Low carrier recombination in polysiloxane gel electrolyte for high-performance DSSC

Previous research on quasi-solid-state DSSC (QSS-DSSC) that utilized polysiloxane-based polymer gel electrolytes (PGE) showed that the functional performance of the cells was highly affected by electrolyte infiltration into the TiO2 nano-porous layers. This study evaluated the efficiency enhancement in siloxane-based cells by introducing a TiCl4 pre-treatment process twice. We compared the impedance spectrum of PGE-based DSSC without (type-1 PGE) and with (type-2 PGE) with the addition of propylene carbonate, measured under dark and light illumination. The impedance spectra of both cells showed different characteristics at different condition measurements, especially in the high-frequency region. Unlike the type-2 PGE-based DSSC, the type-1 PGE-based DSSC did not show the transmission line characteristic, which indicated less charge carrier diffusion inside the TiO2 nano-porous layer. Under light illumination, the interfacial charge transfer between electrons inside TiO2 layers with the electrolyte (Rct), and the electron lifetime inside TiO2 layers before it is recombined (τ r), became smaller for type-2 PGE-based DSSC and larger for type-1 PGE-based DSSC. This indicated that the recombination rate increased as the PGE became more vicious. This result supports the photovoltaic characteristics that yield current density and efficiency values of 16 mA cm−2 and 5.37% for type-2 PGE-based DSSC, 13.4 mA cm−2, and 4.72% for type-1 PGE-based DSSC. The challenge for further improvement in DSSC that employs PGE is to elevate the wetting capability of the gel inside the TiO2 layer without additional solvent since additional solvent eventually can reduce ionic concentration and consequently increase the Rct value as shown in the analysis of the impedance spectrum of TiO2 layer without dye.


Introduction
Since Gratzel and O'Regan first introduced dye-sensitized solar cells (DSSC) in 1991, many studies have been carried out to improve the performance of this kind of solar cell [1]. Although the new emerging metal halide perovskites show much larger conversion efficiency, DSSC has better stability than other types of organic and hybrid organic-inorganic materials-based solar cells. So far, many efforts have been made to improve the performance and efficiency of the resulting solar cell through the development of photo-anode TiO 2 materials [2, 3], dye materials as well as electrolytes [4], resulting in the recent high efficiency of 13% [5]. Unfortunately, the utilization of liquid electrolytes in a dye-sensitized solar cell is known to suffer from the problems of leakage, and corrosion, in addition to the issues of precipitation of salts in the electrolyte at low dissociation of the adsorbed dye at high temperatures. These detrimental effects on the long-term stability of the cell [6][7][8][9] have led to the development of DSSCs employing solid-state p-type inorganic semiconductors [7,10], as well as organic and inorganic hole conducting materials [7,[11][12][13]. However, the energy conversion efficiencies of these solidstate DSSCs are generally undesirably low due to the low ionic mobility and imperfect electric contact.
There have been efforts to replace the electrolytes with polymer gel electrolytes (PGE), forming a quasi-solidstate DSSC (QSS-DSSC), which is supposed to eliminate leakage problems without essentially reducing the conversion efficiency [14][15][16]. The polymer gel electrolytes (PGE) commonly consist of a three-dimensional polymer network with the liquid electrolyte filling the pores of the network [7,17,18]. The first fabricated DSSC using polymer gel electrolyte was reported by Cao et al [19] using poly(acrylonitrile), the cells characterized by open circuit voltage close to 0.6V and short-circuit currents more significant than 3 mA cm −2 under 30 mW cm −2 light illumination. Several investigations on several polymers and polymer hybrids followed this. Kubo et al introduced a low molar mass gelator to solidify the liquid electrolyte in DSSC that showed efficiencies up to 3% at AM 1.5 conditions [20]. The optimum energy conversion efficiency of 5.91% measured at AM 1.5 conditions was obtained by the same group for a DSSC cell using PGE containing amide and urethane groups [21]. Recently, Lee et al reported the preparation of urethane-based UV-curable gel electrolytes and their deposition on a dye-adsorbed TiO 2 film via a screen-printing process, resulting in an overall cell efficiency of about 6.1% in comparison to 6.6% for the reference cell using liquid electrolyte [22]. Jayaweera et al [23] have significantly improved the DSSC based on polymer gel electrolyte using poly(acrylonitrile), resulting in cell efficiency of 8.44%. Dong et al, by using a new elastomeric type of copolymer poly (oxyethylene)-block-imide copolymer (POE-PAI), has reached a conversion efficiency of 9.8% [8]. The best conversion efficiency of PGEbased DSSC so far is reported by Victoria et al by using artificial hydrotalcite nano-clay as an electrolyte additive [24], reaching overall efficiency of 10.9% compared to 9.6% for reference cells.
It has been almost 27 years since the first utilization of PGE in DSSC, but the overall conversion efficiency is still lower compared to the DSSC based on liquid electrolyte, in which the observations of lower efficiency in QSS-DSSC is often to be attributed to low ionic diffusion [25]. Our previous research extensively investigated the cause of photovoltaic performance reduction in QSS-DSSC using electrochemical impedance spectroscopy (EIS). In that study [26] we found that the working performance reduction of DSSC cells is primarily limited by a lack of electrolyte penetration into the mesoporous TiO 2 layer. Therefore, the objectives of the current study were to optimize the photovoltaic performance of DSSC based on the siloxane PGE and evaluate the cause of change in efficiency.

