TiO2 nanofiber photoelectrochemical cells loaded with sub-12 nm AuNPs: Size dependent performance evaluation

Incorporation of gold nanoparticles (AuNPs) into titanium dioxide (TiO2) photoelectrodes has been used traditionally to increase the performance of photoelectrochemical cells (PECs) through their tailored optical properties. In contrast to larger AuNPs, previous studies have suggested that smaller AuNPs are the most catalytic or effective at increasing the photovoltaic (PV) performance of TiO2 photoelectrodes based on PECs. Despite this, AuNPs are often only compared between sizes of 12e300 nm in diameter due to the most common synthesis, the Turkevich method, being best controlled in this region. However, the optimum radius for citrate-capped AuNPs sized between 5 and 12 nm, and their influence on the PV performances has not yet been investigated. In addition to using AuNPs in the photoelectrodes, replacing traditional TiO2 NPs with one-dimensional nanofibers (NFs) is a promising strategy to enhance the PV efficiency of the PECs due their capability to provide a direct pathway for charge transport. Herein, we exploit the advantages of two different nanostructured materials, TiO2 NFs and sub-12 nm AuNPs (5, 8, 10, and 12 nm), and fabricate composite based photoelectrodes to conduct a size dependent performance evaluation. The PECs assembled with 8 nm AuNPs showed ~20% improvement in the average power conversion efficiency compared to the control PECs without AuNPs. The highest performing PEC achieved a power conversion efficiency of 8%, which to the best of our knowledge, is among the highest reported for scattering layers based on pure anatase TiO2 NFs. On the basis of our comprehensive investigations, we attribute this enhanced device performance using 8 nm AuNPs in the TiO2 NF photoelectrodes to the improved spectral absorption, decreased series resistance, and an increase in electron transport and injection rate leading to an increase in current density and fill factor. © 2018 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
Titanium dioxide (TiO 2 ) is the most extensively studied metal oxide semiconductor and is well-known for its chemical stability, environmental compatibility and low production cost [1,2]. With electrons as its majority carriers, the importance of TiO 2 in photocatalysis relies on its ability to oxidize oxygen or organic materials [3,4]. The high photocatalytic activity of TiO 2 is based upon the reduction and oxidation potentials situated within its wide band-gap. This allows for its use in a wide range of applications including artificial photosynthesis [5] and photoelectrochemical cells (PECs) [6]. For TiO 2 and its application in PECs, one of the most important discoveries is the dye-sensitized solar cells (DSSC); one of the promising third generation solar cells [2,7,8]. DSSCs are a PEC which typically consist of a TiO 2 photoelectrode, and a platinum counter electrode, separated by a liquid electrolyte containing either iodide/triiodide I À /I 3 À or cobalt(II/III)based redox mediators [2,9]. While several groups have now replaced the liquid electrolyte with solid state hole transport materials such as Spiro-MeOTAD [10], solid-state DSSCs are typically less efficient than liquid electrolyte-based DSSCs with the highest performance to date being 11% [8]. Although state-of-the-art liquid electrolyte-based DSSCs have achieved remarkable efficiencies of 13% for porphyrin-sensitized [11], and 14% for silyl-anchor and carboxy-anchor dyes [12], TiO 2 nanoparticulate-based PECs suffer from the following drawbacks: (i) limited charge transport [13], (ii) poor absorption of low energy photons [14], and (iii) high charge recombination [15]. The limited charge transport of nanoparticulate TiO 2 PECs can be reduced by substituting nanoparticles (NPs) for one-dimensional (1D) nanofibers (NFs), which have been previously shown to reduce the scattering of free electrons by reducing the grain boundaries of the interconnected TiO 2 NPs [13].
Our previous work has shown that the incorporation of TiO 2 NFs (and other 1D inorganic nanostructures) in both DSSCs and solidstate perovskite solar cells enables fast electron transport by offering enhanced directionality from the 1D structures [16,17].
