As shown in Figure 1, D1–D3 are cyanoacetic acid-based dyes with carbazole, iminodibenzyl, and phenothiazine unit as the donor, respectively. D4 is an acrylic acid-based dye with a phenothiazine unit. The aldehydes 3, 6, 9 were prepared by a Vilsmeier reaction of N-ethylcarbazole, N-ethyliminodibenzyl, and N-ethylphenothiazine, respectively, with POCl₃ in DMF [22]. The final reaction for D1–D4 was the condensation of the respective aldehyde with cyanoacetic acid (or malonic acid) by the Knoevenagel reaction in the presence of piperidine. The optimized structures of D1–D4 were studied to compare the structure differences between these photosensitizers (Figure 1). The dihedral angles (∠1) of the D1 carbazole ring, the D2 iminodibenzyl ring, and the D3 phenothiazine ring are 108.7°, 123.7°, and 122.1°, respectively, implying that the addition of a dimethylene bridge and sulfur units significantly increases the dihedral angle of nitrogen atoms. The dihedral angle between two phenyl units inside the carbazole unit (∠4) of D1 is 106.4°; however, the dihedral angle between the phenyl unit and dimethylene bridge inside the iminodibenzyl unit (∠4) of D2 is 126.1°. This can be attributed to the incorporation of a dimethylene bridge between two phenyl units increasing the steric hindrance, increasing the dihedral angles of ∠1 and ∠4. The dihedral angle (∠8) of the D3 phenothiazine sulfur atom is 99.2°, which is smaller than that of the phenothiazine nitrogen atom (∠1: 122.1°). This may be attributed to the difference between nitrogen and sulfur atoms; nitrogen atom has a lone pair whereas sulfur atom has two lone pairs. For the cyanoacetic acid and acrylic acid-based dye, the dihedral angles (∠9) of the D3 and D4 are 123.8° and 121.6°, respectively, whereas dihedral angles (∠10) of the D3 and D4 are 121.8° and 125.1°, respectively implying that the incorporation of a electron-withdrawing cyano group decreases the dihedral angle (∠10).

Figure 2 shows the UV-Vis absorption and emission spectra for the four dyes in CH₃CN and the λ_max are listed in Table 1, together with the UV–Vis spectra of the corresponding dyes adsorbed on TiO₂ film. The absorption spectrum of D1 in CH₃CN has three distinct absorption bands at around 288, 320 and 385 nm, respectively. The absorption peaks at around 320 nm correspond to the π→π* electron transition of the conjugated molecule; and the absorption peaks at around 385 nm can be assigned to an intramolecular charge transfer between the carbazole-based donor and the cyanoacetic acid [23], providing efficient charge-separation at the excited state. Under similar conditions, the D2 sensitizer had absorption peaks at 258 and 368 nm that were blue-shifted relative to the peaks of D1, implying that the incorporation of dimethylene bridge destroys the conjugation between two phenyl groups. Compare the maximum absorption wavelength of D1–D3 in CH₃CN solution. D3 (400 nm) shows a red shift relative to D1 (385 nm) and D2 (368 nm), which can be attributed to the extra electron-donating sulfur atom in D3. Such red-shifting in the absorption spectra implies a more effective utilizing of solar light. For the cyanoacetic acid and acrylic acid-based dyes (D3 and D4, respectively), when an electron acceptor (-CN) was linked to the vinyl bridge, the maximum absorption was red-shifted from 381 to 400 nm. The red shift of the maximum absorption peak arises from the fact that one more electron acceptor (-CN) enhances the electron-withdrawing ability of electron acceptors and lowers the LUMO, thus reducing the gap between HOMO and LUMO.

Figure 2(b) shows the UV-Vis absorption spectra for the dye loaded TiO₂ films with a bare TiO₂ film as the reference. Upon dye adsorption onto the TiO₂ surface, the maximum absorption respectively red-shifted by 23, 46, 26, and 31 nm for D1, D2, D3, and D4 as compared to the spectra in solution, implying that dyes adsorbed on the TiO₂ surface had partial J-type aggregates. However, the stronger red shift of D2 (368 nm→414 nm) compared to that of D1 (385→408 nm) from the solution to TiO₂ film may imply that the iminodibenzyl-containing sensitizer has a higher tendency to form surface binding with TiO₂ than that of carbazole-containing sensitizer, as chelating to metal cation, it may enhance the intramolecular electron transfer (ICT), leading to spectrum shift. Under similar conditions, the D3 and D4 sensitizers showed a slight red shift (400→426 nm and 381→412 nm, respectivity) after being adsorbed on the TiO₂ surface.

