Carbon-Nanotube-Integrated Multilayer Titanium Dioxide/Tin Dioxide Photoanodes for Enhanced Dye-Sensitized Solar Cell Performance

Abstract

Dye-sensitized solar cells (DSSCs) remain attractive as low-cost photovoltaic devices; however, their practical efficiency is still constrained by electron-transport losses, interfacial recombination, and incomplete light harvesting in conventional titanium dioxide (TiO₂) photoanodes. The effects of TiO₂ film thickness, multi-walled carbon nanotube (MWCNT) incorporation, and multilayer oxide interface engineering on DSSC performance were examined. Degussa P25-TiO₂ photoanodes were first optimized with respect to thickness, after which controlled MWCNT loadings and sequential compact sol–gel TiO₂ and tin dioxide (SnO₂) sublayers were introduced. The optimum pristine P25-TiO₂ photoanode thickness was 9.11 μm, yielding an open-circuit voltage of 0.74 ± 0.01 V, a short-circuit current density of 14.10 ± 0.40 mA/cm², a fill factor of 56.24 ± 1.00%, and a power-conversion efficiency of 5.93 ± 0.20%. The incorporation of 0.025 wt% MWCNTs increased the efficiency to 6.04 ± 0.20%, corresponding to an absolute gain of 0.11 percentage points. The best performance was obtained with the sol–gel SnO₂/sol–gel TiO₂/P25-CNT multilayer photoanode, which delivered 0.74 ± 0.02 V, 16.22 ± 0.40 mA/cm², 57.59 ± 1.00%, and 6.89 ± 0.30%, respectively. FE-SEM, EIS, XRD, Heated Ultrasonic Cleaner and UV–visible analyses indicate that the multilayer architecture preserves porosity, enhances light harvesting, and suppresses interfacial recombination, while the CNT network facilitates charge transport.

1. Introduction

Since the seminal report by O’Regan and Grätzel in 1991, dye-sensitized solar cells (DSSCs) have been regarded as one of the most promising third-generation photovoltaic technologies because they combine relatively low fabrication costs, process simplicity, and favorable performance under low-irradiance conditions. Their continued relevance is further supported by the broader development of solar photovoltaic electricity and the sustained research interest in sensitized nanocrystalline oxide systems [1,2,3].

Among the key components of a DSSC, the photoanode is particularly critical because it governs dye uptake, light harvesting, electron injection, charge transport, and recombination. Titanium dioxide is the most widely used photoanode material because of its chemical stability, suitable band alignment, and strong dye-adsorption capability. Nevertheless, conventional nanoporous TiO2 films still suffer from inefficient electron transport and interfacial back-electron transfer, both of which limit device efficiency. Carbon nanotube incorporation and blocking-layer engineering have each been shown to mitigate these limitations: CNTs provide conductive pathways through the porous oxide network, whereas compact TiO2 or SnO2 sublayers suppress recombination at the transparent-conductive-oxide/electrolyte interface [4,5,6,7]. In parallel, photoanode-thickness optimization has repeatedly been shown to affect dye uptake, light harvesting, electron-transport distance, and photovoltaic output in DSSCs [8,9,10]. Earlier studies on bilayer photoanodes, TiO2 blocking layers, sol–gel interfacial layers, and CNT-assisted TiO2 electrodes further demonstrated the importance of interfacial engineering for suppressing back-electron transfer and improving charge collection [11,12,13,14,15,16,17]. Recent DSSC studies continue to show that mesoporous TiO2 thickness and compact interfacial layers remain key design variables for improving charge collection and suppressing recombination, while the composition and thickness of blocking layers, together with multilayer photoanode architectures, can markedly influence device efficiency [18,19].

Unlike earlier CNT-modified photoanodes, which either dispersed CNTs throughout the entire porous layer or combined CNTs with only a single blocking sublayer, the present study integrates a CNT-rich porous overlayer with sequential sol–gel TiO₂ and SnO₂ interlayers. This design deliberately separates porosity retention from recombination blocking: the upper P25-CNT layer preserves dye-adsorption area and electrolyte accessibility, whereas the lower compact oxides improve substrate coverage and suppress back-electron transfer. The principal contribution of this work is therefore the demonstration of a simple dual-interlayer/CNT-assisted architecture that enhances J_sc and overall efficiency without sacrificing V_oc.

