1. Introduction
Since the dye-sensitized solar cell (DSSC) was first developed in 1991 with an initial efficiency of 7.12% [
1], considerable efforts have been made to improve its power conversion efficiency (PCE) [
2,
3,
4,
5,
6,
7,
8]. Despite considerable advancements over the past two decades, the highest efficiencies reported remain at approximately 12% [
2,
3,
7], and achieving efficiencies above 10% is uncommon without highly optimized fabrication conditions. Given the considerable advantages of DSSCs, such as their manufacturing cost being only one-fifth that of conventional silicon solar cells [
1,
2,
3], and their versatility in applications, including the production of richly colored and transparent products, the potential of DSSCs remains substantial. The research results will remain valuable. To enhance the efficiency of DSSCs, various techniques have been proposed, including the deposition of a thin tunneling barrier layer on the substrate [
9,
10] or the oxide surface [
11], co-sensitization using different dyes [
12], and post-treatment with a TiCl
4 precursor [
13,
14]. Additionally, surface texturing has been widely adopted because it confines the incident light in the electrode, thereby increasing the photocurrent density [
15].
To date, titanium oxide (TiO
2), a well-known photocatalytic semiconductor, has typically been used as the active layer in the working electrode of DSSCs owing to its relatively large band gap and suitable conduction band energy. It is well established that achieving high PCE requires the energy band structure of TiO
2 to be well-aligned with that of the dye. Specifically, if the conduction band energy of TiO
2 is higher than the LUMO energy of the dye, electron injection from the dye becomes challenging. Given the properties of commonly used ruthenium-based dyes (N3, N719), the selection of compatible oxides is extremely limited. This makes TiO
2 an ideal material because its conduction band energy is approximately 0.2 eV lower than the LUMO of dyes, facilitating efficient electron injection. Additionally, TiO
2 is well-suited for use as a scattering layer owing to its chemical stability and dye absorption capability. As a result, many DSSCs are constructed usingTiO
2 nanoparticle films with a TiO
2 scattering layer on top of the active layer. For instance, Grätzel et al. especially used a mesoporous TiO
2 film layer to achieve a 1000-fold increase in surface area [
1,
16]. By incorporating a scattering layer, the overall PCE can be further enhanced.
The thickness of the TiO
2 active layer is also a crucial factor influencing the PCE of DSSCs. As the thickness increases, the surface area of the TiO
2 layer expands, leading to greater dye adsorption [
17,
18,
19]. However, this also lengthens the electron diffusion time, which can compete with the electron lifetime, thereby impairing electron transport properties. Additionally, the increased thickness reduces light transmittance, preventing the concentration of photoelectrons from increasing in proportion to the increased dye adsorption [
20]. If we can control both surface area and thickness of the TiO
2 active layer by changing crystal structure, particle size, and morphology, a high PCE will be achieved. For this purpose, our research group has synthesized TiO
2 nanoparticles, nanowires and nanorods continuously for the last 15 years [
21,
22,
23], even though we did not obtain the result expected. Despite this situation, in this paper, we aim to present a systematic study of the impact of TiO
2 active layer thickness on the PCE of n-type DSSCs. To achieve this, we synthesized other types of TiO
2 nanoparticles using the hydrothermal method and examined the effect of their thickness on the PCE of the fabricated n-type DSSCs. The performance was then compared to that of DSSC based on commercially available P-25 TiO
2.
2. Results and Discussion
The solar energy power (P) can be calculated using the following equation:
The energy conversion efficiency (η) of DSSCs is defined as the ratio of the maximum electrical power output (P
max) to the solar energy power input (P). P
max is calculated by multiplying the maximum current (J
mp) by the maximum voltage (V
mp). The solar energy power input is determined as the product of the irradiance of the incident light, measured in W/m
2, according to the following equation:
The ideal power density of DSSCs is a product of the maximum short-circuit current J
sc and maximum open-circuit voltage V
oc. The fill factor (ff) is a measure of the quality of the solar cell, and it can be calculated by comparing J
sc and V
oc to J
mp and V
mp by the following equation:
Therefore, we can obtain the power conversion efficiency (η) of DSSCs by the following equation.
where P (100 mW/cm
2) is the solar cell testing standard under terrestrial conditions with air mass AM1.5.