The absorption spectra of D1–D4 on the TiO₂ electrode are broader than those in CH₃CN. The broadening of the absorption spectra is due to an interaction between the dyes and TiO₂ [24]. The absorption redshift of D3 relative to D2 is due to its stronger intramolecular charge transfer because phenothiazine is a stronger electron-donating ring than iminodibenzyl. The molar extinction coefficients of D1, D2, D3, and D4 in CH₃CN solution are 22,975 M⁻¹ cm⁻¹ (at 385 nm), 29,846 M⁻¹ cm⁻¹ (at 368 nm), 14,096 M⁻¹ cm⁻¹ (at 400 nm), and 12,639 M⁻¹ cm⁻¹ (at 381 nm), respectively; they are comparable or larger than that of N3 (cis-bis(isothiocyanato)bis(2,2′-bipyridyl-4,4′-dicarboxylato)-ruthenium(II)) (14,200 M⁻¹ cm⁻¹) [25], indicating that the novel dyes have good light harvesting ability.

As shown in Figure 2 (c), the PL maxima of D1 are located at 504 nm in CH₃CN, whereas those of D2 shifted to 470 and 506 nm, and those of D3 and D4 shifted to 573 and 558 nm, respectively. D3 and D4 exhibited a relatively large Stokes shift (173 and 177 nm, respectively), which could be attributed to the geometrically relaxed structure of the phenothiazine center upon excitation [26].

To judge the possibilities of electron transfer from the excited dye molecule to the conductive band of TiO₂ and the dye regeneration, redox potentials of D1–D4 in acetonitrile were obtained using cyclic voltammetry (as shown in Figure 3). The electrochemical data are listed in Table 2. The excited-state oxidation potentials were obtained from the first oxidation potential E_ox (vs. NHE) measured by cyclic voltammetry, the ferrocenium/ferrocene (Fc/Fc⁺) redox couple was used as an internal potential reference. The excited-state oxidation potential E(S⁺/S*) was calculated using Equation 1 [27]. (1) E ( S + / S * ) = E ( S + / S ) − E 0 − 0where E(S⁺/S) is the ground state oxidation potential of D1–D4 and E_0–0 is the zeroth–zeroth transition value obtained by the intersection of the normalized lowest energy absorption peak and highest energy fluorescence peak. All E_0–0 values in our experiment were obtained by the absorption and fluorescence spectra measured in acetonitrile which was also used in the preparation of sandwich cell.

When D1 was swept in the anodic direction, there was an oxidation peak at 1.25 V, which corresponded to the first oxidation steps (formation of the radical cation). However, there was no reduction peak, indicating that the electrochemical behavior in the anodic direction was irreversible. For D2, the oxidation was also irreversible, with peaks appearing at 1.00 and 1.42 V, which corresponded to the first and second oxidation steps (forming the radical cation and dication), respectively. They are higher than those of D1, which could be attributed to the relatively higher electron density on the nitrogen atoms of D2, which allowed it to be more easily oxidized than a carbazole derivative.

