized to identify this design strategy, which was achieved by simple one-step nitridation of La2TiO5 under ammonia flow (NH3). On the surface of LaTiO2N, the La2O3 and OV defects acting as active sites could effectively activate CO2 and H2O into CO32− and OH species, which lowered the CO2 reduction reac tion barriers and improved the proton release from kinetically sluggish water oxidation. It should be noted that the OV sites

Activating CO2 molecule and promoting proton release from kinetically sluggish water oxidation are two important half-reaction processes for achieving efficient solar-driven conversion of CO2 to fuels. Here, an effective route is proposed that uses a solid base to modify photocatalyst with defects, aiming to simultaneously accelerate the two reaction processes. To verify this hypothesis, the La2O3 is decorated on surface of LaTiO2N with oxygen vacancies, achieving a twofold increase in CH4 yield rate for CO2 reduction. The prominent activity results from the following two effects: (1) The O2− in La2O3 as basic sites favors CO2 chemisorption in the form of CO32− species, greatly contributing to both the bending of OCO bond and the decrease of the lowest unoccupied molecule orbit energy of CO2 molecule. (2) The oxygen vacancies on LaTiO2N are beneficial in activating H2O to adsorbed OH, thus effectively decreasing the reaction barriers of water oxidation to release protons. The design concept of simultaneously activating the CO2 and H2O at different spatial sites may offer a new strategy to suppress the reverse reactions for efficient solar energy conversion.

1. Introduction Photocatalytically converting CO2 using H2O as reducing agent into highly value-added solar fuels has been regarded as a promising strategy for a sustainable energy future.[1–5] In this reaction process, the four-electron oxidation of H2O by photogenerated holes to produce oxygen and protons,[6] which is considered to be the rate-determining step for water split ting due to high reaction barriers,[7–9] is needed to occur first. L. Lu, Dr. B. Wang, Dr. S. Wang, Z. Shi, Prof. S. Yan, Prof. Z. Zou Eco-Materials and Renewable Energy Research Center (ERERC) National Laboratory of Solid State Microstructures College of Engineering and Applied Sciences Nanjing University Nanjing 210093, P. R. China E-mail: [email protected] L. Lu, Dr. B. Wang, Dr. S. Wang, Z. Shi, Prof. S. Yan, Prof. Z. Zou Collaborative Innovation Center of Advanced Microstructures Nanjing University Nanjing 210093, P. R. China Dr. B. Wang, Z. Shi, Prof. Z. Zou Jiangsu Key Laboratory for Nano Technology Department of Physics Nanjing University Nanjing 210093, P. R. China The ORCID identification number(s) for the author(s) of this article can be found under https://doi.org/10.1002/adfm.201702447.

DOI: 10.1002/adfm.201702447

Adv. Funct. Mater. 2017, 27, 1702447

1702447 (1 of 11)

© 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

www.advancedsciencenews.com

www.afm-journal.de

used to activate H2O played a significant role on the capacity and stability of the photocatalytic CO2 reduction over its activa tion of CO2. The separated activation sites induced the separa tion of the redox reactions, thus helping to reduce the recom bination of photogenerated charges and suppress the reverse reaction. Optimizing the La2O3 and OV contents on the surface of LaTiO2N to simultaneously activate CO2 and H2O can accel erate and lead to rate-matching half reactions. As a result, the optimal CH4 yield rate for CO2 reduction on La2O3/LaTiO2N showed a twofold higher value than the pure-LaTiO2N. The design concept of simultaneously activating the reactants of CO2 and H2O at different spatial sites may offer a new strategy for an efficient solar energy conversion.

2. Results and Discussion 2.1. Photocatalyst Synthesis and Characterization To obtain our designed model catalyst, the solid La2TiO5 pre cursors were first prepared by heating stoichiometric mix tures of La2O3 and TiO2 in an eutectic salt of NaCl and KCl at 1373 K. X-ray diffraction (XRD, Figure 1a) pattern indicated that the as-prepared product is single-phase orthorhombic La2TiO5 (JCPDS No. 15-0335). Scanning electron microscope

(SEM, Figure 1b) observations showed that the single-crystal La2TiO5 microrods with 100–500 nm diameter and 1–3 µm length was obtained during the flux growth. After nitriding the as-prepared La2TiO5 under NH3 flow at 1223 K, the according XRD analysis of the products were well indexed to mixtures of La2O3 (JCPDS No. 05-0602) and LaTiO2N (JCPDS No. 48-1230), meaning a phase transformation via reaction of 2La2TiO5 + 2NH3 → 2LaTiO2N + La2O3 + 3H2O.[23] Usually, the nitriding reaction driven solid-state oxide phase transformation is pseudomorphic and topotactic, and a porous structure in a single-crystal particle could be formed, resulting from the volume shrinkage during replacing the three O atoms by two N atoms.[24] In our case, SEM image of the products also showed porous structures with apparent profile of La2TiO5 precursor (Figure 1c), though the phase transformation pro cess from La2TiO5 to LaTiO2N involved a phase segregation of La2O3. Transmission electron microscope (TEM) showed that the porous structure was composed of nanocrystals with 50–100 nm diameter and 100–200 nm length (Figure 1d). And high-resolution TEM (HR-TEM) lattice images and selected area electron diffraction (SAED) observations presented that the La2O3 was distributed on the surface of LaTiO2N nanocrys tals, and a compact interface between La2O3 and LaTiO2N could be observed (Figure 1e). This indicated that the apparent porous structure was composed of columnar-like LaTiO2N, and

Figure 1. a) XRD patterns for La2TiO5, xLa2O3/LaTiO2N (x = 0.5, 0.39, 0.3, and 0.16), and pure-LaTiO2N. b) The SEM image of La2TiO5. c) SEM image, d) TEM image, and e) HR-TEM lattice image for 0.5La2O3/LaTiO2N. f) HR-TEM lattice image of 0.3La2O3/LaTiO2N. g) HR-TEM of pure-LaTiO2N. The insets in (e–g) show the SAED patterns.

