Electron competitive migration regulating for dual maxima of water photolysis

Abstract

Electron competitive migration between the conduction band and charge trap centre is the key in governing the catalytic activity and the relevant applications of semiconductor nanomaterials, which is however poorly understood yet. Herein, we systematically studied the electron competitive migration in defective SnO₂ nanoparticles through hybridizing with a polymer electron donor, graphitic carbon nitride (g-C₃N₄). When varying the mass ratio of defective SnO₂ from 5% to 70%, an increase of surface-charge trapping centres (oxygen vacancies) in SnO₂ effectively regulated the electron competitive migration. As a consequence, dual catalytic activity maxima were observed in hydrogen generation from water splitting under visible light irradiation (λ > 420 nm). For instance, the relative mass ratios at 10% and 40% yielded maximum hydrogen generation rates of 54.3 μmol h⁻¹ g⁻¹ and 44.3 μmol h⁻¹ g⁻¹, respectively, far beyond that of 27.9 μmol h⁻¹ g⁻¹ for pure g-C₃N₄. Strikingly, the photon–hydrogen conversion efficiency also showed dual maxima values as SnO₂ mass ratio changes. These abnormal observations were comparatively investigated via XPS, EPR and photoluminescence spectra in solid state and aqueous environments. It is demonstrated that electron competitive migration was primarily caused by oxygen vacancies on the SnO₂ surface, which plays a key role in creating the dual catalytic activity maxima in water splitting.

---

1 Introduction

Hydrogen energy gained from water splitting over supported semiconductors under sunlight has attracted world-wide attention. The conversion efficiency directly determines whether this conversion scheme can be applied to solving the energy shortage crisis. Much effort has been taken to develop photocatalysts with an aim to solve the fundamental problems of energy conversion efficiency from photons to hydrogen since the first report by Fujishima and Honda.¹ Since then, multiple modification of photocatalysts has become an important method to improve the photon utilization, which includes (i) improving light response range through doping or introducing bulk defect state/surface disorder, such as nitrogen-doped TiO₂,² and disorder-engineered black TiO₂,³ (ii) introducing surface defects as charge trap centres to enhance the charge separation rate. For instance, Tang and coworker⁴ have introduced oxygen vacancies in ZnO which enhances the photocatalytic activity; and (iii) semiconductor-based hybrids via hybridizing different semiconductors to enhance the photon–hydrogen conversion efficiency. For example, hybridizing graphitic carbon nitride (g-C₃N₄) with metal oxides or sulfides, like ZnO,^5–7 CdS,^8–10 TiO₂,^11–15 WO₃,^16,17 Bi₂WO₆,¹⁸ and other kind of hybrids have been investigated widely.^19–21 In this regard, photo-generated carrier's migration has shown impacts on surface photocatalytic redox reaction, and photocatalytic ability could be thus enhanced by increasing the separation rate of photo-generated carriers. Though the methods mentioned above have given rise to the enhanced light response ability, introducing defective states in solids could be a direct useful mean to improve the carrier's separation rate. This is because defects can be easily achieved just by controlling materials crystallinity. Most investigations have been performed to introduce surface defects that enhance the charge separation rate only referring to the organic photo-degradation. It is well established that the generation process of hydrogen and oxygen from water photolysis under light irradiation is totally different from the organic degradation process, because more requirement for semiconductors in water splitting than in organic photo-degradation. For instance, photocatalysts for water splitting have to possess a more negative conduction band bottom position than the redox potential of H⁺/H₂ (0 eV vs. NHE), and the top position for valence has to be more positive than the redox potential of O₂/H₂O (1.23 eV vs. NHE).²² When defects are introduced as the electron traps onto materials surface, electron competitive migration could happen between electron trap level and conduction band, and therefore the competitive migration will influence the photocatalysts activity. Unfortunately, it still remains unclear how electron traps influence carrier's migration in water photolysis process. Therefore, it is highly important to study such an issue, most likely through a semiconductor-based hybridization via a systematic control over the defect amount.

SnO₂ is a prototype semiconductor with an appropriate energy band position (i.e., conduction band bottom at −0.11 eV and valence band top position at 3.59 eV)²³ essential for photocatalytic water splitting, while it cannot response to visible light for its wide energy band gap. A recent study has demonstrated the potential of hybrid, g-C₃N₄/SnO₂ as a hybrid photocatalyst for water splitting under visible light,²⁴ which is based on the sole stoichiometric SnO₂ without lattice defect. It is then very difficult to effectively tune the electron migration just via varying the corresponding stoichiometric content ratio in hybrid photocatalysts, and this kind of methods usually gives a single photocatalytic maximum, just referring to the electron migration between different photocatalysts' conduction bands. As a result, electron competitive migration mechanism caused by defects is poorly understood in hybrid photocatalysts. If defects are involved in the semiconductor-based hybrids, the relevant electron migration could be effectively tuned. It is highly possible to understand the photocatalytic mechanism in a systematic method which expects to promote the development of new photocatalysts and many other surface-sensitive technologies. We propose defective SnO₂ as the target semiconductors to study, which was prepared via annealing of SnO₂ quantum dots, since (i) SnO₂ quantum dots can be easily fabricated, (ii) quantum dots usually have much of surface defect and low crystallinity just when prepared using SnCl₄ hydrothermal method,²⁵ and (iii) annealing of SnO₂ quantum dots under given temperature could tune the defect concentration and vary the crystallinity. All these help one to find the appropriate methods of obtaining defective SnO₂, with the amount of defects being regulated just by tailoring the mass ratio of SnO₂.

Here, we initiated a systematic study on the electron competitive migration in defective tin dioxide by hybridizing with different amounts of polymer electron donor, g-C₃N₄. The amount of oxygen vacancy was tailored after subsequent solid calcinations. When tested as a photocatalyst, the hybrids showed dual maxima of visible-light catalytic activity in hydrogen generation from water under visible light irradiation, which totally differ from the only one catalytic activity maximum ever reported. The nature for this abnormal observation was comparatively studied by XPS, EPR and photoluminescence data in terms of an electron competitive migration mechanism.

