A Comparative Study of Organic Dye-Sensitized Solar Cells Based on Anatase TiO₂ and Amorphous Free Mixed Phase’s Anatase/Rutile P25 TiO₂ Photoanodes

Abstract

Dye-sensitized solar cells (DSCs) remain an interesting photovoltaic concept, although recent times have seen their envisioned broad-scale applications being replaced with more niche ones. Nevertheless, as a key component of DSCs, titanium(IV) oxide (TiO₂) must be produced in a large volume, low cost, and highly reproducible manner. Degussa P25 remains a benchmark TiO₂ product, addressing the first two of the above points very well. Post-treatment processes that may also be carried out on a large scale give some hope to addressing the reproducibility issue. This paper builds on our previous works wherein mixed-phase P25 (anatase + rutile + amorphous TiO₂) was converted into an amorphous free form by selectively dissolving and recrystallizing the amorphous component. Here we investigated the performance of metal-free organic dye (D149)-based DSCs with three different TiO₂ films: (1) as-received P25 (TiO₂-P25), (2) amorphous-free P25 (TiO₂-HP25), and (3) anatase nanoparticles obtained from Dyesol (TiO₂-DSL). DSCs based on TiO₂-HP25 showed comparable performance (5.8 ± 0.2% PCE) to DSCs based on the TiO₂-DSL (5.8 ± 0.4% PCE) and substantially higher than for devices based on the as-obtained P25 nanoparticles (3.9 ± 0.4% PCE). The enhancement resulting from the post-processing of P25 derives from simultaneous increases in photo-current density (J_sc), open-circuit voltage (V_OC), and the fill factor (FF), due to enhancing the dye-loading capability and the charge-transport efficiency (suppressing the electron recombination) as a result of the removal of amorphous barriers and associated defect states. This is in line with enhancing DSC performance based on the organometallic N719 dye we reported previously. However, the photoanode material based on abundant P25 TiO₂ sensitized with high-extinction-coefficient organic D149 dye can be adopted as a cost-effective DSC as an alternative to relatively high-cost DSCs based on the commercial anatase TiO₂ sensitized with organometallic N719 dye.

1. Introduction

Dye-sensitized solar cells (DSCs) are a second-generation photovoltaic technology first reported and developed by O’Regan and Gratzel in 1991 [1,2]. DSCs offer the prospect of high-volume production, as their manufacture is amenable to roll-to-roll coating processes [3]. In recent years the cost of silicon-based photovoltaic (PV) panels has dropped substantially, limiting interest in further development of DSCs for applications such as utility-scale PV. On the other hand, DSCs are well suited to domestic and indoor applications due to their colourful designs, semitransparency, and ability to retain operating efficiencies even in lowdiffuse-light conditions as well as under partial shading [4,5,6,7]. Furthermore, the judicious selection of organic components allows for spectral matching, as is desired with artificial light sources.

A typical DSC consists of a porous-oxide semiconductor photoanode, an organic/organometallic sensitizer, a redox-active electrolyte, and a non-photoactive cathode [1,2]. The photoanode is typically made with 2–20 micrometer-thick film semiconductor nanoparticles (such as titanium dioxide (TiO₂)) deposited on a transparent conducting oxide-coated glass (TCO). The loading and absorption coefficient of the sensitizer largely dictates the thickness of TiO₂ film required to attain high light-harvesting efficiency. One other significant factor here is the prevention of charge recombination, which scales with the interfacial area between the semiconductor and the electrolyte (typically either iodide-triiodide or cobalt(II/III) based electrolyte) and hence becomes a more substantial problem as film thickness increases. The aforementioned mediator is regenerated at a catalytic surface such as platinum [1,2].

