Enhanced photocatalytic activity in Ag-nanoparticle-dispersed BaTiO3 composite thin films: Role of charge transfer

Optical absorption and photocatalytic activity can be enhanced by surface plasmon resonance (SPR) effect, but the charge transfer (CT) mechanism between the dispersed noble metal nanoparticles (NPs) and the semiconductor matrix has been ignored. Herein, we adduce a direct and strong evidence in Ag-nanoparticle-dispersed BaTiO3 (Ag/BTO) composite films through X-ray photoelectron and photoluminescence spectra which reveals the CT from BTO trapped by Ag NPs under UV light and from Ag NPs to BTO under visible light. Owing to the broadened optical absorption and efficient CT from Ag NPs to BTO, the Ag25/BTO film manifests the optimal photocatalytic activity under the irradiation of visible light rather than UV–Vis light. Our work provides a helpful insight to design highly efficient plasmonic photocatalyst through considering the synergetic effect of the CT between metal and semiconductor on the enhanced photocatalytic activity.


Introduction
Semiconductor photocatalysis, which provides a relatively simple and environment-friendly route for directly converting and harvesting sunlight into chemical energy, has received considerable attention to tackle the environmental pollution issues [1][2][3]. However, a number of intertwined factors limit the efficiency of photocatalysts. One bottleneck in photocatalysis is the low charge transfer (CT) rate and/or high recombination rate of electron-hole (e  -h + ) pairs, resulting in low quantum yield [4]. In addition, on account of the broad band gap (E g ) of commonly used semiconductors, e.g., TiO 2 (3.26 eV) [5], ZnO (3.2 eV) [6], and ZnS (3.66 eV) [7], the absorption in visible wavelength range is confined, which seriously weakens the photocatalytic efficiency.
Tremendous studies have been carried out to improve the photocatalytic activity, including ion doping [2,8], metal oxide complexes [9], and dispersion of noble metals [10][11][12], etc. Among these methods, dispersing metal (e.g., Au, Ag, and Pt) nanoparticles (NPs) has been proven as an effective way to improve the photocatalytic efficiency of semiconductor [13,14]. Two main mechanisms can be responsible for the enhanced photocatalytic property of semiconductor dispersed with metal NPs. On the one hand, the metal NPs can behave as an e  trap to facilitate the CT from semiconductor to metal in such heterojunction system  [15][16][17]. On the other hand, it has also been discovered that surface plasmon resonance (SPR) effect of metal NPs could broaden the absorption of semiconductor to the visible light region, and promote the CT from metal to semiconductor leading to the increased separation efficiency of e  -h + pairs [4,5,18,19]. Although the effective influence of the CT on the enhanced photocatalytic performance has been confirmed, the detailed CT mechanism between metal and semiconductor remains a matter of debate throughout the field, and direct experimental evidence needs to be disclosed.
BaTiO 3 (BTO) has attracted much attention with respect to its high dielectric constant and refractive index [20,21], chemical stability, environmental benignity, and excellent optical property [22]. However, the photocatalytic activity of BTO is also suffered by its wide E g (~3.18 eV) [23], which has response only to UV light. To date, further study about BTO has not ceased, and a few studies reveal that there are possibilities to improve the photocatalytic property of BTO via dispersing Ag NPs. For instance, Liu et al. [24] prepared flower-like BTO nanotube arrays by a hydrothermal method, and concluded that the photocatalytic performance of BTO nanotube arrays can be improved by dispersing Ag NPs. A recent study on Ag-modified BTO NPs [25] has investigated the effects of the ferroelectricity and loading of Ag on the photocatalytic property of BTO, and demonstrated that the photocatalytic efficiency could be significantly enhanced by attaching Ag onto the surface of ferroelectric BTO. Despite of a few emerging published works, the detailed effect of CT between Ag NPs and BTO on the photocatalytic activity has not been addressed up to now. This motived us to explore the possibility of using BTO thin films dispersed with Ag NPs for understanding the detailed CT mechanism in such a system and for further developing highly efficient photocatalysts.
Herein, Ag/BTO thin films were prepared by a sol-gel and spin-coating method. The effect of Ag loading on the structure, optical absorption property, along with the photocatalytic property of the thin films was investigated systematically. The considerably enhanced optical absorption property and improved photocatalytic efficiency compared with the monolithic BTO were obtained in the Ag/BTO thin films due to the SPR effect. Furthermore, by systematically comparing the photocatalytic performance of the Ag/BTO system under UV, visible, and UV-Vis light, we provided a comprehensive understanding of the SPR effect of the Ag NPs on the photocatalytic activity and the CT mechanism between Ag and BTO in the composite films, which is indispensable in designing highly efficient photocatalysts.

