Sol–Gel Electrophoretically Deposited TiO₂–Multiwalled Carbon Nanotube–SiO₂ Thin-Film Electrode with High Photoelectrochemical Activity

Abstract

Hydrogen production via water splitting has been extensively researched for its environmental friendliness, energy efficiency, and renewability. This study describes the development of TiO₂–multiwalled carbon nanotube (MWCNT)–SiO₂ composite thin-film electrodes via electrophoretic deposition (EPD) from a 2-propanol solution of MWCNTs including TiO₂ and SiO₂ gels. The TiO₂ and SiO₂ gels were prepared via the sol–gel method and by mixing in varying weight ratios to enhance the efficiency of photoelectrochemical water splitting. Dual sol–gel EPD incorporates MWCNTs with a C/TiO₂ molar ratio of ≥0.25 while varying the TiO₂/SiO₂ molar ratio from 5 to 14; the electronic conductivity is improved owing to the pristine graphene structure of the MWCNTs along with hydrophilicity imparted by SiO₂. In addition, the volume of SiO₂ sol influences the anatase-to-rutile ratio, the TiO₂ crystal size, and chemical bonds, thereby affecting the formation of new energy levels. The optimal volume of SiO₂ sol results in elevated ultraviolet–visible absorbance, attributed to midgap states generated by a high anatase-to-rutile ratio and Ti–O–Si formation, further leading to a substantial effective carrier density for the photoelectrochemical water-splitting reaction. Furthermore, the valence band maximum (VBM) and conduction band minimum, estimated using ultraviolet photoelectron and ultraviolet–visible spectroscopies, exhibited a downward shift with increasing SiO₂ sol volume, followed by an upward shift; meanwhile, the Fermi level in a Na₂SO₄ solution under stimulated solar light deepened. The highest photoelectrochemical performance is achieved at the optimal SiO₂ sol volume, where the VBM is deep enough to minimize the water-splitting overpotential, and the flat-band potential aligns with the set potential, thereby reducing band bending with a negligible hole depletion layer at the TiO₂–solution interface. The best TiO₂–MWCNT–SiO₂ composite exhibits a photocurrent ∼7.4 times higher than that of a TiO₂–MWCNT electrode.

1. Introduction

The TiO₂ electrode has been extensively studied for its photocatalytic properties since Honda and Fujishima demonstrated the photocatalytic water-splitting phenomenon using a photoelectrochemical cell in the early 1970s.¹ Over the past decades, TiO₂ has attracted widespread attention owing to its excellent photocatalytic properties, low cost, nontoxicity, and high stability^2,3 in various fields, including environmental remediation,⁴ air purification,⁵ and solar energy conversion.⁶ Despite its numerous advantages, the wide bandgap of TiO₂ (3.0–3.2 eV) limits its photocatalytic efficiency to only the ultraviolet (UV) light region, which comprises only a small fraction of the solar spectrum.⁷ Additionally, the fast recombination of photogenerated electron–hole pairs further impedes its photocatalytic performance.⁸ To address these limitations, various strategies have been employed, including doping with metal and nonmetal elements,⁹ coupling with other semiconductors,¹⁰ and designing novel nanostructures.^11,12 Recent studies have reported promising results from compositing TiO₂ with multiwalled carbon nanotubes (MWCNTs), including enhanced visible light absorption and improved charge separation, leading to its enhanced photoelectrochemical performance for the water-splitting reaction.^12,13