Adv. Funct. Mater. 2017, 27, 1702447

1702447 (2 of 11)



www.afm-journal.de

the La2O3 was acted as solid binder for the porous framework, which was much different to the previous reports on formation of porous structure in a single-crystal particle with profile of oxide precursor. The formations of LaTiO2N nanocrystals may result from the phase segregation of La2O3 in the oxide pre cursor framework, thus obstructing LaTiO2N particle connec tion among these nanocrystals during their growing process. Energy dispersive X-ray detector (EDX) analysis demon strated that the La/Ti atomic ratio for the nitriding product from La2TiO5 was about 2.01:1, confirming no metal element volatili zation during the nitriding reaction. After acid impregnation by hydrochloric acid (HCl) solutions (pH = 1) with different molar amounts to adjust the La2O3 content, several La2O3 modified LaTiO2N samples were obtained (denoted as xLa2O3/LaTiO2N, where x represented the mole amount of La2O3). EDX analysis indicated that the La/Ti ratio decreased from 1.82:1 (x = 0.39) to 1.53:1 (x = 0.3), and to 1.19:1 (x = 0.16) with increasing of the mole amount of acid. The removed La2O3 content was much close to that of the mole amount of the acid used (Figure S1 and Table S1, Supporting Information), meaning that the HCl could react with the La2O3 fully. Indeed, a LaTiO2N with La/Ti mole ratio of 1.06:1 could be obtained by adding an excess of acid to remove La2O3 (denoted as pure-LaTiO2N). Compared with the La2O3/LaTiO2N (x = 0.5) from direct nitridation of La2TiO5,

SEM observations showed that the acid treatment made the LaTiO2N particles with profile of La2TiO5 precursor dissolve into monodispersed nanocrystals (Figure S2, Supporting Information). However, TEM observations showed no visible changes in profiles and particle size of LaTiO2N nanocrystals after the acid treatment with different mole contents (Figure S3, Supporting Information). The XRD patterns exhibited that the LaTiO2N maintained well crystallinity after the acid impregna tion, indicating that the LaTiO2N was stable in acid solution. The difference was that the XRD diffraction peaks of La2O3 dis appeared once with the acid treatment, although only with the partial removal of La2O3. It was found that the polycrystalline diffraction ring in SAED gradually weakened with decreasing x value from 0.5 to 0.3 and disappeared when La2O3 was com pletely removed (Figure 1e–g). And high-resolution TEM lattice images of La2O3 gradually disordered with increasing of the used amount of acid, indicating the formation of amorphous La2O3 due to the acid-induced destruction of crystal lattice. The partial removal of La2O3 coating layer would be beneficial to the formation of monodispersed LaTiO2N nanocrystals. As shown in the ultraviolet–visible (UV–vis) absorption spectra (Figure 2a), the 0.5La2O3/LaTiO2N showed an absorp tion edge up to 600 nm. With decreasing the amount of La2O3, the absorption onsets of the samples gradually shifted to longer

Figure 2. a) UV–vis absorption spectra for various LaTiO2N samples. b) EPR, c) Ti 2p, and d) O 1s core-level XPS spectra for xLa2O3/LaTiO2N (x = 0.5 and 0.3) and pure-LaTiO2N.

Adv. Funct. Mater. 2017, 27, 1702447

1702447 (3 of 11)



www.afm-journal.de

particular, OV related peak intensity in LaTiO2N with complete removal of La2O3 is about 33%, higher than that in LaTiO2N with La2O3 phase. These evidences indicated that such passi vated surface lattice oxygen was unstable and could be released again by acid treatment. Therefore, compared with the LaTiO2N with partially removed La2O3, the one with complete removal of La2O3 exhibited about 20 nm redshift in main band absorp tion due to the full release of surface defect states. High-density surface defects were demonstrated to be able to affect the main band absorption due to the formation of local defect energy levels.[37] Usually, the energy levels of the surface state traps fall within the band gap of the semiconductor. Many experi mental evidences have verified that surface OV defects are an electron donor with an energy level slightly below the conduc tion band edge of an n-type conducting semiconductor. Under irradiation, the photo generated electrons from valence band of semiconductor can be trapped in the OVs, thus inducing a redshift of main band edge. The electrons trapped in OVs were easily thermoexcited to conduction band of semiconductor to take part in the subsequent chemical reactions. Compared with 0.5La2O3/LaTiO2N, the LaTiO2N without La2O3 modification from directly nitriding La2Ti2O7 (denoted as S-LaTiO2N) showed similar adsorption edge redshift and absorption trail to the pureLaTiO2N, attributing to the formation of reduced Ti species.[9] After acid treatment on S-LaTiO2N, no visible changes in XRD patterns, SEM morphologies, and light absorption (Figure S5, Supporting Information) confirmed that the LaTiO2N was stable in the acid solution when no La2O3 formed on its surface. This further demonstrated that the well-regulated increase in oxygen vacancies originated from the acid-released surface lattice oxygen when the surface La2O3 was gradually removed.

wavelength, and an obvious absorption trail in the region above 600 nm was observed for the LaTiO2N with completely removed La2O3. Usually, the absorption trail was strongly asso ciated to the light absorption process of defect states.[25,26] Elec tron paramagnetic resonance (EPR) signal at g = 2.003 could be found in all the as-prepared samples (Figure 2b), which was typically assigned to unpaired electrons trapped by OVs.[27] Meanwhile, the deconvolution of Ti 2p3/2 X-ray photoelectron spectroscopy (XPS) peaks rendered two peaks located at about 457.4 and 456.3 eV (Figure 2c), respectively, assigned to Ti4+ and Ti3+ species.[28] Therefore, the obvious absorption trail may be attributed to the light absorption from oxygen vacancies or Ti3+. A similar light absorption phenomena was observed in the hydrogen reduced black TiO2.[29,30] In our case, during the hightemperature nitriding, the surface Ti4+ of LaTiO2N nanocrys tals was relatively easy to be reduced into Ti3+ species due to the strong reductive capacity of NH3 at high temperature, thus the OVs as an associated defect was formed inevitably due to the thermodynamics requirement and charge balance for stabiliza tion of such defects.[9,29] Such a mechanism of coexisting Ti3+ defect and OVs was well demonstrated in the TiO2.[31–33] Decreasing the amount of La2O3 by acid washing, the sig nals of Ti3+ in XPS and OVs in EPR both increased. This indi cated that the La2O3 could passivate the surface defect states of LaTiO2N nanocrystals, probably resulting from a defect pas sivation process of xTi3+ + 0.5x OVs(catalyst) + La2O3 → xTi4+ + 0.5xO2−(catalyst) + La2O3–0.5x. Negligible differences for La 3d and N 1s XPS spectra were observed for these prepared LaTiO2N samples with different La2O3 amounts (Figure S4a,b, Sup porting Information), indicating the almost identical chemical environment for La and N elements. The O 1s XPS spectra for the LaTiO2N with different La2O3 amounts could be deconvo luted into four peaks (Figure 2d). The binding energies located at 532.4 and 529.2 eV were, respectively, assigned to surface adsorbed MOH species and TiOTi lattice oxygen.[34] XPS peak at 530.6 eV was attributed to lattice oxygen of LaOLa in La2O3 (Figure S4c, Supporting Information) and peak at 531.2 eV belonged to lattice oxygen atoms in the vicinity of OVs.[35,36] Evidently, the OV related peak intensity at 531.2 eV increased with decrease of the peak intensity of LaOLa. In

2.2. Photocatalytic CO2 Reduction

1.2 1.0

0.98

-1

-1

b CH4 yield rate (µ mol gcat h )

The CO2 photoreduction tests over these as-prepared LaTiO2N samples were performed under irradiation by visible light (λ ≥ 420 nm) in the presence of H2O. As shown in Figure 3a, all LaTiO2N samples exhibited increased CH4 yields with pro longing of the irradiation time, while no CH4 could be detected

0.8

0.73 0.62

0.6 0.4

0.40

0.34

0.2 0.0

l l 0.5 mo 0.39 mo

0.3 mo

l

0.16 m

ol

0 m ol

Solid base amount

Figure 3. a) CH4 generation over LaTiO2N with different mole amounts of La2O3 as a function of irradiation time. b) The average CH4 production rate over LaTiO2N with different mole amounts of La2O3 under another 7 h irradiation except the first hour.