On the other hand, spherical noble metal nanostructures such as gold nanoparticles (AuNPs) show several beneficial effects related to light trapping (scattering), fluorescence and local field enhancement, facilitating charge separation and charge transfer, all of which are extremely important in DSSC devices [18e21]. Not surprisingly, all these effects are related to the tailored optical properties and previous studies have suggested three methods of light trapping which result in enhanced DSSC performance using noble metal NPs; 1) trapping incident light inside thin films via scattering elements, 2) energy localization in the near field of the NP, and 3) charge-propagation kinetics at the noble metal NP/TiO 2 interface from the surface plasmon polariton waves [19]. The plasmonic properties in AuNPs are tunable which enables controlled manipulation of the optical properties over the visible spectral range in a variety of photovoltaic devices. AuNPs have previously been incorporated into both the active and scattering layer of TiO 2 in DSSCs [18,22,23]. Regardless of the layer in which AuNPs are incorporated, the observed increase in performance is mainly from an enhanced photocurrent density of the DSSCs containing AuNPs [19]. This performance improvement has been previously related to increased dye-loading due to superior surface area coverage [24], but is more commonly attributed to the plasmonic effect of the AuNPs [19]. Literature has also shown that the amount of AuNPs added to a DSSC system is very important and an excess will result in a decrease in performance or interact with the electrolyte causing problems such as recombination of charge carriers. Heavy loadings (>1.0 wt%) of metal NPs (Ag and Au) have previously been shown to increase the probability of aggregation between the metal NPs in the composite films [21]. In addition to the use of metal NPs, TiO 2 only absorbs in the UV-region and as a result, a light absorbing material (such as a molecular dye) is usually introduced. The most common dye used in DSSCs is the ruthenium-based dye molecule (known as N719) and given its strong absorption band at 530 nm aligns well with the surface plasmonic resonance of AuNPs sized between 5 and 50 nm in diameter (515e540 nm). Notably, red shifting the absorption too far causes dimerization and would no longer contribute to an increase in efficiency. Other than loading amounts, size effects also play a crucial role in performance enhancement. Wang et al. investigated DSSCs loaded with 5, 45, and 110 nm AuNPs and compared their performance [19]. They showed that 40 nm AuNPs show the highest photocurrent due to the perfectly matched absorbance overlap with N719. However, DSSCs loaded with the smallest AuNPs (5 nm in diameter) has the highest power conversion efficiency (PCE). Typically, smaller AuNPs have reduced light scattering effects (scattering cross section) but higher light absorption (absorption cross section). On the other hand, a collection of small AuNPs will scatter more than a single large AuNP. These plasmonic features of small AuNPs (or a collection of AuNPs) can contribute to improved electron collection originating from the increased electron lifetimes and this contributes considerably to the overall performance enhancement [19].
While the combination of plasmonic NPs with TiO 2 to enhance the performance of PECs is well established, small AuNPs are commonly seen as the most catalytic in PECs (for reasons described earlier) [25]. While other synthetic approaches have successfully developed protocols to make sub-12 nm AuNPs, they typically use insulating capping ligands such as polyvinylpyrrolidone (PVP) [25], poly(ethylene glycol) (PEG) [26], polymeric stabilizes [27], strong surfactants [28], and other thiol stabilized moieties [29]. Long chained ligands and hydrophobic thiol coatings have limited capability with TiO 2 PECs, since they will either be insulating or insoluble. An approach by Piella et al. [30] encompasses a kinetically controlled seeded growth strategy to enable the production of sub-12 nm citrate stabilized AuNPs in water, making them fully compatible with TiO 2 photoelectrodes.
Herein, we utilize the advantages of two different nanostructured materials, TiO 2 NFs and AuNPs, and fabricate composite based photoelectrodes on fluorine doped tin oxide (FTO), in which TiO 2 NFs act as strong light scattering layer loaded with size controlled sub-12 nm AuNPs. A schematic of the composite PEC is illustrated in Scheme 1, where a scanning electron microscopy (SEM) image of the TiO 2 NF layer is rendered to visualize the lawnlike architecture. This work is the first example where the PEC performance of TiO 2 NF photoelectrodes has been finely tuned using sub-12 nm AuNPs to determine the smaller size limits for enhancing the performance of NF-based PECs. We describe a simple route to fabricate electrospun electrodes for PECs and explore how the addition of size controlled sub-12 nm AuNPs into the electrospun scattering layer can improve the performance in a PEC system. Scheme 1. Schematic represents the TiO 2 NF photoelectrode loaded with AuNPs and its assembly into a PEC. The assembly is represented by I: FTO, II TiO 2 NPs (active layer), III AuNP-TiO 2 NFs (scattering layer), IV: ruthenium dye molecules (N719), V: iodidebased electrolyte, and VI: FTO/platinum.