In recent years, researchers have attempted to improve the photocatalytic performance of TiO₂ by incorporating SiO₂ into its structure.¹⁴ This strategy was first reported in the late 1990s,¹⁵ and since then, numerous studies have explored the underlying mechanisms and optimized SiO₂ content to enhance photocatalytic performance of TiO₂.¹⁶ The addition of SiO₂ to TiO₂ suppresses the growth of TiO₂ crystals, leading to smaller particle sizes and higher surface areas.¹⁷⁻²¹ Moreover, SiO₂ has been reported to inhibit the phase transformation from anatase to rutile,^18,19,22 which is beneficial for maintaining the photocatalytic properties of TiO₂.²³ The hydrophilic nature of SiO₂^24,25 further enhances the photocatalytic performance of TiO₂ by promoting the adsorption of water molecules and reactants onto the catalyst surface.^17,26 However, excessive addition of SiO₂ can form an insulating layer that impedes the charge transfer process and potentially decreases the photocatalytic efficiency.²⁷ Researchers have also investigated the structural and optical properties of TiO₂–SiO₂ films with different SiO₂/TiO₂ ratios,^28,29 bilayer films,³⁰ core–shell nanostructures,^31,32 porous structures,^33,34 and TiO₂ grafted on SiO₂ spheres.³⁵ These investigations focused on enhancing pigment properties,²¹ optical properties as waveguides,³⁶ photocatalytic activity for the decomposition of methylene blue³⁵ and phenanthrene,³² reduction of nitric oxide,²⁹ removal of acetaminophen³³ and oxytetracycline,³⁴ and oxidation of carbon monoxide.³¹

In photocatalysis applications, carbon nanotubes (CNTs) serve as an advanced material with exceptional electrical conductivity³⁷ to mitigate the insulating effect of SiO₂ and enhance charge separation and transfer efficiency. CNTs also provide a large surface area and promote the adsorption of reactants, thereby improving the overall photocatalytic performance of the composite material.¹² In addition, the excellent hydrophilic properties of SiO₂ can enhance the applicability of TiO₂ in water-based photocatalytic systems, even with hydrophobic CNTs. However, only a few studies have been reported on the photoelectrochemical properties of ternary composite materials of TiO₂, SiO₂, and CNTs.^33,38

Several synthesis methods, such as atomic layer deposition,³⁹ sol–gel,^40,41 high-temperature self-organization,⁴² and electrospinning,⁴³ have been investigated to tailor composites of TiO₂ and SiO₂ for specific applications. The sol–gel process involves the conversion of a colloidal suspension (sol) into a solid network (gel) through chemical reactions such as hydrolysis and condensation, offering precise control over the composition, structure, and properties of the final product. However, the final products often require extensive processing and posttreatment procedures. The dual sol–gel process, which combines the sol–gel synthesis of both TiO₂ and SiO₂, presents a unique approach to produce oxide composites with improved chemical and physical properties.^43,44 This method offers several advantages over traditional synthesis techniques, including simultaneous integration of multiple components in a single step, thereby reducing the complexity of the fabrication process. For example, Czeck et al. synthesized MWCNT–TiO₂–SiO₂ nanocomposites as a photocatalyst using an ultrasonic-assisted sol–gel method for the removal of acetaminophen³³ and phenol/methyl orange.³⁸ In our recent study, TiO₂/MWCNT photoelectrodes were prepared by combining sol–gel and electrophoretic deposition (EPD),^12,45 both of which are simple, clean, and ecofriendly techniques that prevent the use of toxic chemicals and high-energy processes. Herein, TiO₂–MWCNT–SiO₂ ternary composite thin films are synthesized for the first time by combining the dual sol–gel method with EPD for photoelectrode preparation. The TiO₂–MWCNT–SiO₂ ternary composite thin-film electrode demonstrated enhanced photoelectrochemical performance with an optimal mixing ratio, outperforming pristine TiO₂ and binary films of TiO₂–SiO₂ and TiO₂–MWCNT. Moreover, this study first revealed the significant effects of the added SiO₂ sol volume on the band structure of the TiO₂ films in the TiO₂–MWCNT–SiO₂ ternary composite thin films, showing how these changes impact photocatalytic performance by altering both UV–visible light absorbance and charge transfer properties, along with the film morphology, chemical properties, hydrophilicity, and electrical properties, of the composite thin film.