2 Experimental

2.1 Materials

Melamine (C₃H₆N₆), ethanol, SnCl₄·5H₂O and triethanolamine are chemically pure grade, all of the materials were purchased from Sinopharm Chemical Reagent Corp, P. R. China and used without further purification.

2.2 Chemicals for materials preparation

2.2.1 Preparation of g-C₃N₄. Crucible containing 10 g melamine was heated at 540 °C for 4 h in Muffle furnace at a heating rate of 10 K min⁻¹. After cool down to room temperature, a yellow solid product was obtained. To obtain fine powder for the preparation of hybrids, the as-prepared g-C₃N₄ sample was further milled in an agate mortar. Then, the solid powder was washed with water for 3 times and dried at 80 °C over night to get the pure g-C₃N₄ sample.

2.2.2 Preparation of defective SnO₂. There have been many reports on the synthesis of SnO₂ nanoparticles by hydrothermal method.^24,25 Different from these reference work, herein we described the synthesis of defective SnO₂ nanoparticles using SnCl₄·5H₂O as the precursor in our hydrothermal systems. In order to investigate the electron migration and its role in water photolysis, defective SnO₂ was hybridized with g-C₃N₄, just following the hybridization method reported in literature²⁴ but with some modifications: (i) surface oxygen vacancy concentration was controlled via hydrothermal reaction and the subsequent annealing; and (ii) the mass ratio of defective SnO₂ to g-C₃N₄ was regulated in a wide range from 0 to 70%. The details for the sample synthesis were described as follows.

Defective SnO₂ nanocrystals were prepared via two steps. The first one involves the synthesis of SnO₂ quantum dots with a poor crystallinity via a hydrothermal method. Namely, 0.5 mol SnCl₄·5H₂O was dissolved in 100 mL water and stirred for 10 min without adjusting pH value. This solution was poured into two 100 mL volume hydrothermal reactors with equal volume of 50 mL. Two reactors were then placed in an oven at 200 °C for 9 h. When the hydrothermal reactions finished, white slurry was obtained. After washing for several times using water, AgNO₃ solution was used to detect the residual Cl⁻¹ ions in supernatant till no evident white precipitate appeared. Then, the washed white precipitate was dried at 80 °C for 12 h, and milled into powder, which was named as SnO₂-SH.

Further, we employed annealing process to prepare the lower defective SnO₂ so that we can tailor the defect amount just simply changing the mass ratio of SnO₂ essential for achieving the enhanced photocatalytic activity. For this purpose, SnO₂-SH powders were annealed at 700 °C for 2 h in air at heating rate at 10 K min⁻¹. With this process, the lower defective SnO₂, named as SnO₂-AD, was obtained. The amounts of oxygen vacancies were confirmed by XPS and EPR characterization.

2.2.3 Hybridization of g-C₃N₄ with defective SnO₂. The as-prepared SnO₂-AD was milled with g-C₃N₄ at the given mass ratio of SnO₂/(SnO₂ + g-C₃N₄) in a wide range from 0 to 70%, and the mixture mass was set at 1.0 g. These mixtures were put into 20 mL ethanol and ultrasonic for 20 min to form slurries, and then stirred in fume hood to evaporate ethanol, and further calcined at 400 °C in air for 1 h. After cooling down to room temperature, the resulted powders were milled for the subsequent tests and characterization. To express more clearly, the obtained hybrids were named Sn–Hx, here x is the mass ratio of SnO₂ in the hybrid. For example, Sn–H40 is the hybrid sample with a mass ratio of SnO₂ at 40%. To confirm the oxygen vacancies negative role in SnO₂/g-C₃N₄ hybrids, the hybrid in 40% mass ratio of SnO₂-SH was also prepared used the same method, and which was named as “40% SnO₂-SH/g-C₃N₄”.

2.3 Analytical and testing instruments

The particle morphologies of the samples were performed via transmission electron microscopy (TEM, JEM-2010 produced by JEOL). The phase structure and crystalline of samples were determined by X-ray diffraction patterns that were recorded on a Rigaku MinFlex II benchtop X-ray diffractometry (XRD) with Cu Kα irradiation (λ = 1.5418 Å). Thermogravimetric (TG) data of the samples were carried on Netzsch STA449F3 thermal analyzer. The chemical states of the samples were analyzed using X-ray photoelectron spectroscopy (XPS, a monochromatic Al Kα X-ray source operating at 150 W), and binding energies were referenced to the C1s peak at 284.6 eV that arises from adventitious carbon. Electron paramagnetic resonance (EPR) measurement for samples was performed via Bruker-BioSpin, E500 spectrometer to confirm the existence of oxygen vacancy. A frequency of ca. 9.866 GHz was used for a dual-purpose cavity operation. The magnetic field of 0.1 mT was modulated at 100 kHz. Fourier transform infrared (FT-IR) spectra of the samples were recorded on a Perkin-Elmer IR spectrometer using a KBr pellet technique. UV-vis absorbance spectra of the samples were measured by Varian Cary-500 spectrophotometer at ambient condition. Photoluminescence (PL) spectra in solid state and aqueous environment of the samples were carried out at room temperature by solid and aqueous Varian Cary Eclipse Fluorescence Spectrometers. During the aqueous environment PL tests, all samples with the same mass concentration must keep as their suspension state.

