Figure 1 shows the chemical structures of the porphyrin sensitizers used in the present study. Por-1 has two ethyl carboxyl groups at the two β-positions of the porphyrin ring and two vinyl groups at the two β-positions on the opposite side. Here, different central metals, including 2H, Zn, and Cu, have been used as the coordinating metal for Por-1. Por-1 with sodium ions has also been used to study the effect of protons, because the binding mode of the dye onto the TiO₂ surface is a major factor affecting the photovoltaic performance of solar cells [12]. Por-2 is obtained by replacing the vinyl groups of Por-1 by two ethyl groups, and this can lead to a slightly smaller conjugation system that causes Por-2 to be less absorptive in the long-wavelength region.

First, the absorption capabilities of the porphyrin sensitizers were evaluated. Figure 2 shows the electronic absorption spectra of the porphyrin sensitizers dissolved in ethanol. Figure 3 shows the calculated absorption spectra with TD-DFT method in ethanol.

The calculated absorption spectra show excellent agreement with the observed absorptions. The free-base Por-1 gives a broader absorption spectrum from 300–650 nm over these Zn- and Cu-porphyrin sensitizers. Moreover, the Q-bands of the free base Por-1 split into four absorption peaks. The fact that the absorption capability of HPor-1 is greater than that of the metallated Por-1 suggests that if the electron injection efficiencies for these sensitizers are equal, DSSCs based on the former sensitizer may generate more photocurrent than would those based on the latter sensitizers.

Replacing the protons of HPor-1 with sodium ions further shifts the Q-band absorption to the longer-wavelength region. Furthermore, the absorption spectrum of the HPor-1-Na sensitizer shows some characteristics related to dye aggregation in solution. This may be due to the reduced solubility of the dye upon deprotonation. The absorption capability of Por-1 with different central metals can be ranked as follows: HPor-1 > ZnPor-1 > CuPor-1. Meanwhile, the absorption capability of ZnPor-1, over the entire solar spectrum, is very similar to that of ZnPor-2.

To further understand the electron ejection capabilities of these porphyrin sensitizers, their molecular orbitals and energy levels are calculated by the DFT and TD-DFT methods. Figure 4 shows the molecular orbitals, including the HOMO-1, HOMO, LUMO, and LUMO+1, of the porphyrin dyes in ethanol. A comparison of these porphyrin molecules leads to the following conclusions: (1) since the density of the electron cloud at the carboxyl group is similar, the three Por-1 molecules with different central metals have similar electronic coupling abilities with semiconductors; (2) the deprotonation of HPor-1 does not affect the molecular orbital; (3) ZnPor-2 has fewer π-electrons at the β-positions of the porphyrin ring than does ZnPor-1, suggesting that the π-electrons in ZnPor-2 are more delocalized; (4) furthermore, the molecular structure of porphyrin sensitizers does not consist any electron donor groups, and intramolecular charge transfer (ICT) is not reflected by the molecular orbitals.

Figure 5 depicts the energy levels of the four molecular orbitals, together with the conduction band of TiO₂. All the porphyrin sensitizers have the LUMO and LUMO+1 levels above the conduction band edge (CBE) of TiO₂. This suggests that favorable electron injection can take place from these porphyrins into TiO₂. However, the difference in the energy levels of the HOMO orbitals can substantially affect the photovoltaic performance, since the charge recombination kinetics are mainly determined by the difference in the CBE of the semiconductor and the HOMO level (ΔE_R) of the dye. For Por-1 with different central metals, the ΔE_R values can be ranked in the following order: CuPor-1 > HPor-1 > ZnPor-1. The charge recombination will also be in the same order. Since the V_oc value of a DSSC is mainly determined by the charge recombination process, a larger ΔE_R should lead to a smaller V_oc.