Experimental
The DSSC studied in this research consists of two types of PGE, namely type-1 and type-2 PGE. We optimized both cells by repeating the pre-treatment process on the active electrode TiO 2 with TiCl 4 solution. Through this process, we expected that the probability for the carboxylate group in N719 dye to link with the Ti site on TiO 2 increased, improving the charge transfer from the dye to the TiO 2 conduction band and eventually reducing the recombination rate. The pre-treatment process with TiCl 4 was followed by the annealing process up to a temperature of 500°C, and the cell assembly was made simultaneously with the PGE preparation. DSSC was fabricated with PGE prepared from organically modified siloxane gel blended with imidazolium liquid electrolyte (IL) [26].
The synthesis process of hybrid polymer gel (HPG) following the previous report [25], where HPG was synthesized using TMSPMA via a sol-gel route. We put 16 ml of ethanol in a beaker glass and stirred it on a hotplate with a temperature set at 50°C. 4 ml of TMSPMA was then added slowly using a pipette. Distilled water later was added with a 1:1 volume ratio with ethanol. After the bubbles start to show, 0.06 ml of acetic acid was added. The beaker glass was then sealed with parafilm with very tiny holes. The solution was kept stirred for about two-three days. To eliminate the residual water and acetic acid inside the HPG, 20 ml of ethanol was then added to the resulting HPG, followed by stirring and heating at approximately 50°C for several hours until the gel phase was reformed again.
Polymer gel electrolyte (PGE) was made by mixing HPG and IL with a 1:1 ratio in wt% to form type-1 PGE and mixing HPG, IL, and propylene carbonate (PC) with a 1:1:1 ratio in wt% to form type-2 PGE. Both mixtures were then stirred for several hours until the homogeneous gel phase was formed. Freshly made PGE was used on the same day as the DSSC's cell fabrication process.
The simplified illustration at the molecular level of PGE is comprised of siloxane-based hybrid polymer networks with cross-linked oligo-methacrylate chains and imidazolium liquid ions. The wavy lines represent symbolically other segments of this polymer network as presented in figure 1 as adopted from [25].
To prepare the active electrodes, we first cleaned the FTO glass (7.5 × 7.5) cm 2 with soap, millipore water, and ethanol, then subsequently treated it in an ultrasonic bath containing iso-propanol/acetone mixed solution at a (1:1) volume ratio. This was followed by pre-treatment with TiCl 4 and subsequent calcination at the final temperature of 570°C. TiO 2 layers were then screen printed onto FTO glass with an area of 0.5 cm × 0.5 cm to form 5 layers. Each layer was first heated on a hotplate at 80°C for 5 min before being successively stacked by the next layer. After five times, these TiO 2 layers were then calcinated at the final temperature of 470°C for 30 min, producing a stacked transparent TiO 2 layer. Finally, another layer of TiO 2 containing light scattering particles of 10 nm was screen printed on that stacked transparent TiO 2 layer. The annealing and TiCl 4 pre-treatment process was repeated under the same condition as at the beginning step. The TiO 2 film was then immersed in 0.7 mM N719 dye and Chenodeoxycholic acid that was dissolved beforehand in ethanol at room temperature for 24 h. The Pt counter electrode was prepared by spin coating the Platinum paste onto an FTO glass, followed by subsequent heating at 500°C for 30 min.
Those active electrodes were assembled into a sandwich-type cell with a thermoplastic sealing film (Surlyn ® ) of 25 mm in between. Using a vacuum pump, we introduced PGE into the gap between those electrodes. The DSSC was fabricated using two types of PGE. The first one (type-1) was composed only of HPG + IL (1:1). The second one (type-2) was made with the addition of propylene carbonate (PC), which was then composed of HPG: IL: PC (1:1:1). A reference cell was also made using the same materials and fabrication conditions, but using IL only as its ion transport materials. This reference cell is used for comparison as the fabrication procedure in this work may still need to be improved.