2.4 Photocatalytic activity measurements

Photocatalytic hydrogen evolution was conducted in an online photocatalytic hydrogen generation system (Labsolar-IIIAG, Perfectlight Corp, Beijing, P. R. China) at a temperature of 5 °C. A 300 W Xe lamp equipped with a UV-cutoff filter (λ > 420 nm) was used as the light source of Sn–Hx hybrids, g-C₃N₄, SnO₂-AD, and SnO₂-SH/g-C₃N₄ hybrid in 40% SnO₂ mass ratio while the light-cutoff filter (λ > 260 nm) was used alone in the hydrogen generation test of SnO₂-AD to make sure its photocatalytic ability in hydrogen generation from water splitting. In details, in the photocatalytic process, 100 mg samples and 90 mL deionized water were added into the reaction reactor. 10 mL triethanolamine which acted as a sacrificial agent were added. The H₂PtCl₆ solution (1.0 wt% for Pt) was used for Pt deposition. The Pt would act as a co-catalyst to improve the photocatalytic performance of all samples. Prior to the reaction, the whole reaction system was deaerated to remove O₂ and CO₂. Gas evolution was analyzed by an on-line gas chromatograph (Fuli 9790 II) with TCD detector purchased from Fuli Analytical Instrument Corp, Zhejiang, P. R. China. Hydrogen evolution value is calculated based on 100 mg catalyst. Before the formal test, 30 minutes were spent for the light deposition of Pt from H₂PtCl₆, the deposition process could be described as follows:

H₂PtCl₆ → 2H⁺ + PtCl₆²⁻ (1)

PtCl₆²⁻ + 4e⁻ → Pt + 6Cl⁻ (2)

Then the system was evacuated again to avoid the effect of the hydrogen ions of H₂PtCl₆, till it does not influence the hydrogen evolution from water splitting. The optical conversion capacity of g-C₃N₄ was defined by the photon–hydrogen conversion efficiency, as presented by the yield of hydrogen divided by g-C₃N₄ mass ratio in SnO₂/g-C₃N₄ hybrids.

3 Results and discussion

3.1 Hybrids of defective SnO₂ with g-C₃N₄: formation and characterization

3.1.1 XRD analysis. Fig. 1 illustrates XRD patterns of the samples. SnO₂ quantum dot (sample SnO₂-SH) was prepared via a hydrothermal process, and its XRD pattern matched well with the standard diffraction data of rutile-like, cassiterite SnO₂ (JCPDS, no. 41-1445). Here, we use XRD to investigate the crystallinity of SnO₂ quantum dots obtained from hydrothermal reactions. The lower diffraction peak intensity and wide full width at half maximum (FWHM) for SnO₂-SH indicate a poor crystallinity, which suggests that plenty of lattice defects might exist in SnO₂-SH. This inspection is approved by Raman spectra. As indicated in Fig. S1,† SnO₂-SH shows a low asymmetric Raman signal at 574 cm⁻¹, which is caused by oxygen vacancies and local disorder.²⁶ Strong intensity of this Raman shift peak demonstrates a high concentration of surface defect. According to Scherrer formula, the mean grain diameter is calculated to be about 8.3 nm for SnO₂-SH. Therefore, the current hydrothermal process using SnCl₄·5H₂O as starting material has given rise to SnO₂ quantum dots with much defects, which totally differ from the larger mean diameter of 20 nm and higher crystallinity of stoichiometric SnO₂ nanoparticles prepared using precursor K₂SnO₃ as reported in literature.²⁴

Fig. 1 XRD patterns of sample SnO₂-SH prepared directly by hydrothermal reaction, SnO₂-AD obtained by annealing SnO₂-SH at 700 °C for 2 h, g-C₃N₄ and selected hybrids Sn–Hx.

After annealing quantum dots SnO₂-SH at 700 °C for 2 h, the diffraction peaks became sharper and more intensive as for sample SnO₂-AD in Fig. 1, which indicates an enhanced crystallinity and bigger grain size. The enhanced crystallinity would reduce the concentration of defects. It should be noted that calcinations at 700 °C did not alter the phase structure of nanocrystals, because the diffraction pattern of sample SnO₂-AD after 700 °C calcinations also matched well with the standard diffraction data of rutile SnO₂ (JCPDS, no. 41-1445). But its mean grain diameter grew up to 18.0 nm.

The as-prepared g-C₃N₄ shows two diffraction peaks at two theta of 13.0° and 27.7°, which correspond to (100) in-layer structural packing and (002) interlayer-stacking, respectively.⁵ The formation of hybrids between SnO₂ nanoparticles and g-C₃N₄ was displayed in Fig. 1. It is clear that two sets of diffraction peaks for Sn–Hx hybrids are ascribed to SnO₂ and g-C₃N₄, respectively. Because the main peaks for g-C₃N₄ and SnO₂ are close so that double peaks in the 2θ range from 26.0° to 29.0° are overlapped for Sn–H5. Even so, careful data analyses indicate that the overlapped peaks could be distinguished to be associated with (110) diffraction of SnO₂ and (002) peak of g-C₃N₄. With increasing the mass ratio of SnO₂-AD, the diffraction intensity from SnO₂-AD became increased (Fig. 1 and S2†). Furthermore, TG (Fig. S3†) measurements presented that Sn–H5, Sn–H30 and Sn–H60 samples have weight losses of about 95%, 70% and 40%, respectively, confirmed the mass ratios of defective SnO₂ in the corresponding hybrids is compared with our started materials. The formation of Sn–Hx hybrids is also confirmed by FT-IR spectra (Fig. S4†). Different from only one wide and strong absorption for SnO₂-AD appeared at 614 cm⁻¹ in the frequency range of 400 to 2000 cm⁻¹, hybrid samples exhibit three group peaks in the range of 400–800, 800–1000 and above 1000 cm⁻¹. The broad weak peaks at low frequency are attributed to Sn–O–Sn vibration, while the other peaks are characteristic vibration of g-C₃N₄.

