Chun-Ren Ke #, Jyun-Sheng Guo, Yen-Hsun Su, Jyh-Ming Ting *

The effect of silver nanoparticles/graphene-coupled TiO2 beads photocatalyst on the photoconversion efficiency of photoelectrochemical hydrogen production

Chun-Ren Ke #, Jyun-Sheng Guo, Yen-Hsun Su, Jyh-Ming Ting *

Department of Materials Science and Engineering, National Cheng Kung University, Tainan 70101, Taiwan

* Corresponding author. Tel.: +886 6 2757575, ext. 62949

E-mail address: (J.M. Ting)

# Present address: (C.R. Ke)

School of Physics and Astronomy and the Photon Science Institute, University of Manchester, Manchester M13 9PL, United Kingdom

Abstract

In this work, a novel configuration of the photoelectrochemical hydrogen production device is demonstrated. It is based on TiO2 beads as the primary photoanode material with the addition of a heterostructure of silver nanoparticles/graphene. The heterostructure not only caters to a great improvement in light harvesting efficiency (LHE) but also enhances the charge collection efficiency. For LHE, the optimized cell based on TiO2 beads/Ag/graphene shows a 47% gain as compared to the cell having a photoanode of commercial P25 TiO2 powders. For the charge collection efficiency, there is a pronounced improvement of an impressive value of 856%. The reason for the improvement in light absorption is attributed to either the light scattering of TiO2 beads or the surface plasmonic resonance on the Ag nanoparticles/graphene. The photoconversion efficiency (PCE) of the resulting cells is also presented and discussed. The PCE of the TiO2 beads/Ag/graphene cell is approximately 2.5 times than that of pure P25 cell.

Keywords:

Graphene; surface plasmonic resonance; TiO2 beads; hydrogen production; silver nanoparticles; water splitting

1.  Introduction

One of the most promising ways for pure energy generation is hydrogen production [1]. The production of hydrogen can be traced back to the year of 1972 when Honda and Fujishima reported the generation of H2 from a cell containing a TiO2 electrode exposed to ultraviolet (UV) light irradiation [2]. Following this, researchers have made enormous efforts in this field [2-19]. Currently, hydrogen production can be achieved by photosensitized water splitting through photochemical or photoelectrochemical (PEC) reaction. Owing to the fact that in a PEC system an external circuit is connected, allowing the generated photoelectrons to be transferred to the counter electrode via the external connection, the charge separation efficiency is generally better than that of the photochemical type. Therefore separation of product gases is not required in a PEC cell. In the context of the photocatalyst material selection for a PEC cell, there are a wide range of UV active semiconducting materials that have been investigated or developed. Based on the electronic configuration, these UV active photocatalysts can be classified into four groups: (1) d0 metal (Ti4+, Zr4+, Nb5+, Ta5+, W6+, and Mo6+) oxides, (2) d10 metal (In3+, Ga3+, Ge4+, Sn4+, and Sb5+) oxides, (3) f0 metal (Ce4+) oxides, and (4) a small group of non-oxides. Among them, TiO2 has been widely accepted as an efficient photocatalyst, which is low-cost, non-toxic, and photostable [20-22]. However, the problem of this material is that it is only active under the UV light, which merely occupies approximately 4% of the total solar spectrum. In order to utilize solar energy more effectively, a photocatalyst needs to absorb a wider range of the solar spectrum by expanding the absorption to the visible light (around 43% of the solar spectrum). There are many approaches to achieve this goal, of which modifying the band structure and developing composite materials are pertinent examples. One of the most effective ways to develop visible light active photocatalysts is to create impurity levels in the forbidden band through metal ion doping. This enables wide band gap photocatalysts to be active in the visible light region, and such approach has been known for a long time. Over the past decades, there have been numerous reports on the modification of wide band gap photocatalysts, including doped TiO2, SrTiO3, La2Ti2O7, and ZnS. As early as 1982, Borgarello et al. found that Cr5+-doped TiO2 can constantly produce hydrogen and oxygen under visible light (400-550 nm) irradiation [23]. Until now, a lot of different metals, such as V, Ni, Cr, Mo, Fe, Sn, and Mn, have been doped into TiO2 to improve the visible light absorption and photocatalytic activities. Furthermore, non-metal ion doping is another approach to modify UV light active photocatalysts. This approach has been widely used to narrow the band gap and improve the visible light photocatalytic activity. Unlike metal ion dopants, non-metal ion dopants are less likely to form donor levels in the forbidden band but instead shift the valence band edge upward. Although visible light active photocatalysts with proper band structures are thus developed, the efficiency enhancement seems to reach a limit [24].

