Photoinduced electron injection from an organic dye having a pyridyl anchor to Lewis acid site of TiO₂ surface

Abstract

An organic dye with a pyridyl group as an anchor unit (NI4) adsorbs exclusively at the Lewis acid sites of TiO₂ surface, although its adsorbability is very weak compared with that of a similar dye with a carboxyl group anchoring to Brønsted acid sites (NI2). In the present study, photovoltaic features of two similar dyes, NI2 and NI4, are thoroughly investigated by changing dye loadings on TiO₂ surface and the intensity of the illuminated light. A remarkable finding is that photocurrents and photovoltages observed under intense illumination are larger for NI4 than for NI2, suggesting that the adsorbability of dye onto a TiO₂ surface is not necessarily a principal factor leading to a high electron injection probability from dye to TiO₂. Plausible reasons for the difference in photovoltaic features between NI2- and NI4-DSSCs are discussed on the basis of electrochemical impedance spectroscopy and transient photovoltage measurements.

---

1. Introduction

Dye-sensitized solar cells (DSSCs) using nanocrystalline TiO₂ powder and photosensitizing dyes have been studied intensively due to their possible low cost of fabrication and reasonably high power conversion efficiencies.^1–6 A number of studies have been devoted to the design and synthesis of efficient photosensitizing dyes, and in most of the studies, electron-accepting carboxylic or cyanoacrylic acids have been exclusively employed as an anchor unit on the surface of nanocrystalline TiO₂ particles.^7–10 In contrast to this, we have found that a pyridyl group also acts as an anchoring and electron-withdrawing unit in photosensitizing dyes.^11–15 This finding is of great importance from the viewpoint that a possible use of a pyridyl anchor can relieve a restriction that has been imposed on the design and synthesis of organic photosensitizers for a long time. In our subsequent study, details of the adsorption behaviors of an organic dye having a pyridyl anchor (NI4) on a TiO₂ surface were carefully studied.¹⁶ Analysis of adsorption isotherms for dyes on TiO₂ showed that the adsorption equilibrium constant of NI4 is two orders of magnitude smaller than that for a similar dye having a carboxyl anchor (NI2). In addition, elaborate FT-IR studies with NI4 powder and NI4 adsorbed on TiO₂ revealed that a pyridyl dye, NI4, adsorbs on Lewis acid sites of TiO₂ in contrast to Brønsted acid sites for a carboxyl dye, NI2. Coadsorption experiments with a TEMPO radical (4CT: 4-carboxy-2,2,6,6-tetramethylpiperidine-1-oxyl) and either NI2 or NI4, coupled with the measurements of average nearest-neighbor interspin distances of 4CT radicals coadsorbed on TiO₂ by a spin-probe ESR technique, supported this view.¹⁶ Currently, a variety of novel anchor units replacing carboxylic (cyanoacrylic) and pyridyl groups have been proposed: cyano-benzoic acid,¹⁷ phosphinic acid,¹⁸ 8-hydroxyquinoline,¹⁹ 2-(1,1-dicyanomethylene)rhodamine,²⁰ hydroxyl,²¹ TCNE and TCNQ,²² pyridine,²³ benzothienopyridine,²⁴ and triazine.²⁵

It is now of great interest from an academic as well as a practical viewpoint to clarify a relationship between the type of anchor unit in a photosensitizing dye and the electron injection probability from the dye to TiO₂. Our preliminary study has demonstrated that the photocurrents of NI4-DSSCs are greater than those of NI2-DSSCs when comparisons are made at the same dye loadings on TiO₂ surfaces,¹² although the adsorbability of NI4 on TiO₂ is considerably weaker than that of NI2. A similar trend was found for other pyridyl and carboxyl dyes.¹³ To date, however, detailed studies have not been performed to clarify the reasons for this observation, which conflicts with a traditional view that good electron communication between dye and TiO₂ leads to a high electron injection probability. In the present study, photovoltaic properties of DSSCs based on two similar dyes, NI2 and NI4, are studied carefully by changing dye loadings and the intensity of illuminated light, together with the use of electrochemical impedance spectroscopy and a transient photovoltage technique.

