Exploring Different Designs in Thieno[3,4-b]pyrazine-Based Dyes to Enhance Divergent Optical Properties in Dye-Sensitized Solar Cells

Abstract

Two novel organic sensitizers for Dye-Sensitized Solar Cells (DSSC), called TP1 and TP2, based on the electron-poor thieno[3,4-b]pyrazine (TPz) π-bridge and the electron-rich N,N-bis(4-(hexylthio)phenyl)aniline (TPA) were designed following two different approaches: the classical D-A-π-A design and a symmetric structure with double anchoring functions. Both compounds were prepared exploiting short synthetic pathways based on direct arylation strategies and possibly one-pot desymmetrization. The two novel dyes displayed opposite optical properties: a broad and intense light absorption over the entire visible spectrum for TP1, and a localized absorption that peaked in the center of the visible region for TP2, resulting in a pitch-dark coloration and a green tone, respectively. When assembling the photovoltaic devices, different electrolyte compositions were explored to enhance the optical properties of the dyes. Power conversion efficiencies as high as 5.2% under full sun intensity were recorded for small test devices. The composition of the light transmitted through the TP2-containing transparent DSSC fits well with the human eye sensitivity spectrum, thus fulfilling the transparency requirements for building-integrated photovoltaics (BIPV).

1. Introduction

The recent rise in energy prices, together with the ever-growing worldwide energy demand and the shortage of supplies of non-renewable fuels, such as coal, oil, and natural gas, are unassailable reasons which should prompt us to complete the energy transition to renewable and low-impact energy sources as soon as possible. Even the construction industry is starting to realize that a general image of sustainability is not enough, and the aim of zero-energy buildings must be pursued for all new constructions. Photovoltaics should represent one of the leading technologies driving the transition process thanks to the abundance and wide distribution of solar energy. To accomplish this goal, it is necessary to make buildings self-sufficient from an energy point of view through the development of building-integrated photovoltaics (BIPV) [1,2,3,4].

To maximize energy production using BIPV, it would be desirable to take advantage of all the available surfaces in a building, including those designated to receive lighting from the outside, such as windows or glass walls. However, many of the device properties required for such an application, such as semi-transparency, tunable color, and interior visual comfort, are not intrinsic to the classic silicon-based photovoltaic technology, nor they are used in other emerging technologies, such as organic photovoltaics and perovskite solar cells (PSCs) [5,6]. In particular, attention must be paid to the human eye spectral sensitivity, which is not equally distributed along the entire solar spectrum but is centered in the green region at a maximum wavelength of 555 nm under daylight conditions [7]. Technical solutions which fulfill this requirement for BIPV (i.e., average visible transmittance (AVT) > 20%) are already available, such as the employment of amorphous silicon-based (a-Si) solar cells, though the price to pay is a significant reduction in terms of efficiency (≈7%) [8]. On the other hand, among the emerging photovoltaic technologies, Dye-Sensitized Solar Cells (DSSCs) distinguish themselves for their high versatility in terms of color palette and transparency level, having the ideal properties required for an application in the BIPV field, together with the use of non-toxic materials [9].

Optimizing each DSSC component can contribute to enhancing the transparency of the device. Moon et al. [10] substituted the usual diffusive scattering layer with a two-dimensional photonic crystal reflector to maintain high photon-to-electric conversion efficiency (8.23%), but with a significantly higher perceived transparency (52% vs. 0.3%) of the photoanode. Magnesium-doped tin oxide was demonstrated to be a conductive material with higher optical transparency (>80%) than indium-doped tin oxide, as well as good electrical conductivity (421.15 Ω/cm), and was used as a conducting substrate for the fabrication of DSSC [11]. Counter electrodes based on various conducting polymers, such as poly(3,4-ethylenedioxythiophene) (PEDOT) [12] or polyaniline (PANI) [13], can also be used to enhance the perceived transparency without affecting the photovoltaic efficiency of the DSSC device.

However, while modifications of glass substrates, semiconductor, and counter electrode can enhance the light transmittance in the overall visible spectrum, the human eye spectral sensitivity can be enhanced selectively only through modulating the light absorption spectrum of the DSSC photosensitizer. A possible approach is to use a dye cocktail in which two dyes with complementary light absorption properties, found in the UV and the NIR regions, respectively, sensitize the same photoanode to maintain high energy conversion efficiencies and transmit the light only in the intermediate wavelength range. This strategy was pursued by Han et al. [14], who combined the NIR-absorbing squaraine-based dye HSQ5 with the UV-absorbing dye Y1, obtaining green-colored DSSC devices with a maximum light transmittance of 60.3% in the 500–600 nm region and a photon-to-current efficiency of 3.66%. Hagar et al. [15] investigated the theoretical and experimental photo-electrochemical properties of a series of novel diimide and Schiff base UV-absorbers with carboxyl and pyridyl anchoring groups, evaluating their potential co-sensitization with the NIR-absorber VG20 in transparent DSSC. On the other side, Sauvage et al. [16] optimized the structure of a polymethine cyanine-based dye (VG20-C₁₆) to absorb only near-infrared light beyond 700 nm: the resulting colorless DSSC devices displayed a 3.1% power conversion efficiency and a remarkable average visible transmittance of up to 76%.