3.1.2 TEM analysis. Co-existence of g-C₃N₄ and defective SnO₂-AD has been confirmed by XRD. To examine the impacts of defect concentration on the micro-structural characteristics of SnO₂-AD, morphologies for hybrids with given SnO₂-AD mass ratios were firstly studied by TEM and HR-TEM. As illustrated in Fig. S5A,† g-C₃N₄ sheets are stacked together. No distinct lattice fringe was observed in the corresponding HR-TEM image in Fig. S5B,† just like those reported elsewhere.²⁷ From Fig. 2A and C, one can see that SnO₂ particles (black particles) for samples Sn–H10 and Sn–H40 were covered onto the stacked g-C₃N₄ layers (gray colour part), and that several SnO₂ particles are also agglomerated. SnO₂ grain sizes were found to be distributed in a range from 15 to 25 nm, in good agreement with XRD data analyses. Lattice fringes of SnO₂ nanoparticles were given by analyzing the profiles of HR-TEM images (Fig. 2E and F). For hybrids with SnO₂-AD mass ratios of 10% and 40%, the local lattice spacing of SnO₂ nanocrystal was 0.338 nm, slightly larger than the standard d-spacing of 0.335 nm for (110) plane (JCPDS, no. 41-1445). Such a lattice expansion should be induced by defect effect, since the interactions among the surface defects might be strong enough to give a “negative pressure” that produces a lattice expansion as reported for rutile TiO₂ nanocrystals.²⁸

Fig. 2 TEM and HR-TEM images for samples: (A and B) Sn–H10, and (C and D) Sn–H40, profiles of lattice fringes for as-prepared SnO₂-AD in (E) Sn–H10 hybrid and (F) Sn–H40 hybrid.

3.1.3 XPS analysis. To probe the chemical state and interaction between SnO₂ nanocrystals and g-C₃N₄, XPS spectra were recorded for SnO₂-AD and Sn–H40. Compare with the survey XPS data for SnO₂-AD (Fig. S6†), Sn–H40 exhibits two additional photoelectron peaks at 287.2 eV and 398.8 eV, respectively, which could be assigned to C1s and N1s from g-C₃N₄. HR-TEM analysis given above has proved that lattice defect might exist in SnO₂ nanocrystals. More detailed analyses about defects were performed by fitting the fine XPS spectra via Thermo Avantage software.

As illustrated in Fig. 3a, two broad asymmetric photoelectron peaks centred at 486.3 eV and 494.7 eV were observed in Sn3d_5/2 and Sn3d_3/2 region, very closer to those of tin oxide.²⁹ When carrying on the deconvolution of Sn3d, three principles should be followed: (i) the background was corrected according to the smart model; (ii) the peak area ratio of 3d_3/2 to 3d_5/2 was fixed at 2 : 3 with a spin–orbital splitting distance of ca. 8.4 eV; and (iii) the FWHM of Sn3d_5/2 component should be not more than 1.6 eV. Data analyses showed that signal Sn3d is consisted of three components: the strong one at 486.3 eV and 494.7 eV are attributed to Sn⁴⁺ ions, while that at 484.9 eV and 493.3 eV to Sn²⁺ ions that might be induced by oxygen vacancy (V_O) via electron transfer from V_O to Sn⁴⁺ atom.³⁰ The third one Sn(X) at higher binding energy of 487.7 eV is still unknown, which may be caused by residual chloridion in SnO₂-AD. Based on the peak areas as listed in Table 1, the relative ratio of [Sn²⁺] to [Sn⁴⁺] for Sn–H40 hybrid is about 6%.

Fig. 3 Core level spectra of (a) Sn3d and (b) O1s for Sn–H40. C1s and N1s recorded for sample g-C₃N₄ and Sn–H40 hybrid are compared in (c) and (d).

Table 1 Binding energy (B.E.), full width at half maximum (FWHM) and integral areas of Sn3d and O1s core levels for sample Sn–H40

Peak components	Sn3d3/2			Sn3d5/2			O1s
	Sn^2+	Sn^4+	Sn(X)	Sn^2+	Sn^4+	Sn(X)	H2Oads	·OHads	O–Sn^4+	O–Sn^2+
B.E. (eV)	493.3	494.7	496.1	484.9	486.3	487.7	532.7	531.5	530.2	529.6
FWHM (eV)	1.44	1.41	1.58	1.58	1.44	1.58	1.50	1.57	1.43	1.25
Peak area	7519	168 136	14 840	11 245	249 251	22 857	7263	21 763	65 765	1421

---

O1s core level was deconvoluted into four sub-peaks (Fig. 3b). One at the lowest binding energy of 529.6 eV should be corresponding to the oxygen species coordinated with Sn²⁺, while that at 530.2 eV to the oxygen species coordinated with Sn⁴⁺. The other two components with higher binding energies of 531.5 and 532.6 eV are associated with the surface species, such as tin(II) hydroxychloride and surface adsorbed water.³¹ The relative intensity ratio of the component at 529.6 eV to that at 530.2 eV is about 3%. C1s and N1s core levels of pure g-C₃N₄ and Sn–H40 were comparatively studied. In Fig. 3c, the signals at 287.2 and 288.0 eV for pure g-C₃N₄ are well attributed to carbon species of C–N, and C=N. Comparatively, both signals shifted towards 286.6 eV and 288.1 eV for Sn–H40. For N1s spectra in Fig. 3d, pure g-C₃N₄ exhibited three nitrogen species: the main signal at 398.2 eV is corresponding to the N atom with sp² hybrid in N–C=N, while that at 399.5 eV to N atoms in N–(C)₃. The last one at higher binding energy of 400.6 eV contributes from the N species in N–H structure. Owing to the charging effects or positive charge localization in heterocycles, this peak was also observed at 404.3 eV, as reported elsewhere.³² For Sn–H40, N1s signals from N–C=N and N–(C)₃ shifted to 398.4 eV and 399.4 eV, respectively. The slight binding energy shifts of C1s and N1s indicated that a weak chemical interaction was formed between g-C₃N₄ and SnO₂-AD.