The coupled effects of TiO₂ film thickness, MWCNT loading, and multilayer oxide architecture on DSSC performance were investigated, with particular emphasis on how morphology, optical response, and interface design influence current density and overall power-conversion efficiency.

2. Materials and Methods

2.1. Materials

Indium tin oxide (ITO) glass with a sheet resistance of 7 Ω/sq was used as the transparent conductive substrate. Degussa P25 titanium dioxide (TiO₂) served as the primary photoanode material, and D719 was used as the sensitizing dye. MWCNTs were introduced as conductive modifiers. Titanium(IV) isopropoxide (TTIP) and stannic chloride pentahydrate were used as the precursors for the sol–gel TiO₂ and sol–gel SnO₂ layers, respectively. Additional reagents included polyethylene glycol (PEG), 2,4-pentanedione, Triton X-100, nitric acid, isopropyl alcohol, ethanol, iodine, potassium iodide, and 3-methoxypropionitrile (MPN). All reagents were of analytical grade and were used as received. Unless otherwise specified, all materials and reagents listed above were purchased from Uni-Onward Corp. (New Taipei City, Taiwan).

2.2. Preparation of Photoanodes

Four photoanode architectures were fabricated. Electrode (a) consisted of a nanoporous Degussa P25-TiO₂ film. Electrode (b) comprised a nanoporous P25-TiO₂ layer coated with a P25-MWCNT nanocomposite overlayer. Electrode (c) incorporated a compact sol–gel TiO₂ interlayer beneath the P25-CNT structure. Electrode (d) further introduced a sol–gel SnO₂ blocking layer below the sol–gel TiO₂/P25-CNT assembly, thereby forming the fully integrated multilayer photoanode (Figure 1).

For preparation of the P25-TiO₂ paste, Degussa P25 and PEG were dispersed in deionized water containing 2,4-pentanedione and Triton X-100 and then ground to obtain a homogeneous slurry. The resulting paste was spin-coated onto ITO substrates at 1200 rpm for 10 s per cycle, dried at 25 °C under controlled humidity, and finally annealed in air at 450 °C for 30 min. Film thickness was controlled by the number of coating cycles. Before incorporation into the TiO₂ paste, the as-grown MWCNTs were first calcined at 450 °C for 60 min and then treated with 3 M HNO₃ with shaking for 1 h followed by standing for 3 h. The product was gradually diluted with deionized water, filtered through a PTFE membrane, and repeatedly washed to neutral pH. The treated MWCNTs were then thoroughly mixed and ground with the P25 paste in a ceramic mortar. After mortar mixing, the prepared paste/dispersion was further homogenized in an ultrasonic bath (DELTA DC300H., Yuantuo Technology Ltd., Taichung City, Taiwan). The P25-CNT paste was prepared using the same formulation with MWCNT additions of 0.0125, 0.025, 0.1, and 0.2 wt%. The compact sol–gel TiO₂ precursor was prepared from TTIP, PEG, nitric acid, and deionized water, whereas the sol–gel SnO₂ precursor was prepared from stannic chloride pentahydrate in isopropyl alcohol/water. These precursor layers were deposited sequentially beneath the porous P25-CNT overlayer to construct the multilayer photoanodes.

2.3. Cell Assembly

An active area of 0.5 cm² was defined on each sintered photoanode. The electrodes were immersed in a 3 × 10⁻⁴ M D719 solution in ethanol for approximately 12 h to ensure sufficient dye adsorption. Pt-coated ITO glass with a platinum thickness of 160 nm was used as the counter electrode. The electrolyte consisted of 0.34 M KI and 0.01 M I₂ dissolved in MPN. The dye-loaded photoanode, electrolyte, and Pt-coated counter electrode were subsequently assembled into complete DSSC devices.