In the last few years, our research group gained significant experience in the design and synthesis of novel organic dyes for DSSC with light absorption properties tailored to reach high visible transparency [17,18] or attain specific colorations [19,20]. Moreover, we recently prepared a series of thieno[3,4-b]pyrazine-based molecules (TPz) as potential fluorophores in Luminescent Solar Concentrators (LSCs), which showed maximum absorbance well above 500 nm and a higher transmittance in the intermediate wavelength range [21,22]. Similar light absorption properties were displayed by the DSSC dyes featuring the TPz core recently reported by Delcamp et al. [23,24,25]. For this reason, we supposed that the TPz core could be used as the starting point to design novel organic dyes with optical properties highly compatible with the human eye sensitivity spectrum. In particular, we designed compound TP1 with an unsymmetric D-A-π-A structure and TP2 bearing a D-A-D structure and two anchoring moieties on the central core (Figure 1). In both cases, to modulate the photo-electrochemical properties of the dyes, we chose to introduce the N,N-bis(4-(hexylthio)phenyl)aniline as the donor group, since it was demonstrated to enhance dye regeneration and increase DSSC open circuit potential (V_OC) compared to its oxygenated analog [26,27].

Spectroscopic characterization of the resulting dyes showed divergent optical properties, which were selectively exploited to provide DSSC test devices with different predominant features: either enhanced photovoltaic performances, in the case of the more panchromatic absorber (TP1), or improved perceived transparency for that with the most localized absorption profile (TP2).

2. Materials and Methods

2.1. General Information

Unless otherwise stated, all reagents were purchased from commercial suppliers and used without purification. 2,3-diphenylthieno[3,4-b]pyrazine (1) [21], 4-bromo-N,N-bis(4-(hexylthio)phenyl)aniline (2) [26], and thiophene-3,4-diamine (8) [21] were prepared as previously reported. All air-sensitive reactions were performed using Schlenk techniques. Solvents used in cross-coupling reactions were previously degassed using the “freeze-pump-thaw” method. Toluene was dried on a resin exchange Solvent Purification System (MBraun), and ethanol was distilled over magnesium. Petroleum ether, unless specified, was the 40–70 °C boiling fraction. Reactions were monitored via TLC using silica gel 60 F254 aluminum sheet (Merck); the detection was made using a KMnO₄ basic solution or UV lamp. Flash column chromatography was performed using glass columns (10–50 mm wide) and Silica Gel 60 (230–400 mesh). ¹H- and ¹³C-NMR spectra were recorded at 400 and 100.6 MHz, respectively, on Bruker Avance series instruments. Chemical shifts were referenced to the residual solvent peak (CDCl₃, δ 7.26 ppm for ¹H-NMR and δ 77.16 ppm for ¹³C-NMR; CD₂Cl₂, δ 5.32 ppm for ¹H-NMR and δ 53.84 ppm for ¹³C-NMR; THF-d₈ δ, 1.72 and 3.58 ppm for ¹H-NMR, δ 67.21 and 25.31 ppm for ¹³C-NMR). Coupling constants (J) were reported in Hz. ESI-MS analyses were recoded with LCQ-Fleet Ion-Trap Mass Spectrometer (Thermo Fisher Scientific Italy, Milan, Italy). HR-MS analyses were performed at CISM (Mass Spectrometry Center—University of Florence) using an LTQ Orbitrap FT- MS Spectrometer (Thermo). FT-IR spectra were recorded with a Perkin-Elmer Spectrum BX instrument in the range 4000–400 cm⁻¹ with a 2 cm⁻¹ resolution. UV-Vis spectra in different solvents were recorded on diluted solutions of the analyte (approximately 10⁻⁵ M) with a Shimadzu UV-2600 spectrometer. UV-Vis absorption or transmittance spectra of the compounds adsorbed on TiO₂ were recorded with the same instrument in transmission mode after the sensitization of thin, transparent semiconductor films (thickness approximately 5 μm). Fluorescence spectra in solution were measured at room temperature with a Jasco FP-8300 spectrofluorometer equipped with a 450 W Xenon arc lamp.

2.2. Synthesis

2.2.1. Synthesis of 4-(7-(4-(bis(4-(Hexylthio)phenyl)amino)phenyl)-2,3-diphenylthieno[3,4-b]pyrazin-5-yl)benzaldehyde (4)

