Reduced electron recombination of dye-sensitized solar cells based on TiO2 spheres consisting of ultrathin nanosheets with [001] facet exposed

An anatase TiO2 material with hierarchically structured spheres consisting of ultrathin nanosheets with 100% of the [001] facet exposed was employed to fabricate dye-sensitized solar cells (DSCs). Investigation of the electron transport and back reaction of the DSCs by electrochemical impedance spectroscopy showed that the spheres had a threefold lower electron recombination rate compared to the conventional TiO2 nanoparticles. In contrast, the effective electron diffusion coefficient, Dn, was not sensitive to the variation of the TiO2 morphology. The TiO2 spheres showed the same Dn as that of the nanoparticles. The influence of TiCl4 post-treatment on the conduction band of the TiO2 spheres and on the kinetics of electron transport and back reactions was also investigated. It was found that the TiCl4 post-treatment caused a downward shift of the TiO2 conduction band edge by 30 meV. Meanwhile, a fourfold increase of the effective electron lifetime of the DSC was also observed after TiCl4 treatment. The synergistic effect of the variation of the TiO2 conduction band and the electron recombination determined the open-circuit voltage of the DSC.


Introduction
In the past two decades, dye-sensitized solar cells (DSCs) have received substantial attention from both academic and industrial communities as one of the most promising low-cost, highefficiency third-generation photovoltaic devices [1,2]. A typical DSC consists of a dye-coated TiO 2 electrode, which is deposited on a fluorine-doped tin oxide (FTO) conductive-glass substrate, a I − /I 3 − redox-couple-based electrolyte and a platinum counter electrode. Upon illumination, a photon with high energy (higher than the energy difference between the HOMO and LUMO level of the dye molecule) excites an electron from the ground state of the dye molecule to its excited state. The electron is then injected to the conduction band of the adjacent TiO 2 material, owing to a favorable alignment of the energetics. The electron goes through a series of trapping/detrapping process in the TiO 2 film before reaching the current collector, which is based on the conductive fluorine-doped tin oxide (FTO) substrate. Meanwhile, a parallel reaction, which involves transfer of the hole from the oxidized state of the dye (dye + ) to the surrounding I − ions of the redox couple of the electrolyte, occurs to regenerate the dye molecule, resulting in the formation of I 3 − ions. The electrical circuit is completed through transfer of the electron, which arrives at the Pt counter electrode through the external circuit, to the I 3 − ions of the electrolyte.
Apparently, the operation of a DSC depends on several reactions that occur at the interface between different materials [3].
In particular, the process of electron injection at the TiO 2 /dye interface and the electron recombination reaction at the TiO 2 / dye/electrolyte interface are critical because they control both the short-circuit current and open-circuit voltage of the DSC. The surface properties of the TiO 2 material play an important role in both processes. The process of electron injection in DSCs is controlled by the energy difference between the conduction band of the TiO 2 material and the LUMO level of the dye, and the process of electron recombination is mainly dominated by the interaction between the electron at the surface of TiO 2 and I 3 − ions in the electrolyte. Generally, the TiO 2 used in DSCs is based on the anatase phase with the [101] facet exposed, due to the robust stability of this surface compared to other crystal facets [4]. It has been reported that the average surface energies of the different facets of anatase [001] facet in these key processes of electron transport and recombination of DSCs is of great importance for both practical applications and basic research.
In this work, anatase TiO 2 spheres with a hierarchical structure consisting of ultrathin nanosheets with 100% of the [001] facet exposed were synthesized and applied in dye-sensitized solar cells (DSCs). The photovoltaic performance of the DSCs with different concentrations of the hierarchically structured TiO 2 spheres was evaluated. The kinetics of electron transport and back reaction of the DSCs with the spheres were investigated by electrochemical impedance spectroscopy. In addition, the effect of treatment by an aqueous solution of TiCl 4 on the performance of the DSCs with the TiO 2 spheres was discussed.

