Enhanced light-harvesting by plasmonic hollow gold nanospheres for photovoltaic performance

A ‘sandwich'-structured TiO2NR/HGN/CdS photoanode was successfully fabricated by the electrophoretic deposition of hollow gold nanospheres (HGNs) on the surface of TiO2 nanorods (NRs). The HGNs presented a wide surface plasmon resonance character in the visible region from 540 to 630 nm, and further acted as the scatter elements and light energy ‘antennas' to trap the local-field light near the TiO2NR/CdS layer, resulting in the increase of the light harvesting. An outstanding enhancement in the photochemical behaviour of TiO2NR/HGN/CdS photoanodes was attained by the contribution of HGNs in increasing the light absorption and the number of electron-hole pairs of photosensitive semiconductors. The optimized photochemical performance of TiO2NR/HGN/CdS photoanodes by using plasmonic HGNs demonstrated their potential application in energy conversion devices.

In the past decades, considerable efforts to augment photovoltaic performance in QDSSCs by perfecting the counter electrode [23], electrolytes [24] and sensitizers [7][8][9][10][11][12][13] have been implemented. Noble metal nanoparticles (NPs), especially Au and Ag nanocrystals, have been given significant attention for their ability to couple incident photons in the visible to near IR region through the creation of surface plasmon resonances (SPRs) [25,26]. Plasmonic nanostructures can act as subwavelength scattering elements to fold light into a thin absorber, and also be used antennas to concentrate light energy surrounding them, which finally act as the electron sink for facilitating charge separation [27]. The plasmonic properties can be controlled by the type of the metal, size, shape, surrounding dielectric medium, distance between neighbouring objects and configuration of their ensemble [28][29][30]. For instance, Zhang et al. decorated CdSe-TiO 2 photoanode with Au NPs and found that the photocurrent increased to 85% and the photoconversion efficiency increased to 167% due to the electron sink of Au NPs [31]. Bardhan et al. demonstrated enhanced light harvesting in DSSCs with silica-coated Au nanocubes, resulting in a power conversion efficiency of 7.8% [32]. We also clarified the enhanced photovoltaic performance of CdS-or CdSe-sensitized photoanodes with plasmonic Au NPs as the efficient light scattering elements to couple and trap the sunlight [33,34].
Compared with the solid Au NPs, the hollow gold nanoparticles (HGNs) have stronger plasmon resonance and wider plasmonic tuning range [35,36]. However, as far as we know, there is no investigation of photovoltaic performance of DSSCs by employing HGNs as light scattering centres. In this paper, we introduce the HGNs into the TiO 2 NR (nanorod) structure by a facile electrophoretic technology (figure 1a). The CdS sensitizer was deposited on the TiO 2 NR/HGN surface by the successive ionic layer adsorption and reaction (SILAR) to construct the 'sandwich'-structured photoanodes (figure 1b). The physico-chemical properties of TiO 2 NR/HGN/CdS photoanodes and their photovoltaic performances had been systematically studied, which would enable us to better understand SPR behaviour of HGNs in a photovoltaic material with adjustable light absorption and photovoltaic response.  were supplied from Eagle Chemical Reagent Co. Ltd. All chemicals were of analytical grade and used as received. All aqueous solutions were prepared using deionized (D.I.) water with a resistivity of 18.2 MΩ cm prepared by Millipore Q purification system.

Preparation of citrate-stabilized hollow gold nanospheres
The citrate-stabilized HGNs are fabricated by using Co NPs as sacrificial templates with starting material of HAuCl 4 [37,38], as reported by Zhang et al. [39]. In brief, 75 ml of D.I. water was firstly placed into a three-necked Erlenmeyer bulb with 0.1 mol l −1 TCD in a 35°C water bath, and the solution was deoxygenated by ultrapure argon. After 30 min, 100 μl of CoCl 2 (0.4 mol l −1 ) was added into the flask under vigorous magnetic stirring (2500 r.p.m.), and a certain amount of fresh NaBH 4 (1 mol l −1 ) was added into the mixture solution. The colour of solution was changed from pale pink to brown within tens of seconds, indicating the reduction of Co 2+ to cobalt NPs. After the hydrolysis reaction of NaBH 4 was finished, 30 ml Co NP colloidal solution was transferred to aqueous solution of HAuCl 4 (0.6 mmol l −1 , 10 ml) under continuous stirring. Finally, the as-prepared HGNs solution was centrifuged twice and redispersed into D.I. water. The solid gold nanoparticles (SGNs) used as a reference were synthesized by the classic Frens' method [35].