On the other hand, a similar prediction may be made when comparing ZnPor-1 with ZnPor-2. Although the ΔE_R value of ZnPor-2 is larger than that of ZnPor-1, the change in the molecular structure at the β-positions can substantially affect the charge recombination process. It has been found earlier that linear alkyl groups can function as a hydrophobic shell to prevent contact between the semiconductor and the electrolyte [8]. The ethyl groups of ZnPor-2 have a similar effect, and therefore, solar cells based on ZnPor-2 should generate higher photovoltage than those based on ZnPor-1.

The results from our experiments showed that our predictions were correct. We measured the photocurrent-photovoltage (I-V) curves and incident photon-to-current conversion efficiency (IPCE) profiles of the solar cells based on the porphyrin sensitizers (Figure 6). Table 1 lists the relevant parameters obtained from the I-V curves. For Por-1 with different central metals, the short-circuit current (J_sc), open-circuit voltage (V_oc), and η are in the following order: ZnPor-1 > HPor-1 > CuPor-1. This order is in excellent agreement with our principle shown above. The difference in the photovoltaic performances of these metal-based Por-1 sensitizers is mainly due to their different HOMO energy levels.

On the other hand, ZnPor-2 gives lower J_sc, but higher V_oc and η values than does ZnPor-1 for the following reasons: the vinyl groups of ZnPor-1 extend the π-electrons out of the porphyrin core, and this improves the light-harvesting capability of this sensitizer, as shown in Figure 2 and Figure 3.The light-harvesting capability of a sensitizer is a key parameter that determines the amount of solar energy can be captured by the solar cell and thus affects the photocurrent generated by the solar cell. The higher V_oc value of the ZnPor-2-based solar cell is mainly due to the effect of the ethyl groups that prevent the leakage of electrons from TiO₂ to the electrolyte. As a result, the overall photovoltaic performances of ZnPor-1 and ZnPore-2 are practically the same (2.9% and 2.8%, respectively).

Moreover, the deprotonated HPor-1 shows much poorer photovoltaic performance than does the pristine cell. This could be due to the formation of aggregates, which can cause serious exciton annihilation, as suggested in previous studies [7,8].

The chlorin macrocycle forms the core structure in natural chlorophylls. This leads to chlorins having better light-harvesting capabilities than the corresponding porphyrins. Chlorin-based chlorophylls therefore are more suitable for solar cell applications. In the present study, we used chlorin sensitizers that can be easily derived from their natural chlorophyll analogs, i.e., from chlorophylls a and b. Figure 7 shows the chemical structures of the chlorin sensitizers. The difference between Chl-1 and Chl-2 is in the number of carboxyl groups, i.e., three for the former and two for the latter. Similar to porphyrins, Chl-1-Na can be obtained by the deprotonation of Chl-1 with sodium ions. The basic chlorin structure of both Chl-1 and Chl-2 is similar to that of their natural analog chlorophyll a, which has a methyl group at the C7 position. For comparison, we also tested Chl-3-Na with an aldehyde group at the C7 position, since the natural chlorophyll b molecule has this structure.

Figure 8 shows the electronic absorption spectra of the chlorin sensitizers in ethanol solution. In order to reproduce the observed absorption spectra with molecular simulation, Figure 9 shows the calculated absorption spectra with TD-DFT method in ethanol.

It is clear that a change in either the number of carboxyl groups or the number of protons does not strongly affect the absorption spectrum. The Chl-1, Chl-2, and Chl-1-Na sensitizers have identical absorption patterns. This phenomenon is different from that observed for porphyrin sensitizers. In the case of porphyrin sensitizers, the deprotonated porphyrins aggregate in solution. In addition, the absorption region of the Chl-3-Na synthesizer is narrower than that of the Chl-1-Na sensitizer.

DFT calculations are performed for the chlorin sensitizers as well. Figure 10 shows the frontier molecular orbitals of the chlorin sensitizers in ethanol solution. The molecular orbitals of Chl-1, Chl-1-Na, and Chl-2 are very similar to each other. This is because a change in the unconjugated carboxyl groups does not alter the distribution of electrons over the cyclic tetrapyrrole ring. However, because of the aldehyde group at the C7 position, the LUMO of Chl-3-Na shows extended electrons in the direction of the Q_x transition dipole.