Synthesis of TiO 2 nanosheet particles
Hierarchically structured TiO 2 spheres of the nanosheets were synthesized by following the method originally reported by Chen et al. [6]. Briefly, a precursor solution containing titanium isopropoxide (Sigma-Aldrich; 1.15 mL) and diethylenetriamine (DETA; 0.02 mL) in 32 mL isopropanol was prepared by vigorous magnetic stirring of the mixture of the three components at room temperature. The precursor solution was then transferred to a Teflon-lined stainless steel autoclave (45 mL volume, Parr Instrument Co.) for the hydrothermal reaction. The hydrothermal process was carried out at 200 °C for 24 h in an electric oven. After that, the autoclave was allowed to cool to room temperature naturally. The as-collected white powder was washed with deionized water and then ethanol several times to remove the organic residues. The powder was then dried at 80 °C for 5 h and finally sintered at 400 °C for 3 h to improve the crystallinity.

Assembly of dye-sensitized solar cells
The procedure for the fabrication of the dye-sensitized solar cells was reported in our previous work [11,12]. Briefly, a sub-strate based on fluorine-doped tin oxide (FTO) conductive glass (TEC15, Pilkington) was thoroughly washed with detergent water, distilled water, acetone, isopropanol and ethanol in sequence under sonication for 15 min. The cleaned FTO substrate was first coated with a compact layer of TiO 2 film by spray pyrolysis to reduce the electron back reaction at the interface between the bare FTO and the electrolyte. The substrate was then deposited with the as-prepared TiO 2 paste or the commercial paste by a doctor-blading method using adhesive tape as a spacer to control the thickness of the film.

Characterization
The morphology and the crystal structure of the as-prepared TiO 2 powder were investigated by scanning electron microscope (SEM, FEI Quanta 200) and powder X-ray diffraction (XRD, PANanalytical Xpert Pro), respectively. Transmission electron microscopy (TEM, Philips CM 200) was used to monitor the detailed structure of the TiO 2 powder. The thickness of the TiO 2 films for the DSCs was determined by a profilometer (Dektak 150). The photocurrent density-voltage (J-V) characteristics of the DSCs were obtained by using a Xe lamp (150 W) based solar simulator (Newport), by recording the current produced by the cells as a function of the applied bias under AM1.5 illumination (100 mW/cm 2 ) with a computercontrolled digital source meter (Keithley 2420). The illumination intensity of the incident light from the solar simulator was measured with a silicon photodiode, which was calibrated with an optical meter (1918-C, Oriel). Aluminum foil with a size comparable to the active area of the TiO 2 film was used as a reflector on the counter electrode side of the DSCs during the J-V measurement.
The electrochemical impedance spectroscopy (EIS) of the DSCs was measured in the frequency range of 50,000-0.1 Hz at room temperature by a Versa-stat 3 electrochemical workstation (Princeton Applied Research). The EIS measurement was carried out under illumination, which was provided by a light emitting diode (LED, 627 nm) at open-circuit. The intensity of the incident illumination on the front side of the DSC (TiO 2 side) was adjusted by using a combination of neutral density filters. The EIS spectrum was analyzed with a Zview software, by using a transmission-line-based equivalent circuit to obtain the information of chemical capacitance, electron-recombination resistance and electron-transport resistance of the DSCs [12,13]. Figure 1a shows the image of the as-prepared TiO 2 powder by SEM. The material consists of microsized particles with spherical shape. The surface of the sphere is very rough and appears fluffy. The diameter of the sphere is around 1.6 μm as determined by TEM (Figure 1b). TEM images (Figure 1b, Figure 1c) also illustrate that the sphere has a substructure, which consists of ultrathin nanosheets packed together. It is speculated that the sphere is formed through self-assembly of the nanosheets to realize a minimum surface energy. Some spheres have pits on the surface, which may be due to the insufficient reaction duration. The measurement of the N 2 adsorption/desorption isotherms of the TiO 2 powder shows that the specific surface area of the TiO 2 spheres is 82 m 2 /g, which is slightly higher than the specific surface area of the film made from the commercial TiO 2 paste (DSL-18NR, Dyesol. Surface area: 72.9 m 2 /g) [14]. The large surface area of the material suggests that the nanosheets are probably loosely packed such that a greater surface area is exposed. The XRD pattern of the material (Figure 1d) shows that the as-prepared TiO 2 powder is anatase with a tetragonal structure and space group I4   The SEM image of the TiO 2 film consisting of the spheres is shown in Figure 1e and Figure 1f. Apparently, the TiO 2 particles are connected to each other in the film. Figure 1f shows that the film contains a large number of small pores. However, the sphere of the TiO 2 particles is rarely seen in the film. This indicates that the mechanical force of grinding and sonication employed in the preparation of the film broke up the spheres into small particles, probably in the form of nanosheets. Never-