2. Experimental Methods

2.1. Electrode Preparation

Two sol–gel samples were prepared in advance. Initially, TiO₂ gel was synthesized using titanium isopropoxide as a precursor, as described in our previous study.¹² Next, the SiO₂ gel was prepared by mixing 1 mL of tetraethyl orthosilicate (TEOS) (Sigma-Aldrich, 86578) as a precursor with 4 mL of anhydrous ethanol. After the mixture was stirred for 1 h, 0.5 mL of ultrapure water and 0.2 mL of nitric acid were added, followed by an additional 30 min of stirring to obtain the SiO₂ gel. In this process, TEOS acted as the precursor for the formation of SiO₂ gel, anhydrous ethanol served as the solvent, ultrapure water promoted hydrolysis, and nitric acid adjusted the pH, similar to the preparation of the TiO₂ sol. To prepare the MWCNT dispersion solution, 20 mL of 2-propanol was used as the solvent, into which 0.00070 g of MWCNTs (Sigma-Aldrich, 659258) with lengths of 5–9 μm and radii of 110–170 nm, along with 0.0067 mL of benzyl alcohol (BA), were sequentially added. After the addition of 256 μL of TiO₂ sol to the MWCNT dispersion solution, SiO₂ sol was added in arbitrary proportions, ranging from 10 to 80 μL, followed by 30 min of each ultrasonication and magnetic stirring. When 10 μL of SiO₂ sol was added, the weight ratio (molar ratio) of Ti and Si to C in the bath was approximately 1 (0.25) and 0.032 (0.014), respectively.

For the EPD, an indium tin oxide (ITO) substrate was used as the cathode, and a titanium/platinum (Ti/Pt) plate served as the anode.¹² A voltage of 10 V was applied for 10 min, which is shorter than that in our previous study (20 min)¹² using 2-propanol as a solvent but longer than that using mixed solvents of acetone and 2-propanol (5 min).⁴⁵ The bath temperature during EPD was 23 ± 1 °C. After EPD, the deposited composite was air-dried and annealed at 400 °C for 1 h. This temperature was chosen to avoid MWCNT decomposition, considering the higher thermal conductivity of SiO₂ compared to TiO₂.

2.2. Analysis

Various analytical techniques were employed to provide a comprehensive understanding of the material properties and photoelectrochemical performance. Scanning electron microscopy (SEM) images and energy-dispersive X-ray spectroscopy (EDS) spectra were obtained by using a JSM-7800F instrument (JEOL Ltd.). The water contact angle (WCA) was measured using a B100 instrument (Asumi Giken Ltd.) to determine the surface wettability. UV–visible (vis), X-ray diffraction (XRD), Raman, and X-ray photoelectron spectroscopy (XPS) analyses were conducted under the same conditions as in our previous study¹² to analyze the optical properties, crystal structures, film morphology, and chemical bindings. Ultraviolet photoelectron spectroscopy (UPS) was conducted to analyze the valence band maximum (VBM) and work function (W.F.) using a KRATOS ANALYTICAL spectrometer (SHIMADZU Corp.), using a He I light source operated at a power of 21.22 eV, with a bias voltage set at −6 V under ultrahigh-vacuum conditions to ensure data accuracy. Data processing and calibration, including baseline correction, peak fitting, and determination of peak positions and intensities, were performed using MultiPak (Ulvac-PHI Inc.).

Electrochemical measurements, including Mott–Schottky (MS) analysis, electrochemical impedance spectroscopy (EIS), photoresponse measurement, and linear sweep voltammetry (LSV), were performed using an SP-150e-TM electrochemical workstation (Bio-Logic Science Instruments Ltd.) or VersaStat 3 (Amtek, Inc.). Here, the TiO₂ composite thin-film electrode served as the working electrode; Ag/AgCl (saturated KCl) served as the reference electrode; and Pt wire was used as the counter electrode in a three-electrode cell. MS analysis was performed in a 1 M Na₂SO₄ solution by measuring the impedance at six points per decade in logarithmic spacing between 10 Hz and 1 MHz across different potentials from −2 V to +2 V in 10 mV increments under continuous illumination from a PECL01 solar simulator light source (Peccell Technologies, Inc.). EIS, photoresponse measurements, and LSV were conducted under the same conditions as in our previous study.^12,45 For EIS, photoresponse measurements, and LSV, 0.1, 1, and 0.5 M Na₂SO₄ solutions were employed as electrolytes, respectively. EIS and photoresponse measurements were carried out by applying 0.245 V vs Ag/AgCl (saturated KCl). LSV was measured from −2.5 to +2.5 V vs Ag/AgCl (saturated KCl) at a scan rate of 5 mV/s. During the electrochemical measurements, the intensity of the incident light on the photoelectrodes was maintained at ∼90 mW/cm², while aluminum foil was used to block stray light and prevent light leakage from sources other than the solar simulator.