TPz 1 (66 mg, 0.23 mmol, 1.0 eq.), Pd(OAc)₂ (2.6 mg, 0.011 mmol, 5.0 mol%), P(t-Bu)₃HBF₄ (6.6 mg, 0.023 mmol, 10 mol%), pivalic acid (12 mg, 0.11 mmol, 50 mol%), Cs₂CO₃ (0.112 g, 0.34 mmol, 1.5 eq.) and bromide 2 (0.108 g, 0.19 mmol, 0.85 eq.) were dissolved in toluene (1.0 mL). The reaction mixture was stirred for 2.5 h at 90 °C; 4-bromo-benzaldehyde (3, 85 mg, 0.46 mmol, 2.0 eq.) and an additional portion of Cs₂CO₃ (0.112 g, 0.34 mmol, 1.5 eq.) were then added. The resulting mixture was stirred for 5 h at 100 °C, before being cooled down to room temperature and filtered over Celite^®. The filter cake was washed with chloroform (30 mL), and the organic phase was washed with brine (30 mL). After anhydrification over Na₂SO₄, filtration, and evaporation of the solvent under vacuum, the crude was purified via flash column chromatography (toluene) to afford pure 4 (42 mg, 0.052 mmol, 21% yield) as a dark purple solid. ¹H–NMR (400 MHz, CD₂Cl₂): δ = 9.97 (s, 1H), 8.48 (d, J = 8.4 Hz, 2H), 8.18 (d, J = 8.8 Hz, 2H), 7.91 (d, J = 8.4 Hz, 2H), 7.54 (d, J = 7.3 Hz, 2H), 7.49 (d, J = 7.0 Hz, 2H), 7.29–7.41 (m, 6H), 7.26 (d, J = 8.4 Hz, 4H), 7.10 (d, J = 8.8 Hz, 2H), 7.07 (d, J = 8.4 Hz, 4H), 2.91 (t, J = 7.3 Hz, 4H), 1.66 (q, J = 7.3 Hz, 4H), 1.38–1.48 (m, 4H), 1.28–1.35 (m, 8H), 0.90 (t, J = 6.8 Hz, 6H) ppm; ¹³C–NMR{1H} (100 MHz, CD₂Cl₂): δ = 191.8, 153.9, 153.0, 148.2, 145.5, 140.8, 139.8, 139.7, 139.1, 135.21, 135.17, 132.3, 131.0, 130.6, 130.4, 130.3, 129.6, 129.4, 129.3, 128.7, 128.6, 127.9, 127.7, 127.2, 125.9, 125.6, 123.1, 34.7, 31.9, 29.8, 29.0, 23.1, 14.4 ppm; ESI–MS: m/z = 867.3 [M]⁺.

2.2.2. Synthesis of (E)-3-(4-(7-(4-(bis(4-(Hexylthio)phenyl)amino)phenyl)-2,3-diphenylthieno[3,4-b]pyrazin-5-yl)phenyl)-2-cyanoacrylic Acid (TP1)

Aldehyde 4 (45 mg, 0.052 mmol, 1.0 eq.) was reacted with cyanoacetic acid (13 mg, 0.155 mmol, 3.0 eq.) and piperidine (31 mg, 36 μL, 0.364 mmol, 3.0 eq.) in CHCl₃ (2.0 mL) for 8 h at 80 °C; the heating was then interrupted. The reaction mixture was diluted with CHCl₃ (20 mL) and washed with HCl_(aq.₎ 0.3 M (20 mL); after anhydrification of the organic phase over Na₂SO₄, the solvent was removed under vacuum. The crude was purified through washing with portions of ethanol (2 × 10 mL) and pentane (2 × 10 mL), affording TP1 in pure form (41 mg, 0.044 mmol, 84% yield) as a green–black solid. ¹H–NMR (400 MHz, THF-d₈): δ = 8.59 (d, J = 8.4 Hz, 2H), 8.27–8.36 (m, 2H), 8.25 (s, 1H), 8.16 (d, J = 8.4 Hz, 2H), 7.57 (d, J = 6.6 Hz, 2H), 7.53 (d, J = 7.7 Hz, 2H), 7.20–7.41 (m, 11H), 7.04–7.15 (m, 5H), 2.92 (t, J = 7.3 Hz, 4H), 1.65 (q, J = 7.3 Hz, 4H), 1.41–1.51 (m, 4H), 1.29–1.37 (m, 8H), 0.87–0.93 (m, 6H) ppm; ¹³C–NMR{1H} (100 MHz, THF-d₈): δ = 163.8, 154.3, 153.5, 153.4, 148.6, 145.8, 141.2, 140.2, 140.1, 139.5, 138.8, 135.1, 132.9, 132.2, 131.2, 130.6, 130.5, 129.6, 129.5, 128.7, 128.6, 128.0, 127.6, 126.8, 126.1, 124.3, 124.2, 123.1, 116.4, 103.3, 34.5, 32.2, 30.0, 29.2, 23.3, 14.2 ppm; HRMS (ESI): m/z calculated for C₅₈H₅₅N₄O₂S₃: 935.34817 [M]⁺. Found: 935.34815.