J-V characteristics of the DSCs
The J-V characteristics of the DSCs with the TiO 2 film made from paste A, which contained 13 wt % TiO 2 spheres with and without TiCl 4 post-treatment, is shown in Figure 2a. The DSC solely based on paste A without TiCl 4 treatment (curve A) produced a short-circuit current density (J sc ) of 8.79 mA/cm 2 and open-circuit voltage (V oc ) of 0.76 V. In contrast, when the TiO 2 film was subjected to TiCl 4 -solution treatment, the J sc of the DSC (curve B) increased to 12.1 mA/cm 2 , which is 37.5% higher than that of curve A. Compared to curve A, it was found that the ratio of the J sc of curve C to that of curve A ((15.6 mA/cm 2 )/(12.1 mA/cm 2 ) = 1.77) is very close to the ratio of the concentration of the TiO 2 spheres in the two pastes ((TiO 2 wt % in paste B)/(TiO 2 wt % in paste A) = 25/13 = 1.92). This suggests that the higher J sc of the DSC made from paste B is due to the availability of more TiO 2 particles in the film, which can absorb more dye molecules, leading to a stronger light absorption. The J sc of the DSC made from paste B was further increased from 15.6 mA/cm 2 (Figure 2b, curve C) to 18.2 mA/cm 2 when the TiO 2 film was processed with TiCl 4 solution (Figure 2b, curve D). Meanwhile, the V oc of curve D was 20 mV higher than that of curve C, suggesting a beneficial effect of the TiCl 4 post-treatment on V oc as well. The best performance was obtained in the case of curve D with power conversion efficiency = 7.57% (Figure 2b), which is comparable to the efficiency (η = 7.52%) of the DSCs made from the commercial paste (I-V curve is not shown). The detailed characteristic parameters of the performance of the DSCs with different TiO 2 pastes are shown in Table 1.

Electrochemical impedance spectroscopy
Information on the charge-transfer and charge-transport process in DSCs can be measured by small-perturbation-based transient methods, such as electrochemical impedance spectroscopy (EIS) or intensity modulated photocurrent spectroscopy (IMPS) and intensity modulated photovoltage spectroscopy (IMVS) [13,15]. Compared to IMPS and IMVS, the advantage of the EIS method for characterization of DSCs lies in the fact that both the effective electron lifetime, τ n , and the effective electron diffusion coefficient, D n , can be obtained in one measurement. This is achieved by fitting the EIS spectrum using a suitable equivalent circuit that mimics the physical process in the device. The equivalent circuit that depicts the process of electron trapping/detrapping in DSCs is shown in Figure 3a. It contains a series resistance, R s , a capacitance at the Pt electrode/ electrolyte interface, C Pt , and a resistance for the chargetransfer process between electrons at the Pt electrode and I 3 − ions of the electrolyte, R Pt . Z w is the Warburg resistance arising from the ion transport in the electrolyte and Z tl is a distribution line describing the electron transport and recombination in the mesoporous TiO 2 film [13,16]. A typical EIS spectrum of a DSC is shown in Figure 3b for the Nyquist plot and Figure 3c for the Bode plot. The corresponding fitting results (green line) using the equivalent circuit are also shown in Figure 3b and Figure 3c. The distorted semicircle in the high frequency range (above 10 Hz) is ascribed to the electron transfer process at the interface of Pt counter electrode/electrolyte combined with the electron-transport process in the TiO 2 film (the semicircle corresponding to the electron transport process in TiO 2 is buried in the semicircle of the charge-transfer process at the Pt/electrolyte interface in the spectrum) [3]. The large semicircle in the lower frequency range (10-0.1 Hz) is due to the electron recombination process in the TiO 2 film. Under a high incident illumination intensity, the density of the photogenerated electron in the TiO 2 film is very high (up to 10 18 /cm 3 ) and the TiO 2 film becomes conductive [17]. In this case, the resistance corresponding to the electron-transport process becomes too small to be observed in the EIS spectrum. Consequently, the EIS spectrum is mainly dominated by the electron recombination process. Nevertheless, under low incident illumination intensity, the conductivity of the TiO 2 film is very low due to a low density of photogenerated electrons. In this case, the main feature of the EIS spectrum is due to the transport of electrons in the TiO 2 film. Hence, an accurate fitting of the EIS spectrum of a DSC using the equivalent circuit is normally obtained in the illumination range in which both the electron-transport resistance and the electron-recombination resistance are substantial [13]. Only the results of good fits are shown in this work.