Photovoltaic characterizations
The fabricated DSSCs were characterized by the current density-voltage (J-V) measurements using a solar simulator AM 1.5 (Newport, Model no. 11000) with average light power of 100 mW cm −2 on the DSSCs fabricated with different compositions of PGE and the IL based DSSC, namely type-1 and type-2.

Electrochemical impedance spectroscopy (EIS) measurement
We measured EIS under dark and light illumination of 100 mW cm −2 to investigate the charge transport mechanisms using Gamry 3000 reference, with the frequency range from 0.01 Hz to 1 × 10 5 Hz and an ac modulation voltage of 10 mV. EIS measurements were done on the cell without the dye (FTO/TiO 2 /IL/Pt on FTO) with the variation of the volume of the IL to PC, namely IL + PC, 1:0, 1:1, 1:2, and 1:4, and on the complete DSSC cells (FTO/TiO 2 /Dye/IL or PGE/Pt on FTO), measured at the bias voltage from 0V to 0.6V. The measured impedance spectra were fitted with the equivalent circuit using Gamry Echem Analyst.

Results and discussions
3.1. Photovoltaic performances of the fabricated DSSC From the J-V characterizations, we found that type-2 PGE was >0.6% more efficient than type-1 PGE but was still about 0.4% lower than the reference cell ( figure 2, table 1). This result was consistent with our previous result [26] but with higher efficiency. The Jsc value of type-1 and type-2 PGE that was 3.3% and 23% higher than the reference cell suggests that the gelation did not significantly affect the diffusion coefficient of the PGE as shown in [26] in which the diffusion coefficient of type-1 PGE, type-2 PGE, and reference cell was in the same order namely 10 −6 cm 2 s −1 . More detailed, the diffusion coefficient of type-1 PGE decreased by a factor of two compared to the reference cell, specifically from 4 × 10 −6 cm 2 s −1 to 1.6 × 10 −6 cm 2 s −1 , while type-2 PGE almost increased by two and a half factors compared to that reference cell, specifically from 4 × 10 −6 to 9.8 × 10 −6 cm 2 s −1 . The main reason that caused the efficiency of the reference cell to be slightly better than the type-2 PGE, was because the reference cell showed 12.1% and 6.8% higher V oc than type-1 and type-2 PGE.
The V oc in DSSC was the energy difference between the Fermi level of the electron in TiO 2 porous film and the redox potential of the electrolyte [4] that strongly correlated with the conductivity of the semiconductor layer, the conductivity of the electrolyte, and the recombination of the electrons in the TiO 2 layer into the redox potential of the electrolyte [27]. From the photovoltaic characteristics only, we could not determine any factors that caused the performance difference on both cells, and therefore will be discussed more by analyzing the impedance spectra of the cells.

Impedance characteristics
The impedance spectrum of DSSC generally consists of three semicircles. The first semicircle at the highfrequency region originated from the coupling between the charge transport at the Pt/electrolyte interface assigned as R pt and the chemical capacitance of Pt. The second semicircle originated from the coupling between the charge transfer (interface recombination) between active electrode TiO 2 and the electrolyte and the chemical capacitance of the TiO 2 layers assigned as R ct and C ct , respectively. The third semicircle at the lowest frequency region originated from the Nernst diffusion layer, which was a mass-transport-limiting process and was assigned as R dif .

Impedance characteristics of the TiO 2 layers without dye
The impedance spectrum consisted of a perfect semi-circle at the bias voltage of 0V, and with the increasing of the bias voltage, the impedance spectrum got smaller, and there was deformation in the shape of the semi-circle ( figure 3(a), table 2). Based on the relevant results, it was found that the significant contribution of the  Table 1. Photovoltaic parameters of the DSSC using type-1, type-2 of PGE, and the reference cell were measured using solar simulator AM 1.5 with a light intensity of 100 mW cm −2 .