According to Equation (4), achieving high ƞ in DSSCs requires maximizing both Jsc and Voc, as well as ff. Thus, researchers have focused on integrating metal oxide nanostructures into DSSCs to increase dye adsorption and thereby enhance the photocurrent (Jsc).The fundamental operating principle of DSSCs mirrors that of photocatalytic reactions, where the generation of hole–electron pairs is crucial for both high PCE and efficient catalysis. Among various metal oxide materials, TiO2 nanostructures are particularly favored for DSSC and photocatalysis applications owing to their stability and low production cost.
Figure 1 shows a schematic diagram of the doctor-blade method used for coating TiO
2 active layers. The thickness of the active layer can be precisely controlled by adjusting both the number of 3M tape layers and the number of doctor-blade coating cycles. Using these techniques, the TiO
2 active layers were accurately fabricated in a thickness range of 6–24 µm. Detailed procedures for the fabrication and characterization of TiO
2 based DSSCs have been previously published [
21,
22].
Figure 2a presents a typical XRD pattern of TiO
2 nanoparticles synthesized and annealed at 700 °C. The prominent diffraction peak at 25° corresponds to the (101) rutile plane, indicating that the primary crystal growth direction is along the {101} direction. A smaller diffraction peak at 38° is attributed to the (101) plane of anatase phase. Additionally, several other peaks corresponding to either (200) and (211) rutile or (211) anatase crystallites are also observed. By analyzing the peak areas (and considering their FWHM values), we calculated the crystalline phase ratio of rutile to anatase, which is very similar to that of P-25, with values of 78% rutile and 22% anatase. Park et al. reported that anatase crystals generally have smaller particle sizes compared to rutile crystals, which results in a larger surface area for anatase. Consequently, more dye molecules can be adsorbed on the anatase surface, leading to improved light absorption and increased photocurrent. Additionally, the interactions between particles in anatase crystals are stronger than those in rutile crystals, which enhance electron mobility and boost photocurrent [
24]. Conversely, rutile crystals are more stable and more effective at scattering light compared to anatase crystals, and finding the optimal combination of these phases remains a crucial research objective.
Figure 2b shows an atomic force microscopy (AFM) image of the same film depicted in
Figure 2a, with a scan size of 10 × 10 µm
2. The image reveals craters approximately 0.5 µm in size on the surface of the fluorine-doped tin oxide (FTO) glass substrate, along with small nanosized clusters [see the enlarged image in
Figure 2b]. The average surface roughness was measured at 0.274 nm, indicating a relatively larger surface area compared to a film coated with commercially available P-25 powder. This rough surface increases surface haze owing to a decrease in particle size. Therefore, this rough surface can produce more electrons in photovoltaic cells, resulting in the increase of photocurrent (J
sc) from our synthesized TiO
2 nanoparticles.
Figure 3 presents (a) scanning electron microscopy (SEM) and (b) electron dispersion spectroscopy (EDS) data for the TiO
2 nanoparticles synthesized after calcining at 450 °C and annealing at 700 °C. As shown in
Figure 3a, the SEM image reveals spherical nanoparticles with some aggregate morphologies and an average particle size of 25 nm (see SEM inset). To determine the composition ratio of the synthesized TiO
2 nanoparticles, we measured the EDS spectrum.