2.4. Characterization and Measurements

The photovoltaic performance of the DSSCs was evaluated using a Keithley 2400 (Hsinchu County, Taiwan) source meter under simulated solar irradiation from a 500 W Xe lamp equipped with an AM 1.5 filter. The surface morphology and cross-sectional microstructure of the photoanodes were examined by field-emission scanning electron microscopy (FE-SEM; JSM-7401F, JEOL, Tokyo, Japan) operated at 4.0 kV. The thickness of the P25 layer was determined from cross-sectional FE-SEM images by averaging measurements taken at three representative positions for each sample. For the 1–8 coating series, the average thickness increment was approximately 1.67 μm for each additional spin-coating cycle, thereby enabling the effect of film thickness on device performance to be assessed systematically.

Electrochemical impedance spectroscopy (EIS) measurements were carried out using a CH Instruments electrochemical analyzer (Model 6081B, Austin, TX, USA). A sinusoidal perturbation of ±10 mV was applied about the open-circuit voltage (V_oc) over a frequency range of 10² to 6.5 × 10⁴ Hz. Measurements were performed under illumination from a 50 W halogen lamp (Philips) equipped with a UV-cut filter. The resulting impedance spectra were fitted with an equivalent-circuit model using ZView software (Scribner Associates, Inc., Southern Pines, NC, USA), providing further insight into charge-transfer and recombination processes in the multilayer photoanodes.

The crystalline structure of the samples was characterized by X-ray diffraction (XRD; XRD-6000, Shimadzu, Kyoto, Japan), whereas optical transmittance and absorption properties were analyzed by UV–visible spectroscopy. For consistency, the photovoltaic parameters are reported throughout this revised manuscript as V_oc and J_sc.

3. Results and Discussion

3.1. Effect of TiO₂ Film Thickness on DSSC Performance

The photovoltaic performance of the DSSCs was highly sensitive to the thickness of the P25-TiO₂ photoanode. As shown in Figure 2 and Figure 3, and

Table 1. Photovoltaic performance of photoanode a with different P25-TiO₂ film thicknesses.

Thickness (μm)	V_oc (V)	J_sc (mA/cm2)	FF (%)	η (%)
2.59	0.72 ± 0.01	4.93 ± 0.20	61.47 ± 1.00	2.2 ± 0.40
3.45	0.71 ± 0.01	7.64 ± 0.30	62.33 ± 1.00	3.39 ± 0.20
4.23	0.73 ± 0.02	9.59 ± 0.60	60.51 ± 3.00	4.25 ± 0.50
5.91	0.72 ± 0.01	13.28 ± 0.30	54.74 ± 3.00	5.26 ± 0.30
7.01	0.72 ± 0.01	13.59 ± 0.50	57.04 ± 2.00	5.55 ± 0.40
9.11	0.74 ± 0.01	14.10 ± 0.40	56.24 ± 1.00	5.93 ± 0.20
12.52	0.72 ± 0.02	14.16 ± 0.40	57.56 ± 3.00	5.89 ± 0.30
14.25	0.72 ± 0.01	13.66 ± 0.30	54.29 ± 1.00	5.33 ± 0.20

, increasing the film thickness from 2.59 to 9.11 μm raised the short-circuit current density from 4.93 ± 0.20 to 14.10 ± 0.40 mA/cm² and increased the power-conversion efficiency from 2.20 ± 0.40% to 5.93 ± 0.20%. This behavior indicates that, within an appropriate thickness window, thicker TiO₂ films provide a larger effective surface area for dye adsorption and improved light harvesting, thereby promoting photocurrent generation. Across the measured 1–8 coating series, the thickness increased approximately linearly, with an average increment of about 1.67 μm per cycle.

3.2. Morphological Characteristics and Effect of MWCNT Incorporation

As shown in Figure 4, FE-SEM observations showed that both the pristine P25 film and the P25-MWCNT composite film exhibited a porous surface morphology composed of relatively coarse TiO₂ grains with characteristic sizes of approximately 25–40 nm. Although localized aggregation of MWCNTs was observed, the overall pore network of the TiO₂ matrix remained largely intact after nanotube incorporation, indicating that an appropriate MWCNT loading did not substantially disrupt the porous architecture required for efficient dye adsorption.

As shown in Figure 5, the J–V characteristics further demonstrated that MWCNT incorporation improved DSSC performance when the loading level was carefully controlled. As summarized in

Table 2. J–V characteristics of DSSCs employing P25–MWCNT photoanodes.