Adv. Funct. Mater. 2017, 27, 1702447

1702447 (4 of 11)



www.afm-journal.de

either from experiments performed over Table 1. BET surface areas, pore volumes, and the total amounts of adsorbed CO2 on the pure La2O3 or a dark reaction in the presence as-prepared photocatalysts. of La2O3/LaTiO2N. When performing isotope Pore volume CO2 adsorption quantity CO2 adsorption quantity Surface area labeling experiment by using 13CO2 as carbon Sample [× 10−2 cm3 g−1] [m2 g−1] [mg g−1] [mg m−2] source, 13CH4 signal was detected by gas 6.09 2.65 2.38 0.39 chromatography–mass spectrometry in the 0.50 La2O3/LaTiO2N gas products (Figure S6, Supporting Infor 0.39 La2O3/LaTiO2N 14.27 7.57 2.67 0.19 mation). This indicated that CH4 formation 0.30 La2O3/LaTiO2N 15.03 7.93 2.97 0.20 originated from photocatalytic reduction of 15.66 7.97 3.08 0.20 0.16 La2O3/LaTiO2N CO2 over LaTiO2N. All the samples presented 15.65 7.64 3.12 0.20 Pure-LaTiO2N a relatively high CH4 yield in the first hour and a stable CH4 yield in the subsequent 7 h. (BET, Figure S8, Supporting Information) plot ranging from The fast CH4 generation in the first hour probably resulted P/P0 = 0.05 to 0.15 was increased from 6.09 m2 g−1 for 0.5La2O3/ from the sufficient CO2 adsorption on the fresh surface of the catalysts. Given that the reaction of CO2 + 2H2O → CH4 + 2O2 LaTiO2N to 14.27–15.65 m2 g−1 for the La2O3/LaTiO2N samples was involved in eight-electron transfer process, accompanying after acid treatment. Correspondingly, after acid treatment, the formation of CH4, some intermediate species with high the pore volume calculated from the nitrogen adsorption iso therm increased from 0.0265 to 0.0757–0.0797 cm3 g−1. The requirement in overpotentials may form on the surface of catalysts, thus affecting the reaction kinetics. The stable CH4 increase in both specific surface area and pore volume was ascribed to the removal and amorphization of La2O3 during the generation rate was obtained from the slope of the yield curves shown in Figure 3a within the later 7 h irradiation after the acid treatment, as confirmed by the SEM and TEM observa first hour. As shown in Figure 3b, with decreasing the con tions. However, CH4 generation rate (0.4 µmol gcat−1 h−1) over tent of La2O3, CH4 evolution rate revealed volcanic type dis the 0.5La2O3/LaTiO2N (SBET = 6.09 m2 g−1) was slightly higher tribution. The 0.3La2O3/LaTiO2N exhibited the highest CH4 than that (0.34 µmol gcat−1 h−1) of the pure-LaTiO2N (SBET = yield rate of 0.98 µmol gcat−1 h−1, almost 2 times higher than 15.65 m2 g−1), indicating that the increased specific surface area 0.34 µmol gcat−1 h−1 for the pure-LaTiO2N. This meant that the was not responsible for the increase in CH4 yield of catalysts. La2O3 on the surface of LaTiO2N affected the light-driven CH4 This fact may support a standpoint that the activity difference among these xLa2O3/LaTiO2N samples could be attributed to generation over LaTiO2N. It should be noted that during the CO2 reduction reaction, the removal of La2O3. no other products were detected. And the O2 yield over these Indeed, the theoretical calculations indicated that the basic O2− sites on La2O3 were thermodynamically and kinetically as-prepared LaTiO2N was about 1.9 times higher than CH4 yield (Figure S7, Supporting Information), much close to the favorable for CO2 capture and chemisorption into CO3 spe theoretical mole ratio of 2:1 for O2 to CH4 for the reaction of cies,[22] thus achieving CO2 activation for enhanced CO2 reduc CO2 + 2H2O → 2O2 + CH4. These results confirmed that the tion activity. In our case, the amount of CO2 adsorbed on these CH4 was the main reduced product. Indeed, as described in as-prepared samples determined by BET method at 273 K gradually increased with the increase of La2O3 removal amount. the previous reports, the CH4 was the observed main product for the gaseous CO2 reduction reaction.[16–19] This was prob During the acid treatment, formation of the amorphous La2O3 ably attributed to the fact that the CH4 was the most stable would contribute to the increased CO2 adsorption capacity. In product, as well demonstrated in the degradation of alkane. In addition, as known by EPR analysis, the OVs on LaTiO2N were the case of degradation of alkane, a mixture of five different gradually released during removing of La2O3. The highest CO2 hydrocarbons CH4, C2H4, C2H6, C3H6, and C3H8 was exposed adsorption amount on pure-LaTiO2N meant that the OVs may to the photocatalyst in the presence of light and water vapor.[38] be favorable for adsorbing the CO2 molecule. However, the CO2 After 120 min of illumination, the hydrocarbons except CH4 adsorption amount was not positively associated with the CH4 degraded significantly. This indicated that the CH4 was more generation rate when the mole amount of La2O3 was lower than stable during the gaseous photocatalytic reaction. Other prod 0.3. This indicated that the activity difference among these asucts such as the HCOOH or H2 were usually observed during prepared samples cannot be completely attributed to the differ ence in CO2 adsorption. the CO2 reduction on an electrode in liquid media due to the separated oxidation and reduction reaction chamber by proton To check the surface adsorbed states of CO2, the quantita exchange member.[39] However, for gaseous photocatalytic reac tive Fourier transform infrared (FT-IR) spectroscopy analysis was carried out on the 0.5La2O3/LaTiO2N, 0.3La2O3/LaTiO2N, tion, the oxidation and reduction reactions occur in the same reaction chamber, in which metastable reduced products were and pure-LaTiO2N after 2 h CO2 adsorption–desorption equi oxidized. librium under dark condition. A band at 2347 cm−1 in the three Although it was reported that acid treatment could enhance samples was assigned to linearly adsorbed CO2 (Figure 4a).[41] It activity of photocatalyst by eliminating surface defects or pro was obvious that the band intensity at 2347 cm−1 increased with 3+ moting formation of surface OH groups and OV/Ti spe the increase in concentration of OVs, revealing that the CO2 cies,[9,40] we cannot see obvious changes in CH4 yields over the was physically adsorbed on the OV sites due to the electrostatic S-LaTiO2N before and after acid treatment (Figure S5d, Sup absorption effect. The physically adsorbed state was not ben eficial to the activation of CO2 molecule. Therefore, although porting Information). The specific surface area (Table 1) cal culated from the linear part of the Brunauer–Emmett–Teller the high-density OVs induced high CO2 adsorption amount, the Adv. Funct. Mater. 2017, 27, 1702447

1702447 (5 of 11)



www.afm-journal.de

a

b

Intensity (a.u.)