Materials
All chemicals were purchased from Sigma Aldrich (United Kingdom) unless otherwise noted. The sizes of the nanoparticles quoted always refer to the diameter of the particles.

Synthesis of AuNPs
5 nm AuNPs were synthesized following a procedure previously reported by Piella et al. [30] In a typical synthesis, a 150 mL aqueous solution of sodium citrate (SC, 2.2 mM) containing 0.1 mL of tannic acid (2.5 mM) and 1 mL of potassium carbonate (K 2 CO 3 , 150 mM) was heated to 70 C with vigorous stirring. Upon reaching 70 C, 1 mL of tetrachloroauric acid (HAuCl 4 , 25 mM) was quickly injected which resulted in an immediate color change from clear to black, and then to soft pink. The solution was kept at the same temperature for 10 min to ensure complete reduction of the gold precursor. A small aliquot of the solution was removed from the reaction vessel and used for UVeVis, TEM, and for the incorporation into TiO 2 photoelectrodes.

Growth of AuNPs
AuNPs of 8, 10, and 12 nm were then subsequently grown from the 5 nm AuNP solution (described in the previous section). The 5 nm AuNP solution was diluted by extracting 55 mL and adding another 55 mL of SC (2.2 mM). The solution was then heated again to 70 C, and two injections of 0.5 mL HAuCl 4 (25 mM) were added (in 10-minute intervals). This growing step was repeated until the desired particle sizes were reached. The particles had a concentration of 7 Â 10 13 NP mL À1 and were stored in a dark cool place.

Preparation of TiO 2 NFs
TiO 2 NFs were prepared according to our previous reports [16,17]. Briefly, TiO 2 NFs were electrospun from a solegel solution containing titanium(IV) butoxide (TBO, 0.5 g), poly(vinyl pyrrolidone) (PVP, 1 g), and absolute ethanol (10 mL). The TiO 2 solegel was electrospun at a voltage of 10 kV with a flow rate of 1 mL per hour and the collection plate was maintained at a distance of 20 cm. Once enough fiber mats had been obtained, the PVP coated TiO 2 NFs were subject to pyrolysis at 500 C to remove PVP.

Preparation of TiO 2 NF photoelectrodes
For control samples, pure TiO 2 NFs (0.5 g) were mixed with ethanol (20 mL) and the mixture was sonicated for at least 2 h, until no fiber agglomerates could be seen. Once a uniform white solution was achieved, 2 mL of terpineol was added and the mixture was sonicated for a further 15 min. Finally, ethanol was removed by rotary evaporation to yield a white paste of TiO 2 NFs. The TiO 2 NF paste was then stored until it was required. Transparent TiO 2 photoanodes were prepared by doctor blading TiO 2 nanoparticle paste (GreatCell Solar, DSL-18NRT) onto fluorine-doped tin oxide (FTO). The transparent photoelectrodes were sintered at 500 C to remove all organic binders from the paste and form a uniform transparent TiO 2 NP anatase active layer. The photoelectrodes were then subject to treatment in titania tetrachloride (TiCl 4 , 40 mM) for 45 min at 70 C. The photoelectrodes were then thoroughly washed with Milli-Q water and sintered at 500 C. A thinner opaque scattering layer of TiO 2 NFs was then added on top of the transparent layer, using the same technique. The photoelectrodes were again sintered at 500 C to remove all organic binders from the NF paste.
Notably, the TiCl 4 treatment was only used on the transparent layer, not the scattering layer. For samples that incorporated AuNPs, 0.5 wt% of each NP size variation was added to the TiO 2 NF pastes (scattering layer only) and the same fabrication procedure was followed. Note, 0.5 wt% was calculated on the pure mass of TiO 2 NFs and not the paste mass. The loading of 0.5 wt% was chosen as previous work has demonstrated loadings of 0.5 wt% to be the most effective in similar systems, where exceeding 1 wt% loading of AuNPs has been shown to decrease the performance of PECs [21]. The loading of 0.5 wt% AuNPs was kept constant for all photoelectrodes used in this work.