1 Preparation
The Ag/BTO thin films with different Ag/(Ag+BTO) ratios (abbreviated as Ag x /BTO, x = 0, 5,10,15,20,25,30, where x is the mole percent (mol%) of Ag) were prepared by a sol-gel and spin-coating method starting with barium acetate (Ba(CH 3 COO) 2 4 was added to the solution, followed by stirring for another 2 h. Several drops of nitric acid were added before stirring as hydrolysis catalysis for adjusting the pH value of the solution between 2 and 3. Then, a mixed solution of C 3 H 8 O 3 and AgNO 3 was added into the above mentioned precursor solution with a designed chemical composition of Ag x /BTO. To form the Ag x /BTO films, the ultimate solution was spin-coated onto glass substrate with size of 2 cm × 2 cm at 500 rpm for 5 s and subsequently 3000 rpm for 15 s. Finally, the thin films were dried at 150 ℃ for 30 s and successively heated to 500 ℃ for 1 h. These sequences of coating and pyrolysis treatments were repeated for 10 times.

2 Characterization methods
The phase and crystal structure of all samples were investigated by X-ray diffraction (XRD, DMAX-RB, Rigaku Inc., Japan) with Cu K radiation ( = 1.5406 Å) filtered through an Ni foil. The microstructure was observed by transmission electron microscopy (TEM, JEM-100CXП, JEOL, Japan). The chemical binding state of the constitution elements of the prepared films was characterized by X-ray photoelectron spectroscopy (XPS, PHI-5300, PHI, USA) using Al K radiation ( h = 1486.6 eV) as an X-ray source. The energies of all spectra were adjusted by normalizing with respect to the C 1s peak at 284.8 eV. The Ag 3d, Ba 3d, Ti 2p, and O 1s spectra were fitted using the XPS data which were determined by deconvolution with the program of XPS Peak 4.0. The absorption spectra were obtained by UV-Vis spectrophotometer (UV-2800, UNICO Instruments Co., Ltd., China). Steady state photoluminescence (PL) spectra of the thin films were measured using F-4500 FL spectrophotometer with an excitation wavelength of 335 nm at room temperature.

3 Photocatalytic activity assessment
The photocatalytic activity was evaluated by degradation of Rhodamine B (RhB) under UV, visible, and UV-Vis light. A 20 W UV light with emission wavelength of 365±5 nm was used as the UV light source, and a 500 W xenon lamp (Beijing Institute of Electrical Light Sources, China) was used as UV-Vis and visible light source (with a 420 nm optical filter). 20 mM RhB solution was spin-coated onto the surface of the thin films. The reaction temperature was kept at room temperature to prevent any thermal catalytic effect. The films with RhB layer were statically placed in the dark to reach the adsorption/desorption equilibrium between the photocatalyst and RhB before illumination. The degradation rate was monitored by the intensity of absorption peak of RhB (580 nm) relative to its initial intensity under various illumination times. Figure 1 shows the XRD patterns of the Ag x /BTO thin films. The standard diffraction peaks of Ag (PDF No. 04-0783) are indicated by vertical lines for comparison. A broad diffraction peak for all the samples was observed around 25°, corresponding to the amorphous BTO matrix and/or the glass substrate [26]. The absent peak of crystallized BTO after annealing indicates that the matrix is amorphous up to 500 ℃. Besides, the diffraction peaks at 2θ values of 38.1°, 44.3°, 64.4°, and 77.5° correspond to (111), (200), (220), and (311) crystal planes of the fcc Ag phase, respectively, whose peak intensity increases as x increasing, suggesting an improved crystallinity and grown grain.