2-

CO3

-

HCO3 CO2

0.5La2O3/LaTiO2N

T (%)

CO3

2-

C 1s

0.3La2O3/LaTiO2N

{

pure-LaTiO2N

0.5La2O3/LaTiO2N

62.1% 12.6% 25.3%

60.4% 27.7%

0.3La2O3/LaTiO2N

chemisorption species physisorption species

3200

2800

2400

85.9% 14.1%

pure-LaTiO2N

2000

1600

1200

-1

294

292

290

Wavenumbers (cm )

288

d

12.1%

O 1s

0.5La2O3/LaTiO2N

69.6%

29.7%

31.6% 13.4% 0.3La2O3/LaTiO2N

25.3%

23.6%

61.5% pure-LaTiO2N

280

23.1%

24%

73.7%

10.1% 0.3La2O3/LaTiO2N 20.3%

282

21%.5

Intensity (a.u.)

Intensity (a.u.)

11.2%

284

31.4%

C 1s 15.1%

286

Binding energy (eV)

c 0.5La2O3/LaTiO2N

11.9%

87.9%

15%

296 294 292 290 288 286 284 282 280 278 276

536

534

532

530

528

pure-LaTiO2N

526

524

Binding energy (eV)

Binding energy (eV)

Figure 4. FT-IR and XPS spectra for xLa2O3/LaTiO2N (x = 0.5 and 0.3) and pure-LaTiO2N: a) FT-IR spectra for the photocatalysts after CO2 adsorption for 2 h under the dark. b) The C 1s core-level XPS spectra for the photocatalysts after adsorption of CO2 and H2O under the dark. Total content of carbon-containing species does not count the content of the adventitious standard reference carbon (graphitic carbon). c) The C 1s and d) O 1s corelevel XPS spectra for the photocatalysts after the photocatalytic reaction.

pure-LaTiO2N exhibited the low CH4 generation rate still. The band at 1556 cm−1 was both observed on 0.5La2O3/LaTiO2N and 0.3La2O3/LaTiO2N and belonged to the symmetric vibra tion of bidentate carbonate species (vs(OCO)).[15,41] And band at 1658 cm−1 in the two La2O3-modified LaTiO2N was assigned to the symmetric vibration of HCO3− species (vas(OCO)), which was commonly formed via the reaction of CO2 with surface OH groups.[41] Both the band intensities at 1556 and 1658 cm−1 increased when the amorphous La2O3 formed, implying an enhanced interaction between amorphous La2O3 and CO2. These evidences verified that the CO2 was chemisorbed on the La2O3. Indeed, such a chemisorption effect has been well dem onstrated in the previous works of CO2 capture and fixation with the use of various alkali metal oxides.[10] Only a low-inten sity band for the vs(OCO) of carbonate species at 1410 cm−1 was observed in the pure-LaTiO2N, probably resulting from the interaction between CO2 and negatively charged OVs on LaTiO2N. The chemically adsorbed state will lower the OCO bond angle as well as decrease the lowest unoccupied molecule orbit (LUMO) of CO2, thus improving the conversion efficiency with lower energy input for the excited electrons transferring to the activated CO2.[6] To understand the activation of CO2 in presence of H2O, carbon-based species adsorbed on the 0.5La2O3/LaTiO2N, 0.3La2O3/LaTiO2N, and pure-LaTiO2N under dark condi tion were further determined by XPS (Figure 4b). The C 1s

Adv. Funct. Mater. 2017, 27, 1702447

core-level XPS spectra for 0.5La2O3/LaTiO2N and 0.3La2O3/ LaTiO2N could be deconvoluted into four peaks. A main peak at 284.6 eV was assigned to the adventitious standard refer ence carbon (graphitic carbon). The peak at 288.4 eV was due to the chemisorbed CO2δ− species, an activated-state species of CO2 after gaining one electron.[42] A possible route for the for mation of CO2δ− species is that the CO2 get one electron via chemisorbing on OVs with one electron, as demonstrated in reduction of CO2 into CO by gaining electrons from OVs.[21] Indeed, in our case, the CO2δ− species could be observed for all the three LaTiO2N samples with or without La2O3, further suggesting that formation of this species was associated with the OVs on LaTiO2N surface. Interestingly, density of CO2δ− species showed a negative correlation with the OV density. As is well-known, the CO2δ− species could be transformed into methoxy species in the presence of protons.[42] This meant that for the pure-LaTiO2N with high-density OVs, the forma tion of protons from H2O dissociation on OVs would promote the conversion of CO2δ− species. Correspondingly, CHx hydro carbon fragments at 285.6 eV were due to the reaction of CO2 with H2O by activation of OVs since they exhibited well positive relation with the OV content.[42] After irradiation (Figure 4c), the CHx hydrocarbon fragments disappeared, suggesting that they were intermediate species during CO2 reduction. The peak at 289.6 eV was assigned to CO32− species on the La2O3modified LaTiO2N samples.[43–45] No CO32− species was found

1702447 (6 of 11)



