3D Structural Optimization of Zinc Phthalocyanine-Based Sensitizers for Enhancement of Open-Circuit Voltage of Dye-Sensitized Solar Cells

: We designed and synthesized two zinc phthalocyanine sensitizers ( PcS27 and PcS28 ), substituted with branched or cyclic alkoxy chains, to investigate the structural e ﬀ ect of peripheral alkyl chains on the performance of dye-sensitized TiO 2 solar cells. The bulky cyclic alkyl chains of PcS28 decreased the adsorption density of PcS28 on the TiO 2 electrode, while the terminal branches of alkoxy chains of PcS27 did not inﬂuence the adsorption density in comparison to the previously published PcS20 with linear alkoxy chains. Under one sun conditions, PcS27 cells exhibited higher open-circuit voltage but a slightly lower energy conversion e ﬃ ciency, 6.0% less than PcS20 . These results suggest that the small alternation of alkoxy chains resulted in decreasing electron pushing ability of peripheral phenoxy units, giving lower short-circuit current. groups in PcS28 . The split width between two Q band positions for PcS27 was narrower than that of PcS28. Both ZnPcs emitted a ﬂuorescence at 698 nm upon exciting at the Soret band, indicating that the zero-zero excitation energy ( E 0-0 ) obtained by the intersection of the normalized absorption and ﬂuorescence spectra for PcS27 and PcS28 was almost the


Introduction
Dye-sensitized solar cells (DSSCs) have drawn special attention as a promising alternative to conventional silicon-based solar cells [1,2]. The solar-to-electronic power conversion efficiency (PCE) of DSSCs has been enhanced by several approaches such as the systematic structural exploration of light-harvesting dyes, the reduction of mismatch between the oxidation potential of the dye and the redox potential of redox couples in electrolyte solutions, and the optimization of the device structures [3]. The light-harvesting dyes, which consist of chromophore units and adsorption sites, play the following three roles: the absorption of light energy, the injection of electrons from the excited dye to the semiconductor electrodes such as TiO 2 and ZnO, and as a barrier to unfavorable contact of redox shuttles with the surface of nanoparticles [4]. Whereas several chromophores with absorption bands in red and near-IR light regions have been applied as light-harvesting dyes for DSSCs, these chromophores have not shown high PCE values. This is partially because of the difficulty of matching HOMO and LUMO energy levels of dyes with two potentials in the DSSC system (the conduction band of the semiconductor electrode and the redox potential of the redox shuttle) [3,5].
Phthalocyanines (Pcs), and especially their metal complexes (MPcs), have been widely used as blue and green dyes and pigments for various applications [6]. The Pcs exhibit intense Q bands in red and near-IR light regions, whose positions can be tuned by the introduction of substituents on the aromatic core and/or modification of central metal species [6]. Thus, the MPcs are good candidates to be the light-harvesting chromophore in DSSCs to harvest red and near-IR light regions. The PCEs of DSSCs sensitized by Pc-based dyes have been improved by several approaches [7][8][9][10][11][12][13]. In 2013, we achieved a PCE of 6.4% from a DSSC employing PcS20 under one sun conditions [13]. The decoration of short propoxy chains around the Pc core resulted in the formation of a dense adsorption layer of dyes on the surface of TiO 2 without the formation of aggregates. While the incident photon-to-current conversion efficiency (IPCE) at the Q bands for DSSCs based on PcS20 dye reached 85%, the open-circuit voltage (V oc ) (0.60 V) was lower than the reported V oc values for the other dyes because of short electron lifetime in the electrodes of the DSSCs [3]. The attachment of alkyl chains around chromophores has been reported as an effective approach to enhance V oc values in DSSCs [14]. The alkyl chains of sensitizers form a dense barrier layer at the interface of dye-adsorbed TiO 2 surfaces and electrolytes, and the formation of this barrier leads to an increase in electron lifetime. Therefore, further exploration of interfacial engineering for Pc-based dyes is required to achieve high PCE values. In this paper, we examined the structural effect of alkyl chains around the Pc core on the photovoltaic properties.