It is evident that the iminodibenzyl unit enhanced the hole transporting ability of the photosensitizer more than that of the carbazole unit. The oxidation of D3 was quasi-reversible, with peaks located at 0.72 and 1.35 V. Both the first and second oxidation potentials are smaller than those of D2, indicating that the phenothiazine units are much more effective in lowering the ionization potential than iminodibenzyl units. The oxidation of D4 was also quasi-reversible, with three peaks located at 0.63, 1.29, and 1.76 V, comparison of these results with previously reported redox properties of phenothiazine-containing oligomers [28] allow assignment of the three waves to the phenothiazine radical cation, dication, and trication, respectively. The irreversibility of the second and third oxidation waves likely stems from the highly reactive nature of the dication and trication [28]. Both the first and second oxidation potentials of D3 (0.72 and 1.35 V) are larger than those of D4 (0.63 and 1.29 V), indicating that the incorporation of electron-withdrawing cyano groups at the vinyl bridge enhanced electron affinity significantly. Moreover, introducing a phenothiazine group into the photosensitizer framework (D3 and D4) resulted in a largely negative shift of the HOMO level, to 0.52–0.54 V vs. NHE. The HOMO levels of D1 and D2 (1.14 and 0.77 V vs. NHE, respectively) are more positive than that of D3, indicating more efficient dye regeneration for D1 and D2. For the cyanoacetic acid and acrylic acid-based dyes (D3 and D4, respectively), the incorporation of electron-withdrawing cyano groups at the vinyl bridge increases the LUMO level [−2.04 V (D4)→−1.92 V (D3)]. The LUMO level of D2 (−2.14 V vs. NHE) is more negative than those of D1, D3, and D4 (−1.70, −1.92 V, and −2.04 vs. NHE, respectively), indicating that D2 dye has more efficient electron injection.

Figure 4 shows the schematic energy diagram of these dye-sensitized TiO₂ electrodes. All the excited state oxidation potentials (LUMO) of D1 (−1.70 V), D2 (−2.14 V), D3 (−1.92 V), and D4 (−2.04 V) are more negative than the conduction band edge of TiO₂ [−0.5 V (vs. NHE)] [29]. Provided that an energy gap (between dye LUMO and TiO₂ CB) of 0.2 eV is necessary for efficient electron injection [30], the driving force is sufficient for efficient charge injection. Thus, the electron injection process from the excited dye molecule to the TiO₂ conduction band and the subsequent dye regeneration are energetically permitted. Figure 4 also indicates that the energy levels of the ground state (HOMO) of D1 (1.14 V), D2 (0.77 V), D3 (0.52 V), and D4 (0.54 V) are sufficiently more positive than the I₃⁻/I⁻ redox potential (0.42 V (vs. NHE)) [29], indicating that the oxidized dye formed after electron injection into the conduction band of TiO₂ could accept electrons from I⁻ ions in the electrolyte thermodynamically. Such electronic structures thus ensure a favorable exothermic flow of charges throughout the photo-electronic conversion.

The geometrical and electronic properties of D1–D4 were studied with density functional theory (DFT) using the Gaussian 03 program package [31], as shown in Figure 5. The calculations were done on a B3LYP/6-31G(d) level for geometry optimizations in the ground state. B3LYP is a hybrid function modified from the three-parameter exchange-correlation functional of Becke [32], while the gradient-corrected exchange and correlation functions are calculated according to Becke [33] and Lee et al. [34]. The frontier MOs of D2–D4 reveal that HOMO–LUMO excitation moves the electron density distribution from the iminodibenzyl, phenothiazine moiety to the cyanoacrylic acid (or acrylic acid) moiety; however, the frontier MOs of D1 reveal that the electrons at the ground state (HOMO) are homogeneously distributed in both the electron donor group and the methylene bridge of the cyanoacrylic acid. Comparing the frontier MOs of D2 and D3 (or D4) at the ground state (HOMO), the electrons of D2 are dominantly distributed in the benzene ring and the nitrogen atom; however, there are few electrons in the dimethylene bridge of the iminodibenzyl unit, the electrons of D3 (or D4) are homogeneously distributed in the phenothiazine unit. The LUMO electron density geometry distributions of D1–D4 are located over the cyanoacrylic (or acrylic) group through the right phenyl group and the nitrogen atom of carbazole, iminodibenzyl and phenothiazine units, respectively. However, the LUMO + 1 of D1–D4 is centered on the phenyl group of carbazole, iminodibenzyl, and phenothiazine units. At the excited state (LUMO) with light illumination for D1–D4 dyes, intramolecular charge transfer occurs, resulting in electron movement from the donor groups to the acceptor groups (cyanoacrylic and acrylic groups). Furthermore, the location of the LUMO at the side of the TiO₂ surface and the HOMO at the opposite end of the molecule makes the electron injection into the semiconductor easier and prevents back regeneration of the dye with injected electrons.