Photoelectrochemical cell fabrication
Directly after sintering, the warm photoelectrodes (TiO 2 NP/ AuNP-TiO 2 NFs) with a total cell area of 0.125 cm 2 were sensitized with an absolute ethanol solution of ruthenizer 535-bisTBA (Solaronix, 0.5 mM) for 20 h in the dark. The sensitized photoelectrodes were dark purple in color and were then carefully washed with ethanol to remove any unanchored dye molecules from the surface. Sensitized photoelectrodes were eclipsed onto platinum-coated FTO (GreatCell Solar) and sealed with Low Temperature Thermoplastic Sealant (GreatCell Solar, MS004610). The electrodes were heated to 100 C for 15 min to seal the sealant. The iodide-based redox mediator (iodine (I 2 , 0.05 M), Lithium Iodide (LiI, 0.1 M), 1,2-dimethyl-3-propylimidazolium iodide (DMPII, 0.6 M), and 4-tert-Butylpyridine (TBP, 0.5 M) in acetonitrile) was injected into the freshly sealed electrodes by vacuum filling.

Materials chemistry and characterization
All details for the materials chemistry and photovoltaic characterization can be found in the Supporting information.

Results and discussion
Each batch of AuNPs was characterized using UVevisible (UVeVis) absorbance spectroscopy and transmission electron microscopy (TEM) (Fig. 1). A UVeVis absorbance spectrum of each of the different sized AuNPs is shown in Fig. 1a. The localized surface plasmon resonance (LSPR) peak for 5, 8, 10 and 12 nm were 512, 515, 517, and 519 nm respectively, where the red shift in absorbance is consistent with an increase in particle size [30]. All LSPR peaks had similar full-width half maximums (FWHM), and the AuNP concentration was~7 Â 10 13 NPs mL À1 . Furthermore, the absorbance edge of the LSPR peak, slightly overlaps with the peaks for our light absorbing dye-molecule N719 (in ethanol), which can be seen in Fig. 1a. This suggests that any improvement in light harvesting can be partially due to the plasmonic effect of the AuNPs. The size distribution of the AuNPs was calculated by generating a series of histograms from measuring the size of many AuNPs over several areas on the TEM grid (Fig. 1b). The particle size distribution for 5, 8, 10 and 12 nm was 0.6 (FWHM: 1.3), 0.7 (FWHM: 1.6), 0.8 (FWHM: 1.9), and 1.0 (FWHM: 2.5) nm respectively, where the standard deviation of the AuNP size increases with the size of the nanoparticles [30], which is a typical incremental pattern for seedinitiated growth of AuNPs. The TEM images used for the particle counting are shown in Figs. S1eS4 in the Supporting information. The corresponding TEM, depicted in Fig. 1c, shows that all AuNPs are spherical in shape and highly crystalline with lattice separation of 0.21 and 0.23 nm, indicative of the {200} and {111} planes respectively [31]. The AuNPs were stored in a cool dark place and remained freely dispersed in water for several months without showing any signs of aggregation. TiO 2 NFs were synthesized and worked into a screen printable paste via our previous method [16]. Additional TiO 2 NF pastes were prepared and individually doped with 5, 8, 10, and 12 nm citrate stabilized AuNPs. This NF-based paste was then used as a scattering layer in our photoelectrode. All photoelectrodes were assembled on FTO with an underlying active layer of TiO 2 NPs. One dimensional (1D) nanostructures often suffer from poor dye loading, where in NFs, dye uptake decreases as a function of fiber diameter [32]. Therefore, the active layer was prepared using TiO 2 NPs to ensure the sufficient attachment of N719 dye molecules for our PECs. A schematic of the complete PEC assembly can be seen in Scheme 1. The cross-sectional scanning electron microscopy (SEM) image of the TiO 2 NP/AuNP-TiO 2 NF photoelectrodes is shown in Fig. 2a. The total thickness of the electrodes was 13 mm, verified using a stepprofile meter. The cross-section is made up of three main parts, 'I' represents the FTO, 'II' shows the 8 mm active layer of TiO 2 NPs, and 'III' illustrates the 5 mm scattering layer of AuNP-TiO 2 NF photoelectrodes. These thicknesses align with the highest performing PECs, where the thicker active layer provides effective dye adsorption [33], and the thinner scattering layer enables efficient light scattering. To image the AuNPs in the TiO 2 NF layer, highresolution (HR) SEM of the TiO 2 NP/AuNP-TiO 2 NF photoelectrodes is shown in Fig. 2b. The TiO 2 NFs had diameters between 200 and 500 nm and were several microns in length. Despite the use of HR-SEM, the AuNPs were not visible in Fig. 2b, which we attribute to the low concentration of AuNPs added (0.5 wt%) [19]. In order to visualize the AuNPs in the TiO 2 NFs, HR-TEM was used to characterize the composite material (Fig. 2c). While the contrast makes it difficult to pin-point a single AuNP within the NFs, elemental mapping confirmed the presence of the AuNPs within the TiO 2 NFs (see Fig. 2c).