3. Results and Discussion

3.1. SEM and EDS Images

Figure 1a–k shows SEM images at 5,000× magnification of pristine TiO₂, TiO₂–MWCNT, TiO₂–SiO₂-30, TiO₂–MWCNT–SiO₂-x, hereby x is the added volume of the SiO₂ sol. Figure 2a depicts the molar ratios of C, TiO₂, SiO₂, and In₂O₃ in each sample. Meanwhile, Figure 2b displays the molar ratios of TiO₂ and SiO₂ to C and TiO₂ to SiO₂ in each sample. In the pristine TiO₂ sample, darker particles were observed on the substrate. Based on the EDS mapping image shown in Figure S1, the potential darker particles are TiO₂ because Ti and O were detected almost over the entire surface. However, a greater amount was detected in the darker areas with an atomic ratio of approximately 1:2. In contrast, Na and C as impurities from the electrolyte were evenly detected over the entire surface. In the EDS image, the contrast tends to be dark for elements with a low atomic number. The dark contrast for TiO₂ is possibly due to the larger content of O than ITO, which has a relatively smaller atomic number than other components such as In in ITO and Na in impurities. The TiO₂–SiO₂ sample displayed a smooth film with noticeable cracks, likely caused by stress during the drying or annealing process.⁴⁶ For the TiO₂–MWCNT sample, the substrate was almost entirely covered by well-distributed MWCNTs, exhibiting a tubular structure with lengths of 5–7 μm, which is consistent with the MWCNT size used in this study and no visible cracks on the surface. All TiO₂–MWCNT–SiO₂-x samples exhibited complete film coverage on the substrate with visible MWCNT tubular structures embedded within. The samples with 10–20 μL of SiO₂ displayed substantially large cracks, measuring up to ∼10 μm, with residual MWCNT structures observed within the cracks. As the SiO₂ volume increased, the size of the cracks gradually decreased, reaching ∼2 μm in size. The decrease in the size of the cracks above 30 μL is attributed to decreased stress resulting from the reduced film thickness⁴⁵ and a less-packed structure, as suggested by the increased molar ratio of In₂O₃ detected via EDS (Figure 2a) and the film thickness (Figure S2) measured using the cross-sectional SEM image. Figure 2a shows that the carbon content in TiO₂–MWCNT–SiO₂ is more than twice that of the composite prepared by coprecipitation of the gel prepared via the sol–gel method, while SiO₂ content is intermediate compared to reported values for similar composites.^33,38 The In₂O₃ content in TiO₂–MWCNT–SiO₂-30 is considerably lower than those in the other samples, suggesting a more complete surface coverage of the ITO substrate with a thick composite film (Figure S2). As shown in Figure 2b, as the volume of added SiO₂ sol increases, the ratio of SiO₂ to C slightly increases when the volume is above 40 μL, while the TiO₂ to C ratio decreases above 0 μL. The molar ratio of TiO₂ to C for TiO₂–MWCNT–SiO₂-x is more than 4.4 times that of the electrophoresis bath (0.25) and 3.9 times that of the film prepared without SiO₂ sol (0.28), suggesting that SiO₂ sol facilitates the deposition of TiO₂ sol more than MWCNT on the ITO substrate. The positive effect of SiO₂ sol on the deposition rate of TiO₂ sol is discussed in terms of particle size and zeta potential of the bath in the next paragraph.

Figure 1.

SEM images at 5,000× magnification of (a) TiO₂, (b) TiO₂–SiO₂-30, (c) TiO₂–MWCNT (TiO₂–MWCNT–SiO₂-0), and (d–k) TiO₂–MWCNT–SiO₂-x, where x is the added volume of SiO₂ sol.

Figure 2.

(a) Molar ratios of C, TiO₂, SiO₂, In₂O₃, and C in each sample. (b) Molar ratios of TiO₂ and SiO₂ to C, and TiO₂ to SiO₂ in each sample.