2.2.3. Synthesis of Dimethyl 4,4′-(1-Hydroxy-2-oxoethane-1,2-diyl)dibenzoate (6) [28]

Thiamine hydrochloride (0.24 g, 0.71 mmol, 5.0 mol%.) was dissolved in a H₂O:MeOH = 1:3 mixture (8.7 mL) and a 2.0 M solution of aqueous NaOH (0.6 mL) was added dropwise at 0 °C. The colorless solution turned bright yellow, and methyl 4-formylbenzoate (5) (2.00 g, 12.18 mmol, 1.00 eq.) was added. The suspension was heated to 80 °C and stirred for 1 h. The reaction mixture was cooled down to RT and quenched with H₂O (6.0 mL) to precipitate the product (6) as a white solid, which was filtered and washed with H₂O and MeOH (1.56 g, 4.75 mmol, 78% yield). Spectroscopic data were in agreement with those already reported [28]. ¹H–NMR (400 MHz, CDCl₃) δ = 8.01 (d, J = 8.4 Hz, 2H), 7.94 (d, J = 8.3 Hz, 2H), 7.90 (d, J = 8.4 Hz, 2H), 7.38 (d, J = 8.3 Hz, 2H), 6.00 (s, 1H), 4.60 (br, 1H), 3.87 (s, 3H), 3.83 (s, 3H) ppm; ¹³C–NMR{1H} (100 MHz, CDCl₃) δ = 198.3, 166.4, 165.8, 143.1, 136.7, 134.7, 130.5, 129.9, 129.0, 127.8, 76.2, 52.6, 52.3 ppm.

2.2.4. Synthesis of Dimethyl 4,4′-Oxalyldibenzoate (7) [29]

In a round-bottom flask equipped with a reflux condenser, benzoin derivative 6 (1.56 g, 4.75 mmol, 1.0 eq.) was dissolved in DMSO (26.7 mL), and HBr 48% (9.34 mL, 8.25 mmol, 1.74 eq.) was added. The yellow solution was stirred for 16 h at 55 °C, and the reaction progress was controlled via TLC (PE:EtOAc = 1:1). The bright yellow suspension was cooled down to RT, quenched with H₂O (50.0 mL), and filtered. The bright yellow precipitate was suspended in H₂O and filtered again to afford product 7 (1.53 g, 4.70 mmol, 99% yield), with no need for further purification. Spectroscopic data were in agreement with those already reported [29]. ¹H–NMR (400 MHz, CDCl₃) δ = 8.18 (d, J = 7.5 Hz, 4H), 8.05 (d, J = 7.4 Hz, 4H), 3.96 (s, 6H) ppm; ¹³C–NMR{1H} (100 MHz, CDCl₃) δ = 193.0, 166.0, 135.9, 135.7, 130.3, 130.0, 52.8 ppm.

2.2.5. Synthesis of Dimethyl 4,4′-(Thieno[3,4-b]pyrazine-2,3-diyl)dibenzoate (9) [30]

Oxalyldibenzoate (7) (0.15 g 0.46 mmol, 1.00 eq.) and thiophene-3,4-diamine (8) (54 mg, 0.48 mmol, 1.05 eq.) were dissolved in dry EtOH (4.0 mL) under an inert atmosphere of N₂. The resulting suspension was stirred and refluxed under an inert atmosphere for 16 h, monitoring the reaction progress via TLC (PE:EtOAc = 1:1). The resulting dark green suspension was cooled down to RT and filtered off. The solid was washed with EtOH and purified via short column chromatography (gradient from DCM to DCM:MeOH = 100:1) to obtain 9 (0.145 g, 0.36 mmol, 78% yield) as a dark green powder. Spectroscopic data were in agreement with those already reported [30]. ¹H–NMR (400 MHz, CDCl₃) δ = 8.13 (s, 2H), 7.98 (d, J = 8.0 Hz, 4H), 7.50 (d, J = 8.1 Hz, 4H), 3.92 (s, 6H) ppm; ¹³C–NMR{1H} (100 MHz, CDCl₃) δ = 166.7, 152.1, 143.2, 141.6, 130.6, 129.8, 129.7, 118.5, 52.4 ppm.

2.2.6. Synthesis of Dimethyl 4,4′-(5,7-bis(4-(bis(4-(Hexylthio)phenyl)amino)phenyl)thieno[3,4-b]pyrazine-2,3-diyl)dibenzoate (10) [21]

TPz 9 (30 mg, 0.07 mmol, 1.00 eq.), bromide 2 (103 mg, 0.17 mmol, 2.50 eq.), Pd(OAc)₂ (2.4 mg, 0.01 mmol, 15 mol%), P(tBu)₃HBF₄ (6 mg, 0.02 mmol, 30 mol%), Cs₂CO₃ (84 mg, 0.25 mmol, 3.7 eq.), and pivalic acid (3.7 mg, 0.036 mmol, 50 mol%) were dissolved in dry N,N-DMF (0.3 mL). The brown mixture was warmed up to 90 °C and maintained under magnetic stirring for 16 h, during which time the reaction progress was monitored via TLC (PE:EtOAc = 4:1). The dark blue mixture was diluted with DCM, filtered on Silica gel, and purified via flash column chromatography (gradient from PE:DCM = 1.1 to pure DCM) to obtain the desired product 10 (69 mg, 0.05 mmol, 73% yield) as a dark green solid. ¹H–NMR (400 MHz, THF-d₈) δ = 8.25 (d, J = 8.8 Hz, 4H), 7.95 (d, J = 8.3 Hz, 4H), 7.62 (d, J = 8.3 Hz, 4H), 7.27 (d, J = 8.6 Hz, 8H), 7.11 (d, J = 8.8 Hz, 4H), 7.06 (d, J = 8.6 Hz, 8H), 3.85 (s, 6H), 2.91 (t, J = 7.3 Hz, 8H), 1.69–1.60 (m, 8H), 1.52–1.39 (m, 8H), 1.36–1.28 (m, 18H), 0.90 (t, J = 6.8 Hz, 12H) ppm; ¹³C–NMR{1H} (100 MHz, THF-d₈) δ = 166.7, 152.3, 148.4, 146.3, 144.7, 139.4, 132.9, 131.7, 131.6, 130.9, 130.2, 129.5, 128.4, 126.2, 123.8, 52.3, 34.9, 32.5, 30.4, 29.5, 23.6, 14.5 ppm.