3.1.4 EPR analysis. To further reveal the defect species, EPR experiment was performed on SnO₂-AD sample and the reference SnO₂ annealing in 900 °C, as shown in Fig. 4. Only one signal was detected at g = 2.004. Isolated Sn⁴⁺ and Sn²⁺ ions are EPR silent due to their electron configuration of [Kr]4d¹⁰ and [Kr]4d¹⁰5s². However, when oxygen vacancies are introduced into SnO₂ nanocrystals, these oxygen vacancy species easily trap single electron to form a negative charge centre. Therefore, the observed signal at g = 2.004 is due to the single-electron trapped in oxygen vacancies located at the surface of the target oxides.³⁰ It should be noted that the electron traps (ETs) caused by oxygen vacancy in SnO₂ has an important influence on the electron migrate property tailoring in hybrids. It is obvious that EPR signal intensity of SnO₂-AD is stronger than that of SnO₂ annealed at 900 °C, which indicates amount of oxygen vacancies defects decreased with increasing the annealing temperature in air.

Fig. 4 EPR curves of the as-prepared sample SnO₂-AD and the reference SnO₂ annealing in 900 °C for 2 hours.

3.1.5 UV-vis spectral analysis. Comparing to the components of semiconductors and g-C₃N₄, hybrids formed between different types of semiconductors may have impacts on the light response. UV-vis absorbance spectrum is an effective technique to evaluate the light response ability of semiconductor and their hybrids. As shown in Fig. 5 and S7,† the as-prepared SnO₂-AD showed single absorption peak at 290 nm and absorption edge at about 390 nm, while g-C₃N₄ exhibited two absorption peaks at 280 nm and 370 nm and edging at 460 nm. All hybrids gave two absorption peaks. Moreover, with increasing the relative content of SnO₂-AD, the absorption edge displayed blue-shift, and the absorption intensity at 400 nm deceased. The energy band gaps were calculated from equation (αhν)ⁿ = K(hν − E_g), where α and E_g are the absorption coefficient and energy band gap of semiconductor, respectively. K is a constant and hν is photon energy. n is 2 or 1/2 for direct and indirect band gap semiconductor, respectively. For pure g-C₃N₄, the value of n is set at 1/2, while that for SnO₂-AD is set at 2. The energy band gap of g-C₃N₄ is thus determined to be 2.7 eV. Comparatively, a big gap value of 3.5 eV is obtained for the as-prepared SnO₂-AD, which is somewhat lower than that of 3.7 eV for the standard SnO₂ reported by Yin.²³ This observation could be closely related to the quantum confined effect.

Fig. 5 (A) Comparison of UV-vis absorbance spectra for sample SnO₂-AD, g-C₃N₄ and selected Sn–Hx, and (B) calculation of energy band gap for samples g-C₃N₄ and SnO₂-AD.

3.2 Dual maxima of photocatalytic activity for the hybrids

So far, many experiments have proved that most of the hybrid photocatalysts, consisting of different semiconductors, give the higher photocatalytic activity comparing to their corresponding components. The basic enhancement mechanism is ascribed to the improvement of the photo-induced charges separation via electron migration among different semiconductors. Even so, majority of these hybrids usually exhibit only one photocatalytic activity maximum when tailoring the mass ratio of the component semiconductors. If surface defects are introduced as the charge trap centres onto one of component surfaces in hybrid photocatalysts, the competition for photo-generated electrons between conduction band and charge trap centre will be expected, which could also affect the photocatalytic activity in water photolysis. As a consequence, the photocatalytic activity may show a different tendency with two or more activity maximum. More importantly, it may bring us new understanding about electron transportation mechanism in hybrid photocatalysts. In the following, we initiated a study on the photochemical behaviour of hybrids that are constructed by defective SnO₂ (SnO₂-AD) and g-C₃N₄ at different SnO₂ mass ratios, with a hope to get an efficient hydrogen generation from water splitting under visible light irradiation (λ > 420 nm).

As shown in Fig. 6A and Table S1,† pure g-C₃N₄ deposited by 1 wt% Pt showed a hydrogen yield of 8.37 μmol when irradiated under visible light for 5 h. When g-C₃N₄ was hybridized with SnO₂-AD, the hydrogen yield increased. Under the same test condition, Sn–H10 gave a first activity maximum with a hydrogen yield of 16.29 μmol, almost double of that for pure-C₃N₄. Further increasing the content of SnO₂-AD led to the decrease in photocatalytic activity, as indicated in Fig. 6A. For Sn–H30, the hydrogen yield decreased to 7.74 μmol, a minimum value among all hybrids. Beyond 30% content of SnO₂-AD, hydrogen yield of hybrids rose again and reached to 13.38 μmol for Sn–H40, the second activity maximum. That is, two photocatalytic activity maxima were presented for SnO₂-AD/g-C₃N₄ hybrids by tailoring the mass ratio of SnO₂-AD. It should be emphasized that, the activity of as-prepared SnO₂ in water photolysis is really poor to yield hydrogen of 3 μmol and 0 μmol (Fig. S8†) after irradiation for 5 h under the ultraviolet (λ > 260 nm) and visible (λ > 420 nm) light even though it had appropriate energy band position as well as narrower gap (3.5 eV). To confirm the negative role of oxygen vacancies in SnO₂/g-C₃N₄ hybrids, the hybrid in 40% mass ratio of SnO₂-SH was also tested in the same condition, the results indicated that the photocatalytic activity became weakened when compared with that of Sn–H40 (Fig. S9†). Such weakening in photocatalytic activity of hybrid in water splitting confirms that there should be a competitive migration between conduction band and oxygen vacancy.

Fig. 6 (A) Variation of hydrogen yields from water splitting under visible light (420 nm) irradiation for 5 h with varying SnO₂-AD content in Sn–Hx hybrids; (B) photon–hydrogen (P–H) conversion efficiency of these hybrids; and (C) comparison of the photocatalytic stability for sample Sn–H10 (red lines) and Sn–H40 (blue lines).