Figure 3b shows the EDS spectrum along with elemental mapping data (inset), which shows peaks corresponding only to Ti and O, with a weight percent ratio of Ti to O at 49.88:50.12. This indicates that a slightly excess of oxygen likely attributed to water adsorption or further oxidation in the air exhibits. Considering the differences in atomic sensitivity in EDS, the atomic percent indicates 75% O and 25% Ti, which falls within acceptable error limits. Using theseTiO
2 nanoparticles, we fabricated the active layers of n-type DSSCs with varying film thicknesses by adjusting the number of 3M tape layers and doctor-blade coating cycles, ranging from 1 to 5 layers.
Figure 4 shows the SEM surface morphologies (a–c) and cross-sectional images (d–g) of TiO
2 nanoparticle film layers with varying numbers of doctor-blade coatings such as 1 layer (d), 2 layer (e), 3 layer (f), and 5 layer (g), respectively. By adjusting both the numbers of 3M tape layers and doctor-blade coating cycle, we controlled the thickness of the TiO
2 nanoparticle film layers. The thickness of the TiO
2 active layers varied as follows: 6 µm (1 layer), 12 µm (2 layers), 17 µm (3 layers), and 24 µm (5 layers) with an increasing number of doctor-blade coating cycles. As the thickness increased, the surface roughness also increased, indicating an increase in surface area (
Figure 4a–c). When we fabricated films thicker than 17 µm (3 layers), both optical transmittance and PCE of the n-type DSSCs decreased, largely owing to minimal changes in dye absorption [see
Figure 5b as well]. Therefore, in this study, we excluded TiO
2 nanoparticle films thicker than 17 µm (3 layers). The cross-sectional images (d–g) clearly illustrate the variations in TiO
2 nanoparticle film thickness with different numbers of doctor-blade coating cycles. Based on this study, it is evident that, while light efficiency increases with thickness up to a certain point, it subsequently decreases when the thickness exceeds this optimal level. This decrease is likely attributable to increased electron diffusion time and reduced light transmittance, as described earlier. As the layer thickness increases, however, the distance that electrons generated by the dye must travel to reach the electrode becomes too great, which can lead to recombination and reduced electron injection efficiency into the electrode.
Figure 5a shows UV-Visible absorption spectra measured after N719 dye adsorption on the TiO
2 active layers, as shown in
Figure 4. The baseline UV-Visible absorption spectrum was recorded from a bare FTO glass substrate without dye adsorption. The intensity of the dye adsorption peaks (as well as peak area), particularly around 500 nm, which correspond to the π–π* transition in the N719 dye molecule, increases with the thickness of the TiO
2 active layers. This suggests that a thicker TiO
2 layer, with its larger surface area, allows for greater dye adsorption. Using Beer’s law, we calculated the relative amounts of adsorbed dye on the TiO
2 active layers based on the data presented in
Figure 5a.
Figure 5b shows the variation in thickness (left, black color) and dye adsorption amounts (right, red color) with the number of doctor-blade coating cycles. As the number of coating cycles increases, both thickness and dye adsorption amounts increase from 1.2 × 10
−7 mol/cm
3 (6 µm, 1 layer) to 2.9 × 10
−7 mol/cm
3 (24 µm, 5 layers). However, the rate of increase begins to level off after three coating cycles. This indicates that the amount of dye molecules increases with the thickness of the active layers up to a critical thickness. Additionally, we observed that the PCE of DSSCs with active layer thicknesses exceeding 17 µm (3 layer) did not improve and even declined at 24 µm (see
Table 1). This suggests that, as the TiO
2 active layer becomes thicker, the distance electrons generated by the dye must travel to reach the electrode becomes too long. Consequently, recombination may occur, preventing efficient electron injection into the electrode. Therefore, achieving high surface roughness (i.e., a large surface area) combined with an optimal TiO
2 active layer thickness can absorb more dye molecules, enhancing the light-trapping or scattering effect, particularly in the long-wavelength region. This leads to an increase in the PCE of the photovoltaic device. This result indicates that TiO
2 nanoparticle-based DSSCs are highly effective in enhancing light-trapping or light-scattering properties in photovoltaic cells by adjusting thickness and surface area. However, it is important to note that, while nanoparticle films demonstrate considerably higher PCE compared to other nanostructures, they also face a clear disadvantage in charge transport owing to the excessive number of adsorbed dyes.