MWCNT Content (wt%)	V_oc (V)	J_sc (mA/cm2)	FF (%)	η (%)
0%	0.74 ± 0.01	14.10 ± 0.40	56.24 ± 1.00	5.93 ± 0.20
0.0125	0.72 ± 0.01	14.64 ± 0.20	56.51 ± 1.00	5.99 ± 0.10
0.025	0.74 ± 0.01	13.96 ± 0.50	58.77 ± 3.00	6.04 ± 0.20
0.1	0.76 ± 0.02	12.56 ± 0.30	63.05 ± 3.00	6.02 ± 0.30
0.2	0.75 ± 0.01	10.42 ± 0.60	62.79 ± 4.00	4.88 ± 0.50

,relative to the pristine P25 electrode, which exhibited an efficiency of 5.93 ± 0.20%, devices containing 0.0125, 0.025, 0.1, and 0.2 wt% MWCNTs delivered efficiencies of 5.99 ± 0.10%, 6.04 ± 0.20%, 6.02 ± 0.30%, and 4.88 ± 0.50%, respectively. The optimum performance was obtained at 0.025 wt% MWCNTs, corresponding to a V_oc of 0.74 ± 0.01 V, a J_sc of 13.96 ± 0.50 mA/cm², and a fill factor of 58.77 ± 3.00%. This represents an absolute efficiency gain of 0.11 percentage points relative to the pristine P25 reference. The improvement is attributed to the conductive nature of the nanotube network, which facilitates electron transport through the TiO₂ framework. By contrast, excessive MWCNT loading most likely caused nanotube agglomeration, generated charge-trapping sites, partially screened the porous oxide surface, and reduced effective charge collection. This behavior is consistent with prior reports showing that limited CNT incorporation can promote charge transport, whereas excessive addition often introduces aggregation and recombination losses [5,6,11].

3.3. Photovoltaic Performance of Composite Multilayer Photoanodes

To further enhance charge transport and suppress interfacial recombination, composite multilayer photoanodes were fabricated by integrating a P25-CNT overlayer with a compact sol–gel TiO₂ layer and a sol–gel SnO₂ blocking layer. As shown in

Table 3. Short-circuit current density (J_sc), open-circuit voltage (V_oc), fill factor (FF), and power-conversion efficiency (η) of photoanodes a–d.

Photoanode	V_oc (V)	J_sc (mA/cm2)	FF (%)	η (%)
a	0.74 ± 0.01	14.10 ± 0.40	56.24 ± 1.00	5.93 ± 0.20
b	0.74 ± 0.01	13.96 ± 0.50	58.77 ± 3.00	6.04 ± 0.20
c	0.75 ± 0.01	15.22 ± 0.40	58.73 ± 1.00	6.74 ± 0.20
d	0.74 ± 0.02	16.22 ± 0.40	57.59 ± 1.00	6.89 ± 0.30

and Figure 6, the photovoltaic data reveal a clear stepwise improvement with increasing structural sophistication. The pristine P25 photoanode exhibited an efficiency of 5.93 ± 0.20%, which increased to 6.04 ± 0.20% after MWCNT incorporation. Introduction of the sol–gel TiO₂ compact layer further increased the efficiency to 6.74 ± 0.20%. The best performance was achieved with the sol–gel SnO₂/sol–gel TiO₂/P25-CNT multilayer photoanode, which delivered a V_oc of 0.74 ± 0.02 V, a J_sc of 16.22 ± 0.40 mA/cm², a fill factor of 57.59 ± 1.00%, and an overall efficiency of 6.89 ± 0.30%. Relative to the pristine P25 photoanode, this corresponds to an efficiency enhancement of approximately 16%.

Figure 7 illustrates the FE-SEM surface morphologies of the annealed sol–gel TiO₂ and sol–gel SnO₂ films. Both films exhibit uniform and compact surface coverage, supporting their function as interfacial blocking layers rather than porous dye-loading matrices. Because these plan-view FE-SEM images primarily reveal surface compactness and continuity, they provide qualitative rather than direct structural evidence of blocking-layer behavior. Importantly, Table 1 and Figure 2 show that increasing thickness alone does not guarantee higher photoelectric conversion efficiency. Instead, the superior device performance arises from the rational division of functions within the multilayer structure. Specifically, the lower sol–gel SnO₂ and sol–gel TiO₂ layers act as compact interfacial barriers that reduce substrate-side recombination, whereas the upper P25-CNT layer preserves the mesoporous network necessary for dye uptake and electrolyte diffusion. Thus, performance enhancement is more reasonably attributed to the cooperative effect of recombination-blocking compact underlayers and a porous photoactive overlayer than to thickness increase alone.