As displayed in Fig. 6B, the photon–hydrogen (P–H) conversion efficiency of hybrids also showed two peaks, i.e., Sn–H10 gave the first P–H conversion efficiency peak of 14.15 μmol g_C3N4⁻¹, and the second one of 23.39 μmol g_C3N4⁻¹ belongs to Sn–H45, also the highest conversion efficiency value among all hybrids. Sn–H30 and Sn–H60 exhibited P–H conversion efficiency minima. The high SnO₂-AD content dependence of photocatalytic activity for SnO₂-AD/g-C₃N₄ hybrids, especially for dual maxima observed in hydrogen yield (Fig. 6A) and P–H conversion efficiency (Fig. 6B), should be associated with the trapping of excited electron by defects of SnO₂-AD component.

In addition to the hydrogen yield and P–H conversion efficiency, photochemical stability is also a key factor in evaluating photocatalyst. As illustrated in Fig. 6C, photocatalytic activity of H₂ yield for Sn–H10 and Sn–H40 hybrids can be maintained at the same level after 5 times cycle tests, demonstrating that the obtained SnO₂-AD/g-C₃N₄ hybrids are stable in the photocatalytic process and could be potentially used in practical clean energy fields.

3.3 Origin of dual catalytic activity maxima

Introducing electron traps in SnO₂ component results in dual activity maxima for SnO₂–Hx hybrids. Therefore, there should be a different photo-generated carrier's migration mechanism from other routine hybrids with single maximum. As well known, water photolysis is a reaction that involves the contributions from photo-generated electrons/holes. The photo-generated charges have two ways of retuning back to the ground state: one is via the radiative recombination caused by electron–hole pair recombination. This process will release photon. Another one is non-radiative recombination via complex electron–phonon interaction, which would release heat. Photoluminescence (PL) test is often employed to evaluate the radiative recombination rate of photo-generated electrons/holes. Usually, the higher PL intensity indicates a higher radiative recombination rate of photo-generated charges under the same excitation wavelength, which means photo-generated charges cannot be separated completely. In the present investigation, PL spectra of hybrids were recorded and will be discussed as follows.

Fig. 7A and S10A† displayed the PL spectra of samples measured for the solid state powder directly. All samples exhibit a wide band emission peaking at about 460 nm. As a whole, PL integral intensity decreased as SnO₂-AD mass ratio increases (Fig. 7C), which means that SnO₂-AD could efficiently improve the photo-generated charges separation rate, and PL delay measurement of g-C₃N₄ and SH-5 also approved that SnO₂-AD can decrease the intrinsic charges recombination rate of g-C₃N₄ (see Fig. S12†). Pure g-C₃N₄ had the highest PL intensity among all samples, which is associated with its highest radiative recombination rate of photo-generated carries that govern its low P–H conversion efficiency in Fig. 6B. Even though sample Sn–H60 gave a weaker PL emission as indicated by its lower PL integral intensity (Fig. 7C), it did not display a high P–H conversion efficiency. This observation infers that photo-generated electrons in sample Sn–H60 do not take effectively part in the photochemical reaction. That is to say, photo-generated electrons could be trapped by oxygen vacancy in SnO₂-AD. Moreover, these trapped electrons might deliver energy to lattice vibration via phonon-electron interactions rather than the recombination with holes via radiative emission. It should be mentioned that PL test for solid state powders can only partly reveal some information about carrier separation rate at the absence of solution environment.

Fig. 7 PL intensity spectra in (A) solid state and (B) aqueous state of g-C₃N₄, selected Sn–Hx (C) PL integral areas of Sn–Hx under 390 nm excitation wavelength.

In water splitting experiment, photocatalyst participates in two photo-energy converter processes: one is the light-to-chemical photoenergy converters that yield hydrogen, and another one is light-to-light photoenergy converters that produce PL emission. Capturing PL signal in aqueous solution has been reported by Yeh et al.³³ for understanding water splitting. In fact, water photolysis reaction occurs in aqueous environment. This aqueous state would provide a quasi-in situ environment owing to the co-existence of ·OH and H⁺ in solution. Even though the absence of co-catalysts and react agents in PL test might reduce the reaction rate of splitting water into hydrogen and oxygen, PL test in aqueous state, much different from the solid state emission, is still a effective method to give more detailed information about charge action among water photolysis reactions, especially for defective photocatalysts.

PL spectra recorded for Sn–Hx hybrids in suspension aqueous environment in the absence of sacrificial agent and co-catalyst are shown in Fig. 7B and S10B.† During the test, the mass concentration of sample was kept the same so as to evaluate PL intensity of hybrids. Three different features could be easily found for solution emission from those for solid state powders. Firstly, the intensity decreases rapidly with increasing SnO₂-AD content when proceeding PL test in aqueous environment comparing to that in solid state when maintaining the test parameters, which means that more free electrons were separated by SnO₂ in aqueous state than in solid state. Secondly, PL peak position exhibited apparent blue shifts from 464 nm in solid state emission to 458 nm in aqueous emission. One may thus expect that some other kinds of emissions in hybrids might contribute to the whole PL intensity except for those in g-C₃N₄, such as the interface recombination caused by oxygen defects.³⁴ Thirdly, pure g-C₃N₄ and sample Sn–H70 gave the strongest and the weakest emission. Moreover, sample Sn–H30 exhibited much more intensive emission when comparing to the neighbouring composition (Fig. 7C). This observation is in good agreement with their P–H maxima. Even though the hybrid Sn–H20 has shown a higher aqueous PL intensity, excellent P–H conversion efficiency is still indicated, superior to those for Sn–H25 and Sn–H30. Therefore, there exist a higher radiative recombination rate of photo-generated carriers on the interface and lower electron–phonon interaction rate in oxygen vacancy. PL test in solid state and suspension environment displayed a paradox between PL intensity and P–H conversion efficiency. What should be noted, after deposited 1 wt% Pt for all of the samples, the PL spectra integral areas trend in triethanolamine solution environment is same to the PL properties in pure water (Fig. S11†). Which approved the deposition of Pt didn't change the PL properties trend in a very short period of time.