To further verify the presence of the AuNPs in the TiO 2 NF layer of the photoelectrodes, X-ray Photoelectron Spectroscopy (XPS) was performed and the results are shown in Fig. 3a and b. To ensure AuNPs were not only attached to the surface, a monatomic depth profile of the TiO 2 NP/AuNP-TiO 2 NF photoelectrodes was measured. Depth profiling involved using a calibrated ion beam to etch layers of the photoelectrode, revealing subsurface information.
The etching was performed for 400 s which corresponded to a penetration of~100 nm into the surface of the electrode. The depth profile analysis was performed to review sub-surface information and confirm the presence of the AuNPs embedded inside the TiO 2 NFs. Fig. 3a verified the TiO 2 by the Ti2p peaks present at binding energies 464 eV and 458 eV, indicative of Ti2p (1/2) and Ti2p (3/2) , respectively. The presence of AuNPs was verified in Fig. 3b at binding energies of 83.7 eV, and 87.5 eV, indicative of Au4f (7/2) and Au4f (5/2) , respectively. AuNP peaks in 4f regions correspond to a spin orbital splitting of 3.8 eV. Previous studies of AuNPs on TiO 2 show that for Au þ 4f (7/2) and Au 3þ 4f (7/2) , there is a significant shift towards higher binding energies 84.4 eV and 86.5 eV [34,35]. Fig. 3b clearly shows only one species for the Au4f (7/2) 83.7 eV that is most closely matched with metallic Au 0 which is typically present at 84.0 eV. The negative shift of 0.3 eV for Au 0 4f (7/2) has been attributed to strong interactions between the Au in the TiO 2 framework [35]. Measuring the XPS of AuNPs at such low concentrations is challenging and we attribute their low concentration (0.5 wt%) to be the reason for the noisy Au4f spectrum, which has been reported before [36]. Surface scans for the Au4f peaks, confirming the presence of AuNP at the surface, can be seen in the Supporting information, Fig. S5. The surface scan measurements confirm the presence of AuNPs on the surface and the depth profile measurements confirm that the AuNPs are also within the porous TiO 2 NF matrix. The anatase phase of the TiO 2 was maintained throughout doping. This was investigated using X-ray diffraction (XRD) where the JCPDS card number was confirmed for anatase TiO 2 (JCPDS card number 21-1272). The peaks not assigned to the anatase phase of TiO 2 can be assigned to the FTO (Fig. 3c). While no shifting in the diffraction pattern was seen, the TiO 2 NP/AuNP-TiO 2 NF photoelectrodes showed slight broadening of the XRD peaks which can be attributed to the incorporation of AuNPs into the TiO 2 lattice [13]. Furthermore, mesoporous TiO 2 structures which have low AuNP content (<0.5%) have previously been shown not to exhibit any Au characteristic diffraction, whereby changes are typically observed with increasing Au content and/or increasing particle size [35]. Since the XRD pattern is typically only associated with long range order of structural arrangements, UVeVis measurements confirmed that the absorbance edge for both TiO 2 and TiO 2 /AuNP NF electrodes did not change, Fig. 3d. No SPR peak was visible in the TiO 2 NP/AuNP-TiO 2 NF photoelectrodes, and this was again attributed to the low concentration of AuNPs present in the NFs. The inset of the UVeVis measurements shows the Tauc plot, confirming the band gap of the electrode is 3.15 eV for both TiO 2 NF control and AuNP-TiO 2 NF, which are expected to exhibit similar open-circuit voltage (V oc ) values in the PECs. To gain further insight to the optical features of our TiO 2 NP/AuNP-TiO 2 NF photoelectrodes, Raman spectroscopy was also performed. The Raman spectra of both pure TiO 2 and TiO 2 NP/AuNP-TiO 2 NF photoelectrodes are shown in Fig. S6 (SI). Both TiO 2 and TiO 2 NP/AuNP-TiO 2 NF photoelectrodes show the TiO 2 Raman active vibration modes E g (143 cm À1 ), A 1g (395 cm À1 ), B 1 (514 cm À1 ), and an overtone at E g (639 cm À1 ). Vibrational mode E g (143 cm À1 ), distinctly identifies the anatase phase for both TiO 2 NP/TiO 2 NF control and TiO 2 NP/AuNP-TiO 2 NF photoelectrodes where the decrease in intensity and slight peak broadening for the AuNP containing photoelectrode can be attributed to surface adsorbed AuNPs [23]. This was confirmed over several points on the photoelectrode, where the height of the sample was kept constant for both TiO 2 NP/TiO 2 NF control and TiO 2 NP/AuNP-TiO 2 NF photoelectrodes. The results for the Raman measurements suggest a slight increase in crystalline defects where the AuNPs incorporated into the TiO 2 NFs provide additional photoelectron trapping centers which helps to reduce charge recombination, a major problem in TiO 2 based PECs [35].