2.2.7. Synthesis of 4,4′-(5,7-bis(4-(bis(4-(Hexylthio)phenyl)amino)phenyl)thieno[3,4-b]pyrazine-2,3-diyl)dibenzoic acid (TP2) [24]

Ester 10 (55 mg, 0.04 mmol, 1.00 eq.) was dissolved in THF (20.0 mL), and a freshly prepared 2 M solution of LiOH·H₂O in MeOH (20.0 mL) was added. The dark green thick solution was stirred vigorously for 2 h at RT. The resulting dark blue mixture was poured into ice-cold water, and the dark blue suspension was neutralized with a 3 M HCl aqueous solution till pH = 5, at which point a blue precipitate formed. The solid was recovered via filtration and washed with water, affording the desired product TP2 (44 mg, 0.03 mmol, 83% yield) as a dark green solid. ¹H–NMR (400 MHz, THF-d₈) δ = 8.24 (d, J = 8.6 Hz, 4H), 7.96 (d, J = 8.2 Hz, 4H), 7.60 (d, J = 8.2 Hz, 4H), 7.27 (d, J = 8.5 Hz, 8H), 7.10 (d, J = 8.6 Hz, 4H), 7.06 (d, J = 8.4 Hz, 8H), 2.91 (t, J = 7.3 Hz, 8H), 1.69–1.60 (m, 8H), 1.49–1.40 (m, 8H), 1.37–1.27 (m, 16H), 0.90 (t, J = 6.7 Hz, 12H) ppm; ¹³C–NMR{1H} (100 MHz, THF-d₈) δ = 167.3, 152.4, 148.3, 146.3, 144.4, 139.4, 132.8, 132.2, 131.6, 130.8, 130.4, 129.5, 128.5, 126.2, 123.8, 34.9, 32.5, 30.4, 29.5, 23.6, 14.6 ppm; HRMS (ESI): m/z calculated for C₈₀H₈₇N₄O₄S₅: 1327.53254 [M]⁺. Found: 1327.53135.

3. Results and Discussion

3.1. Synthesis of the Dyes

The synthetic pathways followed to prepare the two dyes are reported in Scheme 1. The synthesis of dye TP1 required the employment of diphenyl-substituted TPz 1 and 4-bromo-N,N-bis(4-(hexylthio)phenyl)aniline (2), which were prepared as previously reported by our group [21,26]. Thus, TPz 1 was subjected to a one-pot double arylation reaction (DA) to break the symmetry of the molecule and introduce the donor group (via reaction with 2) and, after that, the acceptor group (via reaction with commercially available bromide 3). The reaction conditions (stoichiometric ratio of the reagents, the choice of pre-catalyst, ligand, and base, as well as temperature and time) were adapted from our previous reports of one-pot DA reactions for the construction of other DSSC dyes [19,20], allowing us to isolate the desired D-A-π-A intermediate 4 with a 21% overall yield. The overall synthesis was carried out in just one step, resulting in a much shorter sequence and a reduced production of waste in comparison with classic cross-coupling protocols, such as Suzuki–Miyaura and Stille–Migita reactions [31].

Regarding the preparation of TP2, access to 1,2-diketone 7 was required, since it is not commercially available. Unlike the process reported in the literature [32], we were not able to prepare 7 via palladium-catalyzed cross-coupling on vinylene carbonate acting as a synthetic equivalent of glyoxal; therefore, some alternative routes were investigated. Initially, we attempted a cyanide-free benzoin umpolung self-condensation of commercially available methyl 4-formylbenzoate (5) in a basic environment. When the reaction was catalyzed with ionic liquid BMII, using ultrasounds to obtain a homogeneous solution [33,34], the desired product 6 was obtained. Unfortunately, this process yielded unreproducible yields ranging from 6 to 81%, probably due to the non-homogeneity of the ultrasounds mixing method. On the other hand, the thiamin-catalyzed cyanide-free benzoin condensation of benzoate 5 gave more reproducible results, producing the desired α-hydroxyketone 6 in good yield [28]. Acid-catalyzed oxidation of compound 6 with DMSO gave the desired dicarbonyl intermediate 7 in quantitative yield [29,35]. Condensation of 7 with diamine 8 in dry ethanol afforded TPz 9 with high yield, followed by a DA with bromide 2, carrying the donor group, while quantitative ester hydrolysis under basic conditions allowed us to isolate dye TP2 in 83% yield. The yield of the overall five-step TP2 synthesis is 36%.