On the other hand, the issue of photogenerated charge separation is another vital factor strongly affecting the efficiency of photocatalytic water splitting process. In order to increase the use of the photogenerated charges and obtain high photocatalytic water splitting activities, these charges must be effectively separated by transferring the positive and negative charges to separate sites on the surfaces of the photocatalysts, thus restricting the backward reaction of hydrogen and oxygen to form water. There are some methods to reach this goal through co-catalyst loading, semiconductor combinations, and modification of crystal structure and morphology. For co-catalyst loading, taking Pt as an example, adding this kind of noble metal can effectively separate the electron-hole pair since the Fermi energy level of noble metal is always lower than that of semiconductor photocatalysts. For semiconductor combinations, a well-known example is the use of CdS [25, 26]. Under visible light irradiation, the photogenerated electrons in the CdS particles rapidly transfer to TiO2 particles, but photogenerated holes still remain in CdS. By utilizing this method, effective electron-hole separation and charge recombination prevention can be obtained to improve the photocatalytic activity. Although this material is an appealing visible light photocatalyst for hydrogen production, it is unstable in the system, leading to a serious self-oxidation induced by the photogenerated holes in the valence band. For the modification of crystal structure and morphology, it has been found that the anatase TiO2 is more active than the rutile TiO2 [27, 28]. This is because the photogenerated electrons trapped in oxygen vacancies of anatase TiO2 can be easily quenched by the added Pt particles, whereas those in the intrinsic defects of rutile TiO2 are hardly influenced by the existence of Pt [29]. Decreasing the particle size also yields a higher efficiency in photocatalysis [30, 31]. Lee et al. reported that smaller NaTaO3 particles with a higher surface area lead to a higher photocatalytic activity in the overall water splitting owing to the increased probability of surface reactions for gas generation (either O2 or H2) rather than recombination in the bulk [32].

The methods or the materials mentioned above only partially solve the problems such that the improvement of hydrogen production is limited. For this, graphene provides an opportunity to break this limit. Graphene discovered in 2004 has been shown to possess various superior properties, namely, fast room-temperature mobility of charge carriers (200,000 cm2V-1s-1), exceptional conductivity (106 Scm-1) similar to that of silver, large theoretical specific surface area (2,630 m2g-1), and excellent optical transmittance (around 97.7%) [33]. With these unique properties, graphene finds itself a great potential to be used in PEC electrode for hydrogen production for decreasing the recombination rate of photogenerated electrons-holes pairs due to its fantastic conductivity. It transfers the electrons in an ultrafast speed, leading to a rapid decrease in the carrier recombination rate, which is expected to further increase the performance of hydrogen production. The positive contribution of graphene addition in TiO2 photoanode has been demonstrated in, for example, dye-sensitized solar cell [34] and perovskite solar cell [35]. Furthermore, graphene also exhibits advantageous surface plasmonic resonance (SPR) effect as described below.