The vertical excitations were calculated by time dependent density functional theory (TDDFT) with B3LYP/6-31 + G(d) in vacuo and in CH₃CN solution. It was found that the inclusion of solvation effects does not lead to a qualitative change in the electronic structure, even though smaller HOMO-LUMO gaps were calculated in CH₃CN as compared to the gas phase. The origins of these electronic absorptions were found by calculating the singlet electronic transitions with the TDDFT method in Gaussian 03W program suite. Calculations show that these visible bands are mainly attributed to the electronic transition from the highest occupied molecular orbitals (HOMOs) to the lowest unoccupied molecular orbitals (LUMOs). Table 3 shows the calculated absorption maxima in the visible region, their oscillator strengths, and their compositions for D1–D4. To analyze the photophysical properties of these dyes, we performed TD-DFT calculations of the lowest 10 singlet–singlet excitations of dyes in CH₃CN solutions. The three transitions of D1–D4 with oscillator strengths (f) above 0.05 are summarized in Table 3. Compared with the experimental data, the considerable blue-shift of the absorption maximum from the calculations is related to the self-interaction error in TD-DFT arising from the electron transfer in the extended charge-transfer state [35]. The calculated absorption maxima for the excited states seem in good agreement with the measured values, supporting the band assignments in the UV solution plot.

DSSCs were prepared and compared to investigate the relationships between the sensitizing behavior of D1–D4 dye molecules and their structures. The dye-sensitized solar cells were constructed by using these dyes as a sensitizer for nanocrystalline anatase TiO₂.

The incident photon-to-current conversion efficiency (IPCE) of a DSSC was calculated by normalizing the photocurrent densities for incident light energy and intensity according to the Equation 2 [36]: (2) IPCE ( λ ) = 1240 ( eV · nm ) λ ( nm ) J s c ( mA · cm − 2 ) φ ( mW · cm − 2 )where J_sc is the short-circuit photocurrent generated by monochromatic light, λ is the wavelength of incident monochromatic light, and φ is the power of the incident radiation per unit area.

Figure 6 shows action spectra of monochromatic incident-to current conversion efficiencies (IPCEs) for DSSCs using D1–D4. The IPCEs of D1–D4 are more than 50% in the spectral range from 400 to 430 nm. D3 reaches its maximum of 82% at 430 nm, whereas D1, D2, and D4 reach their maximums of 60–66% at 410–420 nm. When reflection and absorption losses of the FTO glass substrate are considered, the net photon-to-electron conversion efficiencies of the D1–D4 would almost exceed 85% in their spectral ranges. The decrease of the IPCE above 600 nm in the long-wavelength region is attributed to the decrease of light harvesting for these dyes. The lower photocurrent response of D1, D2, and D4 sensitized devices is ascribed to the blue shifted dyes with a higher energy band-gap compared to that of the D3 dye.

A typical photocurrent–photovoltage (I – V) curve for cells based on D1–D4 is shown in Figure 7. The detailed photovoltaic parameters are summarized in Table 4. The solar-energy-to-electricity conversion efficiency (η) of the DSSCs is calculated from short-circuit current (J_sc), the open-circuit photovoltage (V_oc), the fill factor (FF), and the intensity of the incident light (P_in) according to the following Equation [36]: (3) η = [ J s c ( mA · cm − 2 ) ] [ V o c ( V ) ] [ FF ] P i n ( mW · cm − 2 )