The photovoltaic performance of TiO 2 NP/TiO 2 NF control and TiO 2 NP/AuNP-TiO 2 NF photoelectrodes (containing 5, 8, 10, and 12 nm AuNPs) were analyzed under 1 sun conditions (AM 1.5 simulated sunlight). The JeV characteristics and incident photon to current conversion efficiency (IPCE) are plotted in Fig. 4 and the corresponding photovoltaic properties of each device have been tabulated in Table 1. The TiO 2 NF control devices achieved a shortcircuit current density (J SC ) of 15.8 mA cm À2 , an open-circuit voltage (V OC ) of 757.9 mV, fill factor (FF) of 57.7% and overall power conversion efficiency (PCE) of 6.7%. Interestingly the highest performing device was that containing 8 nm AuNPs, where the best device achieved a J SC of 17.4 mA cm À2 , V OC of 745.7 mV, FF of 62.4% and PCE of 8.0%. The 19.4% improvement in PCE for the TiO 2 NP/ AuNP-TiO 2 NF (8 nm) device can be attributed to the increase in J SC and FF, consistent with previous reports [20,37]. The J SC and FF for the TiO 2 /AuNP-TiO 2 NF (8 nm) device improved by 10% and 8%, respectively when compared to the control device. The addition of 5 nm AuNPs in the TiO 2 NP/AuNP-TiO 2 NF photoelectrodes also resulted in an enhancement in J SC (16.6 mA cm À2 ) and FF. Surprisingly, additions of 10 and 12 nm AuNPs decreased the J SC (Fig. 4a) but still increased the FF. The increase in the J SC for the 5 and 8 nm TiO 2 NP/AuNP-TiO 2 NF photoelectrodes can be attributed to the more favorable surface area of the sub-12 nm plasmonic NPs [25]. Wang et al. [19] previously showed that photocharging, light scattering, and the LSPR effect all co-exist as factors causing the increase in performance of TiO 2 NP/AuNP-TiO 2 NF photoelectrodes, but are selectively dominant in different sizes of AuNPs, whereby they report 5 nm AuNPs to be the highest performing when compared to 45 and 110 nm AuNPs [19]. While previous results in similar systems have suggested that the V OC increases with decreasing particle size, our results show the V OC stays relatively   constant for all devices. We attribute our result to the amount of AuNPs added being too low to have any upward shift on the Fermi level [19], and given our small size increments, any change in V OC would be difficult to observe. This is in good agreement with what we have seen from our Tauc plot (Fig. 3d). Furthermore, adding low amounts of AuNPs has been shown to be far more beneficial than larger loadings due to the genuine concern of electrolyte attack on metal AuNPs [20,38]. The FF substantially increased for all devices containing the AuNPs. This increase in the FF value is commonplace for TiO 2 NP/ AuNP-TiO 2 NF photoelectrodes and has been frequently attributed to a reduction in the charge recombination [19,20]. Electron effective lifetime and the recombination resistance have been previously shown to decrease as AuNP size increases and there is a suppression of interfacial recombination compared with the control photoelectrodes [19]. Therefore when AuNPs are present, previous studies show longer electron lifetimes which correlates to the longer existence of free electrons in the conduction band of the TiO 2 , thus reducing the chances of recombination in the PEC [19]. Unsurprisingly, the overall PCE was higher for all devices containing AuNPs mainly due to the higher FF value as compared to the control cells. To further investigate the enhancement in J SC , the IPCE for the PECs was measured (Fig. 4b). The addition of 8 nm AuNP resulted in a peak IPCE value of 73%, compared to 60% for the control PECs. This 13% improvement in peak IPCE value shows that when 8 nm AuNPs are incorporated into a TiO 2 NF scattering layer, there is a significant optoelectronic benefit, and this is reflected in the overall enhancement in the efficiency. Furthermore, there is a slight red shift in the peak maximum for the IPCE curve in the TiO 2 NP/AuNP NF PEC. The slight red shift in the IPCE is due to the enhanced light absorption of dye molecules and AuNPs assist in absorption towards the red portion of the solar spectrum. This has also been verified in previous studies where the presence of AuNPs in the scattering layer have been found to enhance the solar radiation absorption in the higher wavelength range [39].