1 Phase structure and morphology
To determine the dispersion and size of the Ag NPs inside the films, TEM analysis was performed for the Ag 25 /BTO thin film as shown in Fig. 2. The in-plane TEM image in Fig. 2(a) reveals that most of the Ag NPs are in spherical structure and uniformly dispersed in the amorphous BTO matrix, apart from some aggregations. The corresponding selected area electron diffraction (SAED) pattern in Fig. 2(b) shows the characteristic diffraction rings of cubic structure Ag (111), (200), (220), (311), and (222) (from inner to outer), which match well with the XRD results ( Fig. 1). The Ag NPs in the Ag 25 /BTO thin film have a wide range of size distribution from 20 to 110 nm with a mean size of ca. 52 nm as shown in distribution histogram ( Fig. 2(c)), which may be due to the partial aggregation of the Ag NPs. Further energy dispersive spectroscopy (EDS) analysis in Fig. 2(d) shows the co-existence of filmrelated Ag, O, Ba, and Ti emission peaks in addition to the Cu emission peak which is attributed to copper microgrid.

2 Chemical binding state and interaction of the involved elements
To further investigate the chemical binding state of the 2θ (°)  The appearance of Na peaks is due to the diffusion of Na atoms onto the surface from the glass substrate after annealing, the Si peak results from the SiO 2 substrate, and the carbon peak arises from an overlying contaminant hydrocarbon layer which is unavoidable for the XPS analysis. High-resolution XPS spectra of Ag 3d, Ba 3d, Ti 2p, and O 1s peaks of the Ag 25 /BTO thin film are displayed in Fig. 4. We show that the spectrum of Ag 3d (Fig. 4(a)) is made up of two characteristic peaks, Ag 0 3d 5/2 at 367.5 eV and Ag 0 3d 3/2 at 373.5 eV. The spin-orbit splitting of 3d doublet is 6.0 eV [27,28]. This spectrum is consistent with the XPS result obtained for metallic Ag, confirming that Ag in the film is in metallic form. Moreover, the binding energy (BE) of Ag 0 3d 5/2 and Ag 0 3d 3/2 for the Ag/BTO sample shifts to lower BE compared to the corresponding value of the synthesized pure metallic Ag (BE values of Ag 0 are about 368.3 and 374.3 eV, respectively) [27]. The high-resolution spectrum of Ba 3d ( Fig. 4(b)) exhibits two peaks at 779.2 (Ba 2+ 3d 5/2 ) and 794.4 eV (Ba 2+ 3d 3/2 ), which are consistent with those at 779.0 and 794.2 eV detected in BTO powders [29] respectively. The BE of 458.0 and 463.7 eV in Fig. 4