Around 40% of scientific publications on DSCs have focused on optimizing photoanode materials [1,2], with the highest efficiencies reported for devices using pure-anatase TiO₂. We previously synthesized high-surface-area anatase TiO₂ nanoparticles with a higher isoelectric point (IEP) compared to the commercial anatase TiO₂ (Dyesol). These synthesized anatase nanoparticles showed higher photovoltaic conversion efficienct when applied as a photoanode compared to commercial Dyesol anatase nanoparticles when using the organic dye D149. The enhancement was attributed to the higher dye loading due to the higher IEP of the synthesized anatase TiO₂ nanoparticles. [8]. On the other hand, the commercially available product P25 from Degussa, a mixture of the anatase and rutile phases of TiO₂, along with some amorphous content, is an attractive alternative due to the simplicity and scale of its synthesis.

DSC performances using P25 are usually quite poor in comparison with the aforementioned pure anatase materials [9], and it has routinely been argued that low surface area and the more positive conduction-band edge of rutile are responsible for the lower photocurrent and voltage, respectively. On the other hand, studies by both Park and Lin showed that rutile TiO₂, with a comparable surface area to anatase TiO₂, achieved a similar performance and as such the rutile component is not itself the reason for the low efficiencies seen for P25-based DSCs [10,11]. Dye loading/binding on TiO₂ nanoparticles is known to differ according to the exposed facet [12]. Amorphous TiO₂ has a flat surface without distinct facets [13]. This morphology leads to poor dye packing, which both decreases the total light harvesting and provides opportunities for charge recombination between the mediator and TiO₂ [14,15,16].

Recognizing this issue, P25 has been modified using a hydrothermal process to selectively convert the amorphous TiO₂ to a crystalline form. This amorphous-free P25 (H-P25) was first applied as a photocatalyst and showed a significant enhancement in activity compared to untreated P25 [17]. More recently we applied this material as a photoanode sensitized with commonly used organometallic N719 dye, seeing a ~65% enhancement in photovoltaic efficiency compared to devices based on untreated P25 [18]. Other authors have reported alternate approaches to the removal of amorphous TiO₂, such as Kurian et al., who used a water-treatment approach, which also led to enhanced DSC efficiencies [19]. Mohammadi et al. compared DSC performance based on either anatase TiO₂ or rutile TiO₂ or composite anatase/rutile nanoparticles as photoanode materials [20]. Anatase TiO₂-based photoanode sensitized with N719 dye showed higher performance than rutile TiO₂, whereas the composite anatase/rutile photoanode showed the highest performance. It can be concluded that the commercially available mixed anatase/rutile TiO₂ (P25) can be adopted as a cost-effective photoanode material based on organic DSCs.

It is noted that for the two studies mentioned above (ours and Kurian et al.’s) ruthenium-based sensitizers were used. These have low extinction coefficients (<2 × 10⁴ M⁻¹ cm⁻¹), which, when combined with the lower surface area of TiO₂-P25, limit the light harvesting and hence the photocurrent. On the other hand, organic dyes typically offer high extinction coefficients (up to 2 × 10⁵ M⁻¹ cm⁻¹). Here we see that using such dyes helps to alleviate this issue. Meanwhile, organic dyes also have other attractive attributes, such as being comprised solely of earth-abundant elements, thus minimizing uncertainties around the future supply of precursors for their synthesis and having more readily tuneable colours to match light sources/aesthetic considerations [21,22,23]. In this work we compare DSC performance based on three photoanode materials: as-received P25 nanoparticles (TiO₂-P25), amorphous-free TiO₂ P25 (TiO₂-HP25), and as-received Dyesol NR18-T TiO₂ nanoparticles (TiO₂-DSL), all sensitized with D149, a model high extinction coefficient-metal free organic dye.