www.afm-journal.de

on the pure-LaTiO2N, well consistent with the FT-IR analysis, implying that the formation of CO32− species was dependent strongly on the La2O3. The increased content of CO32− species on 0.3La2O3/LaTiO2N gave solid evidence that the amorphous La2O3 promoted the formation of CO3 species. After the photocatalytic reaction, the C 1s core-level XPS analysis revealed that, for all the studied photocatalysts, the contents of CO2δ− species (288.4 eV) significantly decreased, the methoxy species at 285.6 eV completely disappeared, and a new XPS signal of oxycarbide species presented at 283.3 eV (Figure 4c).[45] However, CO32− species as the main active spe cies (289.6 eV) from CO2 activation by La2O3 showed inverse changes for the two La2O3 modified LaTiO2N samples after the photocatalytic reaction. The decreased CO32− species dem onstrated the fast consumption of activated CO2 on 0.3La2O3/ LaTiO2N surface, whereas the increased CO32− species on 0.5La2O3/LaTiO2N indicated that the CO2 capture and activa tion rate were faster than its consumption rate. In other words, the CO2 reaction on 0.5La2O3/LaTiO2N showed an unmatched rate in generation and conversion of CO32−. Noting that La2O3 on 0.5La2O3/LaTiO2N is crystalline, the amorphous La2O3 on 0.3La2O3/LaTiO2N would benefit to lower the barrier for the formation and conversion of CO32− species. Indeed, 0.3La2O3/ LaTiO2N exhibited a higher CH4 generation rate. The CO32− species could also be observed in the pure-LaTiO2N, probably because the photogenerated electrons trapped in OVs could change the physisorption state of CO2 into the chemisorbed state to form the active CO32− species. Similarly, under irradia tion, the experimental evidence had indicated that CO2 mole cule could adsorb at OV sites and be activated into formate and formic acid with the help of H2O.[22] The O1s core-level XPS spectra for the 0.5La2O3/LaTiO2N and 0.3La2O3/LaTiO2N after CO2 reduction could be resolved into four peaks (Figure 4d). Binding energy at 529.2 eV was assigned to TiOTi. Peaks at 530.2 and 531.8 eV belonged to the oxygen species in oxycarbides and the oxygen species in carbonates, respectively.[45,46] The peak at 531.8 eV was not visible in the pure-LaTiO2N, meaning that the carbonate spe cies formed on the surface of both 0.5La2O3/LaTiO2N and

0.3La2O3/LaTiO2N from the interaction between CO2 and La2O3. The content of the carbonate species over 0.3La2O3/LaTiO2N was evidently lower than those over 0.5La2O3/LaTiO2N, prob ably implying that the carbonate species was the active species that contributed to the fast CH4 generation. Compared with the fresh pure-LaTiO2N (Figure 2d), the XPS signal for lattice oxygen atoms near OVs (531.2 eV) disappeared (Figure 4d) after the CO2 reduction. Vanishing of OVs may originate from some species filling to the OVs. It should be noted that a strong sur face OH group XPS signal at 532.2 eV was observed in pureLaTiO2N after CO2 reduction, while the same XPS signal was known to be absent in the fresh pure-LaTiO2N. The H2O is a strong polar molecule and is easy to interact with crystal defects if compared with the nonpolar linear CO2 molecule. Under irradiation, the significantly enhanced OH group signal affirmed that the photogenerated charges trapped in the OVs could enlarge the interaction between H2O and OVs, thus pro moting the dissociation of H2O into the OH groups. A similar H2O splitting into OH groups was well observed at OVs on (110) surface of TiO2.[14,47] To further check the activation of CO2 and H2O on the sur face of La2O3 and OV sites of LaTiO2N, the surface models of LaTiO2N (002) facet with OVs and La2O3 (002) facet, well con sistent with the TEM observations, were built for calculations of adsorption energies of CO2 and H2O by density functional theory. Adsorption energy of an adsorbate over a photocatalyst was calculated by the equation of Eads = Eads/pho − Epho − Eadsorbate. Where, Eads is adsorption energy. Eads/pho, Epho, and Eadsorbate are energy of adsorption system, energy of a photocata lyst, and energy of an adsorbate, respectively. The calculated results were illustrated in Figure 5 and Table 2. Eads for CO2 adsorbed on La2O3 (002) facet (−1.7 eV) is much higher than that on LaTiO2N (002) with OVs (−0.5 eV). And the Eads for H2O adsorbed on LaTiO2N (002) with OVs was −3.9 eV, exhibiting more strong adsorption compared to adsorption of CO2 on La2O3 (002) facet (−1.7 eV). After CO2 adsorption on La2O3, the OCO bonding angle was changed from 178.3° into 130.4°, and the CO bonding length (1.182 Å) was increased to 1.265 and 1.273 Å. After adsorption of H2O on the LaTiO2N (002)

2000 0 -2000 Energy (eV)

H2O

CO2 -468.8 eV

-1028.0 eV

-4000

Ti

-6000

N

-8000 -10000 -12000 -14000 -16000

La

O

H2O over LaTiO2N (002) -7265.3 eV

-7820.1 eV

O

CO2 over LaTiO2N (002) La H H

-13203.7 eV

O

CO2 over La2O3 (002)

Figure 5. The energies of CO2 and H2O over La2O3 (002) and LaTiO2N (002) with OVs. The calculated energies of La2O3 (002) and LaTiO2N (002) with OVs were −12174.0 and −6792.6 eV, respectively.

Adv. Funct. Mater. 2017, 27, 1702447

1702447 (7 of 11)



www.afm-journal.de

Table 2. Adsorption energy and geometric parameter change before and after adsorption of H2O and CO2 over the surface of La2O3 (002) and LaTiO2N (002) with OVs. CO2 over La2O3 (002) Before adsorption

CO2 over LaTiO2N (002) with OVs

After adsorption

Before adsorption

After adsorption

H2O over LaTiO2N (002) with OVs Before adsorption

After adsorption

Eads [eV]

–

−1.7

–

−0.5

–

−3.9

dCO [Å]

1.182

1.265, 1.273

1.182

1.161, 2.315

–

–

∠OCO [°]

178.3

130.4

178.3

113.5

–

–

–

–

–

–

0.975

0.977, 3.119

dOH [Å]

facet with OVs, the preferential adsorption sites of H2O are the OVs of LaTiO2N (002) facet. The HO bonding length (0.975 Å) was turned into 0.977 and 3.119 Å. The significantly elongated HO bonding means that splitting of H2O into H+ and OH− occurs due to the fact that H2O reacts with the OVs of LaTiO2N (002) facet. The similar water splitting process was observed on OVs of (110) surface of TiO2.[20] When CO2 adsorbed into OVs of LaTiO2N (002) facet, OCO bonding obviously bent (113.5°) and the CO bond lengths changed to 1.161 and 2.315 Å, implying that the CO2 can also be activated by OVs of LaTiO2N (002) facet. Given the higher Eads and elongated HO bonding, we believe that the activation of H2O by OVs is more favored than activation of CO2 on OVs of LaTiO2N. Although above-mentioned analyses indicated that the H2O was more favorable to be activated by OVs, and competitive molecule adsorption would occur during the reduction reac tion of CO2 by H2O. Therefore, an interest was to know the effect of adsorption order of CO2 and H2O at OVs on CH4 evolu tion. For this purpose, two model catalysts with different con centrations of OVs were prepared: S-LaTiO2N with low-density OVs prepared by directly nitriding La2Ti2O7 and the S-LaTiO2N with high-density OVs obtained by hydrogen treatment at 473 K for 4 h. Coadsorption of CO2 and H2O on the two S-LaTiO2N samples for 2 h was carried out after a preadsorption of CO2 or H2O for 13 h. As shown in Figure 6a, the molecule adsorption order did not affect the CH4 generation over the S-LaTiO2N with low-density OVs. In contrast, about 2 times increase in CH4 generation rate was observed for S-LaTiO2N with highdensity OVs (0.3 µmol gcat−1 h−1) than the one with low-density OVs (0.1 µmol gcat−1 h−1). More importantly, the preadsorption of H2O on S-LaTiO2N with high-density OVs exhibited a con stant CH4 generation rate within 8 h irradiation, however, the preadsorption of CO2 induced slight increase in CH4 yield in the first 4 h irradiation and then showed a significant CH4 yield decrease in the subsequent 4 h irradiation (Figure 6b). These results indicated that the surface OVs acting as H2O acti vation sites could effectively accelerate the conversion of CO2 due to the enhancing H2O splitting to provide protons. FT-IR analysis revealed that OH groups at about 3648 cm−1 could be found for H2O preadsorption on S-LaTiO2N with high-density OVs (Figure S9, Supporting Information). However, no clear OH signal was detected for CO2 preadsorption on S-LaTiO2N with high-density OVs. This evidence further confirmed that the CO2 preoccupying into OVs sites was not beneficial to the activation of H2O molecule. Theoretical calculations had dem onstrated that adsorbing CO2 at OVs showed a bent structure toward the surface, which induced difficulty for protons to