SPR has been observed between graphene and a semiconductor. This means that, in the presence of graphene, a semiconductor photoanode can absorb more light, resulting in increased photocurrent. SPR also occurs in the interface between nanoparticle-Ag and graphene interface under illuminations of 262 and 422 nm light [36]. Furthermore, loading metal nanoparticles (such as Au and Ag) on to, for examples, TiO2 or KNbO3, allows the absorption of visible light also via SPR effect [37-40]. Therefore, the possibility of combining three materials, for example, Ag, graphene, TiO2, for even enhanced light absorption through SPR can be expected. However, there is no or little study that takes the advantage and applies to water splitting cell. The possible positive effects of these two highly conductive materials (Ag and graphene) on the performance of photoanodes are seldom addressed. As a result, we have investigated the effects of Ag and graphene additions into TiO2 photoanode in PEC. Not only the light - harvesting efficiency (LHE) for characterizing SPR but also charge collection performance for evaluating TiO2/G/Ag bridging are discussed. The improvement of photoconversion efficiency (PCE) of PECs by using this heterostructures is also demonstrated. Also, we have used homemade TiO2 beads that give better performance than the commercial P25 TiO2 particles.

2.  Experimental

The active materials used in this study include commercial P25 TiO2 powders (Degussa), homemade TiO2 beads, home-made graphene oxide (GO), and silver nanoparticles. The beads were synthesized using a two-step process [41, 42]. In short, a sol-gel method was first used to form amorphous TiOx particles, followed by a microwave-assisted hydrothermal process for transforming the particles to crystalline anatase TiO2 beads. The GO was prepared using a modified Hummer’s method as described elsewhere [43]. The silver nanoparticles were obtained using a precipitation process where a solution consisting of AgNO3, NaBH4, and polyvinyl pyrrolidone were used. To prepare the photoanode, the paste was first made by mixing P25 powders or beads with desired amounts of GO and/or silver nanoparticles. A mixture of de-ionized (DI) water and t-butanol was used as the solvent for P25-containg paste. For the paste having beads, anhydrous ethanol was used as the solvent with the addition of HCl (37%) for forming a homogeneous paste. The concentration of silver nanoparticles in the resulting paste was 0.001 M. Subsequently, the paste was spin-coated onto indium tin oxide (ITO)-coated glass substrate. Following 10 min drying in ambient air, the coating was heated at 450 oC for 4 hr to form photoanodes, which were all controlled to be around 10 μm. In the PEC water splitting cell, Pt-coated glass was used as the counter electrode, while a mixture of methanol and NaHCO3 (1 M) was used as the electrolyte. The methanol served as a sacrificing agent and facilitate the wetting of photoanodes [33]. The photoanodes and the resulting cells share the same designations, which are, for example, the photoanode made from the paste containing bead (B), 2 weight % graphene oxide (G), and silver nanoparticles (A) is designated as BG2%A. If P25 was used, the designation is PG2%A.

The resulting photoanodes and cells were subjected to various characterizations. The photoanode morphology was examined using scanning electron microscopy (SEM, JEOL 6701F). UV-vis spectrophotometry (UV-vis, PerkinElmer LAMBDA 950) was used to determine the light harvesting efficiency (LHE) of the photoanode. The LHE (or so-called absorption) was calculated via the equation of LHE (%) = 1 – T (%) = 1 – 10-A, where T is transmission and A is absorbance obtained from the UV-vis measurement under transmission mode. X-ray photoelectron spectrometer (XPS, VersaProbe PHI 5000) was used for analyzing the surface chemistry of the photoanode. Incident photon-to-electron conversion efficiency (IPCE, IQE 200) or the so-called external quantum efficiency (EQE) of the cells was also determined. Photochemical hydrogen production was performed in a gas-closed circulation system with a solar simulator (100 mWcm-2, AM1.5G; Newport 91160A) at room temperature and the PCE was thus obtained.