Under the standard global AM 1.5 solar condition, the D1 cell had a short circuit photocurrent density (J_sc) of 5.76 mA·cm⁻², an open-circuit voltage (V_oc) of 0.595 V, and a fill factor of 0.71, corresponding to an overall conversion efficiency of 2.43%. Under similar conditions, the photovoltaic parameters (J_sc, V_oc, and η) of cells with the D2 sensitizer are 8.28 mA·cm⁻², 0.62 V, and 3.49%, respectively, those of the D3 sensitizer are 13.35 mA·cm⁻², 0.669 V, and 5.53%, respectively, and those of the D4 sensitizer are 8.09 mA·cm⁻², 0.628 V, and 3.46%, respectively. The great efficiency of the D3 sensitizer is probably due to the strong electron-donating ability of the phenothiazine unit when electrons transfer from the phenothiazine unit to the cyanoacrylic group. The solar-energy-to-electricity conversion efficiency of the present organic dyes could be improved by extending the conjugated length of organic dyes or incorporating a thiophene π-bridge. The synthesis of high efficiency organic dyes in our lab is still in progress. For the stability or long-term durability of presented organic dyes, the efficiencies of devices don’t show obvious decay till 400 hr, the main contribution of DSSC efficiency decay after 400 hr is the electrolyte. Furthermore, the larger V_oc of D3 is attributed to the retardation of charge recombination at the nanocrystallite/dye/redox electrolyte interface [37], the nonlinear I – V curve observed in Figure 7 is intriguing, the schematic energy level diagram (Figure 4) for a DSSC based on D3 shows less energy gap between HOMO vs. (I⁻/I₃⁻) redox potential than those of D1 and D2, this may result in slow diffusion current of main carrier and nonlinearity in I – V curve.

The conversion of solar energy into electricity is attracting more attention as non-renewable energy reserves diminish. Dye sensitised solar cells consist of a nanostructured semiconductor film (e.g., ~10 μm anatase titania, Figure 8), coated with a monolayer of photoactive dye and an electrolyte bearing a redox couple (e.g., I₃⁻/I⁻). Sunlight enters the cell through the transparent glass plate striking the dye on the surface of the TiO₂. Absorption of a photon of light by the dye promotes an electron is transferred from S° to a higher lying energy level, then the sensitizer is in the excited state S*. The excited electron is injected into the conduction band of the semiconductor. A carboxyl group in D1–D4 can form an ester linkage with TiO₂ surface to provide a strongly bound dye and a good electron communication between them.

The general operating principle of a dye-sensitized solar cell comprising a photoanode (TiO₂) and a passive cathode is depicted in Figure 9. The regenerative process in the dye solar cell consists of the following five steps, the process leads to photovoltaic performance in the solar cell.

Electrons in dye are excited by solar energy adsorption.

All starting materials were purchased from Aldrich, Lancaster, TCI, and Acros and used as received. DMF, dichloromethane, and POCl₃ were distilled from CaH₂ under N₂ atmosphere. Tetrabutylammonium perchlorate (TBAP) and 1,2-dimethyl-3-propylimidazolium iodide (DMPImI) were synthesized and purified according to a procedure described in the literature [38].

The ¹H-NMR spectra were obtained on a Bruker 400 MHz FT-NMR. Chemical shifts are reported in ppm relative to tetramethylsilane δ units. The absorption spectra of the dyes in solution and adsorbed on TiO₂ films were recorded on a Cary 100 UV-vis spectrophotometer. Fluorescence measurements were carried out using a Hitachi F-4500 fluorescence spectrophotometer. Cyclic voltammetry was performed using an electrochemical workstation (CH instruments Inc., CHI, model 750A) and conducted in CH₃CN solution using 0.1 M tetrabutylammonium perchlorate (TBAP) as the supporting electrolyte. The working electrode was a glassy carbon electrode, the auxillary electrode was a Pt wire, and the reference electrode was Ag/Ag⁺. The scan rate was 100 mv·s⁻¹ and the temperature was 25 °C. Ferrocene was added to each sample solution at the end of the experiments. The ferrocenium/ferrocene (Fc/Fc⁺) redox couple was used as an internal potential reference. The potential versus SCE in acetonitrile was calibrated according to a procedure published by Matsui et al. [39].