The superior photovoltaic performance of the multilayer photoanodes is supported by the electrochemical impedance spectroscopy (EIS) results shown in Figure 8 and summarized in

Table 4. Numerical values of the equivalent-circuit elements obtained by fitting the impedance spectra of DSSCs with photoanodes a–d under the constrained condition of P2 = 0.81.

P2	T2 (mF)	R2 (Ω)	τ (ms)	R_D (Ω)	C1 (μF)	R1 (Ω)	R_s (Ω)	Electrode
0.810 (fixed)	24.52 ± 1.48	12.06 ± 0.25	33.926 ± 1.680	10.12 ± 0.28	21.00 ± 2.37	2.697 ± 0.125	27.89 ± 0.10	a
0.810 (fixed)	17.86 ± 1.16	8.65 ± 0.19	4.161 ± 0.186	13.04 ± 0.28	23.31 ± 2.18	3.978 ± 0.245	31.89 ± 0.12	b
0.810 (fixed)	19.87 ± 1.22	13.99 ± 0.30	28.893 ± 1.146	15.38 ± 0.32	23.83 ± 2.42	3.588 ± 0.157	30.19 ± 0.11	c
0.810 (fixed)	12.69 ± 1.14	16.03 ± 0.53	28.304 ± 1.228	21.44 ± 0.58	28.77 ± 3.93	3.553 ± 0.214	26.64 ± 0.15	d

Note: Values are presented as fitted values ± standard error. P₂ was fixed at 0.81 for all four samples. The corresponding reduced Chi-Sqr/adjusted R₂ values were 0.149/0.99966 for photoanode a; 0.237/0.99960 for photoanode b; 0.226/0.99963 for photoanode c; and 0.410/0.99941 for photoanode d.

. Under the constrained fitting condition with P₂ fixed at 0.81, clear differences in the fitted circuit elements were observed among photoanodes a, b, c, and d. Photoanode d exhibited the highest interfacial charge-transfer resistance (R₂ = 16.03 ± 0.53 Ω) together with the largest diffusion-related resistance (R__D = 21.44 ± 0.58 Ω), indicating stronger suppression of back-electron transfer at the TiO₂/dye/electrolyte interface and a more pronounced diffusion contribution than in the other samples. Photoanode c also showed a favorable impedance response, with a relatively high R₂ (13.99 ± 0.30 Ω) and a moderate R__D (15.38 ± 0.32 Ω), suggesting improved interfacial charge retention. By contrast, photoanode b displayed the highest series resistance (R_s = 31.89 ± 0.12 Ω) and the lowest R₂ (8.65 ± 0.19 Ω), implying comparatively less efficient charge collection under the present constrained-fit condition. Photoanode a exhibited the lowest R__D (10.12 ± 0.28 Ω) and an intermediate R₂ (12.06 ± 0.25 Ω), indicating relatively facile electrolyte-related transport but weaker recombination suppression than photoanodes c and d. Taken together, the EIS results indicate that the improved photovoltaic performance of the multilayer architecture is associated with enhanced electron extraction, suppressed interfacial recombination, and modified transport characteristics. In the present design, the compact interfacial oxide layers improve substrate-side charge-collection pathways and retard back-electron transfer, while the CNT-containing porous overlayer provides more effective conductive pathways for electron transport. Consequently, the multilayer architecture enhances charge collection and mitigates recombination losses, consistent with the higher photocurrent density and overall efficiency observed for photoanoded.