3 Optical properties and band energy
The optical absorption spectra of the Ag x /BTO thin films except for the monolithic BTO in Fig. 5 exhibit an absorption peak in the wavelength region of 450-600 nm, which results from the SPR effect of the Ag NPs under visible light. On the one hand, the SPR peak becomes stronger as x increases from 5 to 30, which is explained generally by Mie and relevant "Mie-like" theories [33][34][35]. In this study, the major factors affecting the intensity of SPR peak are the amount and size distribution of the Ag NPs along with a varied x from 5 to 30. The Ag NPs are almost uniformly dispersed in the BTO matrix when x ≤ 30 (a series of TEM micrographs with different x are not given herein), resulting in an upward SPR peak intensity. On the other hand, the broadened SPR peak may be due to the mean free path effect as well as the influence of conduction electron collisions with particle surfaces [36]. Apart from the variation of SPR peak intensity in Fig.  5, another feature is the shift of SPR peak position (named λ SPR ), which shows a slightly red shift trend from 517 to 523 nm as x increasing from 5 to 25 while an obvious blue shift to 498 nm as x further increasing to 30, as listed in Table 2. It is well known that the shift of λ SPR for small free Au NPs as increasing particle size is due to the quantum size effect (QSE) [33]. In the present study, the Ag x /BTO thin films may follow the similar behavior. Herein, it is possible that, the QSE becomes stronger due to grown particle amount and size when x increases from 0 to 25, which leads to a red shift of λ SPR ; while that weakening may be due to the oversize Ag NPs with x further increasing to 30, resulting in a blue shift of λ SPR although the SPR effect increases with an upward SPR intensity; thereby, x = 25 is the dispersion threshold.
To further understand the optical absorption behavior of the nanocomposite thin films, E g was estimated by the Tauc's equation [37]: where , A, and h are the absorption coefficient, the constant, and the incident photon energy, respectively. The exponent m is an index which determines the type of optical transition (m is 2 and 1/2 corresponding to indirectly allowed and directly allowed transitions, respectively [38]). The dependence of Fig.  6(a) as a function of photon energy h indicates that all samples are the directly allowed nature of band-toband transition. Figure 6  gradually narrowed as x increasing up to 25 leading to the red shift, and then widened as x increasing from 25 to 30 resulting in the blue shift of λ SPR , which is in well agreement with the optical absorption spectra mentioned above.

4 Photocatalytic activity
The photocatalytic activity in Fig. 7 indicates that the self-degradation of RhB is extremely weak under UV-Vis light at the absence of the catalyst, and only 2% of RhB is degraded after 5 h irradiation, while ca. 12% of RhB could be photodegraded by the monolithic BTO. The Ag x /BTO films with SPR effect show an enhanced photocatalytic property, whose degradation rates are ca. 19%, 24%, 31%, 36%, 43%, and 40% as x = 5, 10, 15, 20, 25, and 30, respectively. The highest photodegradation rate of RhB solution arrives in 43% as x = 25. The optimized photocatalytic property in Au/Co 3 O 4 [33] system was achieved in the sample with the strongest SPR peak intensity due to the more intensified electric field existed between metal and semiconductor. However, different from the Au/Co 3 O 4 system, the photodegradation rate of RhB solution for the present Ag/BTO system in Fig. 7 reaches the peak value of 43% at x = 25 instead of x = 30 which has the strongest SPR peak intensity. This might be due to the fact that the present Ag 25 /BTO film possesses smaller E g (1.70 eV) than the Ag 30 /BTO (1.76 eV), since smaller E g indicates a possibility of utilizing more visible light for semiconductor photocatalyst, thus generating larger amount of e  -h + pairs [39].