Combining with the PL properties of samples in solid and in aqueous environments, interface recombination and electron–phonon interaction caused by the electron trapping in oxygen vacancy would be expected to play a key role in photo-generated carrier separation rate. Therefore, the existence of oxygen vacancy in SnO₂ nanocrystals should be the dominated factor for the dual maxima in hydrogen yield.

Here, we proposed an oxygen vacancy induced electron competitive migration mechanism in hybrids. It is well documented that photo-generated carrier migration between different semiconductors is determined by the energy band position of component in hybrid, which could be important in affecting photocatalytic activity. As illustrated in Scheme 1, electrons in valence band (VB) of SnO₂ cannot be excited to the conduction band (CB) under visible-light irradiation due to a wide band gap (3.5 eV). All electrons that participated in the water photolysis reaction should come from g-C₃N₄. Under the irradiation of visible light, electrons are excited from the VB of g-C₃N₄ to the CB. These excited electrons would be divided into three parts: (i) the first part is those that participate directly in the reduction of H⁺ on g-C₃N₄ surface, (ii) the second part is those that move to the CB of SnO₂-AD as free electron or trapping by defect level of oxygen vacancy in SnO₂-AD as localized electron, and (iii) the third one is those that recombine with photo-generated holes via radiative or non-radiative transition. On the one hand, the lower valence band position of SnO₂-AD prevents the transporting of holes from valence band of g-C₃N₄ to that of SnO₂-AD. More importantly, less of photo-generated holes appear at VB of SnO₂-AD because the transition of hole from VB of g-C₃N₄ to that of SnO₂-AD is electrical-dipole-forbidden. Therefore, recombination possibility of electrons in the CB of SnO₂-AD with holes would be greatly reduced. These photo-generated electrons move to the CB of SnO₂-AD, and would also take part in the water splitting and enhance the photocatalytic activities.

Scheme 1 Relative band position of SnO₂-AD/g-C₃N₄ hybrid photocatalyst, and the migration of photo-generated carriers in the hybrid. The trap level affected the carriers' migration between defective direct-gap semiconductor SnO₂-AD and indirect-gap semiconductor g-C₃N₄.

The electrons that are trapped by defect level (about 2.1 eV above value band top as reported by ref. 35) of oxygen vacancy could be consumed by two processes. In the first process, trapped electrons deliver their energy to the lattice vibration of SnO₂-AD due to electron–phonon interaction, and in another one, recombining with holes from g-C₃N₄ because oxygen vacancy could become the recombination centre of photo-generated carrier.³⁶ The coulomb attraction between trapped electrons and holes help them overcome the interface potential bias between g-C₃N₄ and SnO₂-AD. Therefore, some of holes in g-C₃N₄ would be attracted to the adjacent defect position and recombine with trapped electrons on SnO₂-AD/g-C₃N₄ interface. With increasing SnO₂-AD mass ratio, more photo-generated electrons would be trapped in oxygen vacancy. In this regard, there might cause a competitive electron migration to SnO₂-AD CB and defective level. When the content of SnO₂-AD is less than 10% in hybrids, the electron migration just happen among different CB and interior of g-C₃N₄. Less of electrons would be trapped by defect level of oxygen vacancy for the little amount of SnO₂-AD. Therefore, the variation of PL intensity detected for suspension environment is the same as the P–H conversion efficiency, and reaches the first photocatalytic maximum at Sn–H10. As more SnO₂-AD component was introduced in hybrids, the amount of trapped electron increases obviously. The coulomb attraction between trapped electrons and holes become stronger at the same time, which resulted in more free holes tunnelling g-C₃N₄/SnO₂-AD interface and recombining with trapped electrons via radiative recombination. Alternatively, the free photo-generated carriers in hybrids decease due to the reducing of g-C₃N₄ mass ratio. When SnO₂-AD mass ratio reaches to 20%, the coulomb attraction would reach to a maximum, giving a higher interface radiative recombination rate. At this composition, the electron–phonon interaction is still weak, resulting in a higher aqueous PL intensity for Sn–H20 when comparing to Sn–H25 and Sn–H30. As SnO₂-AD mass ratio increases beyond 30%, the interface recombination start to decrease evidently, and electron–phonon interaction becomes the predominant energy delivering process for the trapped electron. Owing to the limited recombination rate, the second maximum in photocatalytic activity is still observed for Sn–H45.

4 Conclusions

Defective SnO₂ nanoparticles were fabricated via two steps including hydrothermal reaction and subsequent annealing. By tailoring the mass ratio of defective SnO₂ in SnO₂/g-C₃N₄ hybrids, we investigated the influence of oxygen vacancy on the photo-induced carries competitive migration and photocatalytic activity of hybrids. HR-TEM, XPS and EPR test approved that oxygen vacancies exist on the surface of SnO₂, which acted as the electron trap centre and resulted in competitive electron migration of photo-generated carriers in defective SnO₂. Comparing to other routine hybrids with single photocatalytic activity maximum, double peak value were observed in water splitting test for SnO₂/g-C₃N₄ hybrids, and the hybrids with SnO₂ relative mass ratio of 10 wt% and 40 wt% showed double peak hydrogen generation rate of 54.3 μmol h⁻¹ g⁻¹ and 44.3 μmol h⁻¹ g⁻¹, both are far beyond of pure g-C₃N₄ (27.9 μmol h⁻¹ g⁻¹) under visible light irradiation. Photon–hydrogen conversion efficiency also showed two maxima by tailoring the defective SnO₂ weight ratio. This work develops a method that regulates oxygen vacancy amount by tailoring defective SnO₂ content essential for comprehending the impacts of oxygen vacancies on photocatalytic activity. Moreover, the relevant experimental results demonstrate that the electron competitive migration is caused by oxygen vacancy in water photolysis. Therefore, the findings reported in this work could be highly helpful to deeply understand oxygen vacancy and its impacts in hybrids, and hopefully to give a new idea in designing novel photocatalysts for water splitting.