Electrochemical impedance spectroscopy (EIS) was employed to analyze the electrochemical behavior of photoanodes a, b, c, and d. As commonly observed for DSSCs, three characteristic responses can be identified in the impedance spectra, corresponding to charge transfer at the Pt counter electrode in the high-frequency region, interfacial electron transfer/recombination at the TiO₂/dye/electrolyte interface in the middle-frequency region, and Nernstian diffusion in the electrolyte in the low-frequency region [20]. To improve cross-sample comparability and avoid non-physical drift of the constant-phase exponent during unconstrained fitting, the present analysis adopted a constrained-fit condition in which P₂ was fixed at 0.81 for all four photoanodes. This fixed value is consistent with the P₂ range reported for related multilayer DSSC photoanodes in the Solar Energy reference study and therefore provides a physically reasonable common basis for comparing the fitted circuit elements [20].

As shown in Figure 8 and summarized in Table 4, the constrained-fit results reveal clear differences among the four photoanodes. Under the common P₂ condition, photoanode a exhibits the lowest diffusion resistance (R__D = 10.12 ± 0.28 Ω), whereas photoanode d shows the highest R__D (21.44 ± 0.58 Ω), indicating that electrolyte-related transport limitations become progressively more pronounced across the sample series. According to prior SCI reports, a larger middle-frequency semicircle or higher R2 is generally associated with a lower charge-recombination rate at the TiO₂/dye/electrolyte interface [5,20]. In this regard, photoanode d displays the highest R₂ (16.03 ± 0.53 Ω), followed by photoanode c (13.99 ± 0.30 Ω), suggesting stronger suppression of interfacial recombination than in photoanodes a and especially b.

By contrast, photoanode b exhibits the highest series resistance (R_s = 31.89 ± 0.12 Ω) and the lowest R₂ (8.65 ± 0.19 Ω), implying comparatively less favorable charge collection under the constrained-fit condition. Photoanode c presents a more balanced combination of moderate R_s, moderate R__D, and relatively high R₂, whereas photoanode a combines a comparatively low R__D with intermediate interfacial resistance. Taken together, the EIS results indicate that the electrochemical behavior of these P25-based photoanodes is governed by coupled changes in series transport, interfacial charge transfer, and electrolyte diffusion rather than by any single resistance term alone. This trend is consistent with prior SCI studies showing that porous TiO₂ microstructure, interfacial transport pathways, and blocking-layer design jointly determine recombination kinetics and DSSC efficiency [11,19,20].

3.4. XRD Analysis of the Composite Films

Figure 9 shows the XRD patterns of the bare ITO substrate, the ITO/TiO₂-CNT intermediate film, and the final ITO/sol–gel SnO₂/sol–gel TiO₂/P25-CNT heterostructure annealed at 450 °C. The bare ITO substrate exhibits characteristic reflections at approximately 30.64°, 35.56°, 50.96°, and 60.56°. After deposition of the P25-CNT layer, a dominant diffraction peak appears near 25.64°, which is assigned mainly to anatase TiO₂ (101), together with a weak feature around 27.7° that is consistent with rutile TiO₂-related contributions. In the final multilayer heterostructure, the TiO₂-related peak near 25.60° remains dominant, while a weak shoulder emerges in the 26.6–26.8° region, which is attributable to SnO₂ (110). These results indicate that the multilayer architecture preserves the crystalline TiO₂ framework while successfully incorporating the SnO₂ interlayer without evidence of phase degradation. The XRD evidence therefore supports phase coexistence in the final heterostructure and is consistent with the improved photovoltaic response observed for the multilayer photoanode.

3.5. Optical Properties of the Films

The optical transmittance spectra presented in Figure 10 show that the bare ITO substrate exhibited approximately 80% transmittance in the visible range of 450–800 nm. After deposition of the P25, P25-MWCNT, and sol–gel SnO₂/sol–gel TiO₂/P25-CNT films, the electrodes still retained more than 66% transmittance over 520–800 nm, indicating that the modified photoanodes preserved sufficient transparency for DSSC operation. The stronger attenuation below approximately 520 nm is reasonably attributed to greater intrinsic absorption and scattering from the oxide/CNT composite layers.

In addition, the optical absorption spectra in Figure 11 indicate that the sol–gel SnO₂/sol–gel TiO₂/P25-CNT multilayer film exhibited stronger light absorption than the conventional P25 film. This enhanced optical response is consistent with an increased effective optical path length created by the multilayer porous structure and with additional scattering from the CNT-containing overlayer. The stronger absorption is therefore compatible with the higher photocurrent density observed in the corresponding J–V measurements.