2. Experimental

2.1. Chemicals

Two types of organic dyes, NI2 and NI4, were synthesized as described earlier¹² and their chemical structures are illustrated in Fig. 1. Tetrahydrofuran (THF, Kanto Chemical Co., Inc.) was purified by distillation and used as a solvent for dye adsorption. Acetonitrile (MeCN, Tokyo Kasei) was used as a solvent for DSSCs and was refluxed over P₂O₅ for a couple of hours under a N₂ atmosphere. 1,2-Dimethyl-3-n-propylimidazolium iodide (DMPrII) was obtained from Shikoku Kasei and was used as received. 4-tert-Butylpyridine (TBP) was obtained from Nacalai Tesque. Other chemicals were used without further purification. TiO₂ paste (PST-18NR) was purchased from JGC Catalysts and Chemicals Ltd. Fluorine-doped tin oxide (FTO, 13 Ω □⁻¹) was obtained from Nippon Sheet Glass Co. Ltd.

Fig. 1 Chemical structures of NI2 and NI4.

2.2. Fabrication of DSSCs and evaluation of dye loadings

In most experiments, two identical dye-adsorbed TiO₂ electrodes were prepared under the same experimental conditions: one was used for photovoltaic measurements and the other for the evaluation of dye loading. TiO₂ paste was deposited on an FTO substrate by doctor-blading. The deposited TiO₂ paste was sintered according to a temperature program (from room temperature to 450 °C for 50 min and at 450 °C for 35 min). After cooling the TiO₂ electrode to 45 °C, the sintered electrode was immersed in THF solutions containing various concentrations of NI2 or NI4 to control the NI2 or NI4 loading on the TiO₂ surface. Unless otherwise stated, the adsorption of NI2 and NI4 on TiO₂ was made in THF solutions of 0.2 mM NI2 and 2 mM NI4, where dye loadings for NI2 and NI4 are similar to each other. In all the cases, the dye adsorption was performed for 18 hours in an incubator maintained at 25 °C. DSSCs were fabricated using one of the two as-prepared dye-adsorbed TiO₂ electrodes, Pt-coated glass as a counter electrode, and a redox solution consisting of 0.05 M I₂, 0.1 M LiI, and 0.6 M DMPrII in MeCN. On the other hand, the other dye-adsorbed TiO₂ electrode was cautiously soaked in pure THF for several seconds and dye loading was determined by obtaining absorption spectra of a mixed solution (THF : DMSO : H₂O (1 M NaOH) = 5 : 4 : 1, v/v/v), in which dye molecules deposited on TiO₂ were desorbed for three hours. Time courses of absorption spectra of NI2 and NI4 in the mixed solutions are shown in Fig. S1† to demonstrate that both dyes are stable in the solutions. The thickness of the TiO₂ film (0.5 × 0.5 cm² in the photoactive area) was evaluated to be about 9 μm using a 3D-laser scanning microscope (Keyence VK-9700). A specific surface area of the sintered TiO₂ electrode needed for the evaluation of dye loadings was determined by N₂ adsorption experiments to be 83 m² g⁻¹.

2.3. Photovoltaic measurements

Photocurrent–voltage characteristics of DSSCs were measured with a potentiostat (Hokuto Denko HA-501G) under the irradiation of AM 1.5, 100 mW cm⁻² supplied by a solar simulator (Asahi Spectra HAL-302). Incident photon-to-current conversion efficiency (IPCE) spectra were obtained under monochromatic irradiation from a tungsten–halogen lamp (150 W) and a monochromator with intensities of monochromatic lights at different wavelengths measured in advance using a thermopile (Eppley Type E6). The dependence of short-circuit photocurrent (J_sc) on light intensity (P) was carefully measured, where a monochromatic light of 410 nm obtained from an interference filter (Toshiba Glass KL-41), a bandpass filter (Toshiba Glass B390), and a 500 W Xe lamp was used as a light source and the intensity was attenuated using neutral density filters. Intensities of the attenuated 410 nm light were measured using a Si photodiode (Hamamatsu Photonics S1226-5BQ).

2.4. Spectroscopic measurements

Absorption and fluorescence spectra of NI2 and NI4 in solutions were recorded using a spectrophotometer (Shimadzu UV-3150) and a spectrofluorometer (Hitachi F-4500), respectively.