The average photovoltaic parameters (J SC , V OC , FF, and PCE) and their standard error are depicted in Fig. 5. The average photovoltaic parameters follow the same trend as all the highest performing devices summarized in Table 1, which confirms the high reproducibility of the PECs. The photovoltaic parameters were plotted between the corresponding minimum and maximum values in order to get a better representation of improvement (Fig. 5). For the highest performing TiO 2 NP/AuNP-TiO 2 NF photoelectrodes containing the 8 nm AuNPs, the biggest improvement can be seen from the increase in J SC and FF. Interestingly, the V OC stayed relatively consistent with the somewhat larger standard error being attributed to the volatile and temperature sensitive liquid-based electrolyte. Due to the significant increase in J SC and FF, the average PCE for the control devices was improved by 20% upon addition of 0.5 wt% 8 nm AuNPs. To gain further understanding into the performance enhancement, the average resistance measurements for all devices were compiled into Table S1 in the SI. The series resistance (R S ) for the 5, 8, and 10 nm containing TiO 2 NP/AuNP NF devices was lower than that for the control devices. The lowest R S was found in the TiO 2 NP/AuNP NF device containing the 8 nm AuNPs, which supports it being the highest performing. Typically, lowering the R S results in an increase in FF and while this is consistent for the TiO 2 NP/AuNP NFs devices containing the 5, 8, and 10 nm AuNPs, the 12 nm AuNP devices show a considerable increase in R S . Interestingly, the J SC for this device was lower than all other devices and this can therefore be attributed to the significant increase in R S . It is well known that small AuNPs can have catalytic properties and the lowering of the R S may indicate that the small particles are effective at catalyzing the oxidation of the I À at the photoanode to regenerate the dye molecule. The catalytic effects for AuNPs are very size dependent so we postulate that the catalytic effect of the 12 nm particles is much reduced. No trends were found in the shunt resistance (R SH ) where the R SH remained within an order of magnitude of each other for all devices. Significant power losses due to low R SH values are attributed to fabrication defects and cause a reduction in V OC due to the possibility of an alternative current path. Given the V OC for all devices remained consistent, supports the hypothesis that the increase in PCE can be partly attributed to a decrease in R S for the TiO 2 NP/AuNP NF devices containing 8 nm AuNPs. Furthermore, dye desorption tests were undertaken and all photoelectrodes were found to have similar dye loading (Table S1). This also agrees with previous work where both the size of AuNPs and the concentration has shown little effect on dye adsorption [19,20].