3.2. Spectroscopic and Electrochemical Characterization of Dyes TP1–2

Both dyes TP1 and TP2 were completely characterized from a spectroscopic and electrochemical point of view (see Table 1). Firstly, UV-Vis spectra of the dyes were recorded in CHCl₃ solvent (conc. ≈ 10⁻⁵ M; Figure 2a,b). TP1 solution showed a dark purple color and almost panchromatic light absorption, with three main peaks: the first one was centered in the green region of the spectrum (λ_max = 572 nm), and the other two were located in the blue (λ_max = 420 nm) and UV (λ_max = 337 nm) regions. Molar attenuation coefficients (ε) of the peaks in the visible region fall in the 1.8–1.9 × 10⁴ M⁻¹ cm⁻¹ range, while the UV-absorption intensity is higher (ε = 4.10 × 10⁴ M⁻¹ cm⁻¹). TP2 solution is characterized by a green coloration, due to a fairly intense absorption peak in the 500–700 nm region (λ_max = 597 nm, ε = 3.70 × 10⁴ M⁻¹ cm⁻¹) and a broad UV-Vis band (λ < 450 nm) with a high molar attenuation coefficient, up to 3.0 × 10⁵ M⁻¹ cm⁻¹. These light absorption properties were confirmed through the transmittance spectra of the semi-transparent thin-layer TiO₂ electrodes (thickness: 5 μm) dyed with TP1 and TP2 (Figure 2c), with relative minimum transparencies in the visible region, i.e., among 550 and 600 nm and below 400 nm. The overlay with the CIE photopic spectral luminous efficiency function [36], which represents the average spectral sensitivity of human visual perception of light, showed a significant overlapping with the transmittance spectrum of TP1, while TP2 exhibited a much better matching. The average visible light transmission (AVT) through the photoanode was calculated based on the following equation: [37]

$$AVT\left(\%\right)=\frac{{\int }_{390}^{830}T\left(\lambda \right)\times V\left(\lambda \right)\times S\left(\lambda \right)d\lambda }{{\int }_{390}^{830}V\left(\lambda \right)\times S\left(\lambda \right)d\mathit{\lambda }}$$

where T(λ) is the transmission of the sample, V(λ) is the photopic spectral luminous efficiency function, and S(λ) is the AM 1.5 g solar spectrum (ASTM G-173 Global 37 Deg Tilt standard) [38]. While TP1 photoanode reached 42.3% AVT, TP2 showed a very high AVT of 83.7%, thus appearing as a very promising candidate as a photosensitizer for highly transparent DSSCs in BIPV. This trend in AVT values could be in contrast with the fact that TP2 has a molar attenuation coefficient (ε) in solution larger than that of TP1. The explanation for such opposite behavior can be found in the mismatch of the light absorption spectra of TP2 and the human eye sensitivity V(λ), while TP1 has a maximum absorption that perfectly matches with V(λ). Moreover, it is worth noticing that, despite a larger molar attenuation coefficient in solution, the TP2 transmittance on the TiO₂ layer is higher than that of TP1. This discrepancy could be related to a lower amount of dye anchored onto the semiconductor surface, since each rigid dicarboxylic TP2 molecule occupies at least double as many active sites of the semiconductor layer with respect to TP1.

To complete the spectroscopic characterization, Tauc plots were drawn, starting from the UV-Vis data in CHCl₃ solution (Figures S1 and S2) to estimate the optical bandgaps (E_0–0) of the dyes, whose values were equal to 1.91 and 1.82 eV for TP1 and TP2 respectively, confirming their ability to also absorb light in the visible/near-IR regions. Cyclic voltammetries (CV) of dyes TP1 and TP2 were recorded in CH₂Cl₂ and THF solutions, respectively, using 0.1 M TBAPF₆ as supporting electrolyte and ferrocene as internal standard [39] to evaluate the oxidation potentials of the dyes at both the ground and excited states (see Figures S3 and S4, and Table 1). Both molecules showed quasi-reversible first oxidation peaks and very similar ground-state oxidation potentials (E_ox), which were equal to 1.01 and 1.04 V for TP1 and TP2, respectively. These values should be positive enough to guarantee dye regeneration by the most common redox couples, such as I⁻/I₃⁻, Co^2+/3+, and Cu^+/2+ complexes. Excited-state oxidation potentials (E_ox*) were obtained subtracting E_0–0 to E_ox; while the −0.90 V value calculated for TP1 should be negative enough to ensure a fast electron injection into the TiO₂ conduction band, TP2 exhibited a lower E_ox* (−0.78 V), which could make the electron transfer from the excited state of the dye more troublesome (see Figure S5) [40,41].

Table 1. Spectroscopic and electrochemical data of dyes TP1 and TP2.

	λabs [nm] (ε × 104 [M−1 cm−1]) a	λabs on TiO2 [nm]	AVT [%]	E0–0 [eV] b	Eox [V] c	Eox* [V] d
TP1	572 (1.83); 420 (1.88); 337 (4.10)	557	42.3	1.91	1.01	–0.90
TP2	597 (3.70); 306 (30.49)	591	83.7	1.82	1.04	–0.78

^a In CHCl₃ solution. ^b Calculated from Tauc plots. ^c Working electrode: glassy carbon; counter-electrode: platinum; reference electrode: Ag/AgNO₃ 0.01 M in MeCN; supporting electrolyte: TBAPF₆ 0.1 M in DCM (TP1) or THF (TP2); ferrocene as the internal standard [40]. Values are reported against NHE. ^d E_ox* = E_ox. − E_0–0.