Comparison of electron transport and recombination of the DSC based on TiO 2 spheres and nanoparticles
The electron-recombination process in DSCs is reflected by the effective electron lifetime, τ n , whereas the electron-transport process is manifested by the effective electron diffusion coefficient D n . Bisquert et al. showed that both τ n and D n of a DSC are dependent on the distribution of the density of electrons in the conduction band (free electron) and in the trap states (trapped electron) of the TiO 2 film as well as the lifetime and diffusion coefficient of free electron (τ 0 and D 0 ), through the following consideration: and (2) where n t and n c are the densities of the trapped electron and free electron, respectively [18].
The charge distribution, g(E), in a mesoporous TiO 2 film is described by [18,19]: where n E F is the quasi Fermi level of TiO 2 , E c the conduction band of TiO 2 , E F,redox the potential energy of the redox couple, N t,0 the total density of the trapped electrons, k B is the Boltzmann constant and T 0 the characteristic temperature that reflects the profile of the charge distribution in TiO 2 .
Therefore, comparison of the change of τ n and D n in DSCs due to the different material composition should be made by using the density of charge as the reference, provided that the distribution profile of charge density is the same [20].
The density of charge in the TiO 2 film is reflected by the chemical capacitance, C μ , which is measured by EIS, through the relationship [18]: (4) where E v is the valence band of TiO 2 . Thus, we employ the density of chemical capacitance as a reference for the investigation of the variation of τ n and D n in the following. Figure 4a shows the τ n of the DSCs with the TiO 2 films consisting of the nanosheet-based spheres and the conventional nanoparticles as a function of the chemical capacitance density. It is found the τ n of the nanosheets based DSC is nearly threefold higher than that of the nanoparticles for a constant capacitance density. This suggests that the TiO 2 film with the spheres has a lower electron-recombination reaction rate compared to the film with the nanoparticles. Besides τ n, the effective electron diffusion coefficient, D n , is another important parameter that determines the performance of a DSC. The comparison of the D n of the cell based on the spheres and the nanoparticles is shown in Figure 4b. It is interesting that both materials show the same D n , suggesting that the electron transport is not affected by the morphology and the exposed crystal facet of the TiO 2 material. The identical D n also suggests that the diffusion coefficient of the free electron is the same for the two materials, according to Equation 2 [13]. It also justifies the assumption that the profile of the distribution of charge density is the same in the two types of TiO 2 film. In contrast, the different τ n suggests that the free-electron lifetime of the spheres is different to that of the nanoparticles. The high τ n of the spheres could be related to the properties of the [001] facet, but clarification of this issue requires further investigation. As a consequence, the electron diffusion length, L n , which depends on both the τ n and D n by , is up to 1.6-fold higher for the nanosheetbased TiO 2 spheres compared to that of the nanoparticles (Figure 4c). It is found that the L n of the DSC based on the nanoparticles is only around 16 μm (Figure 4c), which is comparable to the thickness of the TiO 2 film (13 μm). A previous study has shown that the L n of a DSC needs to be at least three times the thickness of the TiO 2 film in order to collect most of the photogenerated electrons [13]. Therefore, the short L n may limit the performance of the DSC. The higher L n of the spheres-based DSC should lead to a higher electron collection efficiency compared to its nanoparticles counterpart.
Besides J sc , V oc is another key performance parameter of a DSC. The maximum voltage of a DSC is determined by the potential difference between the conduction band of TiO 2 and the redox potential of I − /I 3 − in the electrolyte. Obviously, the position of the TiO 2 conduction band edge, E c , has a direct impact on the open-circuit voltage (V oc ) of the DSC. Thus, it is important to know the relative position of the E c of the nanosheet-based spheres relative to the nanoparticles in order to determine the reason for the different V oc . According to Equation 3, the change of E c of TiO 2 can be monitored by the variation of the voltage (V) of the DSC at a constant electron density.
As shown in Figure 4d, the E c of the nanosheet-based spheres is found to be 100 meV lower than that of the nanoparticles. The lower E c of the spheres is probably due to the different dye