Adv. Funct. Mater. 2017, 27, 1702447

reach the C atom of CO2 molecule to form hydrocarbons.[22] Therefore, the preadsorption of CO2 exhibited the unstable CH4 generation, probably due to the increasing difficulty in proton support. To further verify the fact of activating H2O by OVs, the pureLaTiO2N after the first cycle of CO2 photoreduction under irradiation for 8 h was suffered from a vacuuming procedure to remove the physically adsorbed OH groups, and a second illumination test was carried out under CO2 without adding H2O. Under the irradiation by visible light, a slow CH4 gen eration could be observed as well (Figure 6c). This indicated that the OH from chemical activation of H2O by OVs could indeed participate in the photocatalytic CO2 reaction. After reaction, a typical OV EPR signal at g value of 2.003 emerged (Figure 6d), indicating that the OVs filled by OH groups were able to be released by photogenerated hole oxidation. A sim ilar sustainable OV regeneration was found in the water split ting on TiO2.[20] Three cycle tests were performed for checking the stability for the CH4 and O2 generation over 0.3La2O3/ LaTiO2N. As shown in Figure 6e, after three runs, the CH4 yield decreased by 5.4%, from 9.18 to 8.69 µmol gcat−1. And the mole ratio of O2:CH4 was well kept at a constant ratio of about 1.9:1, confirming the high stability of La2O3 modified LaTiO2N photocatalysts. This indicated that activation of H2O on OV sites is sustainable via continuous hole oxidation under irradiation, thus leading to the stable proton support for CO2 reduction. On the basis of the above-mentioned results, the possible mechanism for activation and reduction of CO2 to CH4 was given in Figure 7. Obviously, the CO2 reduction by H2O over La2O3-modified LaTiO2N with OVs was mainly attributed to the fact that O2− in crystal lattice of La2O3 and OVs on LaTiO2N were able to activate CO2 and H2O into CO32− and H+, respec tively (Equations (1) and (2)). During the dark reaction, the O atom of H2O filled into OVs on LaTiO2N to form lattice oxygen Ob2− and released the protons. Under irradiation, the photo generated holes would oxidize the Ob2− to generate OVs and O2 (Equation (4)), while the protons reduced the CO32− to CH4 by recovering the O2− into La2O3. Clearly, such activating CO2 by O2− of La2O3 and H2O at OVs of LaTiO2N was sustainable and greatly lowered the reaction barrier of CO2 conversion. In addition, the separated activation at different spatial sites could effectively restrain the recombination of electrons and holes and the occurrence of reverse reactions. Therefore, we believed that constructing different spatial sites to activate CO2 and H2O provided a novel pathway to enhance photocatalytic activity.

1702447 (8 of 11)



d

0.8

-1

CH4 evolution (µ mol gcat )

c

www.afm-journal.de

Intensity (a.u.)

0.6

0.4

0.2

2.003

LaTiO2N

0.0 0

2

4

6

8

10

2.10

Irradiation time (h)

CH4 and O2 evolution (µ mol/gcat)

e

CH4

20 18

2.05

1.95

1.90

O2

nd

st

rd

2 run

1 run

2.00

g value

3 run

16 14 12 10 8 6 4 2 0 0

2

4

6

8

10

12

14

Time (h)

16

18

20

22

24

26

28

Figure 6. a) The effect of adsorption order for CO2 and H2O on CH4 yields over S-LaTiO2N with low-density OVs obtained by directly nitriding La2Ti2O7 and b) S-LaTiO2N with high-density OVs prepared by hydrogen treatment of LaTiO2N. c) CH4 yields over illuminating pure-LaTiO2N under CO2 atmosphere. The pure-LaTiO2N after 8 h CO2 reduction by H2O was again used in the CH4 generation under CO2 atmosphere. d) EPR spectrum for pureLaTiO2N after irradiation under vacuum. The pure-LaTiO2N after 8 h CO2 reduction by H2O was used in this EPR study. e) The stability tests for CH4 and O2 generation over 0.3La2O3/LaTiO2N.

Figure 7. Possible mechanism for activation and reduction of CO2 by H2O to CH4.

Adv. Funct. Mater. 2017, 27, 1702447

1702447 (9 of 11)



www.afm-journal.de

3. Conclusions In summary, to improve the activity of photocatalytic CO2 con version, we propose a route using a solid base to modify photo catalyst with defects, aiming to simultaneously accelerate the two important half reaction processes. In this work, in situ La2O3 modified LaTiO2N with oxygen vacancies are designed as a catalyst model, achieving a twofold increase in CH4 yield for CO2 reduction. The prominent activity can be attributed to the following effects: (1) Strong basicity of O2− in La2O3 ther modynamically and kinetically favors CO2 chemisorption into CO32− species, greatly contributing to both the bending of OCO bond and the decrease of LUMO energy of CO2 mole cule. (2) The oxygen vacancies on LaTiO2N surface benefits the activating H2O to adsorbed OH, thus effectively decreasing the reaction barriers of water oxidation to providing protons. The simultaneous activation of the CO2 and H2O at different spatial sites for achieving rate-matching half reaction rates may offer a new strategy to suppress the reverse reactions for efficient solar energy conversion.