3.6. Dye-Loading Characteristics of the Photoanodes and Their Implications for Light Harvesting

To further elucidate the dye-adsorption behavior of the different photoanode architectures, the dye loading amount was quantified by a dye-desorption/UV–vis method, and the results are summarized in

Table 5. Absorbance, desorbed dye concentration, dye adsorption amount, and calculated dye loading amount for architectures a–d.

Architecture	Absorbance at 310 nm	Desorbed Dye Concentration (mM)	Dye Adsorption Amount (mg)	Dye Loading Amount (mol cm−2)
a	0.15973	5.047 × 10−3	0.0299	5.05 × 10−8
b	0.19002	5.889 × 10−3	0.0350	5.89 × 10−8
c	0.22819	6.950 × 10−3	0.0413	6.95 × 10−8
d	0.27222	8.175 × 10−3	0.0486	8.18 × 10−8

Note. The dye loading amount was calculated from the desorbed dye concentration using n/A = C × V/A, assuming a desorption solution volume of 5.0 mL and an effective electrode area of 0.5 cm².

and Figure 12. This approach is widely used in dye-sensitized solar-cell (DSSC) studies to evaluate the amount of sensitizer adsorbed on TiO₂-based photoanodes and to relate dye uptake to light harvesting and photocurrent generation [20,21,22].

As shown in Table 5, the absorbance at 310 nm increased monotonically from 0.15973 for architecture a to 0.27222 for architecture d. Correspondingly, the desorbed dye concentration increased from 5.047 × 10⁻³ to 8.175 × 10⁻³ mM, the dye adsorption amount rose from 0.0299 to 0.0486 mg, and the calculated dye loading amount increased from 5.05 × 10⁻⁸ to 8.18 × 10⁻⁸ mol cm⁻². This corresponds to an overall increase of approximately 62.4% from architecture a to architecture d, clearly indicating that the progressively developed architectures provided more accessible sites for dye anchoring and/or improved penetration of the dye solution into the electrode framework. Similar trends have been reported for hierarchical or composite DSSC photoanodes, in which increased accessible surface area and suitable pore characteristics promote higher dye uptake and improved light harvesting [11,21,22].

This trend is mechanistically significant because the working principle of DSSCs relies on a high-surface-area semiconductor film coated with a dye monolayer for photon harvesting. Since the seminal work of O’Regan and Grätzel, it has been well established that the large internal surface area of the TiO₂ film is essential for achieving high dye coverage and efficient light collection [1]. In general, greater dye loading increases the number of photoactive molecules available to absorb incident photons and therefore tends to enhance light-harvesting efficiency and short-circuit photocurrent density, provided that electron transport and interfacial recombination remain properly controlled. For the present architectures, the monotonic increase in absorbance, desorbed dye concentration, and dye loading from a to d suggests that the structural evolution remained within a beneficial regime and had not yet reached a pore-blocking condition. Accordingly, the data in Table 5 and Figure 12 provide direct quantitative evidence that the more advanced architectures favored greater sensitizer uptake, which likely contributed to the higher J_sc and overall conversion efficiency observed for architecture d [20,21,22].

3.7. Overall Discussion

The enhancement in DSSC performance arose from the combined effects of thickness optimization, nanotube-assisted charge transport, and multilayer interface engineering rather than from any single parameter alone. An appropriately controlled TiO₂ thickness increases dye adsorption and photocurrent generation, whereas excessive thickness imposes longer transport distances and more severe recombination. Likewise, a suitable MWCNT loading improves conductivity and electron extraction, whereas excessive addition induces aggregation and performance loss [8,9,10,11].

For clarity, the principal photovoltaic metrics of the representative device configurations are compiled in Table 3. The data show a stepwise improvement from photoanode a to the multilayer architectures, with the most pronounced gain occurring in J_sc and overall efficiency as the compact interfacial layers and CNT-containing overlayer were integrated.