5 CT mechanism for the enhanced photocatalytic activity
To get further understanding of the SPR effect on the photocatalytic activity and the CT mechanism, we carried out PL measurement with an excitation wavelength of 335 nm at room temperature to probe the charge separation behavior in the Ag/BTO system using the sample of Ag 25 /BTO, which shows the highest photocatalytic efficiency, as well as the monolithic BTO film for comparison. Both of the spectra in Fig. 8 are composed of two emission peaks regardless of the Ag NPs dispersion. The UV-range emission at 387 nm is a radiative recombination due to the recombination between e  in the conduction band (CB) and h + in the valence band (VB). The visiblerange emission around 424 nm, called nonradiative recombination, occurs possibly due to the recombination between e  in a defect level and h + in the VB. Similar behavior has been reported in noble metal/ZnO system [40]. The emission intensity of the Ag 25 /BTO increases compared to that of the monolithic BTO film, suggesting the occurrence of CT from Ag NPs to BTO (named CT SPR ) under visible light. Similar phenomena were also observed in the Au/ZnO system [41], which showed a significantly increased intensity of emission in PL spectra compared to that of the monolithic ZnO film, demonstrating the existence of CT SPR from Au NPs to ZnO. Our present work indicates that the SPR effect of the Ag NPs could enhance the photocatalytic activity in the Ag/BTO system under UV-Vis light in Fig. 7. However, the photocatalytic performance and CT mechanism of the Ag/BTO system under UV and/or visible light have not been investigated so far. To determine which CT process between CT trap and CT SPR contributes more to the enhanced reaction rate of photocatalysis, we studied the photodegradation of RhB for the Ag 25 /BTO thin film under UV, visible, and UV-Vis light, as shown in   Fig. 9, along with the monolithic BTO under UV and UV-Vis light for comparison. An approximate photocatalytic property for the monolithic BTO under UV and UV-Vis light is observed. The photocatalytic activity of the Ag 25 /BTO is identical to the monolithic BTO under UV light, which indicates that the CT trap from BTO to Ag NPs is very poor for the Ag 25 /BTO system. Surprisingly, the sample of Ag 25 /BTO shows the fastest degradation rate (ca. 62%) under visible light where the CT SPR from Ag NPs to BTO is existed, much better than that (ca. 43%) under UV-Vis light where the CT trap from BTO to Ag NPs and the CT SPR from Ag NPs to BTO are co-existed. This indicates that the CT SPR from Ag NPs to BTO plays a dominant role on the photocatalytic reaction of the Ag/BTO, along with an opposite role caused by the CT trap from BTO to Ag NPs, when employing UV-Vis light, which will be discussed later. For better explaining the CT mechanism in the Ag/BTO system, the band structure is constructed as illustrated in Fig. 10.
Since the E f(Ag) (4.3 eV [42]) is more positive than E f(BTO) (5.3 eV [43]), when they come into contact, the e  of Ag will diffuse to BTO at the lower energy level to build up an equilibrium state E f(balance) between E f(Ag) and E f(BTO) .
Under UV light in Fig. 10(a), when the photocatalyst is irradiated by photons with energy ( h ) equal to or higher than the E g of BTO, the photoexcited e  in the VB are elevated to the CB with simultaneous generation of the same amount of h + left in the VB. As the CB bottom level of BTO is higher than the E f(balance) , the CT trap from BTO to Ag NPs as indicated by a green arrow would occur actuated by the above potential energy, which can promote the separation and minimize the recombination of e  -h + pairs of BTO to some extent. It has been reported that the enhanced photocatalytic activity of semiconductor dispersed with metal NPs could be ascribed to the possible reason that nanostructured metal (in our case Ag) can act as e  trap, leading to the efficient separation of e  -h + pairs, and in turn to enhance the photocatalytic property [44]. Such CT trap from BTO to Ag NPs behavior in the present work demonstrated by the XPS result (Fig. 4) helps to minimize the e  -h + recombination. Similar CT behavior has previously been reported in the Au/TiO 2 system [45], which is however irrelevant to the SPR effect.