Figure 3a–k shows the light scattering intensity distribution from different equivalent sphere diameters of the electrophoresis bath used for preparing each sample. Figure 3l shows the zeta potential of each bath calculated with the reflectance, viscosity, and dielectric constant of 2-propanol, which are 1.37 (25 °C), 2.32 (actual value),⁴³ and 19.9 (25 °C), respectively. Adding SiO₂ sol to TiO₂ sol increased the light scattering intensity from aggregates with small equivalent sphere diameters <0.5 μm. This is owing to the reduced aggregation of TiO₂ sol in the 2-propanol solvent by interacting with SiO₂ sol, which has a lower dielectric constant of SiO₂ compared to TiO₂. When SiO₂ sol was added to the TiO₂ sol and MWCNT mixture, the light scattering intensity from aggregates with large equivalent sphere diameters above 2 μm increased. The average zeta potential shifted to negative values with SiO₂ sol volumes of 30 and 40 μL, indicating that an equivalent sphere diameter of 2 μm or more is likely due to MWCNTs with a negative zeta potential. SiO₂ sols, with a smaller dielectric constant than the TiO₂ sol, preferentially interact with hydrophobic MWCNTs, forming large agglomerates of SiO₂ and MWCNTs. Thus, when SiO₂ sol is present, the slow electrophoresis rate of these large agglomerates containing a high concentration of MWCNTs toward the cathode and/or the migration of negatively charged particles toward the anode results in a reduced molar ratio of C in the film deposited on ITO serving as the cathode above 30 μL of SiO₂ sol (EDS result). When the volume of SiO₂ sol increased from 40 to 70 μL, the light scattering intensity from aggregates with small equivalent sphere diameters of ∼0.5 μm was detected, similar to the TiO₂–SiO₂ case. The proportion of TiO₂ in the film decreased while SiO₂ slightly increased above 40 μL (Figure 2b), suggesting that these small particles originated from excess SiO₂ sol that interacted minimally with MWCNTs. The average zeta potential shifted toward a positive value with increasing SiO₂ sol volumes above 40 μL, reflecting the increased contribution of the SiO₂ sol to the zeta potential. Excess small SiO₂ sol in the bath possibly resulted in films with a decreased TiO₂/SiO₂ molar ratio. These small SiO₂ sols, deposited at the beginning of electrophoresis, can reduce the electrical conductivity of the substrate surface, slow the rate of film formation, and result in the formation of a thinner film compared to other samples prepared with less SiO₂ sol. Alternatively, the thin ternary composite film may only remain during EPD owing to the poor stability of the film resulting from the large mechanical stress on the SiO₂ deposited underneath.¹⁹

Figure 3.

Light scattering intensity distribution from different equivalent sphere diameters of the electrophoresis bath for preparing (a) TiO₂, (b) TiO₂–SiO₂-30, (c) TiO₂–MWCNT (TiO₂–MWCNT–SiO₂-0), and (d–k) TiO₂–MWCNT–SiO₂-x. (l) Zeta potential of each bath.

3.2. WCA Measurement

Figure 4 shows the WCA values for each sample. WCA was notably influenced by the addition of MWCNTs and SiO₂ sol to the TiO₂ sol. The addition of MWCNTs increased WCA compared to pristine TiO₂, suggesting increased hydrophobicity owing to the hydrophobic nature of MWCNTs and the increased surface roughness⁴⁷ of the film with MWCNTs. In contrast, the addition of SiO₂ to TiO₂ decreased the WCA, indicating increased hydrophilicity. As previously reported using a sol–gel method,¹⁷ the increased Lewis acidity of the surface by partially replacing Ti on TiO₂ with Si induces additional hydroxyl groups on the surface of composite films can increase hydrophilicity. In addition, the absorbed impurities in the atmosphere are attached as carboxylic acids because of their bidentate binding onto the TiO₂ surface with changing WCA.⁴⁸ Therefore, MWCNTs and SiO₂ potentially change the adsorbed species and their orientation depending on the composition via a change in the atomic arrangement of the TiO₂ surface caused by a change in the crystallite size (Section 3.4) and chemical bonding (Section 3.6). These changes have a complex effect on the WCA.

Figure 4.

WCA of TiO₂, TiO₂–SiO₂-30, TiO₂–MWCNT (TiO₂–MWCNT–SiO₂-0), and TiO₂–MWCNT–SiO₂-x, where x is the added volume of the SiO₂ sol.