2. Experiment

2.1. Materials Synthesis

Commercially sourced P25 was hydrothermally treated by our group as previously reported [17,18]. In brief, tetrapropylammonium hydroxide (TPA, 40% in water) (Tokyo Chemical Industry Co., Ltd., Tokyo, Japan), distilled water, ammonium fluoride (NH₄F) (Kanto Chemical Co., Inc., Tokyo, Japan), and P25 (Nippon Aerosil, Tokyo, Japan) were mixed at a molar ratio of 5:4:25:1 (P25:TPA:H₂O:NH₄F). This was then hydrothermally treated for 7 days at 170 °C using an autoclave. TiO₂-HP25 was obtained after washing three times with distilled water and ethanol using a centrifuge before being dried at 90 °C for three hours. TiO₂-DSL was supplied as prepared from Dyesol, Queanbeyan, NSW, Australia, as a paste product (NR18-T). TiO₂-HP25 and TiO₂-P25 photoanodes were deposited as follows: TiO₂-HP25 or TiO₂-P25 powder (1.0 g) was ground in Milli-Q water(1.0 mL, Sartorius Stedim Biotech, Victoria, Australia), absolute ethanol (25 mL, Sigma, Castle Hill, NSW, Australia), and acetic acid (0.2 mL, Sigma, Castle Hill, NSW, Australia). An ethyl cellulose (Sigma, Castle Hill, NSW, Australia) solution was prepared in ethanol (10% w/w) and was added to the nanoparticle suspension, along with terpineol (5.0 g, Sigma, Castle Hill, NSW, Australia). This was then stirred and sonicated at 50 °C for 1.0 h. Following this, a rotary evaporator was employed (50 °C for 2 h) to evaporate water and ethanol from the pastes [24].

2.2. DSC Devices Assembly

Fluorine-doped tin-oxide (FTO) glass substrate was cleaned thoroughly using soapy water, acetone, and ethanol subsequently. A blocking layer of TiO₂ was sprayed on the FTO substrate using a (1:9 v/v) solution of diluted titanium diisopropoxide bis (acetylacetonate) (75% in isopropanol, Sigma, Castle Hill, NSW, Australia) in absolute ethanol. TiO₂-HP25, TiO₂-P25, and TiO₂-DSL films were screen printed (Keywell semi-automatic screen printer, Kang Yuan Industrial Co. Wufeng Shiang, Taiwan ) onto the FTO substrate to obtain film thicknesses of 11.5 ± 0.3 µm and areas of 0.16 cm² (4mm × 4mm). The films were annealed for 3 h at 450 °C to remove the organic binders [24]. A post-treatment step of titanium chloride (TiCl₄) was applied to the sintered films by immersing them in an aqueous solution of TiCl₄ (20 mM, Sigma, Castle Hill, NSW, Australia ) for half an hour at 70 °C before rinsing with ethanol and then again with water before re-sintering for one hour at 500 °C. After being cooled down to 110 °C, the films were soaked overnight in a D149 (1-material, Dorval, Quebec, Canada) dye bath at a concentration of 0.5 mM in a 1:1 v/v mixture of acetonitrile and tert-butanol. After the photoanodes were taken from the dye solution, they were washed with acetonitrile and dried at 110 °C. The counter electrode (cathode) was produced by casting ~10 µL (H₂PtCl₆) (0.01 M, Sigma, Castle Hill, NSW, Australia ), then treating it at 450 °C for 30 min to reduce the platinum to Pt₀. Working electrodes (photoanodes) were affixed to counter electrodes using a 25 μm Surlyn spacer (Solaronix, Aubonne, Switzerland). The electrolyte consisted of iodine (50 mM, Sigma, Castle Hill, NSW, Australia), 1,2-dimethyl-3-propylimidazelium iodide (0.6 M, Solaronix, Aubonne, Switzerland), and lithium iodide (0.1 M, 99% Acros Organics, ThermoFisher, who are based in Scoresby, Vic, Australia ) in 3-methoxypropionitrile (≥ 98% Sigma, Castle Hill, NSW, Australia) and was injected between the two assembled electrodes using a vacuum back-filling approach. Finally, Surlyn-laminated aluminium foil was used to seal the filling port.