Materials and Methods
General. NMR spectra were recorded on a Bruker AVANCE 400 FT NMR spectrometer at 399.65 MHz and 100.62 MHz for 1 H and 13 C in CDCl 3 solution. Chemical shifts are reported relative to internal tetramethylsilane. Absorption and fluorescence spectra were measured on a SHIMAZU UV-2600 and a JASCO spectrophotometer FP-8600, respectively. MALDI-TOF mass spectra were obtained on a Bruker Microflex spectrometer with dithranol as the matrix. Mass spectra with electrospray ionization were obtained on a Bruker Daltonics micrOTOF II. Differencial pulse voltammetry (DPV) data were recorded with an ALS 720C potentiostat, and electrochemical experiments were performed under purified nitrogen gas. Nanoporous TiO 2 electrodes with 6 µm thickness as the working electrodes of DPV measurements were prepared by applying pastes of TiO 2 nanoparticles with 15-20 nm diameter onto fluoride-doped tin oxide (FTO) glass substrates (Asahi Glass) following sintering at 550 • C for 30 min in air. The TiO 2 electrodes were immersed into 50 µM PcS27 or PcS28 solutions in dry toluene for 3 h, and the dye-stained electrodes were used as working electrodes. The reference electrode was Ag/AgCl, which we corrected for junction potentials by referencing it with the ferrocenium/ferrocene (Fc + /Fc) couple.
All chemicals were purchased from commercial suppliers and used without further purification. Column chromatography was performed with activated alumina (Wako, 200 mesh) or silica gel (Wakogel C-200). Recycling preparative gel permeation chromatography was carried out by a Japan Analytical Industry recycling preparative HPLC using CHCl 3 as an eluent. Analytical thin layer chromatography was performed with commercial Merck plates coated with silica gel 60 F 254 or aluminum oxide 60 F 254 .

Synthesis of Phthalocyanine Precursors 1 and 2 (Figure 1)
Energies 2020, 13, x FOR PEER REVIEW 2 of 7 achieved a PCE of 6.4% from a DSSC employing PcS20 under one sun conditions [13]. The decoration of short propoxy chains around the Pc core resulted in the formation of a dense adsorption layer of dyes on the surface of TiO2 without the formation of aggregates. While the incident photon-to-current conversion efficiency (IPCE) at the Q bands for DSSCs based on PcS20 dye reached 85%, the opencircuit voltage (Voc) (0.60 V) was lower than the reported Voc values for the other dyes because of short electron lifetime in the electrodes of the DSSCs [3]. The attachment of alkyl chains around chromophores has been reported as an effective approach to enhance Voc values in DSSCs [14]. The alkyl chains of sensitizers form a dense barrier layer at the interface of dye-adsorbed TiO2 surfaces and electrolytes, and the formation of this barrier leads to an increase in electron lifetime. Therefore, further exploration of interfacial engineering for Pc-based dyes is required to achieve high PCE values. In this paper, we examined the structural effect of alkyl chains around the Pc core on the photovoltaic properties.

Materials and Methods
General. NMR spectra were recorded on a Bruker AVANCE 400 FT NMR spectrometer at 399.65 MHz and 100.62 MHz for 1 H and 13 C in CDCl3 solution. Chemical shifts are reported relative to internal tetramethylsilane. Absorption and fluorescence spectra were measured on a SHIMAZU UV-2600 and a JASCO spectrophotometer FP-8600, respectively. MALDI-TOF mass spectra were obtained on a Bruker Microflex spectrometer with dithranol as the matrix. Mass spectra with electrospray ionization were obtained on a Bruker Daltonics micrOTOF Ⅱ. Differencial pulse voltammetry (DPV) data were recorded with an ALS 720C potentiostat, and electrochemical experiments were performed under purified nitrogen gas. Nanoporous TiO2 electrodes with 6 μm thickness as the working electrodes of DPV measurements were prepared by applying pastes of TiO2 nanoparticles with 15-20 nm diameter onto fluoride-doped tin oxide (FTO) glass substrates (Asahi Glass) following sintering at 550 °C for 30 min in air. The TiO2 electrodes were immersed into 50 μM PcS27 or PcS28 solutions in dry toluene for 3 h, and the dye-stained electrodes were used as working electrodes. The reference electrode was Ag/AgCl, which we corrected for junction potentials by referencing it with the ferrocenium/ferrocene (Fc + /Fc) couple.