Effect of TiCl 4 treatment
The strategy of treating TiO 2 mesoporous films with TiCl 4 aqueous solution has been extensively employed to improve the performance of DSCs. In most cases, it is found that the J sc of the DSC is enhanced, while the V oc is reduced after the TiCl 4 treatment of the film. O'Regan et al. found that TiCl 4 treatment caused 80 meV downward shift of the TiO 2 conduction band, resulting in an increased driving force for the electron-injection process. They reported that the enhanced J sc was owing to an improved electron-injection efficiency of the DSC [21,22]. In the following section, the influence of the TiCl 4 solution treatment on the E c of the TiO 2 -spheres-based film and on the kinetics of electron transport and back reaction of the corresponding DSCs is investigated. Figure 5a illustrates the chemical capacitance density of the DSCs made from paste B with and without TiCl 4 treatment, as a function of the voltage. It is found that, at a constant charge density, the voltage of the cell with TiCl 4 treatment is lower than that of the DSC without the treatment. The maximum difference in voltage between the cells is around 30 mV. Provided that the distribution profile of the charge density is the same for the TiO 2 film with and without TiCl 4 treatment, the reduced potential of the DSC with TiCl 4 treatment means that the TiCl 4 treatment caused a downward shift of the TiO 2 conduction band by 30 meV, which may decrease the maximum voltage the DSC can achieve. This observation is in good agreement with the results reported by O'Regan et al. [21]. However, the V oc of the cell with TiCl 4 treatment is actually 20 mV higher than the DSC without the treatment, as shown in Figure 2b. This indicates that the electron recombination of the DSC is probably affected by the TiCl 4 treatment. Figure 5b shows the τ n as a function of capacitance density of the DSC with and without TiCl 4 treatment. It is found that τ n is enhanced by a factor of 3.8 after TiCl 4 treatment. In contrast, D n of the DSCs is relatively unchanged with the TiCl 4 treatment (Figure 5c). Owing to the enhanced τ n , the electron diffusion length, L n , of the DSC is enhanced by a factor of two through TiCl 4 treatment (Figure 5d). Hence, the improved voltage (20 mV) of the DSC (Figure 2b, curve D) with TiCl 4 treatment compared to the cell without TiCl 4 treatment (curve C in Figure 2b) should be a result of a synergistic effect of the decreased TiO 2 conduction band and the increased electron life-time. Apparently, the beneficial effect of the enhanced electron lifetime on V oc surpasses the negative effect of the downward shift of the E c of TiO 2 , leading to a higher V oc .

Conclusion
Dye-sensitized solar cells with a TiO 2 electrode made from hierarchically structured TiO 2 spheres, consisting of nanosheets with 100% of the [001] facet exposed, were assembled and characterized in terms of the device performance, the kinetics of electron transport and back reaction. It was found that the TiO 2spheres-based DSCs generated an energy conversion efficiency of 7.57%, which is comparable to the conventional TiO 2 nanoparticles. Investigation of the kinetics of electron transport and back reaction of the DSCs showed that the spheres had a threefold higher effective electron lifetime compared to the nanoparticles. However, the effective electron diffusion coefficient, D n , of the DSCs was not affected by the different morphology and exposed crystal facets of the TiO 2 material. Monitoring of the variation of the conduction band of the dyed TiO 2 film disclosed that the E c of the spheres-based TiO 2 electrode was 100 meV lower than that of the nanoparticles.
This work also investigated the influence of treatment with TiCl 4 aqueous solution on the E c of the TiO 2 spheres and on the τ n and D n of the corresponding DSCs. It was found that TiCl 4 treatment caused a downward shift (30 meV) of the TiO 2 conduction band and a fourfold increase of the τ n , whereas the D n of the cell was not significantly affected by the TiCl 4 treatment.