2.5. Cyclic voltammetry

Cyclic voltammograms (CVs) were recorded in a MeCN/TBAP (0.1 M) solution using an air-tight electrolysis cell with a three-electrode system consisting of a Pt sphere as a working electrode, Ag/Ag⁺ as a reference electrode, and a Pt wire as a counter electrode. CVs were recorded with a potentiostat equipped with a functional generator (Hokuto Denko HAB-151). Before electrochemical measurements were made, Ar gas was purged into the electrolyte solution to remove oxygen. Electrode potentials were referred to ferrocene/ferrocenium (Fc/Fc⁺) redox couples to evaluate HOMO and LUMO energy levels. All the electrochemical measurements were carried out at room temperature.

2.6. Electrochemical impedance spectroscopy (EIS)

EIS measurements of DSSCs were performed with an impedance analyzer (VersaSTAT 3, PAR). To measure the impedance, a perturbation amplitude of 10 mV within the frequency range from 10 mHz to 100 kHz was applied in the dark. The observed impedance spectra were analyzed with a software (VersaStudio) to interpret the characteristics of the DSSCs.

2.7. Transient photovoltage technique

A light pulse of 0.1 ms duration generated by a light emitting diode (396 nm) was employed in transient photovoltage measurements in the DSSCs. The light pulse was incident on the FTO side of the DSSC and its intensity was controlled to keep the change of voltage within 2 mV. White bias light supplied by a halogen lamp (150 W) was also illuminated from the FTO side. The transient photovoltage decay was recorded on an oscilloscope (Iwatsu DS-4372) with an RC circuit (τ = 2.2 s) through a voltage amplifier (NF 5307).

3. Results and discussion

Spectroscopic and adsorption properties of NI2 and NI4 obtained earlier are summarized in Table 1 for reference.^12,13,16 Reflecting a similarity of chemical structures, except for the anchor units between the two dye molecules, absorption and fluorescence wavelengths of the two dyes in 1,4-dioxane are similar to each other. Likewise, CV measurements of the dyes in MeCN showed that both NI2 and NI4 exhibit a reversible one-electron transfer process and their redox potentials are close to each other: 0.48 and 0.50 V vs. Ag/Ag⁺ for NI2 and NI4, respectively. Details of the CV measurements are given in the ESI (Fig. S2–S4†). Fig. 2 illustrates the energy level diagram for NI2- and NI4-DSSCs, evaluated with the spectroscopic and electrochemical data shown in Table 1. A small difference in the LUMO energy level is seen between NI2 and NI4. However, the LUMO energy difference of 0.11 eV may not affect the electron injection process in NI2- and NI4-DSSCs because both the LUMO levels are located sufficiently above the E_c of TiO₂. Likewise, HOMO energy levels of the two dyes are close to each other and well below the E_f of the redox couple. These suggest that the combination of NI2 and NI4 used in this study can be an almost ideal example for discussing the difference between the anchor units. On the other hand, there is a marked difference in adsorbability between the dyes: 0.6 × 10⁵ M⁻¹ for NI2 adsorbed on Brønsted acid sites of TiO₂ and 1.1 × 10³ M⁻¹ for NI4 adsorbed on Lewis acid sites.¹⁶

Table 1 Spectroscopic properties of NI2 and NI4, HOMO and LUMO energy levels, and adsorbability of NI2 and NI4 on TiO₂

Dye	λ^absmax^a/nm	Emax^a/M^−1 cm^−1	λ^f1max^a/nm	φ^a^,^b	E^ox1/2^c/V	HOMO^d/V	LUMO^d/V	K^e/M^−1
a In 1,4-dioxane.b Fluorescence quantum yield determined by a calibrated integrating sphere system (Hamamatsu Photonics).c Half-wave potential vs. Ag/Ag^+ for oxidation in MeCN/TBAP (0.1 M).d vs. normal hydrogen electrode (NHE).e Equilibrium constant for adsorption of the dyes on TiO2.
NI2	376	34 300	442	0.87	0.48	1.00	−1.86	0.6 × 10^5
NI4	375	33 000	423	0.84	0.50	1.02	−1.97	1.1 × 10^3

---

Fig. 2 Energy level diagram of DSSCs based on NI2 and NI4.