All chemicals were purchased from commercial suppliers and used without further purification. Column chromatography was performed with activated alumina (Wako, 200 mesh) or silica gel (Wakogel C-200). Recycling preparative gel permeation chromatography was carried out by a Japan Analytical Industry recycling preparative HPLC using CHCl3 as an eluent. Analytical thin layer chromatography was performed with commercial Merck plates coated with silica gel 60 F254 or aluminum oxide 60 F254. 1: 2,6-(3′-methylbutoxy)phenol [13,15] (0.78 g, 2.9 mmol) and 4,5-dichlorophthalonitrile (0.23 g, 1.2 mmol) were dissolved in 2.0 mL of DMSO. Powder of dried K2CO3 (1.23 g, 8.93 mmol) was added to the mixture and stirred at 105 °C for 8 h [12]. After cooling down to room temperature, the reaction mixture was poured into 30 mL of water, and the reaction mixture was extracted with CH2Cl2. The combined organic phases were dried over Mg2SO4, and the solvent was evaporated. The crude product was purified by silica column chromatography (CH2Cl2) and recycling preparative HPLC (CHCl3) to obtain 1. Yield 0.42 g (55%). 1   1: 2,6-(3 -methylbutoxy)phenol [13,15] (0.78 g, 2.9 mmol) and 4,5-dichlorophthalonitrile (0.23 g, 1.2 mmol) were dissolved in 2.0 mL of DMSO. Powder of dried K 2 CO 3 (1.23 g, 8.93 mmol) was added to the mixture and stirred at 105 • Cc for 8 h [12]. After cooling down to room temperature, the reaction mixture was poured into 30 mL of water, and the reaction mixture was extracted with CH 2 Cl 2 . The combined organic phases were dried over Mg 2 SO 4 , and the solvent was evaporated. The crude product was purified by silica column chromatography (CH 2 Cl 2 ) and recycling preparative HPLC (CHCl 3 ) to obtain 1. Yield 0.42 g (55%). 1

Synthesis of Sensitizers PcS27 and PcS28
PcS27: PcS27 was synthesized from 1 (0.42 g, 0.6 mmol), methyl 3,4-dicyanobenzoate (40 mg, 0.2 mmol), and Zn(CH 3 COO) 2 (59 mg, 0.3 mmol) according to the reported method [12,13]. Yield: 50 mg (10%). UV-Vis in toluene λ max/ nm (log ε): 693 (4.99) and 678 (4.89). 1 [13] TiO 2 electrodes (apparent surface area: 0.25 cm 2 (0.5 × 0.5 cm)) were printed onto FTO glass substrates by a screen printing technique using two TiO 2 nanoparticle pastes with different diameters of 15-20 and 400 nm. After sintering at 550 • Cc for 30 min in air, the TiO 2 electrodes were treated with TiCl 4 . The PcS27 or PcS28 sensitizers were adsorbed onto TiO 2 films by immersion of electrodes in 50 µM toluene solutions of the dyes for 24 h at 25 • Cc. The dye-adsorbed TiO 2 electrode and Pt counter electrode were sealed by heating of a 50 µm thick hot melt ring (Surlyn, DuPont). Redox electrolytes (ELA-1: 0.6 M 1,2-dimethyl-3-propylimidazolium iodide, 0.1 M LiI, 0.05 M I 2 , 0.5 M tert-butyl pyridine in acetonitrile) were injected into the space between two electrodes, and then the photovoltaic performance was measured by applying black mask (0.16 cm 2 ) under a standard AM 1.5 solar condition (100 mW cm −2 ) with a solar simulator (Otenso-Sun 3SD, Bunko Keiki). [12,13] The adsorption densities of PcS27 and PcS28 were determined by measuring the absorbance of dyes released from the TiO 2 electrodes (thickness: 6 µm) by immersing into THF containing tetrabutylammonium hydroxide methanoic solution. The absorbance at the Q band of the released dye was converted into the concentration of dye in the TiO 2 electrode. The adsorption density of dye was calculated from the concentration in the TiO 2 electrode, the surface area of mesoporous TiO 2 (61 m 2 /g determined by Brunauer-Emmett-Teller surface area analyzer), and the weight of the TiO 2 electrode on the FTO substrate.