4. Experimental Section Synthesis of La2TiO5 Precursors: To synthesize the solid La2TiO5, a stoichiometric mixture of La2O3 and TiO2 was mixed with an eutectic molten salt of NaCl and KCl (mole ratio of 1:5 for La2TiO5 to NaCl+KCl, a1:1 for NaCl:KCl). After intensive mixing, the resulting mixture was heated to 1373 K with a raise rate of 45 K h−1, and held at this temperature for 10 h. It was subsequently cooled to 773 K at a rate of 150 K h−1 and allowed to cool to room temperature naturally. The fluxgrown La2TiO5 crystals were separated by deionized water washing and dried at 353 K overnight. Preparation of Photocatalyst: To prepare the La2O3/LaTiO2N, the as-prepared La2TiO5 powders (≈0.5 g for each run) were heated at 1223 K for 24 h at a rate of 10 K min−1 under 500 mL min−1 flowing NH3, and followed by a natural cooling process to room temperature. Then, the prepared composite La2O3/LaTiO2N (0.1 g) was soaked in a dilute HCl (0.1 m) with different volumes (0, 1.5, 3.0, 5.3, and 7.6 mL) overnight to adjust the solid base amount, and the final products were collected by deionized water washing and centrifugation. Material Characterization: The derived crystal phases were determined by powder XRD (Rigaku Ultima III, Japan) operated at 20 kV and 40 mA with Cu-Kα radiation. The surface morphology was observed by scanning electron microscopy (FEI Nova Nano SEM 230, USA). High-solution lattice images and SAED patterns were obtained by transmission electron microscopy (FEI Tecnai G2 F30 S-Twin, USA) operated at 200 kV. Diffused reflectance spectrum was scanned by a UV–vis spectrophotometer (UV-2500, Shimadzu Co., Japan) and transformed into absorption spectrum with Kubelka–Munk relationship. The specific surface area was measured from nitrogen (N2) adsorption–desorption isotherm at 77 K by an automatic surface area analyzer (Micromeritics Tristar-3000, USA) after the samples had been dehydrated at 423 K for 3 h in the flowing N2. And the mesopores above 2 nm were determined by the Barrett–Joyner–Halenda method. The FT-IR spectra were measured by an infrared spectrometer (NEXUS870, Nicolet, USA) at ambient condition. The surface chemical species was investigated by XPS on PHI5000 Versa Probe (ULVAC-PHI, Japan) with monochromatized Al-Kα X-ray radiation (1486.6 eV). The energy resolution of the electrons analyzed by the hemispherical mirror analyzer was about 0.2 eV. The C 1s core level at 284.6 eV was taken as an internal reference to correct the shift of the binding energies. The surface paramagnetic species having one unpaired electrons such as oxygen vacancies were measured by EPR with an X-band EPR spectrometer (EMX-10/12, Bruker, German) operating at a microwave frequency of 9.77 GHz. Adv. Funct. Mater. 2017, 27, 1702447

Photocatalytic Evaluation: The gas phase photocatalytic reduction of CO2 with H2O was tested in a glass reactor with an area of 4.2 cm2, and 40 mg of photocatalyst was uniformly dispersed. The light source was a 300 W Xenon arc lamp fitted with a cutoff filter (λ > 420 nm). The volume of the reaction chamber was about 230 mL. Before the irradiation, the reaction system was vacuumed several times, and then high-purity gas of compressed CO2 (purity 99.999%) was introduced into the reaction chamber to achieve an ambient pressure. Subsequently, 0.4 mL of deionized water was injected into the chamber as reactant. Prior to irradiation, the adsorption process was held for 2 h. During the reaction, 1 mL gas was extracted by a sampling needle from the chamber at given intervals for subsequent concentration analysis. The carbon-based products were quantified by a gas chromatography with flame ionization detector (GC-2014, Shimadzu Corp., Japan). The amount of O2 was determined by gas chromatography with thermal conductivity detector (GC-8A, Shimadzu Corp., Japan). The isotope labeling was carried out using 13CO2 as carbon source, and 13CH was identified by gas chromatography–mass spectrometry (Agilent 4 6890N/5973I, Agilent Corp., USA). Computational Methods: The calculations of adsorption energy were conducted through the Cambridge Serial Total Energy Package code. The general gradient approximation with Perdew-Bueke-Ernzerh functional (PBE) was employed to describe the exchange–correlation effects. The attractive energy between nuclear and electrons was calculated via ultrasoft pseudopotential. The convergence threshold of geometric optimization was set at 2.0 × 10−5 eV per atom for total energy, 0.05 eV Å−1 for maximum force, 0.1 GPa for pressure, and 0.002 Å for maximum displacement. The crystal structure of La2O3 (002) was built including 8 atoms of La, 12 atoms of O, and lattice parameter was a = 8.0 Å, b = 8.0 Å, c = 14.7 Å, α = β = 90°, and γ = 120°. The crystal structure of LaTiO2N (002) with O vacancy contained 2 atoms of La, 2 atoms of Ti, 3 atoms of O, and 2 atoms of N, and lattice parameter was a = 6.0 Å, b = 7.8 Å, c = 12.8 Å, α = β = 90°, and γ = 97°.

Supporting Information Supporting Information is available from the Wiley Online Library or from the author.

Acknowledgements This work was financially supported by the National Basic Research Program of China (2013CB632404), the National Natural Science Foundation of China (51572121, 21603098, and 21633004), the State Key Laboratory of NBC Protection for Civilian (SKLNBC2014-09), the Natural Science Foundations of Jiangsu Province (BK20151265, BK20150580, BK20161277), the Natural Science Foundation of Jiangsu Education Department (16KJB610002), the Fundamental Research Funds for the Central Universities (021314380084), and the Postdoctoral Science Foundation of China (No. 2017M611784). The authors are grateful to the High Performance Computing Center (HPCC) of the Nanjing University for doing the numerical calculations in this paper on its IBM Blade cluster system.

Conflict of Interest The authors declare no conflict of interest.

Keywords basic sites, molecular activation, oxygen vacancies, photocatalysis

1702447 (10 of 11)

Received: May 8, 2017 Revised: June 18, 2017 Published online: July 31, 2017 © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim


www.afm-journal.de

[1] T. Inoue, A. Fujishima, S. Konishi, K. Honda, Nature 1979, 277, 637. [2] S. Habisreutinger, L. Schmidt-Mende, J. Stolarczyk, Angew. Chem. Int. Ed. 2013, 52, 737. [3] G. Olah, G. Prakash, A. Goeppert, J. Am. Chem. Soc. 2011, 133, 12881. [4] H. Tong, S. Ouyang, Y. Bi, N. Umezawa, M. Oshikiri, J. Ye, Adv. Mater. 2012, 24, 229. [5] S. Yan, S. Ouyang, J. Gao, M. Yang, J. Feng, X. Fan, L. Wan, Z. Li, J. Ye, Y. Zhou, Z. Zou, Angew. Chem. Int. Ed. 2010, 122, 6544. [6] W. Tu, Y. Zhou, Z. Zou, Adv. Mater. 2014, 26, 4607. [7] J. Feng, W. Luo, T. Fang, H. Lv, Z. Wang, J. Gao, W. Liu, T. Yu, Z. Li, Z. Zou, Adv. Funct. Mater. 2014, 24, 3535. [8] F. Zhang, A. Yamakata, K. Maeda, Y. Moriya, T. Takata, J. Kubota, K. Teshima, S. Oishi, K. Domen, J. Am. Chem. Soc. 2012, 134, 8348. [9] M. Matsukawa, R. Ishikawa, T. Hisatomi, Y. Moriya, N. Shibata, J. Kubota, Y. Ikuhara, K. Domen, Nano Lett. 2014, 14, 1038. [10] S. Wang, S. Yan, X. Ma, J. Gong, Energy Environ. Sci. 2011, 4, 3805. [11] Q. Wang, J. Luo, Z. Zhong, A. Borgna, Energy Environ. Sci. 2011, 4, 42. [12] A. Rodriguez, P. Liu, J. Stacchiola, D. Senanayake, G. White, G. Chen, ACS Catal. 2015, 5, 6696. [13] S. Xie, Y. Wang, Q. Zhang, W. Fan, W. Deng, Y. Wang, Chem. Commun. 2013, 49, 2451. [14] L. Liu, C. Zhao, H. Zhao, D. Pitts, Y. Li, Chem. Commun. 2013, 49, 3664. [15] X. Meng, S. Ouyang, T. Kako, P. Li, Q. Yu, T. Wang, J. Ye, Chem. Commun. 2014, 50, 11517. [16] B. Pan, S. Luo, W. Su, X. Wang, Appl. Catal., B 2015, 168, 458. [17] B. Pan, Y. Zhou, W. Su, X. Wang, RSC Adv. 2016, 6, 34744. [18] B. Pan, Y. Zhou, W. Su, X. Wang, Nano Res. 2017, 10, 534. [19] S. Chen, B. Pan, L. Zeng, S. Luo, X. Wang, W. Su, RSC Adv. 2017, 7, 14186. [20] O. Bikondoa, C. Pang, R. Ithnin, C. Muryn, H. Onishi, G. Thornton, Nat. Mater. 2006, 5, 189. [21] J. Lee, D. Sorescu, X. Deng, J. Am. Chem. Soc. 2011, 133, 10066. [22] Y. Ji, Y. Luo, J. Am. Chem. Soc. 2016, 138, 15896. [23] K. Kawashima, M. Hojamberdiev, H. Wagata, K. Yubuta, S. Oishi, K. Teshima, Cryst. Growth Des. 2014, 15, 333. [24] A. E. Maegli, S. Pokrant, T. Hisatomi, M. Trottmann, K. Domen, A. Weidenkaff, J. Phys. Chem. C 2014, 118, 16344. [25] R. B. Singh, H. Matsuzaki, Y. Suzuki, K. Seki, T. Minegishi, T. Hisatomi, K. Domen, A. Furube, J. Am. Chem. Soc. 2014, 136, 17324.

Adv. Funct. Mater. 2017, 27, 1702447

[26] M. Sluban, P. Umek, Z. Jaglicˇic´, J. Buh, P. Smitek, A. Mrzel, C. Bittencourt, P. Guttmann, M. Delville, D. Mihailovic´, D. Arcˇon, ACS Nano 2015, 9, 10133. [27] T. R. Gordon, M. Cargnello, T. Paik, F. Mangolini, R. T. Weber, P. Fornasiero, C. B. Murray, J. Am. Chem. Soc. 2012, 134, 6751. [28] Y. Kim, P. Woodward, K. Baba-Kishi, C. Tai, Chem. Mater. 2004, 16, 1267. [29] X. Chen, L. Liu, P. Y. Yu, S. S. Mao, Science 2011, 331, 746. [30] J. Chen, W. Song, H. Hou, Y. Zhang, M. Jing, X. Jia, X. Ji, Adv. Funct. Mater. 2015, 25, 6793. [31] G. Liu, L. Yin, J. Wang, P. Niu, C. Zhen, Y. Xie, H. Cheng, Energy Environ. Sci. 2012, 5, 9603. [32] G. Liu, J. Pan, L. Yin, J. T. S. Irvine, F. Li, J. Tan, P. Wormald, H. Cheng, Adv. Funct. Mater. 2012, 22, 3233. [33] X. Xu, C. Randorn, P. Efstathiou, J. T. S. Irvine, Nat. Mater. 2012, 11, 595. [34] Y. Li, X. Cheng, X. Ruan, H. Song, Z. Lou, Z. Ye, L. Zhu, Nano Energy 2015, 12, 775. [35] K. Banger, Y. Yamashita, K. Mori, R. L. Peterson, T. Leedham, J. Rickard, H. Sirringhaus, Nat. Mater. 2011, 10, 45. [36] X. Zhang, J. Qin, Y. Xue, P. Yu, B. Zhang, L. Wang, R. Liu, Sci. Rep. 2014, 4, 4596. [37] S. Hoang, S. P. Berglund, N. T. Hahn, A. J. Bard, C. B. Mullins, J. Am. Chem. Soc. 2012, 134, 3659. [38] M. S. Hamdy, R. Amrollahi, I. Sinev, B. Mei, G. Mul, J. Am. Chem. Soc. 2014, 136, 594. [39] S. Gao, Y. Lin, X. Jiao, Y. Sun, Q. Luo, W. Zhang, D. Li, J. Yang, Y. Xie, Nature 2016, 529, 68. [40] Z. He, J. Tang, J. Shen, J. Chen, S. Song, Appl. Surf. Sci. 2016, 364, 416. [41] S. Bian, J. Baltrusaitis, P. Galhotra, V. H. Grassian, J. Mater. Chem. 2010, 20, 8705. [42] X. Deng, A. Verdaguer, T. Herranz, C. Weis, H. Bluhm, M. Salmeron, Langmuir 2008, 24, 9474. [43] X. Yang, S. Kattel, S. D. Senanayake, J. Anibal Boscoboinik, X. Nie, J. Graciani, J. A. Rodriguez, P. Liu, D. J. Stacchiola, J. G. Chen, J. Am. Chem. Soc. 2015, 137, 10104. [44] J. Graciani, K. Mudiyanselage, F. Xu, A. E. Baber, J. Evans, S. D. Senanayake, D. J. Stacchiola, P. Liu, J. Hrbek, J. F. Sanz, J. A. Rodriguez, Science 2014, 345, 546. [45] M. D. Porosoff, X. Yang, J. A. Boscoboinik, J. G. Chen, Angew. Chem. Int. Ed. 2014, 53, 6705. [46] P. Huang, Y. Zhao, J. Zhang, Y. Zhu, Y. Sun, Nanoscale 2013, 5, 10844. [47] Y. Li, Y. Gao, Phys. Rev. Lett. 2014, 112, 206101.

1702447 (11 of 11)