Preparation of photoanodes for photovoltaic cells
The FTO glass substrates (10 mm × 20 mm) were cleaned ultrasonically with acetone, ethanol and D.I. water, respectively, and placed into the Teflon-liner at an angle against the wall. Then 0.32 ml of Ti(OCH 2 CH 2 CH 2 CH 3 ) 4 was slowly added into the HCl solution within 5 min and transferred into the Teflon-lined stainless steel autoclave (50 ml) for TiO 2 NR growing at 150°C for 12 h. For the electrophoretic deposition of HGNs on the TiO 2 NR anode (figure 1b), a two-electrode system was used with the Pt foil as the cathode and the separation distance of two electrodes was 15 mm. A DC voltage of 35 V was applied for 3-13 min to infuse the HGNs on the TiO 2 NR surface in a 10 ml of HGN dispersion solution (0.04 mM) while ceaselessly stirred. Then the electrodes were dried at 60°C to improve the contact between HGNs and TiO 2 NR. The TiO 2 NR/HGN samples were further sensitized in situ with CdS by SILAR deposition by immersing in 0.1 mol l −1 Cd(NO 3 ) 2 ethanol solution and 0.1 mol l −1 Na 2 S methanol : D.I. water solution (1 : 1 by volume) several times. Various quantities of HGNs and CdS were easily controlled by changing the deposition time of HGNs and SILAR cycles for CdS to optimize the photovoltaic performance of photoanodes. In addition, the FTO/TiO 2 NR/CdS photoanode without HGNs and the FTO/TiO 2 NR/SGN/CdS photoanode were also prepared as control to explore the effect of HGNs in the photoelectrochemical behaviours of photoanodes.

Characterization
The morphologies of different samples were characterized by transmission electron microscopy (TEM, Philips CM 300 FEG) at 200 kV, and scanning electron microscopy (SEM, Hitachi S4800, Japan) with an energy-dispersive X-ray (EDX) spectrometer. X-ray diffraction (XRD) pattern of sample was determined by a diffractometer (Bruker AXS D8) with Cu Kα radiation (λ = 0.15418 nm) at an accelerating voltage of 40 kV and applied current of 40 mA. The extinction spectra of samples were measured by the UV-vis spectrophotometer (Hitachi U3900). Photoelectrochemical performances of samples were performed on a CHI 630D electrochemical system (Chenhua Instruments, Shanghai, China). The system consisted of three electrodes and a transparent quartz cell, which was filled with Na 2 S (0.2 mol l −1 ) and Na 2 SO 3 (0.2 mol l −1 ) electrolyte (25 ml). A Pt foil was used as a counter electrode with Ag/AgCl/KCl as a reference electrode. All the samples (10 mm × 20 mm) with impregnated area of 1.5 cm 2 were employed as working electrodes. The curves of photovoltage and photocurrent versus times were recorded by the electrochemical system using a 150 W Xe lamp (filtered, λ ≥ 300 nm) as the light source. The illumination intensity near the sample surface was about 100 mW cm −2 . Incident photon conversion efficiency (IPCE) was measured on an IPCE measurement system (Solar Cell Scan 100, Zolix) with a 150 W Xe lamp aligned into a monchromator scanning. All measurements were carried out in ambient air and at room temperature without encapsulation.
The  Figure 6 displays the UV-vis absorption spectra of different photoanodes. The absorption of bare TiO 2 NR is featureless above 380 nm, but TiO 2 NR/HGN shows an obvious absorption in the visible region between 550 and 700 nm with respect to TiO 2 NR, which is attributed to the plasmon resonance absorption peak arising from HGNs [35,43]. As compared to the HGNs colloid in D.I. water (figure 2c), the plasmon band of HGNs for TiO 2 NR/HGN is red-shifted and broadened owing to the high refractive index of rutile TiO 2 (2.55-2.76) surrounding HGNs. Since the plasmon resonance frequency of HGNs is sensitively dependent on the surrounding matter, the local electromagnetic field enhancement around the HGNs might cause the red shift of absorption and increase light absorption of the surrounding photosensitizer [34]. This means that the optical feature of TiO 2 NR/HGN is a synergistic effect between TiO 2 NR and HGNs. The broad peak from 380 to 480 nm corresponds to the absorption of CdS QDs. Compared with TiO 2 NR/CdS, there is a significant absorption enhancement of TiO 2 NR/HGN/CdS photoanode in the region from 550 to 750 nm, which is attributed to the behaviour of HGNs. When considering the TiO 2 NR/HGN and TiO 2 NR/HGN/CdS photoanodes, the obvious red shift of the plasmon band of HGNs could be attributed to the HGNs-mediated near-field enhancement effect and near-field coupling with photosensitive semiconductor [34]. Mahmoud's study indicated that HGNs possessed stronger plasmonic electromagnetic field and wider plasmon resonant wavelengths than the special SGNs with same size via the finite difference time domain (FDTD) calculations, which hinted that the HGNs might be more suitable as the scatter centre to concentrate photons and enhance the light absorption when compared to the SGNs [44]. To explore the light-harvesting capabilities of HGNs, the optimum amount of the HGNs was found by varying the density of HGNs on the TiO 2 NR surface with the electrophoretic deposition time from 3 to 13 min (the number of SILAR cycles applied to deposit CdS was 4). It can be seen that as the amount of HGNs in the photoanodes increases (from 0.6 to 3.3 wt% obtained from EDS results), the absorption of TiO 2 NR/HGN/CdS in the visible region between 500 and 700 nm is intensified by the impact of SPR absorption peak arising from HGNs when the electrophoresis times of HGNs increases from 3 to 9 min (figure 7a), and tends to stabilize with the electrophoresis times rising up to 13 min. Besides, we observe a slight red shift in the optical absorption of TiO 2 NR/HGN/CdS, which may be attributed to an increase in HGNs clusters at a certain high amounts [45].