Results and Discussion
Two ZnPc-based dyes, PcS27 and PcS28, bearing branched 3-methylbutoxy and cyclic cyclohexylmethoxy chains at the 2 and 6 positions of three peripheral phenoxy units, were designed ( Figure 2). We expected the branched and cyclic alkoxy chains to improve the barrier property of the dye monolayer on the TiO 2 surface. Phthalocyanine precursors 1 and 2 were synthesized from resorcinol and 1-bromo-3-methylbutane or (bromomethyl)cyclohexane according to the same procedures as PcS20 [12,13]. Target ZnPcs PcS27 and PcS28 were prepared by a mixed cyclotetramerization between phthalocyanine precursor and methyl 3,4-dicyanobenzoate in the presence of Zn(OAc) 2 , and a following hydrolysis [13]. Two ZnPcs and their corresponding intermediates were characterized by using standard spectroscopic techniques.
Energies 2020, 13, x FOR PEER REVIEW 4 of 7 cyclotetramerization between phthalocyanine precursor and methyl 3,4-dicyanobenzoate in the presence of Zn(OAc)2, and a following hydrolysis [13]. Two ZnPcs and their corresponding intermediates were characterized by using standard spectroscopic techniques.  Figure 3a,b shows absorption spectra of PcS27 and PcS28 in toluene. The spectrum of PcS28 exhibited split Q bands at 675 and 695 nm, which were in fair agreement with the previously reported peak positions of PcS20 [13]. The appearance of split Q bands implies that the degeneracy of LUMO and LUMO+1 energy levels is broken by the asymmetrical substitution of electron-withdrawing carboxylic acid and electron-donating 2,6-dialkoxyphenoxy groups in PcS28. The split width between two Q band positions for PcS27 was narrower than that of PcS28. Both ZnPcs emitted a fluorescence at 698 nm upon exciting at the Soret band, indicating that the zero-zero excitation energy (E0-0) obtained by the intersection of the normalized absorption and fluorescence spectra for PcS27 and PcS28 was almost the same. The TiO2 electrodes were immersed in toluene solutions of PcS27 and PcS28 (50 μM) for 24 hr to give dye-stained TiO2 electrodes. FT-IR spectra of TiO2 films stained with PcS27 or PcS28 did not show absorption peaks corresponding to carboxylic acid, implying the formation of bidentate binding between carboxylic acid in the dyes and the TiO2 surface. The absorption spectra of PcS27 and PcS28 adsorbed onto TiO2 films exhibited a single Q band at 692 and 697 nm (Figure 3a,b). The fluorescence emissions from PcS27 and PcS28 were quenched on the surface of TiO2. The absorption spectral changes and fluorescence quenching suggest a good electronic communication between  Figure 3a,b shows absorption spectra of PcS27 and PcS28 in toluene. The spectrum of PcS28 exhibited split Q bands at 675 and 695 nm, which were in fair agreement with the previously reported peak positions of PcS20 [13]. The appearance of split Q bands implies that the degeneracy of LUMO and LUMO+1 energy levels is broken by the asymmetrical substitution of electron-withdrawing carboxylic acid and electron-donating 2,6-dialkoxyphenoxy groups in PcS28. The split width between two Q band positions for PcS27 was narrower than that of PcS28. Both ZnPcs emitted a fluorescence at 698 nm upon exciting at the Soret band, indicating that the zero-zero excitation energy (E 0-0 ) obtained by the intersection of the normalized absorption and fluorescence spectra for PcS27 and PcS28 was almost the same.