The working TiO₂ electrode (photoanode) and counter electrode for dye-sensitized solar cells were prepared as follows. F-doped tin oxide (FTO) glass plates (3 mm thickness, 7 Ω cm⁻²) were first cleaned in a cleaning detergent aqueous solution with an ultrasonic bath for 15 min, and rinsed with water and ethanol. Then, the FTO electrodes were immersed into 40 mM TiCl₄ (aqueous) at 70 °C for 30 min and rinsed with water and ethanol. Two kinds of TiO₂ paste, containing nanocrystalline (~25 nm) TiO₂ (Degussa P25, paste A) and submicroparticle TiO₂ (paste B, 500 nm), respectively, were prepared using a previously reported procedure [40]. To prepare paste A, commercial titania powder (3 g, Degussa P25) was ground in a mortar with a small amount of water (1 mL) containing acetylacetone (0.1 mL), which was added to prevent reaggregation of the particles. After the TiO₂ particles were dispersed, the paste was diluted by the slow addition of water (3 mL) under continued grinding and a surfactant, Triton X-100 (0.05 mL), was added to facilitate the spreading of the colloid on the substrate. Paste A was sonicated for about 24 h at 28 °C. The spin coating procedure for paste A was repeated to get the appropriate thickness of TiO₂ films (12 μm). After paste A was dried at 125 °C, paste B was coated two more times, which made the thickness of the TiO₂ films with 500 nm particles for the scattering layer ~4 μm. The electrodes coated with TiO₂ pastes were gradually heated (5 °C min⁻¹) under airflow up to 450 °C, which was kept for 30 min. The electrodes were treated with 40 mM TiCl₄. The TiO₂ films were then rinsed with water and ethanol and sintered again at 450 °C for 30 min. An active area of 0.5 cm × 0.5 cm was selected from a sintered electrode. The electrodes were immersed in a 3 × 10⁻⁴ M solution of dye containing acetonitrile and isoporpanol (in a volume ratio of 1:1). Dye coatings were applied at room temperature for 24–30 h. The dye-adsorbed TiO₂ films were taken out and rinsed with dry ethanol. The rinsing process was repeated several times to remove unbound dyes completely. Finally, the dye-adsorbed TiO₂ films were dried in air. A counter electrode was prepared by sputtering a thin platinum layer on an FTO substrate. The thickness of the Pt layer was controlled to be 50 nm. The thicknesses of the TiO₂ films were determined by profilometry. Platinum sputtering was carried out using a Hitachi E 1045 instrument. The platinum layer thickness of on the glass substrates was estimated using the amount of sputtered platinum recorded on a quartz thickness monitor.

The dye adsorbed TiO₂ electrode and Pt-counter electrode were assembled into a sandwich sealed type cell by heating them with hot-melt ionomer film (25 μm thickness, Solaronix) as a spacer. A drop of electrolyte solution [0.1 M LiI, 0.05 M I₂, 0.6 M DMPII, 0.5 M tert-butyl pyridine (TBP) in acetonitile (ACN)] was injected through a hole in the counter electrode, which was then sealed with hot-melt ionomer film and glass. The electrolyte was introduced into the cell and sealed with AB epoxy for 30 min. The working area of the electrode was 0.25 cm².

The photovoltaic measurements of the DSSCs were performed using a Newport M-66907 450 W xenon light source through an infrared blocking filter and a Keithley 2,400 digital source meter linked to a computerized control and data acquisition system. The light intensity was 1000 W·m⁻² under an AM 1.5 light source. Cell temperatures were kept at 25 °C during the illumination. Light intensity was calibrated using a mono-Si reference solar cell (PVM134). The incident photon-to-current conversion efficiency (IPCE) as a function of excitation wavelength was measured using an incident light 300 W xenon lamp. A 300W xenon arc lamp solar simulator (#91160A, Oriel) with an AM 1.5 Globe filter (#59044, Oriel) was used to measure the I-V characteristics of the quasi-solid-state DSSC. The illumination was fixed at 100 mW·cm⁻² using a reference solar cell and meter (#91150, Oriel).

D1–D4 are the carbazole, iminodibenzyl, and phenothiazine-containing dyes, respectively. The geometric and electronic properties of sensitizers D1–D4 were obtained using the Gaussian 03 program package [31]. The calculation was optimized using B3LYP (Becke three parameter hybrid function with Lee–Yang–Perdew gradient correlation functions) with the Pople 6–31 + g(d) atomic basis set. The excitation transitions of D1–D4 were calculated using time-dependent density functional theory (TD-DFT) calculations with B3LYP/6–31 + g(d). Molecular orbitals were visualized using Gaussview.