3.  Results and discussion

The mesoporous TiO2 beads have diameters ranging from 300 to 400 nm and are composed of numerous nanoparticles (~20 nm in diameter), which are slightly smaller than the commercial P25 powders (~25 nm in diameter), as shown in Figures S1A in the Supplementary Materials for the P25 and S1B for the mesopourous TiO2 beads photoanodes, similar to our previous research [41]. The bead sizes allow light scattering to occur as shown later. The beads and P25 powders were respectively mixed with GO for making photoanodes with or without Ag silver particles. The characteristics of the used RGO were examined using Raman spectroscopy and transmission electron microscopy (TEM) in our previous work [43]. We have found that the Raman peaks were broadened, indicating the multilayer feature of the obtained GO [44, 45], and the number of the graphene layers in the RGO was 12 based on the TEM analysis. After the 450 oC heat treatment for making into photoanode, the GO became reduced GO (RGO), as shown in Fig. 1 for Sample BG2%. The XPS C1s spectrum of BG2% before the 450 oC heat treatment is shown in Fig. 1A, while that after the heat treatment is shown in Fig. 1B. After deconvolution using XPS PEAK 41 software, the various bonding obtained and their concentrations are summarized in Table 1. After the deconvolution, four distinct peaks were found in the samples before and after heat treatment of 450 oC. The peak positions are almost identical before and after the heat treatment. For the sample without heat treatment, the four characteristic peaks are C=C (graphitic sp2 network, 284.6 eV), C-H (localized hydrocarbon, 285.5 eV), C-O (286.3 eV), and COOH (288.8 eV) [34, 46]. Among these, C-O shows the highest concentration (near half, 47.3%) and COOH also has a high percentage of 11.5%, both representing the characteristic of GO. However, after the heat treatment, there is a significant reduction in the C-O concentration from 47.3% to 15.1%. This reduction reaction also results in the increase in the C-H ratio, from 1.0% to a pronounced percentage of 35.7%. Surprisingly, C=O (287.4 eV) has a concentration 6.7% that cannot be ignored, possibly stemming from the COOH. In addition the concentration of C=C slightly rises from 40.2% to 42.5%, which is not obvious due to the insufficient heat treatment temperature. The transformation during heat treatment generally matches previous study regarding the XPS difference between GO and reduced GO (RGO) [46].

In order to evaluate the light utilization of the photoanodes, the light-harvesting efficiency (LHE) was determined using the relevant UV-vis measurement data and is shown in Fig. 2, where Fig. 2A and Fig. 2B represent the P25 and bead group, respectively. Although the LHE decreases when the wavelength exceeds 400 nm, differences can be found due to the use of beads and addition of RGO or RGO/silver nanoparticles. For comparison, the integrated LHE values normalized to that of the P25 sample (LHEnorm) are shown in Table 2. In the P25 group, the LHEnorm of PG1% (0.92) is lower than that of pristine P25 (1.00). Due to the fact that graphene is more optical transparent than TiO2, the LHEnorm is therefore lower in PG1%. Should this optical effect continue, the LHEnorm would continue to decrease as the amount of GO doubles (PG2%). However, the LHEnorm of PG2% (0.97) is better than that of PG1% (0.92). This suggests that there is another phenomenon happening. We believe that this is due to the fact that the addition of a sufficient amount graphene, for example, 2%, generates SPR [47], which therefore enhances the light absorption. For the same reason, the LHEnorm of PG2%A is higher than that of PG1%A. Furthermore, the LHEnorm of PG1%A (0.93) and PG2%A (1.00) are respectively higher than that of the counterpart samples PG1% and PG2% without the addition of Ag. It was further observed that the LHE improvement due to the addition of graphene or Ag in the P25 group occurs primarily between 400 nm and 470 nm, as shown in Fig. 2C, centering near 430 nm. It has been reported that combining graphene and silver nanoparticles, more significant resonance at a typical wavelength of 422 nm can be generated [36], and the advantage gained from the Ag nanoparticles without the addition of graphene has also been demonstrated in ZnO nanocomposites [48]. For the bead group, the addition of RGO leads to the same effect as in the case of the P25 group. However, the LHEnorm of the bead group is significantly better than that of the P25 group. As shown in Fig. 2D, more improvement of LHE occurs from 700 nm and beyond. It has been reported that the TiO2 beads induce a strong light scattering in long wavelength region [49]. The scattering effect leads to an increased light travelling distance, allowing the light to travel among bead, graphene, and silver nanoparticles for a longer time. As this happens, a very important consequence is that the aforementioned SPR would have an increased probability to occur. As a result, the LHE of BG2%A is nearly 50% higher than that of pristine P25.