Energies 2020, 13, x FOR PEER REVIEW 4 of 7 cyclotetramerization between phthalocyanine precursor and methyl 3,4-dicyanobenzoate in the presence of Zn(OAc)2, and a following hydrolysis [13]. Two ZnPcs and their corresponding intermediates were characterized by using standard spectroscopic techniques. Figure 2. Structure of phthalocyanine sensitizers PcS20, PcS27, and PcS28. Red: structural differences with PcS20. Figure 3a,b shows absorption spectra of PcS27 and PcS28 in toluene. The spectrum of PcS28 exhibited split Q bands at 675 and 695 nm, which were in fair agreement with the previously reported peak positions of PcS20 [13]. The appearance of split Q bands implies that the degeneracy of LUMO and LUMO+1 energy levels is broken by the asymmetrical substitution of electron-withdrawing carboxylic acid and electron-donating 2,6-dialkoxyphenoxy groups in PcS28. The split width between two Q band positions for PcS27 was narrower than that of PcS28. Both ZnPcs emitted a fluorescence at 698 nm upon exciting at the Soret band, indicating that the zero-zero excitation energy (E0-0) obtained by the intersection of the normalized absorption and fluorescence spectra for PcS27 and PcS28 was almost the same. The TiO2 electrodes were immersed in toluene solutions of PcS27 and PcS28 (50 μM) for 24 hr to give dye-stained TiO2 electrodes. FT-IR spectra of TiO2 films stained with PcS27 or PcS28 did not show absorption peaks corresponding to carboxylic acid, implying the formation of bidentate binding between carboxylic acid in the dyes and the TiO2 surface. The absorption spectra of PcS27 and PcS28 adsorbed onto TiO2 films exhibited a single Q band at 692 and 697 nm (Figure 3a,b). The  The TiO 2 electrodes were immersed in toluene solutions of PcS27 and PcS28 (50 µM) for 24 hr to give dye-stained TiO 2 electrodes. FT-IR spectra of TiO 2 films stained with PcS27 or PcS28 did not show absorption peaks corresponding to carboxylic acid, implying the formation of bidentate binding between carboxylic acid in the dyes and the TiO 2 surface. The absorption spectra of PcS27 and PcS28 adsorbed onto TiO 2 films exhibited a single Q band at 692 and 697 nm (Figure 3a,b). The fluorescence emissions from PcS27 and PcS28 were quenched on the surface of TiO 2 . The absorption spectral changes and fluorescence quenching suggest a good electronic communication between ZnPc and TiO 2 by anchoring carboxylic acid in dyes on the TiO 2 . Moreover, the sharp Q bands for both dyes indicate the prevention of molecular aggregation among the ZnPc sensitizers on the TiO 2 surface. The first oxidation potentials (E ox ) of PcS27 and PcS28 were determined by differential pulse voltammetry measurements using dye-stained TiO 2 electrodes in acetonitrile containing 0.1 M Bu 4 NClO 4 as a supporting electrolyte [12]. The E ox values of PcS27 and PcS28 were 0.92 and 0.90 V vs. normal standard electrode (NHE), which were attributed to the phthalocyanine ring-based oxidation process. The E ox value of PcS27 was slightly higher than that of PcS28 because of the difference in electron-donating properties between 3-methylbutoxy and cyclic cyclohexylmethoxy chains on the peripheral phenoxy units. The  Table 1. The PcS27 cell exhibited a J sc of 14.1 mA cm −2 , a V oc of 610 mV, and an FF of 0.70, yielding a PCE of 6.0%. In contrast, the PcS28 cell displayed a low PCE value in comparison to the PcS27 cell. Figure 4c shows the IPCE spectra of PcS27 and PcS28 cells. The maximum IPCE values for PcS27 and PcS28 cells were 82% and 71% at 600-750 nm corresponding to the Q band of ZnPc sensitizers, and the IPCE spectrum of the PcS28 cell was narrower than that of the PcS27 cell. In order to examine the performance difference in the PcS27 and PcS28 cells, the adsorption densities of dyes on the TiO 2 surface were determined. The adsorption densities were determined to be 8.2 × 10 −11 and 4.9 × 10 −11 mol cm −2 for PcS27 and PcS28, respectively [17]. The projected molecular area of PcS27 on the TiO 2 surface calculated from the adsorption density was 2.0 nm 2 , and this value agreed with the area (2.1 nm 2 ) estimated from density funrional theory (DFT) and the Corey-Pauling-Koltun (CPK) space-filling model ( Figure 5), implying that