Photovoltaic performance of TiO 2 NR/HGN/CdS photoanode
Furthermore, the photocurrent and photovoltage of TiO 2 NR/HGN/CdS photoanodes were examined by constructing a photoelectrochemical cell system with the Ag/AgCl/KCl as a reference electrode and platinum foil as a counter electrode in the Na 2 S/Na 2 SO 3 electrolyte (0.2 mol l −1 ). The   [33,46]. This could favour the strong local-field enhancement around the HGNs and near-field coupling with the CdS semiconductor, increasing the absorption of CdS and the number of electronhole pairs [27]. However, the intensity of localized electric field is weakened with the reduced 'hot spots' when the HGNs greatly overlap [33] (as the recombination centres of the electrons and holes), resulting in the dramatic drop in the photovoltaic response for the TiO 2 NR/HGN9/CdS and TiO 2 NR/HGN13/CdS systems. The IPCE result measured in the 300-600 nm wavelength range (figure 7d) follows a similar trend to that observed in the photocurrent responses spectra (figure 7b          decrease when the number of SILAR cycles is up to 6 and 7. This phenomenon may be attributed to the following reasons: firstly, the excessive SILAR cycles would lead to the conglomeration of CdS QDs, resulting in the poor charge injection efficiency due to the large QDs; secondly, the aggregation of CdS QDs may form excessive grain boundaries between CdS NPs, which would act as potential barriers and enhance the recombination possibilities of electron-hole pairs, leading to a decrease in photovoltaic behaviours [34,47,48]; thirdly, the conglomeration of excess CdS crystal nucleus may hinder the diffusion of the electrolyte into the TiO 2 NR/HGN/CdS photoanode, which also limits the efficiency of charge separation and charge extraction, as shown in the IPCE responses (figure 8d).

Conclusion
In summary, the 'sandwich'-structured TiO 2 NR/HGN/CdS photoanode was fabricated by using plasmonic HGNs as light concentrators to effectively enhance photovoltaic performance. The light scattering and trapping of plasmonic HGNs and the near-field coupling of plasmonic HGNs with the photosensitive layers increase the absorption and the number of photoproduced electron-hole pairs of semiconductor. The TiO 2 NR/HGN/CdS photoanode with optimized design presents an obvious predominance in the improvement of photovoltaic performances such as photocurrent, open-circuit voltage and incident photon-to-current conversion efficiency when compared to the TiO 2 NR/CdS one. The plasmonic light-trapping concept would have potential applications in QDs-sensitized film solar cells and related energy conversion devices.