2.3. Material and Device Characterizations

The X-ray diffractometer (GBC Scientific Equipment LLC, Hampshire, Illinois, USA) was used to examine the crystalline structure of materials (Scan range = 20°–80°, Voltage = 40 kV, Current = 30 mA, Target = Cu Kα radiation, Wavelength = 1.54 Å). Transmission electron microscopy (TEM) (JEOL JEM-6500F, Tokyo, Japan) was used to examine the nanostructure of the materials. Brunauner–Emmet–Teller (BET) analysis was conducted on N₂ adsorption–desorption data, obtained using a MicrotracBel Belsorp-mini,Osaka, Japan) to quantify the specific surface area. D149 dye was desorbed from material films after being immersed in NaOH (0.4 M) in methanol for 5 min. Absorption spectroscopy (UV-Vis, Shimadzu UV-1800, Rydalmere, NSW, Australia) was used to calculate the dye loading on the photoanodes by measuring the absorption spectra of the aforementioned solutions. Film thicknesses were measured by Surface Profiler-Veeco Dektak 150 (NewSpec, Myrtle Bank, SA, Australia). The photocurrent density–voltage (J-V) measurements under 1 sun condition (100 mW/cm²) with an AM1.5 filter were conducted with a solar simulator (PV Measurements, Boulder, Colorado, USA). A QEX10 system from (PV Measurements, Boulder, Colorado, USA) was used to measure incident photon to current conversion efficiencies (IPCEs). Electrochemical impedance spectroscopy measurements for three devices were measured with a Gamry Reference 600 workstation (Gamry Instrument,Warminster, USA)) under illumination at V_oc in the frequency range of 0.1–10⁶ Hz and using an AC voltage of 10 mV.

3. Results and Discussions

3.1. Material Structure Analysis

X-ray diffraction (XRD) traces, shown in Figure 1d, indicate that, as expected, TiO₂-P25 and TiO₂-HP25 contained a mixture of anatase and rutile, whereas TiO₂-DSL only showed anatase. This is in line with what we reported in our previous work [18]. We also note that the amorphous TiO₂ could not be directly observed by XRD. The amount of amorphous TiO₂ in TiO₂-P25 and TiO₂-HP25 before and after the hydrothermal treatment was investigated and quantified in our previous report as 76:13:11 anatase: rutile: amorphous wt.% and 81:19:0 anatase: rutile: amorphous wt.% [17,18]. Average grain sizes were obtained using the Scherrer formula (D = 0.89λ/β cosθ, where D is the grain size, λ the wavelength of the X-ray, β the full-width half-maximum in radians, and θ the diffraction angle). From this, the crystal size was estimated to be 20 nm in TiO₂-DSL, whereas both TiO₂-HP25 and TiO₂-P25 revealed values of around 30 nm and 34 nm for the anatase and rutile phases, respectively.

As can be seen in the TEM in Figure 1a–c, TiO₂-HP25 was more aggregated compared to TiO₂-P25. This was presumably a result of the hydrothermal treatment resulting in crystal growth focused at the junction of particles, in a similar manner to necking occurring during the sintering process. These images show that TiO₂-DSL had particle size of 15 ± 7 nm, whereas TiO₂-HP25 and TiO₂-P25 had a similar particle size of 28 ± 8 nm (see insets in Figure 1a–c, along with Figures S1 and S2), which were within the range of the above values, calculated from the XRD data. The average particle size of the materials was quantified by a Digital Gatan Micrograph and Image J software based on measuring the aspect ratio of the width to the height of the particle image, as shown in Figures S1 and S2.