PcS27 formed a dense packing layer on the TiO 2 surface. Despite the projected molecular area of PcS28 (2.1 nm 2 ) being the same as that of PcS27, the steric crowding among rigid cyclic cyclohexane side chains in PcS28 may result in poor packing among dyes on the TiO 2 surface. Therefore, the lower PCE of the PcS28 cell was due to the lower adsorption density on the TiO 2 surface. with the area (2.1 nm 2 ) estimated from density funrional theory (DFT) and the Corey-Pauling-Koltun (CPK) space-filling model ( Figure 5), implying that PcS27 formed a dense packing layer on the TiO2 surface. Despite the projected molecular area of PcS28 (2.1 nm 2 ) being the same as that of PcS27, the steric crowding among rigid cyclic cyclohexane side chains in PcS28 may result in poor packing among dyes on the TiO2 surface. Therefore, the lower PCE of the PcS28 cell was due to the lower adsorption density on the TiO2 surface.   Since there is no difference in adsorption density between PcS27 and our previously reported PcS20 decorated with linear propoxy chains [13], the branching terminal units do not influence the packing density on the surface of TiO2. The Voc value of the PcS27 cell was 10 mV higher than that of the PcS20, and the offset potential of dark current of the PcS27 cell was higher than that of the PcS20 cell. This suggests that covering the ZnPc core with branched alkyl chains in PcS27 may retard the interfacial charge recombination from the conduction band of TiO2 to I3 -. The narrower split width of the Q band and the higher Eox for PcS27 revealed the poor electron-pushing ability of peripheral units as compared to PcS20. This poor electron-pushing ability of PcS27 relative to PcS20 leads to the inferior Jsc and IPCE values in DSSCs.

Conclusions
The performances of DSSCs based on ZnPcs decorated with branched or cyclic alkoxy chains at the peripheral phenoxy units were examined. The PcS27 cell with 3-methylbutoxy chains showed 6.0% efficiency when used as a light-harvesting sensitizer on a TiO2 electrode under one sun conditions. We found that the terminal branches of alkoxy chains did not influence the adsorption Since there is no difference in adsorption density between PcS27 and our previously reported PcS20 decorated with linear propoxy chains [13], the branching terminal units do not influence the packing density on the surface of TiO 2 . The V oc value of the PcS27 cell was 10 mV higher than that of the PcS20, and the offset potential of dark current of the PcS27 cell was higher than that of the PcS20 cell. This suggests that covering the ZnPc core with branched alkyl chains in PcS27 may retard the interfacial charge recombination from the conduction band of TiO 2 to I 3 -. The narrower split width of the Q band and the higher E ox for PcS27 revealed the poor electron-pushing ability of peripheral units as compared to PcS20. This poor electron-pushing ability of PcS27 relative to PcS20 leads to the inferior J sc and IPCE values in DSSCs.

Conclusions
The performances of DSSCs based on ZnPcs decorated with branched or cyclic alkoxy chains at the peripheral phenoxy units were examined. The PcS27 cell with 3-methylbutoxy chains showed 6.0% efficiency when used as a light-harvesting sensitizer on a TiO 2 electrode under one sun conditions. We found that the terminal branches of alkoxy chains did not influence the adsorption density on the TiO 2 surface, and the PcS27 cell exhibited a higher V oc value than the PcS20 with linear alkoxy Energies 2020, 13, 2288 7 of 8 chains. The obtained V oc value of the PcS27 cell was still lower than the potential difference (0.9 V) between the conduction-band-edge potential of TiO 2 and the redox potential of the I -/I 3 redox shuttle.
The covering of the ZnPc core with alkyl chains should be effective to prevent unfavorable charge recombination. However, the excess decoration of the ZnPC core with alkyl chains leads to a decrease in the adsorption density of sensitizers on the TiO 2 surface. The achievement of high V oc values for ZnPc-based DSSCs requires further substituent design of a large π-surface of the ZnPc core to enhance the blacking function to the approach of I 3 in electrolytes, while keeping high adsorption density on