3.3. UV–vis Spectroscopy

The UV–vis absorption spectra of synthesized TiO₂ and its composites with MWCNTs and SiO₂ (Figure 5a) demonstrate noticeable alterations in optical properties upon nanomaterial integration. The pristine TiO₂ displays strong absorption in the UV region and a wide bandgap characteristic of the anatase phase of TiO₂. As shown in Figure 5a,b, the absorbance remarkably increases with the addition of MWCNTs owing to the high loading of TiO₂, consistent with our previous study.⁴⁵ For TiO₂–MWCNT–SiO₂-x, large UV absorbances are observed between 20 and 40 μL of SiO₂ sol under all conditions, where bandgaps were small. The addition of MWCNTs to the TiO₂ and TiO₂–SiO₂ induces a red shift in their respective absorption edges (Figure 5a), suggesting the formation of midgap states that can improve photoelectrochemical performance under sunlight and visible light.^12,45 However, a decrease in the main bandgap energy of TiO₂,^33,38,46 owing to the introduction of MWCNTs was not confirmed (Figure 5c,d). This is due to the minimal chemical bonding at the interface between TiO₂ and MWCNTs. This is consistent with the fact that MWCNTs have few graphene structural defects based on the low I_d/I_g ratio in the Raman spectrum (Section 3.5). However, the main bandgap values might not be accurately measured for TiO₂–MWCNT and TiO₂–MWCNT–SiO₂-40, which have wide frequency ranges with measured absorbance values exceeding 3 (refer to page 4 and Figure S3 for more details). Conversely, an increase in bandgap owing to SiO₂^17,19,22 was confirmed for TiO₂–SiO₂-30 and TiO₂–MWCNT–SiO₂-x when x exceeded 50 μL. Beyond this concentration, the bandgap stabilized, indicating saturation of the bandgap modification capacity of SiO₂ within the composite. This behavior may be attributed to the saturated coverage of active sites on TiO₂ by SiO₂, leading to maximal alteration of the electronic properties at this specific SiO₂ concentration. Further discussion of the importance of 50 μL as a turning point is provided in Sections 3.6 and 3.7, focusing on the chemical features and band structure.

Figure 5.

(a) UV spectra of TiO₂, TiO₂–SiO₂-30, TiO₂–MWCNT (TiO₂–MWCNT–SiO₂-0), and TiO₂–MWCNT–SiO₂-x, where x is the added volume of SiO₂ sol; (b) peak absorbance of each sample; (c) Tauc plot for each sample; and (d) bandgap of each sample calculated from Tauc plot.

3.4. XRD Analysis

Figure 6a shows the XRD patterns of the pristine TiO₂, TiO₂–SiO₂-30, TiO₂–MWCNT, and TiO₂–MWCNT–SiO₂ samples prepared with different SiO₂ sol volumes. Figure S4a–k displays an enlarged XRD pattern of each sample in the 22–29° range. Peaks attributed to the anatase(101) and rutile(110) phases are observed in all samples within this range. Furthermore, the addition of MWCNTs increased peak heights, which reflects the increased amount of TiO₂, consistent with UV spectra results. For samples prepared with SiO₂ sol, another peak indexed to SiO₂(100) appeared in the 22–24° range. When the SiO₂ addition increased above 50 μL, the peak intensity of anatase(101) decreased, reflecting the reduction of the TiO₂ amount, as supported by UV spectra (Figure 5b) and EDS data (Figure 2a).

Figure 6.

(a) XRD patterns of TiO₂, TiO₂–MWCNT, TiO₂–SiO₂-30, and TiO₂–MWCNT–SiO₂-x, where x is the added volume of SiO₂ sol. (b) Crystallite sizes of anatase and rutile phases of TiO₂ and SiO₂ in each sample. (c) Anatase-to-rutile ratios for each sample.