Fig. 3 depicts typical IPCE spectra and J–V curves of DSSCs based on NI2 and NI4, where the dye loadings on TiO₂ (C_ad) were 0.99 × 10¹³ cm⁻² for NI2 and 0.92 × 10¹³ cm⁻² for NI4. IPCE spectra measured with the same DSSCs as those used in Fig. 3b are similar in shape and the maximum IPCE value at ca. 400 nm is greater for NI2 than for NI4. It is interesting to note that in Fig. 3b, the J_sc for NI4 is higher than that for NI2, which is in agreement with our previous observation,^12,13 although a larger J_sc value should be expected for NI2 because IPCE values of NI2 are greater than those obtained for NI4 over the whole wavelength range. Another salient feature to be noted here is a difference in open-circuit photovoltage (V_oc) between NI4 (560 mV) and NI2 (500 mV), irrespective of there being no clear difference in J–V curves measured in the dark. This photovoltaic feature will be discussed later in detail.

Fig. 3 (a) IPCE spectra and (b) current density (J)–voltage (V) curves of DSSCs based on NI2 and NI4. Solid J–V curves in Fig. 3b are measured under illumination, while broken curves are obtained in the dark.

To confirm the smaller IPCE but greater J_sc values for NI4 than those for NI2, detailed photovoltaic measurements for NI2- and NI4-DSSCs were conducted with C_ad as a parameter. Fig. 4 summarizes changes of IPCE at 400 nm, J_sc, V_oc, and energy conversion efficiency (η) with a change of C_ad. As seen in Fig. 4a, IPCE values for both NI2 and NI4 increase gradually with the increase of C_ad and tend to level off. Over the entire C_ad range studied, IPCE values for NI2 are greater than those for NI4, which is in accordance with the IPCE spectra of Fig. 3a. Fig. 4b depicts plots of J_sc against C_ad for NI2 and NI4. Similar to the plots of IPCE shown in Fig. 4a, J_sc values for both dyes increase with C_ad and tend to saturate. Importantly, however, J_sc values for NI4 are obviously greater than those for NI2. This trend in J_sc is also consistent with the J–V curves shown in Fig. 3b under illumination. We also note that J_sc values reported earlier¹³ (filled circles in red and blue) do not deviate considerably from the experimental points obtained in this study. As shown in Fig. 4c, V_oc values do not change much with C_ad and they are ca. 50 mV greater for NI4 than for NI2 in the high C_ad region. Moreover, the values of η calculated with the data of J_sc, V_oc, and fill factor (ff, not shown here as there was no obvious difference between NI2 and NI4) as a function of C_ad are illustrated in Fig. 4d, which show higher photovoltaic performances for NI4 weakly anchoring to Lewis acid sites of TiO₂ than for NI2 bonding strongly to Brønsted acid sites.

Fig. 4 (a) Plots of IPCE at 400 nm and (b) J_sc, (c) V_oc, and (d) η against C_ad for DSSCs based on NI2 and NI4. Filled circles in blue and red in (b) denote J_sc values for NI2 and NI4, respectively, cited from Fig. 7 of ref. 13.²⁶

We will now discuss some intriguing photovoltaic features of DSSCs based on NI2 and NI4. At first, to make the problem clearer, J_sc values are plotted against IPCE values at 400 nm, where the J_sc and IPCE data in Fig. 4a and b are used for making the plots. It is evident from Fig. 5 that the slope of the plot for NI4 is larger than that for NI2. This demonstrates that the photocurrent generation under intense illumination (J_sc, white light of 100 mW cm⁻²) relative to that under weak monochromatic illumination (IPCE, 400 nm light of ca. 1 μW cm⁻²) is efficient for NI4 compared with that for NI2. The dependencies of J_sc on the intensity of 410 nm light (P) for NI2- and NI4-DSSCs were examined over a wide range of P and the results are depicted in Fig. 6 as double logarithmic plots. Herein, J_sc values observed at the lowest intensity of 0.8 μW cm⁻² are set to be equal. Both plots appear to fit straight lines with light exponents (γ in J_sc ∝ P^γ) almost close to unity over a change in P ranging from ca. 10⁻³ to 10 mW cm⁻². On taking a closer look at the high intensity regime, shown in the inset of Fig. 6, however, we see that the data points for NI4 lie slightly above those for NI2. This indicates that the recombination loss in photocurrent generation is less enhanced by increasing the light intensity in the case of NI4 anchoring to Lewis acid sites than in the case of NI2 anchoring to Brønsted acid sites.