Furthermore, we performed electrochemical impedance spectroscopy (EIS) to provide a better understanding about the dynamics of electron transport, and reveal the difference in the interfacial characteristics (such as recombination) of the 8 nm AuNPs in the PECs [40]. Fig. 6 shows the Nyquist plots for the TiO 2 NP/TiO 2 NF (control) and TiO 2 /AuNP-TiO 2 NF (8 nm AuNPs) measured under dark conditions, ranging from 0.01 Hz to 1000 kHz at an applied bias of À0.700 V. The inset of Fig. 6 represents the equivalent circuit which was adopted to fit the Nyquist spectra. The results reveal two distinguishable semicircles, where the small semicircle is related to the electrochemical reaction at the counter electrode (R ct1 ) and the large semicircle is related to the TiO 2 photoelectrode/electrolyte interface (R ct2 ) [40e42]. As depicted in the equivalent circuit, R S represents the series ohmic resistance which exists in the external circuit, where C m is the constant phase element representing the capacitor components in the solar cell. Given the same counter electrode and electrolyte were used in both PECs, there is no apparent difference in R ct1 and this can be seen by the almost identical diameters of the first small semicircle for both photoelectrodes (Fig. 6). Interestingly, the electron transport resistance (R ct2 ) dramatically decreases from 121.8 U for the control photoelectrode to 79.0 U for the photoelectrode containing the 8 nm AuNPs ( Table 2). The decrease in resistance implies improved and faster electron transport and injection rate for the photoelectrodes containing the 8 nm AuNPs. The electron lifetime (t e ¼ RC m ) [41,43] values were calculated to be 38.0 and 27.1 ms for the control and 8 nm AuNPs PECs, respectively ( Table 2). The shorter electron lifetime for the photoelectrode with the 8 nm AuNPs suggests this PEC is slightly more susceptible to recombination between the photoelectrons in the conduction band of the TiO 2 and the redox mediator (I 3 À ). This is reflected in the decrease in V OC which was seen for the 8 nm AuNP PECs compared to the control PECs (Fig. 5b) [41]. However due to the superior reduction in resistance which was evident under both light and dark conditions, and increase in J SC and FF, the photoelectrodes containing the 8 nm AuNPs had the highest PCE (Fig. 5d).
The benefits of adding sub-12 nm AuNPs to the scattering layer of TiO 2 NF PECs has clearly been shown where improvements can be seen across the entire spectral range. In terms of performance, the PECs are to the best of our knowledge one of the highest ever reported for scattering layers based on 100% pure anatase TiO 2 NFs. The current record for NF-based scattering layer PECs was achieved by Chuangchote et al. [44] where they obtained an impressive 8.14% with an active area of 0.25 cm À2 and 10.3% for a small active area of 0.05 cm À2 . However, their PECs were over 15 mm thick where only 1 mm was that of TiO 2 NFs. This ultrathin scattering layer suggests that some of the performance enhancement was also due to the large amount of high surface area NPs in the PEC. In another report, Naphade et al. [20] fabricated AuNP containing TiO 2 NF DSSCs and achieved a PCE of 7.8%, improving their control NF PECs by 15%, which was similar to the improvement margin demonstrated in this work. However, they did not control the particle size of their AuNPs. In contrast to existing reports, we provide a compressive performance evaluation for precisely size controlled sub-12 nm AuNPs in TiO 2 NF photoelectrodes and determine the upper and lower size limits of incorporating small AuNPs into assembled PECs. Interestingly, the dominating roles of each different sized AuNP was clearly different. For the 5 and 8 nm AuNP based PECs, benefits were seen from increased light harvesting and improved charge transport which resulted in the increased J SC , FF and decrease in the series resistance. On the other hand, the dominating role of the 10 and 12 nm AuNPs was less likely to be due to the light harvesting due to a decrease in the J SC . Despite this, the consistent increase in FF for all AuNP based PECs suggests a reduction in parasitic resistive losses. While the overall PCE was consistently higher for all AuNP based PECs, the highest performing cells were those that contained the 8 nm AuNPs. Therefore, the dominant role of the 8 nm AuNPs is the improvement in the light harvesting from the plasmonic features, decrease in series resistance and increase in electron transport and injection rate.

Conclusion
In this paper, we have demonstrated the benefits of adding sub-12 nm AuNPs into the scattering layer of TiO 2 NF photoelectrodes. The composite photoelectrodes were assembled into PECs and the gold loaded photoelectrodes show improvements in the current density and the fill factor which is attributed to increased spectral absorption, decrease in series resistance, and increase in electron transport and injection rate. Compared to the control PECs, devices loaded with 8 nm AuNPs show 20% improvement in average PCE, where the highest performing device obtained an PCE of 8% which is among the highest reported for TiO 2 NF PECs. We show that the precise control over NP size is critical to performance enhancement in TiO 2 NF PECs and we believe that this knowledge will help shape future studies into work incorporating plasmonic NPs into renewable energy applications.   Electron microscopy was completed by T.J.M, D.Y, and C.C. The manuscript was written by T.J.M with contributions of all authors. All authors have given approval to the final version of the manuscript.