Table 1. Spectroscopic and electrochemical data of dyes TP1 and TP2.

	λabs [nm] (ε × 104 [M−1 cm−1]) a	λabs on TiO2 [nm]	AVT [%]	E0–0 [eV] b	Eox [V] c	Eox* [V] d
TP1	572 (1.83); 420 (1.88); 337 (4.10)	557	42.3	1.91	1.01	–0.90
TP2	597 (3.70); 306 (30.49)	591	83.7	1.82	1.04	–0.78

^a In CHCl₃ solution. ^b Calculated from Tauc plots. ^c Working electrode: glassy carbon; counter-electrode: platinum; reference electrode: Ag/AgNO₃ 0.01 M in MeCN; supporting electrolyte: TBAPF₆ 0.1 M in DCM (TP1) or THF (TP2); ferrocene as the internal standard [40]. Values are reported against NHE. ^d E_ox* = E_ox. − E_0–0.

3.3. Photovoltaic Characterization of Dyes TP1–2

The prepared dyes TP1–2 were employed as sensitizers in the construction of small DSSC test devices to evaluate their photovoltaic properties. All the details on the device construction are reported in the supplementary information (see Section S4). Aiming for devices with high perceived transparency, all prepared samples featured a thin semi-transparent screen-printed layer of TiO₂ as the semiconductor for the photoanode and a semitransparent screen-printed Pt layer as the cathode.

The prepared dyes and a reference compound (TTZ5, Figure S6) previously reported by our group [42] were tested together with the commercially available iodine-based electrolyte solution EL-HSE (GreatCell Solar, entries labeled HSE in

Table 2. Photovoltaic characterization of DSSC devices sensitized with TP1–2 or reference dye TTZ5 and featuring iodine-based electrolytes.

Dye	Electrolyte	JSC [mA cm−2]	VOC [mV]	ff [%]	η [%]
TP1	HSE	10.38 ± 0.27	680 ± 6	73.0 ± 0.5	5.16 ± 0.12
	LIE	8.44 ± 0.48	695 ± 9	72.5 ± 2.9	4.24 ± 0.04
TP2	HSE	8.67 ± 0.41	680 ± 5	68.4 ± 1.4	4.03 ± 0.14
	LIE	6.97 ± 0.38	689 ± 3	69.6 ± 1.5	3.35 ± 0.19
TTZ5	HSE	14.08	711	74.4	7.44
	LIE	13.48	751	69.0	6.98

All devices were measured two days after construction. Values for TP1–2 samples are reported as average of three samples with their standard deviation.

and Figure 3a). Aiming again for high-transparency devices, other electrolytic solutions characterized by a weaker visible light absorption spectrum and/or a lower molar attenuation coefficient were also tested (entries labeled LIE in Table 2 and Figure 3a). In particular, an electrolyte solution with low-iodine content [43] was formulated via mixing iodide salts with no iodine (details on the electrolyte formulation are reported in Section S4). Indeed, a low amount of iodine, which was necessary to collect charges at the cathode, was produced in situ using the photo-oxidated dye once the DSSC devices were assembled and exposed to simulated sunlight. Despite that, the overall transparency of the device was preserved and a remarkable transmittance value, higher than 30% in almost the whole visible spectrum, was obtained for a test DSSC device containing the TP2 dye and the LIE electrolyte (see Figure S7).

The photovoltaic performances of all devices improved constantly for a few days, during which time it is believed the electrolytes intercalated into the mesoporous structure of the semiconductor and a small amount of iodine (triiodide) was produced in situ in the low-iodine content electrolyte to further improve device efficiency. Table 2 and Figure 3a report the photovoltaic characterization on the second day after device construction, and clearly show how all samples benefit from the use of the low-iodine content electrolyte, in terms of not only transparency but also V_OC. This behavior is in line with a lower concentration of the oxidized species in the electrolyte, making its redox potential more positive and, thus, enlarging its gap compared to the Fermi level of the semiconductor, which correlates with the V_OC of the device. On the other hand, the short-circuit photocurrent density (J_SC) decreases when replacing HSE with LIE electrolyte, probably due to the mass transport limitation [43]. Interestingly, the changes in J_SC and power conversion efficiency (η%) across different dyes follow the same trend of the light-harvesting properties of the dyes employed. Despite the loss in terms of η%, which was limited to 16–17% of the initial efficiency value for TP1–2, the employment of the LIE electrolyte can be encouraged for BIPV applications; this finding is especially relevant when TP2 is used as the dye, since the increment in terms of device transparency is remarkable (see Figure S7). To quantify such added value, we recorded the transparency of the active area of full devices sensitized with TP2 and calculated an AVT value of 27.5% when using HSE as the electrolyte, compared to 36.9% when using LIE. The product between AVT and η is best known as the light utilization efficiency (LUE) and can be used to better evaluate the photovoltaic device, simultaneously taking into account both the perceived transparency and the energy conversion efficiency [37]. The LUE value for TP2-HSE was computed to be 1.42, while TP2-LIE devices reached 1.57, proving that the gained transparency justifies the slight loss in power conversion efficiency.