The surface area and pore-size distribution of three types of nanoparticles were investigated by Brunauer–Emmett–Teller (BET) measurements. As shown in Figure 1e and its inset, TiO₂-DSL nanoparticles had the highest surface area of the three materials investigated here, resulting from their smaller particle size and good dispersion. TiO₂-HP25 nanoparticles had a slightly smaller surface area than TiO₂-P25 nanoparticles, which was primarily ascribed to the aggregation, as discussed above. The type IV isotherm indicated pore diameters of 2–50 nm, which represents a mesoporous structure of all the materials [25]. The inset in Figure 1e shows the pore-size distribution range of the materials and their average pore diameters obtained from the Barrett–Joyner–Halenda (BJH) analysis. In general, when the particle size increased, the surface area increased and the average pore size decreased. From the BET-based nitrogen isotherm adsorption, the more nitrogen adsorption on nanoparticles means a higher surface area, resulting in a smaller average pore size, depending on the voids or pores among the nanoparticles. The different pore-size distributions of the three materials indicate the different surface areas and determine the capability of dye loading on the materials.

The amount of D149 adsorbed on the photoanode films was quantified by a dye-desorption method [26]. As shown in Figure 1f, TiO₂-DSL possessed the highest loading, as was to be expected, given that it also possessed the highest specific surface area. Meanwhile, TiO₂-HP25 showed greater dye adsorption than TiO₂-P25, despite having a lower surface area.

The dye loading can be normalized against the specific surface area to give a better understanding of the way dyes pack onto the surface. From this, it can be seen that the dye packing on TiO₂-HP25 was closer to that of TiO₂-DSL than that of TiO₂-P25, indicating the degree to which the amorphous content disproportionately impacts surface chemistry, despite only representing around 11 wt% of P25. The obtained BET data, the specific surface area (S_A), the average pore diameter, and the amount of desorbed dye of TiO₂-DSL, TiO₂-P25, and TiO₂-P25 are listed in

Table 1. The specific surface area (S_A), the average pore diameter (D_P), and the amount of desorbed dye per unit area/volume of materials.

Material	SA (m2·g−1)	DP (nm)	Dye Loading (µmol·cm−3)	Dye Loading (nmol·cm−2)
TiO2-DSL	79	26	99 ± 1	64 ± 1
TiO2-HP25	47	37	42 ± 2	50 ± 1
TiO2-P25	57	41	27 ± 1	33 ± 1

.

3.2. Photovoltaic Performance of DSCs

Photocurrent-voltage measurements performed under 1 sun illumination showed very similar performances between TiO₂-DSL|D149 and TiO₂-HP25|D149, with both showing around 45% higher overall photovoltaic conversion efficiency (PCE) than for TiO₂-P25|D149 devices. As shown in

Table 2. Photovoltaic measurements of TiO₂-DSL|D149, TiO₂-HP25|D149, and TiO₂-P25|D149 devices.

Devices	Jsc (mA·cm−2)	VOC (V)	FF (%)	PCE (%)
TiO2-DSL|D149	15.3 ± 0.9	0.608 ± 0.01	62 ± 2	5.81 ± 0.4
TiO2-HP25|D149	15.4 ± 0.4	0.620 ± 0.02	60 ± 1	5.78 ± 0.2
TiO2-P25|D149	13.3 ± 1.3	0.600 ± 0.02	49 ± 4	3.89 ± 0.4

as well as in Figure 2a, there was a simultaneous improvement in photocurrent density (J_sc), open-circuit voltage (V_oc), and fill factor (FF) for TiO₂-HP25|D149 compared to the untreated P25. Figure 2b shows the IPCE measurements, which agreed with the trend in J_sc values, with very similar responses for TiO₂-DSL|D149 and TiO₂-HP25|D149. Two regions (430–460 nm and >650 nm) were noted to show a slight difference in favour of the TiO₂-DSL devices, which may be explained by the higher overall dye loading; however, it was not enough to lead to a difference in J_SC above the experimental variability (one standard deviation, based on multiple devices). The absorbance spectra of D149 dye adsorbed on TiO₂-DSL, TiO₂-P25, and TiO₂-HP25 photoanodes and IPCEs values of their corresponding devices to estimate the absorbed photon to charge carrier efficiencies (APCE). The APCE values had a similar trend to the IPCE values, which explains the difference in device performance. (see Figure S3a,b).