Figure 6b compares the crystallite sizes of the anatase and rutile phases of TiO₂ and SiO₂ in each TiO₂–MWCNT–SiO₂-x sample, calculated from the half-width of each peak using Scherrer’s formula.¹² The correlation between the crystallite size and photocatalytic performance can be attributed to several factors. Small crystallites provide a large surface area with numerous active sites for photoelectrochemical reactions and may contain many defect levels in the bandgap, enhancing electron–hole pair separation and photoelectrochemical performance.^12,45 The addition of MWCNTs decreased the crystallite sizes of SiO₂, as well as the anatase and rutile phases of TiO₂, owing to preferential nucleation over its growth, similar to previous reports.^33,38,45 Variations in SiO₂ contents in the TiO₂–MWCNT–SiO₂ samples markedly affected the crystallite sizes of SiO₂ and the rutile phase of TiO₂ but not the anatase phase of TiO₂. With an increase in the amount of added SiO₂ sol, the crystallite size of the rutile phase increased up to 50 μL and then decreased, whereas that of SiO₂ decreased up to 40 μL and then increased. The suppression of TiO₂ crystallite growth by the addition of SiO₂ was reported to reduce the crystallite size of rutile TiO₂ in TiO₂/SiO₂ composite films by increasing the SiO₂ content in a microsize range.^17,19 Herein, a decrease in the TiO₂/SiO₂ ratio above 40 μL of SiO₂ sol led to an increased SiO₂ crystallite size above 6 nm, inhibiting the growth of the rutile phase. Moreover, the crystallite size of anatase TiO₂ in this study is between 3 and 5.5 nm, smaller than the 5 and 8 nm range reported for ternary composites prepared via coprecipitation of the gel prepared via sol–gel methods.^33,38 Owing to the higher content of MWCNTs in the ternary composite than in other reports,^33,38 numerous MWCNTs interacted with BA via pie–pie interaction, enabling many hydroxy groups to face outward. Thus, several hydroxy groups interacted with TiO₂ via hydrogen bonds during the crystallization stage to accelerate nucleation, resulting in the small crystallite sizes of TiO₂, as reported in previous studies.^45,49,50

Figure 6c shows the anatase/rutile ratios for each sample. SiO₂ is known to inhibit anatase-to-rutile phase transformation.^18,19,22,51 Herein, TiO₂–MWCNT–SiO₂-40 showed the highest anatase ratio, indicating that 40 μL of SiO₂ effectively inhibited the anatase-to-rutile phase transformation. However, further addition of SiO₂ sol to the mixture of MWCNT and TiO₂ sol decreased the anatase ratio of TiO₂. The anatase ratio remained high when the volume of the added SiO₂ sol was between 20 and 50 μL, where the small SiO₂ particle size provided a large surface area for the interaction with TiO₂.

3.5. Raman Spectroscopy

Raman spectroscopy was employed to investigate chemical bindings in the TiO₂–MWCNT–SiO₂ composite films. Raman spectra revealed distinct features of the different components in the composites (Figure 7a–k). For TiO₂, the anatase phase exhibited dominant peaks at approximately 143 cm^–1 (E_1g mode), 197 cm^–1 (E_2g mode), and 396 cm^–1 (A_1g mode), whereas the rutile phase showed relatively weak peaks at 144 cm^–1 (E_g mode) and 447 cm^–1 (A_1g mode).⁵² The increased presence of the anatase phase and suppression of the rutile phase, especially in films with high SiO₂ contents, suggests that SiO₂ plays a role in inhibiting the anatase-to-rutile phase transformation, as also suggested by XRD.

Figure 7.

(a–k) Raman spectra of TiO₂, TiO₂–SiO₂-30, TiO₂–MWCNT (TiO₂–MWCNT–SiO₂-0), and TiO₂–MWCNT–SiO₂-x, where x is the added volume of SiO₂ sol. (l) I_d/I_g ratio for each sample.

For SiO₂, the Raman feature appeared around 265 cm^–1, corresponding to Si–O–Si bending vibration mode,⁵³ particularly in the 40 μL sample. This observation suggested that the successful incorporation of SiO₂ into the composites as also suggested by the XRD results.