The photocurrent (J
sc) and photovoltage (V
oc) of the solar cell devices, with an active area of 0.25 cm
2, were measured using simulated sunlight at AM-1.5 produced by solar simulator.
Figure 6 shows the J–V curves for TiO
2 nanoparticle film-based DSSCs with varying active layer thicknesses. The photovoltaic characteristics of these n-type DSSC devices are summarized in
Table 1. As illustrated in
Figure 6 and detailed in
Table 1, the TiO
2 nanoparticle film-based DSSCs with active layer thicknesses of less than 12 µm (2 layers) exhibit considerably lower PCE values compared to the commercially available P-25 based DSSC, which has a PCE of 3.77%. This decrease is likely attributable to the relatively low crystalline quality (poor polycrystallinity) of the synthesized TiO
2 nanoparticles. Despite the lower PCE values compared to P-25 based DSSCs, the experiment shows that J
sc gradually increases with the thickness of TiO
2 nanoparticle active layers, from 1 to 3 layers (see the y-axis changes in
Figure 6). This increases in J
sc results in an increase in PCE from 2.86% to 4.02%, which surpasses the P-25 based DSSC. In the case of 5 layer, however, the PCE was dropped to 3.91% due mostly to the decrease of J
sc and ff rather than V
oc. This indicates that if a thickness of active layer has over optimum value, the distance electrons generated by the dye should long travel to reach the electrode owing to relatively prolonged recombination lifetime (This will be confirmed later). From the
Table 1, we concluded that the considerable increase in PCE by 41% is primarily attributed to a large improvement in photocurrent density (at least 52%) in the n-type DSSCs, although the values of V
oc and ff remained relatively constant. Ohno et al. reported the analysis results of composition ratio between rutile and anatase crystalline phase, as well as average particle sizes with commercially available P-25 (Degussa, Frankfurt, Germany) TiO
2 powder of 70–80% anatase and 20–30% rutile, with average particle sizes ranging from 20 to 30 nm, respectively [
25]. Since our synthesized TiO
2 nanoparticles have a similar composition ratio (78% anatase and 22% rutile, with little high amount of anatase due possibly to relatively low annealing temperature [
26]) and average particle size (25 nm) with P-25, they can be comparable to each other even though we obtained a relatively low PCE value.
To understand this result in detail, we performed an analysis using electrochemical impedance spectroscopy (EIS).
Figure 7a shows the Nyquist plots of the EIS data for the same DSSCs shown in
Figure 6. The EIS data reveal four components: Z
1 (high-frequency region, 100 kHz–500 Hz), Z
2 (middle-frequency region, 500 Hz–1 Hz), Z
3 (low-frequency region, 1 Hz–100 mHz), and the Ohmic series resistance of the TCO (R
h). In
Figure 7a, three distinct semicircles are observed in the measured frequency range of 100 mHz–100 kHz. R
h in the high-frequency region is associated with the resistance of the electrolyte and the FTO, while the impedances (or resistances) Z
1, Z
2, and Z
3 correspond to different aspects of the charge transfer and diffusion processes. Z
1 is the charge transfer interface resistance between the Pt counter electrode and the electrolyte (R
1), Z
2 is the charge transfer resistance at the TiO
2/electrolyte interface (R
2), and Z
3 is the Warburg diffusion resistance in the electrolyte (R
3) [
27]. As we can see in the
Figure 7b, however, in the analysis of a traditional
equivalent circuit model for DSSCs, a constant photo-generated current source, a series parasitic resistance (i.e., series resistance, R
s) and a parallel parasitic resistance (i.e., parallel resistance, R
p) are generally included. For an understanding of the electronic behaviors of the solar cell, therefore, two types of resistance values, series and parallel resistance, were identified to measure the change in charge-carrier mobility with the energy level of the transport layer due to changes inside the solar cell. Because internal impedance is inversely proportional to solar cell performance, DSSCs with thicker TiO