Fig. 5 Correlation between J_sc and IPCE at 400 nm for DSSCs based on NI2 and NI4.

Fig. 6 Double-logarithmic plots of J_sc against P for DSSCs based on NI2 and NI4, where illumination is made with a monochromatic light of 410 nm and J_sc values at P = 0.8 μW cm⁻² for NI2- and NI4-DSSCs are equated. Inset denotes a magnified view of the main figure in a high light-intensity regime.

Second interesting factor in the photovoltaic features of NI2- and NI4-DSSCs is the difference in V_oc between NI2 and NI4, as illustrated in Fig. 3b and 4c: V_oc values for NI4 are greater than those for NI2 when compared at the same C_ad values. It is well known that the addition of TBP to the electrolyte solution leads to an increase in the V_oc of the DSSC, which is sometimes followed by a considerable decrease in J_sc.^27–29 The reason for the effect of additive TBP on the photovoltaic properties of DSSCs has been discussed several times to date.^30–32 Currently, the influence of TBP adsorbed on TiO₂ is accounted for in terms of an upward shift of the conduction band of TiO₂ due to the formation of a surface dipole layer or a suppression of dark currents by blocking the approach of I₃⁻ to the TiO₂ surface. TBP, having a pyridyl group, may adsorb on Lewis acid sites of TiO₂ like NI4. The influence of TBP is seemingly related to the larger V_oc value for NI4 than that for NI2. However, it is worth noting that the J–V curves in the dark, as shown in Fig. 3b, do not differ considerably between NI2- and NI4-DSSCs. In addition, the adsorption of NI2 or NI4 on the TiO₂ surface did not affect the J–V curves observed with the bare TiO₂ electrode, although the addition of TBP to the electrolyte shifted the J–V curve by ca. 100 mV in a negative direction, as shown in Fig. 7. To further investigate properties of the TiO₂ surfaces, EIS measurements of the DSSCs were carried out and the spectra were simulated, as shown typically in Fig. 8, using a simple equivalent circuit shown in the inset of Fig. 9.³³ Here, R_s represents a series resistance due to the sheet resistance of the FTO substrate and the electrical contact at FTO/TiO₂ interfaces, R₁ is a charge transfer resistance at the Pt counter electrode, and R₂ denotes a charge-transfer resistance related to the recombination of electrons at the TiO₂/electrolyte interface. CPE1 and CPE2 are constant phase elements corresponding to R₁ and R₂, respectively. Fig. 9 summarizes R₂ values for the NI2- and NI4-DSSCs measured at various bias voltages in the dark. The R₂ values for the two DSSCs decrease with the increase of the bias voltage with a negligible difference between them except at a bias voltage close to a flat-band potential of the TiO₂ electrodes in contact with the I⁻/I₃⁻ redox solution. These findings demonstrate that the reason for NI4 to exhibit larger V_oc values than NI2 is clearly different from the one for the enhancement of V_oc due to the addition of TBP. Moreover, we note that near the flat-band potential of TiO₂, the back reaction of electrons injected from the LUMO level of the dye into the conduction band of TiO₂ with I₃⁻ in solution is less probable on the NI4-adsorbed TiO₂ electrode than on the NI2-adsorbed TiO₂, and this may be responsible for the difference in V_oc between NI2- and NI4-DSSCs. To examine this possibility, the transient photovoltage technique was employed to evaluate an electron lifetime in the conduction band (τ) with photovoltage as a parameter.^34,35 A typical photovoltage transient for the NI4-DSSC is depicted in Fig. 10a and the τ values obtained with the two dyes at different photovoltages are summarized in Fig. 10b. It is seen that the τ values for the NI2-DSSC are slightly greater than for the NI4-DSSC in the small photovoltage region, whereas they decrease with an increase of the photovoltage and the plots intercept at 250 mV. At higher photovoltages, the electron lifetimes for the NI4-DSSC are longer than for the NI2-DSSC, which is in accordance with our speculation that the back electron transfer for the NI4-DSSC is slow compared with that for the NI2-DSSC.