To further investigate the charge transport in these devices, electrochemical impedance spectroscopy (EIS) measurements were conducted and are represented by Nyquist and Bode plots in Figure 2c,d. These were collected at V_OC under 1 sun illumination with a perturbation voltage of 10 mV. Key parameters taken from these are shown in

Table 3. EIS measurements of TiO₂-DSL|D149, TiO₂-HP25|D149, and TiO₂-P25|D149 devices.

Devices	RS (Ω·cm−2)	Rct1 (Ω·cm−2)	Rct2 (Ω·cm−2)	Cµ (µF·m−1)
TiO2-DSL|D149	5.5 ± 0.1	2.2 ± 0.2	24.1 ± 0.3	1450 ± 100
TiO2-HP25|D149	5.0 ± 0.1	2.5 ± 0.1	30.5 ± 0.3	540 ± 40
TiO2-P25|D149	5.2 ± 0.1	2.5 ± 0.1	33.5 ± 0.5	210 ± 20

. The Nyquist plots (Figure 2c) all showed two depressed semicircles, with the inset showing the equivalent circuit model used to interpret these data. In this, R_ct1 and R_ct2 are the electrochemical impedances, which refers to the charge-transfer resistances at high and mid-range frequencies fitted by the first and second semicircles, respectively. CPE₁ and CPE₂ refer to the constant-phase elements (considered a double-layer capacitor at a solid–liquid interface with non-ideal or “leaky” capacitance), with R_ct1 and R_ct2 at cathode–electrolyte and photoanode–electrolyte interfaces, respectively [25]. R_s is the overall series resistance of the device.

As R_ct1 refers to the charge transfer at the cathode–electrolyte interface and R_s is the series contact resistance, their values were expected to be similar for all three systems, as shown in Table 3. TiO₂-DSL|D149 showed the lowest R_ct2, whereas, TiO₂-P25|D149 showed the highest R_ct2, indicating the lowest and highest charge recombination at the photoanode–electrolyte interface, respectively. R_ct2 was decreased slightly in TiO₂-HP25|D149 devices compared to TiO₂-P25|D149 due to the absence of amorphous TiO₂. Amorphous content was expected to lead to recombination due to (1) defect sites and (2) lower dye-packing density, directly exposing more of the semiconductor to the electrolyte. The chemical capacitance (C_µ) of the devices showed an opposite trend to R_ct2, but to a much greater degree. The higher values of C_µ indicates that there was a higher density of shallow traps in the TiO₂-DSL and TiO₂-HP25 nanoparticle surface, resulting in a greater capability to receive electrons compared to TiO₂-P25. The Bode plots in Figure 2d provide us with important frequency-based information. The low-range maxima show the values of electron lifetimes obtained from (τ_n = 1/2$\pi f$_max) [26], with similar values for TiO₂-DSL|D149 and TiO₂-HP25|D149 devices, indicating longer electron lifetimes (lower electron recombination rate) compared to TiO₂-P25|D149 devices, and consistent with the difference shown in FF values in photovoltaic measurements.

4. Conclusions

Dye-sensitized solar cells based on an organic dye and a modified form of one of the cheapest sources of TiO₂ (P25) were produced and compared against both untreated P25 and a benchmark commercially available product specifically designed for this application, NR18-T (Dyesol). The amorphous content-free P25 (TiO₂-HP25|D149) devices showed higher solar performance than untreated P25 (TiO₂-P25|D149), which was very similar to those using the Dyesol product (TiO₂-DSL|D149). The enhanced performance resulting from the treatment of P25 is explained by higher dye-loading and packing densities, as well as a decrease in recombination sites. The process used to treat TiO₂-P25 is time-consuming; however, it does suggest that with further optimization this may be a viable pathway to making photoanodes. This amorphous content-free P25 can be considered as a cost-effective photoanode material for organic dye-sensitized or perovskite solar cells.