2 active layers exhibit relatively lower impedance compared to those with thinner TiO
2 layers [see the maximum values of each second semicircle in
Figure 7a]. Thus, our data in
Figure 7a confirm a decreasing trend in all impedances with increasing thickness of the TiO
2 active layer. Among the resistances R
1, R
2, and R
3, the R
2 value (32.13 Ω) for the DSSC with a thick TiO
2 active layer of 17 µm (3 layers) showed lower impedance compared to those of DSSCs with thinnerTiO
2 layers (see also
Table 2). However, the EIS data of thickness shows an abnormal trend for the relatively high second semicircle and low impedance compared with other data. This suggests a longer electron diffusion time, indicating that the DSSC with the thickerTiO
2 layer below critical value can has a higher electron density and reduced electron recombination, which improves charge transport. Consequently, the DSSC with the 17 µm thick TiO
2 active layer demonstrates the highest photocurrent (8.42 mA/cm
2) and the best cell performance (η = 4.02%) in this study. However, the PCE begins to decline when the thickness exceeds 17 µm (3 layers) owing to extended electron diffusion times and reduced light transmittance in thicker layers of lower-quality TiO
2 nanoparticles. This indicates the need for further experiments to enhance light harvesting through improved light-trapping or light-scattering techniques. Exploring alternative active layers, such as 1-dimensional TiO
2 nanostructures, may offer better performance compared to nanoparticles.
In order to understand in detail the main reason for the thickness effect on the J
sc and PCE, calculations of electron lifetime (
τe) and recombination lifetime (
τr) were carried out as following procedures. The electron diffusion length (
Ln) is determined by the diffusion coefficient (
De) and the electron lifetime (
τe) [
16].
According to Equation (5), we can realize that the
Ln becomes longer as the
τe increases. The electron lifetime (
τe) is calculated by Equation (6) using the peak frequency (
fp,max) of the second semicircle from the Nyquist diagram shown in
Figure 7a [
28].
The peak frequency (
fp,max) of the second semicircle and the electron lifetime (
τe) for the different TiO
2 thicknesses are shown in
Table 2, together with the values of R
2 and recombination lifetime (
τr). The peak frequency (
fp,max) decreases and the electron lifetime (
τe) increases as the thickness of TiO
2 increases. Based on our obtained EIS data shown in
Figure 7a, relative impedance values are obtained from the DSSCs with different thickness of TiO
2 active layers [
29]. We could also obtain a value of electron diffusion time (
τe) theoretically by using Equation (6) and its results are summarized in
Table 2. The obtained
τe values are 5.24 × 10
−3 (6 µm), 5.71 × 10
−3 (12 µm), 6.04 × 10
−3 (17 µm), and 5.53 × 10
−3 s (24 µm), respectively. The increase in the
τe makes the electron diffuse and transfer more easily due to the increase in the
Ln, since the
τr is closely related to both electron lifetime (
τe) and electron diffusion length (
Ln). Recombination lifetimes (
τr) were therefore determined from the photovoltage (V
oc) decay curve (see
Figure 8) and realized a tendency to oppose between
τe and
τr (see
Table 2). As shown in
Figure 7a (see also
Table 1) and
Figure 8, the highest efficiency cell showed the lowest
τr owing to lowest recombination. Therefore, in this work, the J
sc and PCE are increased as the TiO
2 thickness increases up to 17 μm because the internal resistance (R
2) related to the electron transport in the TiO
2/dye/electrolyte interface is decreased and the electron lifetime (
τe) is increased, reflecting a very short electron recombination time (
τr). The performance of the DSSC with a thickness above 17 μm, however, is decreased because the recombination of the electron is strengthened. Finally, it is demonstrated that the optimal TiO
2 thickness is 17 μm for the best performance of the DSSC.