Fig. 7 J–V curves of bare TiO₂ electrode in the dark in MeCN/(0.05 M I₂, 0.1 M LiI, and 0.6 M DMPrII) (a) with and (b) without 0.25 M TBP.

Fig. 8 Nyquist plots and fitting curves of NI2 and NI4-DSSCs biased at (a) 300 mV, (b) 400 mV, (c) 500 mV, and (d) 600 mV.

Fig. 9 Charge recombination resistances (R₂) of NI2- and NI4-DSSCs measured in the dark with the EIS technique. Inset is an equivalent circuit for the simulation of DSSCs.

Fig. 10 (a) Typical photovoltage transient signal for the NI4-DSSC at a photovoltage of 250 mV and (b) electron lifetime (τ) of NI2- and NI4-DSSCs measured with the transient photovoltage technique.

More elaborate studies will be needed for an in-depth understanding of the mechanisms of charge-injection processes through Brønsted and Lewis acid sites. However, it is now quite clear that the adsorbability of a photosensitizing dye onto a TiO₂ surface is not necessarily a principal factor for a high electron injection probability from dye to TiO₂.

4. Conclusions

Photovoltaic features of DSSCs based on NI4 anchoring weakly to Lewis acid sites of a TiO₂ surface are compared with those based on NI2 anchoring strongly to Brønsted acid sites by changing dye loadings on the TiO₂ surface. In contrast to a conventional view, it is found that photocurrents observed under illumination of intense light are greater for NI4 than for NI2, although this is not the case under reduced illumination. The finding demonstrates that the adsorbability of a dye on a TiO₂ surface is not necessarily a major factor for attaining a high injection probability of electrons from the dye to the conduction band of TiO₂. Another feature to be noted for NI4-DSSCs is its photovoltage, which is greater than that of NI2-DSSCs. The reason for the difference in photovoltage is ascribed to longer electron lifetimes for NI4-DSSCs than for NI2-DSSCs under intense illumination, on the basis of transient photovoltage measurements and electrochemical impedance spectroscopy.

Acknowledgements

This study was supported in part by a Grant-in-Aid for Scientific Research (B) (No. 25288085) from the Ministry of Education, Science, Sports and Culture of Japan.