Acknowledgements

This work was financially supported by NSFC (21025104, 21271171, 21401190 and 91022018).

References

1. A. Fujishima and K. Honda, Nature, 1972, 238, 37.
2. R. Asahi, T. Morikawa, T. Ohwaki, K. Aoki and Y. Taga, Science, 2001, 293, 269–271.
3. X. Chen, L. Liu, P. Y. Yu and S. S. Mao, Science, 2011, 331, 746–750.
4. Y. Tang, H. Zhou, K. Zhang, J. Ding, T. Fan and D. Zhang, Chem. Eng. J., 2015, 262, 260–267.
5. J. X. Sun, Y. P. Yuan, L. G. Qiu, X. Jiang, A. J. Xie, Y.-H. Shen and J. F. Zhu, Dalton Trans., 2012, 41, 6756–6763.
6. W. Liu, M. Wang, C. Xu, S. Chen and X. Fu, J. Mol. Catal. A: Chem., 2013, 368, 9–15.
7. Y. He, Y. Wang, L. Zhang, B. Teng and M. Fan, Appl. Catal., B, 2015, 168, 1–8.
8. J. Zhang, Y. Wang, J. Jin, J. Zhang, Z. Lin, F. Huang and J. Yu, ACS Appl. Mater. Interfaces, 2013, 5, 10317–10324.
9. Z. L. Fang, H. F. Rong, Z. L. Ya and P. Qi, J. Mater. Sci., 2015, 50, 3057–3064.
10. Y. Xu and W. D. Zhang, Eur. J. Inorg. Chem., 2015, 1744–1751.
11. J. Shen, H. Yang, Q. Shen, Y. Feng and Q. Cai, CrystEngComm, 2014, 16, 1868–1872.
12. Y. Zang, L. Li, Y. Xu, Y. Zuo and G. Li, J. Mater. Chem. A, 2014, 2, 15774–15780.
13. H. Chen, Y. Xie, X. Sun, M. Lv, F. Wu, L. Zhang, L. Li and X. Xu, Dalton Trans., 2015, 44, 13030–13039.
14. J. Lei, Y. Chen, L. Wang, Y. Liu and J. Zhang, J. Mater. Sci., 2015, 50, 3467–3476.
15. W. Li, C. Li, B. Chen, X. Jiao and D. Chen, RSC Adv., 2015, 5, 34281–34291.
16. L. Huang, H. Xu, Y. Li, H. Li, X. Cheng, J. Xia, Y. Xu and G. Cai, Dalton Trans., 2013, 42, 8606–8616.
17. S. Chen, Y. Hu, S. Meng and X. Fu, Appl. Catal., B, 2014, 150, 564–573.
18. L. Ge, C. Han and J. Liu, Appl. Catal., B, 2011, 108, 100–107.
19. J. Yu, S. Wang, B. Cheng, Z. Lin and F. Huang, Catal. Sci. Technol., 2013, 3, 1782–1789.
20. Y. Su, J. Lang, N. Cao, T. Wang, B. Zhu and X. Wang, J. Nanopart. Res., 2015, 17, 63.
21. J. Zhang, J. H. Bang, C. Tang and P. V. Kamat, ACS Nano, 2010, 4, 387–395.
22. A. Suryawanshi, P. Dhanasekaran, D. Mhamane, S. Kelkar, S. Patil, N. Gupta and S. Ogale, Int. J. Hydrogen Energy, 2012, 37, 9584–9589.
23. R. Yin, Q. Luo, D. Wang, H. Sun, Y. Li, X. Li and J. An, J. Mater. Sci., 2014, 49, 6067–6073.
24. Y. Zang, L. Li, X. Li, R. Lin and G. Li, Chem. Eng. J., 2014, 246, 277–286.
25. Y. Zhang, L. Li, J. Zheng, Q. Li, Y. Zuo, E. Yang and G. Li, J. Phys. Chem. C, 2015, 119, 19505–19512.
26. A. Das, V. Bonu, A. K. Prasad, D. Panda, S. Dhara and A. K. Tyagi, J. Mater. Chem. C, 2014, 2, 164–171.
27. X. Wang, K. Maeda, X. Chen, K. Takanabe, K. Domen, Y. Hou, X. Fu and M. Antonietti, J. Am. Chem. Soc., 2009, 131, 1680.
28. G. S. Li, J. Boerio-Goates, B. F. Woodfield and L. P. Li, Appl. Phys. Lett., 2004, 85, 2059–2061.
29. A. W. C. Lin, N. R. Armstrong and T. Kuwana, Anal. Chem., 1977, 49, 1228–1235.
30. C. Canevali, N. Chiodini, P. DiNola, F. Morazzoni, R. Scotti and C. L. Bianchi, J. Mater. Chem., 1997, 7, 997–1002.
31. C. Fettkenhauer, G. Clavel, K. Kailasam, M. Antoniettia and D. Dontsova, Green Chem., 2015, 17, 3350–3361.
32. S. Yang, Y. Gong, J. Zhang, L. Zhan, L. Ma, Z. Fang, R. Vajtai, X. Wang and P. M. Ajayan, Adv. Mater., 2013, 25, 2452–2456.
33. T. F. Yeh, C. Y. Teng, S. J. Chen and H. S. Teng, Adv. Mater., 2014, 26, 3297–3303.
34. T. V. Torchynska, A. D. Cano, M. M. Rodriguez and L. Y. Khomenkova, Phys. B, 2003, 340, 1113–1118.
35. A. Kar, M. A. Stroscio, M. Dutta, J. Kumari and M. Meyyappan, Appl. Phys. Lett., 2009, 94, 101905.
36. A. Kar, S. Kundu and A. Patra, J. Phys. Chem. C, 2011, 115, 118–124.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra23361e

---