Cell
Jsc impedance spectrum at the bias voltage of 0 V originated from the recombination process between electrons at the edge of the TiO 2 conduction band with the redox couple in the electrolyte (R ct , table 2). The applied voltage will increase the conductivity of the TiO 2 layers and will, in turn, decrease the resistance that comes from the TiO 2 layer. The increase in the bias voltage also caused the shift of the Fermi level to be closer to the conduction band, resulting in a very high concentration of electrons inside TiO 2 , escalating the rate of the recombination process. It means that the Rct becomes smaller (table 2). The impedance spectrum became more prominent as the electrolyte concentration decreased due to the addition of PC ( figure 3(b)) since the addition of PC will reduce the concentration of I−/I 3 − redox couple in the electrolyte and, therefore, will also reduce the probability of the recombination process between electrons in TiO 2 with the electrolyte, or in other words, will increase the R ct value (table 2).   3.2.2. Impedance characteristics of the complete cells At dark conditions, both of the cells (type-1 and type-2) showed different impedance characteristics in which the DSSC that was fabricated using type-2 PGE exhibited the transmission lines characteristic. This transmission lines characteristic was commonly attributed to the diffusion-controlled charge carrier transport in the porous TiO 2 layer, which is visible in the enlarged plot of figure 5(a). The corresponding equivalent circuit that was used for fitting was presented in figure 6. The presence of light illumination is expected to give rise to an overall effect on the impedance spectrum. Accordingly, it was attractive to present the impedance spectrum and fitting parameters of the DSSC's cells that were measured under light illumination conditions ( figure 5(b), table 3). Under light illumination, the overall impedance spectrum was smaller ( figure 5(b)), which might be caused by the heating process by the lamp that was used during the measurement process and consequently increasing the temperature of the cell and also accelerating the interfacial charge transfer inside the TiO 2 layers [28], which corresponding to the Buttler-Volmer formula.
Fn redox 1 0 R ct was inversely proportional to the temperature (T), therefore, if the temperature increased, the R ct value would be smaller. On the other hand, the increase in the temperature also reduced the viscosity of the PGE used [27] and decreased the diffusion resistance. The increase of the triiodide local ions concentration inside the pores of the TiO 2 layers also contributed to the decrease in the R ct value since it accelerated the recaptured of the electrons at the TiO 2 conduction band by the redox couple in the electrolyte that was indicated by the smaller value of the recombination lifetime (τ r ). The smaller the R ct and τ r values mean, the more recombination happened. Under light illumination, the R ct value of type-1 PGE-based DSSC was larger than type-2 PGE-based DSSC. This was supposedly an indication that the recombination rate in type-1 PGE-based DSSC is smaller, which suppose to result in higher current density. However, from the J-V characteristics (figure 2), the current density of type-1 PGE was smaller than type-2 PGE. This is because the R ct value in type-2 PGE-based DSSC is compensated by the low transport resistance (R t ) in the TiO 2 electrode (table 3), resulting in higher current density.
The ratio of electron diffusion length to the thickness of the active electrode (TiO 2 ) for DSSC based on type-2 PGE can be calculated from the equation [30,32]: t c t We found that the electron diffusion length is larger than the thickness of the TiO 2 electrode (table 3), denoting that the transit time is shorter than the lifetime, and this is a requirement to efficiently collect the charge injected by the dye when the solar cell is illuminating [30]. This result is 231.1% larger than our previous result [25] for DSSC based on type-2 PGE with the same composition HPG + PC + IL = 1:1:1 but without repetition of the TiCl 4 layer deposition, where we obtained (  [30,33]. As for the DSSC based on type-1 PGE, we cannot calculate the ratio of L L , n since its impedance spectrum did not show the appearance of the transmission line characteristics. This implies that the charge transfer process predominantly occurred just near the surface of the TiO 2 layer and not deep inside its mesoporous region [25]. This might explain why its current density was lower that the type-2 PGE-based DSSC.

Summary
We have managed to improve the performance of the DSSC based on PGE by employing the repetition of the TiCl 4 layer deposition (two times) on top of the TiO 2 electrode and compared to our previous report. The performance enhancement was mainly attributed to the high ratio of electron diffusion length to the thickness of the active electrode (TiO 2 ) which is larger than one. This means that the electron diffusion length was larger than the TiO 2 thickness. In addition, we were able to detect the contribution of the charge transfer process originating from the interface between TiO 2 with the electrolyte, which is known as the recombination process in the impedance spectrum of the DSSC, and we found that this recombination process was affected by the applied bias voltage and also by the electrolyte concentration. While from the impedance spectrum of the complete cells, we have revealed that the main factor contributing to the difference in the operating performance between type-1 PGE and type-2 PGE-based DSSC was the absence of the diffusion-controlled charge carrier transport in porous TiO 2 layers. Moreover, the inability of the electrons to diffuse inside the TiO 2 nano-porous layers corresponded to the lack of the type-1 PGE infiltration inside the TiO 2 nano-porous as a consequence of reduced viscosity of this type of PGE.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).