Furthermore, dyes TP1–2 and the reference sensitizer TTZ5 were also tested with three electrolyte solutions based on cobalt complexes with bi- and tri-dentate nitrogen-containing the following ligands: [44,45] 2,2′-bipyridine (Co(bpy)₃), phenanthroline (Co(phen)₃), and 6-(3,5-dimethylpyrazol-1-yl)-2,2′-bipyridine (Co(pzbpy)₂) (Figure 4). The corresponding J/V curves are reported in Figure 3b–d.

Other than a light absorption spectrum located almost only in the UV region (see Figure S8), such electrolytes are also characterized by progressively more positive redox potentials (0.565, 0.614, and 0.767 V vs. NHE respectively) and were, therefore, anticipated to enhance the V_OC of the devices. Indeed, as shown in

Table 3. Photovoltaic characterization of DSSC devices fabricated with dyes TP1–2 or reference sensitizer TTZ5 and three cobalt-based electrolytes.

Dye	Electrolyte	JSC [mA cm−2]	VOC [mV]	ff [%]	η [%]
TP1	Co(bpy)3	7.37 ± 0.52	706 ± 8	75.3 ± 1.8	4.00 ± 0.30
	Co(phen)3	7.20 ± 0.44	770 ± 3	72.4 ± 4.4	4.01 ± 0.36
	Co(bpypz)2	3.55 ± 0.15	844 ± 1	73.6 ± 3.0	2.21 ± 0.13
TP2	Co(bpy)3	4.58 ± 0.21	721 ± 6	75.9 ± 3.7	2.46 ± 0.18
	Co(phen)3	3.59 ± 0.15	731 ± 4	77.5 ± 3.3	2.03 ± 0.05
	Co(bpypz)2	1.20 ± 0.19	806 ± 11	78.3 ± 1.4	0.75 ± 0.10
TTZ5	Co(bpy)3	6.13 ± 0.32	640 ± 7	74.5 ± 1.5	2.92 ± 0.08
	Co(phen)3	6.71 ± 0.45	697 ± 14	75.1 ± 1.0	3.51 ± 0.32
	Co(bpypz)2	4.64 ± 0.29	817 ± 8	72.1 ± 2.0	2.74 ± 0.27

All devices were measured two days after construction. Values for TP1–2 samples are reported as average of three samples with their standard deviation.

, the expected progressive increase in V_OC moving from Co(bpy)₃ to Co(pzbpy)₂ was observed, but was also accompanied by a relevant loss in J_SC, which was generally attributed to a slower regeneration reaction due to a lower driving force for the reaction [46]. In fact, while TP1 features the well-established D-A-π-A structure, which is known to separate the photoinduced charges efficiently, therefore stunting the TiO₂-dye charge recombination, TP2 instead has a branched symmetrical D-A-D structure that seems to be less efficient in separating the charges and, when used in combination with electrolytes with a very low regeneration overpotential, gives rise to TiO₂-dye recombination phenomena with a rate that is competitive with the one of dye–electrolyte regeneration. Similarly, the potential of Co(pzbpy)₂ seems to be excessively positive, leading to dye regeneration and/or charge recombination issues for both TP1–2. This issue is particularly evident with TP2, whose drop in photovoltaic efficiency is solely imputable to a drastic reduction in J_SC, which is symptomatic of slow dye regeneration. Interestingly, when used with Co(phen)₃ transparent electrolyte, dye TP1 still shows good power conversion efficiencies of 4%, overcoming the reference compound TTZ5.

4. Conclusions

In this work, two novel thieno[3,4-b]pyrazine-based organic sensitizers TP1–2 for DSSC were synthesized, characterized, and tested for photovoltaic performance. TP1 features a D-A-π-A design, while TP2 has a symmetric structure with double anchoring functions. The dyes displayed divergent optical properties: TP1 has a broad and intense light absorption in the whole visible spectrum, resulting in an average transparency (AVT) equal to 42.3%, while TP2 presents two localized absorption bands with AVT of 83.7%, suggesting, therefore, a feasible application in BIPV. Cells built with TP1–2 show appreciable photovoltaic performance with iodide-based electrolytes (5.16% and 4.03%, respectively), and the transparency of the whole devices could be improved through exploiting transparent electrolytes with low-iodine formulations, at the cost of a slight efficiency loss (less than 20%). Tests with cobalt-based electrolytes, i.e., Co(bpy)₃, Co(phen)₃, and Co(pzbpy)₂, highlighted a limited compatibility of TP1–2 with redox couples characterized by potentials as high as that of Co(pzbpy)₂ (0.767 V vs. NHE). On the other hand, devices showed a large gain in V_OC compared to those built with the triiodide/iodide-based electrolytes, especially in the case of TP1, which, in combination with Co(phen)₃ reached a PCE of 4.01% with a V_OC of 770 mV, overcoming the reference dye TTZ5 in the same transparent configuration.