As summarized in Table 3, the pristine P25-TiO₂ photoanode delivered an efficiency of 5.93 ± 0.20% at the optimized thickness of 9.11 μm. Incorporation of 0.025 wt% MWCNTs increased the efficiency to 6.04 ± 0.20%, indicating that CNT incorporation alone provided a measurable but modest benefit. The addition of the compact sol–gel TiO₂ interlayer further increased the efficiency to 6.74 ± 0.20%, whereas the complete SnO₂/sol–gel TiO₂/P25-CNT multilayer achieved the highest J_sc (16.22 mA/cm²) and the highest overall efficiency (6.89 ± 0.30%). This progression is fully consistent with the EIS evidence showing that charge collection and interfacial recombination were most effectively improved in the multilayer design.

External benchmarking against representative SCI reports is provided in

Table 6. Representative SCI reports on CNT-modified and multilayer DSSC photoanodes compared with the present study.

Study/Photoanode	V_oc (V)	J_sc (mA/cm2)	η (%)	Key Finding
Lee et al. [5], low-temperature MWCNT/TiO2 porous film	0.781	9.08	5.02	Optimum at 0.1 wt% MWCNT
Ahn et al. [15], MWCNT-embedded TiO2 nanowires	0.67 ± 0.01	10.72 ± 0.21	5.03 ± 0.35	5 wt% MWCNT condition
Lin et al. [11], bilayer MWCNT–TiO2/TiO2 photoelectrode	—	—	~100% increase	Approximate efficiency doubling vs. single-layer TiO2
Present study, P25-CNTs	0.74	13.96	6.04	CNT-only reference in this work
Present study, sol–gel SnO2/sol–gel TiO2/P25-CNTs	0.74 ± 0.02	16.22 ± 0.40	6.89 ± 0.30	Best multilayer architecture in this work

to place the present results in context.

As shown in Table 6, representative CNT-modified photoanodes reported by Lee et al. and Ahn et al. delivered efficiencies of approximately 5.0%, whereas the bilayer MWCNT–TiO₂/TiO₂ photoelectrode reported by Lin et al. showed a substantial relative improvement compared with a single-layer reference [5,11,15]. Against this background, the CNT-only P25-CNT photoanode in the present study achieved 6.04 ± 0.20%, and the fully integrated sol–gel SnO₂/sol–gel TiO₂/P25-CNT architecture reached 6.89 ± 0.30%. These comparisons indicate that CNT incorporation alone is beneficial, but that combining CNTs with compact interfacial oxide engineering produces a stronger gain in photocurrent generation and overall conversion efficiency, in agreement with the multilayer DSSC behavior reported in the Solar Energy reference study [20].

4. Conclusions

Carbon-nanotube-integrated multilayer photoanodes were successfully realized for dye-sensitized solar cells. The performance evolution observed across the electrode series shows that the main improvement did not originate from CNT incorporation alone. The optimized pristine P25-TiO₂ photoanode, with a thickness of 9.11 μm, reached a power-conversion efficiency of 5.93 ± 0.20%, and the addition of 0.025 wt% MWCNTs increased this value only marginally to 6.04 ± 0.20%.

A substantially larger gain was obtained after introduction of the sol–gel TiO₂ compact nanolayer, which raised the efficiency to 6.74 ± 0.20%, while the complete sol–gel SnO₂/sol–gel TiO₂/P25-CNT architecture achieved the highest efficiency of 6.89 ± 0.30%.

The contribution of the present work lies in the deliberate organization of known material components into a nanostructured multilayer photoanode with differentiated interfacial functions. In this configuration, the sol–gel SnO₂ nanolayer facilitates electron extraction, the sol–gel TiO₂ nanolayer improves compact interfacial coverage and suppresses substrate-side recombination, and the porous P25-CNT overlayer preserves dye adsorption while providing conductive pathways for charge transport. The stepwise enhancement from the pristine P25 electrode to the dual-interlayer architecture, therefore, reflects the effect of interfacial functional partitioning rather than a simple additive combination of conventional modification methods.

Equally important, this architecture is achieved through low-cost, solution-processable sol–gel deposition and does not rely on complex fabrication routes. The present results, therefore, support sequential dual-sol–gel nano-interlayer engineering as a practical design strategy for DSSC photoanodes, particularly when simultaneous improvement in charge extraction, recombination control, and sensitizer utilization is required.