Notes and references

1. N. Robertson, Angew. Chem., Int. Ed., 2006, 45, 2338–2345.
2. Z. Ning and H. Tian, Chem. Commun., 2009, 483.
3. A. Mishra, M. K. R. Fischer and P. Bäuerle, Angew. Chem., Int. Ed., 2009, 48, 2474.
4. A. Hagfeldt, G. Boschloo, L. Sun and L. Kloo, Chem. Rev., 2010, 110, 6595.
5. M. Grätzel, R. A. J. Janssen, D. B. Mitzi and E. H. Sargent, Nature, 2012, 488, 304.
6. K. Kakiage, Y. Aoyama, T. Yano, T. Otsuka, T. Kyomen, M. Unno and M. Hanaya, Chem. Commun., 2014, 50, 6379.
7. B. O'Regan and M. Grätzel, Nature, 1991, 353, 737.
8. A. Yella, H.-W. Lee, H. N. Tsao, C. Yi, A. K. Chandiran, M. K. Nazeeruddin, E. W.-G. Diau, C.-Y. Yeh, S. M. Zakeeruddin and M. Grätzel, Science, 2011, 334, 629.
9. E. Miyazaki, T. Okanishi, Y. Suzuki, Y. N. Ishine, H. Mori, K. Takimiya and Y. Harima, Bull. Chem. Soc. Jpn., 2011, 84, 459.
10. A. S. Hart, B. K. C. Bikram, N. K. Subbaiyan, P. A. Karr and F. D'Souza, ACS Appl. Mater. Interfaces, 2012, 4, 5813.
11. Y. Ooyama, S. Inoue, R. Asada, G. Ito, K. Kushimoto, K. Komaguchi, I. Imae and Y. Harima, Eur. J. Org. Chem., 2010, 92.
12. Y. Ooyama, S. Inoue, T. Nagano, K. Kushimoto, J. Ohshita, I. Imae, K. Komaguchi and Y. Harima, Angew. Chem., 2011, 50, 7429.
13. Y. Ooyama, T. Nagano, S. Inoue, I. Imae, K. Komaguchi, J. Ohshita and Y. Harima, Chem.–Eur. J., 2011, 17, 14837.
14. Y. Ooyama, N. Yamaguchi, I. Imae, K. Komaguchi, J. Ohshita and Y. Harima, Chem. Commun., 2013, 49, 2548.
15. Y. Ooyama, Y. Hagiwara, Y. Mizumo, Y. Harima and J. Ohshita, New J. Chem., 2013, 37, 2479.
16. Y. Harima, T. Fujita, Y. Kano, I. Imae, K. Komaguchi, Y. Ooyama and J. Ohshita, J. Phys. Chem. C, 2013, 117, 16364.
17. M. Katono, T. Bessho, S. Meng, R. Humphry-Baker, G. Rothenberger, S. M. Zakeeruddin, E. Kaxiras and M. Grätzel, Langmuir, 2011, 27, 14248.
18. I. López-Duarte, M. Wang, R. Humphry-Baker, M. Ince, M. K. Martínez-Díaz, M. K. Nazeeruddin, T. Torres and M. Grätzel, Angew. Chem., Int. Ed., 2012, 51, 1895.
19. H. He, A. Gurung and L. Si, Chem. Commun., 2012, 48, 5910.
20. J. Mao, N. He, Z. Ning, Q. Zhang, F. Guo, L. Chen, W. Wu, J. Hua and H. Tian, Angew. Chem., Int. Ed., 2012, 51, 9873.
21. J. Zhao, X. Yang, M. Cheng, S. Li and L. Sun, ACS Appl. Mater. Interfaces, 2013, 5, 5227.
22. T. Michinobu, N. Satoh, J. Cai, Y. Lia and L. Han, J. Mater. Chem. C, 2014, 2, 3367.
23. Y. Ooyama, K. Uenaka, Y. Harima and J. Ohshita, RSC Adv., 2014, 4, 30225.
24. Y. Ooyama, T. Sato, Y. Harima and J. Ohshita, J. Mater. Chem. A, 2014, 2, 3293.
25. Y. Ooyama, K. Uenaka and J. Ohshita, RSC Adv., 2015, 5, 21012.
26. Amounts of dye adsorbed on TiO₂ in Fig. 7 of ref. 13 are converted to those for the real TiO₂ surface by using a specific surface area of 83 m² g⁻¹ for PST-18NR.
27. S. Y. Huang, G. Schlichthörl, A. J. Nozik, M. Grätzel and A. J. Frank, J. Phys. Chem. B, 1997, 101, 2576.
28. J. He, G. Benkö, F. Korodi, T. Polıvka, R. Lomoth, B. Åkermark, L. Sun, A. Hagfeldt and V. Sundström, J. Am. Chem. Soc., 2002, 124, 4922.
29. K. Hara, Y. Tachibana, Y. Ohga, A. Shinpo, S. Suga, K. Sayama, H. Sugihara and H. Arakawa, Sol. Energy Mater. Sol. Cells, 2003, 77, 89.
30. K. Hara, Y. Dan-oh, C. Kasada, Y. Ohga, A. Shinpo, S. Suga, K. Sayama and H. Arakawa, Langmuir, 2004, 20, 4205.
31. S. Zhang, M. Yanagida, X. Yang and L. Han, Appl. Phys. Express, 2011, 4, 042301.
32. J.-Y. Kim, J. Y. Kim, D.-K. Lee, B. Kim, H. Kim and M. J. Ko, J. Phys. Chem. C, 2012, 116, 22759.
33. L. Yu, K. Fan, T. Duan, X. Chen, R. Li and T. Peng, ACS Sustainable Chem. Eng., 2014, 2, 718.
34. B. C. O'Regan, I. López-Duarte, M. V. Martínez-Díaz, A. Forneli and J. Albero, J. Am. Chem. Soc., 2008, 130, 2906.
35. X. Wang, L. Guo, P. F. Xia, F. Zheng, M. S. Wong and Z. Zhu, J. Mater. Chem. A, 2013, 1, 13328.

---

Footnote

† Electronic supplementary information (ESI) available: Details of CV measurements of the dyes. Absorption spectra of the dyes in a mixed solution. Transient absorption spectroscopy. See DOI: 10.1039/c5ra12616a

---