It is well known that no e  -h + pairs are generated in the monolithic BTO under visible light because of the lower energy of irradiated photons than the E g of BTO. As a result, there is no photocatalytic performance for the monolithic BTO under visible light. After dispersing Ag NPs in the BTO matrix, the Ag/BTO shows the excitation of the Ag NPs SPR effect under visible light (Fig. 5) and enhanced visible photocatalytic activity   (Fig. 9), which could be ascribed to several possible reasons. Firstly, noble metal NPs, such as Ag and Au, can broaden the optical absorption over a wider visible light wavelength range through the SPR effect [46]. Secondly, the SPR effect could also play an important role in enhancing the rate of photocatalytic reaction, through injecting hot e  from metal NPs to semiconductor, that is, the CT SPR from Ag NPs to BTO in our case as demonstrated by the PL spectra in Fig. 8. The excited SPR effect under visible light induces the efficient CT SPR from Ag NPs to BTO as indicated by blue arrows in Fig. 10(b), which is an opposite direction to the CT trap from BTO to Ag NPs happened under UV light. Such CT SPR from Ag NPs to BTO can enhance the generation of active •O 2  radicals on the BTO surface, while the h + retained in the Ag NPs can be used for the generation of active •OH radicals [47], both of which contribute to the degradation of RhB. Furthermore, the SPR effect could be intensified with the increase of x, resulting in a larger amount of the CT SPR from Ag NPs to BTO. Moreover, the work of Cushing et al. [48,49] pointed out that the SPR effect can also excite e  -h + pairs in semiconductor by transferring the plasmonic energy from metal to semiconductor, and more photoexcited charge carriers will participate in the process of photocatalysis, resulting in the enhanced photocatalytic property.
In the last case, under the irradiation of UV-Vis light (Fig. 10(c)), the CT mechanism for the monolithic BTO is consistent with that under UV light ( Fig. 10(a)), leading to the approximate photocatalytic activity of the monolithic BTO under UV and UV-Vis light (Fig. 9). However, both of the CT trap from BTO to Ag NPs and the CT SPR from Ag NPs to BTO under UV-Vis light would occur after dispersing Ag NPs; that is, the photogenerated e  in the BTO induced by UV light will transfer to the Ag NPs, and the Ag NPs can inject hot e  to BTO due to the SPR effect induced by visible light simultaneously. On the other hand, the Ag NPs and BTO can respectively inject the trapped e  back into the BTO and Ag NPs under UV-Vis light; hence, the time that the e  resides in the Ag NPs and BTO under UV-Vis light is shortened in comparison to that under UV or visible light alone. These two processes will promote the recombination of e  -h + pairs to a certain extent and impair the photocatalytic activity [50]. Meanwhile, collisions would possibly occur in the process of e  transferring from one to another between Ag NPs and BTO in different directions, whereby the migration rate will decrease to some extent. Hence, above pros and cons exist simultaneously in the photocatalytic reaction of the Ag/BTO when employing UV-Vis light, consequently leading to the lower photocatalytic activity of Ag x /BTO under UV-Vis light compared to that under visible light (Fig. 9). A similar phenomenon was observed in the UV-Vis light irradiated Au/TiO 2 photocatalysis system although the authors mainly focused on the pros and cons of localized SPR [50]. Table 2 lists the potential energy (1240/λ SPR ) of the hot e  (named hot e h  ) of the Ag x /BTO thin films through the λ SPR of the absorption spectra (Fig. 5) which is induced by the SPR effect of metal NPs escaping from metal to semiconductor [51]. As long as the hot e  has higher energy than the CB level bottom of the BTO, the CT SPR from Ag NPs to BTO could occur. Here, we introduce the concept of E determined by the difference between the SPR state and the CB of BTO, i.e., E = hot e h   E CB , as shown in Table 2. From the standpoint of energy, e  can more easily transfer between two more closely matched energy levels, implying that the smaller E is more beneficial to the CT SPR from Ag NPs to BTO. We have shown that the Ag 25 /BTO film which has the lowest E possesses the best photocatalytic activity (Fig. 7). These interesting phenomena only observed herein would provide a useful guidance for further designing highly efficient plasmonic photocatalysts.

Conclusions
Ag x /BTO (0 ≤ x ≤ 30) composite thin films are prepared using a sol-gel and spin-coating method. The approximately spherical Ag NPs are dispersed in the amorphous BTO matrix. All the Ag x /BTO thin films except for the monolithic BTO exhibit an absorption peak in the visible wavelength region of 450-600 nm resulting from the SPR effect of the Ag NPs; the SPR