The energy levels of the four molecular orbitals have been compared in Figure 11. The energy levels of the LUMO of the chlorin sensitizers are also sufficiently higher than the CBE of TiO₂. Thus, a major factor that determines the photovoltaic performance must be related to the charge recombination process rather than to the electron injection process. Similar to porphyrin dyes, charge recombination should be more pronounced in chlorin sensitizers with a deeper HOMO energy level. Chl-1, with the deepest HOMO level, seems to be the best suited for the reverse electron transfer process and thus should give the lowest V_oc and η values.

In order to test the theory stated above regarding HOMO levels to determine the photovoltaic performance, we fabricated DSSCs based on the chlorin sensitizers. Figure 12 shows the I-V curves and IPCE profiles of DSSCs based on the chlorin sensitizers, and Table 2 lists the relevant parameters obtained from the I-V curves. In contrast to its effect on the porphyrin sensitizers, deprotonation of the Chl-1 sensitizer substantially improves the photovoltaic performance. Except in the case of Chl-3, which has a different chlorin frame, the J_sc, V_oc and η values were 2.5 mA·cm⁻², 0.47 V, and 0.9% for Chl-1, 4.9 mA·cm⁻², 0.61 V, and 2.1% for Chl-2, and 5.7 mA·cm⁻², 0.65 V, and 2.7% for Chl-1-Na. This effect of protons can be attributed to the change in the HOMO energy levels, as predicted in the previous section.

The poorer photovoltaic performance of Chl-3-Na as compared to that of Chl-1-Na is attributable to the following reasons: (1) the light-harvesting capability of Chl-3-Na is the lowest among that of all the chlorin sensitizers; (2) the HOMO level of this dye is favorable for charge recombination; and (3) most importantly, the LUMO+1 and LUMO of this particular dye are close to each other, making the excitation relaxation too slow to facilitate electron injection from the LUMO level.

We also critically examined the porphyrin and chlorin sensitizers as potential candidates for real-world applications. The photovoltaic performance of the best sensitizer studied in the present investigation is still lower than that of the state-of-art sensitizers currently used in DSSCs. However, the sensitizers that we studied are much more inexpensive and environment-friendly. Moreover, the photovoltaic performance of these commercially available dyes might be improved by selective co-sensitization [13].

Dye sensitizers were obtained from Frontier Scientific (Logan, UT, USA) and used as received.

The OTE with 0.25 cm² working area contains 18 nm and 400 nm TiO₂ nanoparticles with thickness of 10 μm and 4 μm obtained from CCIC, for light-harvesting and light scattering, respectively. The details of fabrication of DSSCs were described before [14]. Each DSSC that was fabricated by the use of this OTE, the counter electrode of Pt-sputtered FTO glass (10 Ω·cm⁻², Nippon Sheet Glass, Osaka, Japan), and the electrolyte A containing 0.1 M LiI, 0.05 M I₂, and 0.6 M 1-propyl-3-methylimidazolium iodide in a mixture of acetonitrile and valeronitrile (1:1, v/v). The experimental set-up for measurements of I-V characteristics and IPCE profiles was described before [14].

All systems were optimized by the use of DFT calculation with the CAM-B3LYP [15,16,17] exchange-correlations functional and the 6-31G(d,p) [18] basis set, and with solvation effect for the free base sensitizer described by CPCM (ethanol) [19]. The TD-DFT [20] with the CAM-B3LYP exchange-correlations functional and the 6-31G(d,p) basis set calculations are based on the optimized structure with solvation effect described by CPCM (ethanol) for the set of sensitizers. All calculations were done by Gaussian 09 [21] using the Research Center for Computational Science, Okazaki, Japan.