All samples containing MWCNTs exhibited two main characteristic peaks: the D peak at ∼1,350 cm^–1, representing sp³ carbons from impurities and defects in the graphene structure, and the G peak at ∼1,580 cm^–1, representing the graphene structure of carbon materials. As shown in Figure 7l, the addition of SiO₂ sol decreased the I_d/I_g ratio, indicating a decreased defect density in the graphene structure of the MWCNTs. The I_d/I_g ratio typically decreases when MWCNTs are introduced into SiO₂ nanocomposites.^33,54 With increasing SiO₂ sol volume, the I_d/I_g ratio initially decreased and then slightly increased with excessive volumes above 50 μL. The lowest I_d/I_g ratios were observed for the TiO₂–MWCNT–SiO₂-30, 40, and 50 samples, suggesting the fewest defects in MWCNTs for these compositions, which may enhance charge transfer and improve photoelectrochemical performance owing to the excellent conductivity of the graphene structure. Between 30 and 50 μL, the small SiO₂ crystallites (Section 3.4) provided a large contact area, enhancing their interaction with TiO₂ and introducing a Lewis acidic surface with abundant hydroxyl groups.¹⁷ Consequently, the surface may interact via the hydroxyl group of BA, while the graphene structure of MWCNTs interacts with its benzene group, avoiding Ti–C and Ti–O–C bond formation by breaking sp² bonds in MWCNTs. The I_d/I_g ratio in this study (<1) is lower than that of a ternary composite prepared via coprecipitation of the gel prepared via sol–gel methods,³³ partially because of a high content of MWCNTs in the film, resulting in an abundant graphene structure without breaking π bonds by TiO₂.

3.6. XPS

In the wide XPS spectra of pristine TiO₂, TiO₂–MWCNT, TiO₂–SiO₂, and TiO₂–MWCNT–SiO₂-x, all 11 samples exhibit four distinct peaks corresponding to Ti 2p, Si 2p, O 1s, and C 1s (data not shown). Notably, the Si 2p peak is highly pronounced in the TiO₂–SiO₂-30, TiO₂–MWCNT–SiO₂-20, and TiO₂–MWCNT–SiO₂-30 samples, indicating the successful incorporation of SiO₂ into several nanometers deep from the surface of these samples. The observed O1s peak can be ascribed to oxygen atoms within the TiO₂ lattice, whereas the Ti2p peak originates from the Ti atoms in the TiO₂ lattice. The C1s peak, present in all samples, may originate from the adsorption of carbon-containing species on the sample surfaces during synthesis, processing, or handling. In the pristine TiO₂ and TiO₂–SiO₂ samples, the C1s peak serves as a calibration reference for the binding energy scale.

The high-resolution Ti 2p spectra shown in Figure 8a reveal two characteristic peaks corresponding to the Ti 2p_3/2 and Ti 2p_1/2 orbitals, confirming the presence of Ti⁴⁺ in all samples. The incorporation of MWCNTs increased the binding energies of the Ti–O bonds, indicating enhanced electronic interactions of Ti–C between TiO₂ and MWCNTs.^12,55 With MWCNTs, the addition of SiO₂ sol caused a shift of Ti 2p peaks to high energies, with the largest shift observed at 20 μL, attributed to the replacement of Ti on TiO₂ by Si⁵⁶ was confirmed under all conditions. In samples with low I_d/I_g ratios (SiO₂ sol volumes between 30 and 50 μL), the shift in Ti 2p binding energy was minimal, indicating reduced Ti–C bonding, which introduces defects into the graphene structure of the MWCNT, as discussed in Section 3.5. Among these samples with low I_d/I_g ratios, the bandgap determined from UV spectra narrowed at 30 and 40 μL but not at 50 μL. This suggests that the moderate number of Ti–C bonds at the interface has minimal effect on the bandgap, as determined from the Tauc plot, as the bandgap reflects both surface and bulk TiO₂ properties.

Figure 8.

(a) High-resolution Ti 2p spectra, (b) Si 2p spectra, (c) O 1s spectra, and (d) C 1s spectra of TiO₂, TiO₂–SiO₂-30, TiO₂–MWCNT (TiO₂–MWCNT–SiO₂-0), and TiO₂–MWCNT–SiO₂-x, where x is the added volume of SiO₂ sol.

The high-resolution C 1s spectra (Figure 8b) for the pristine TiO₂ and TiO₂–SiO₂-30 samples reveal a dominant peak owing to the presence of carbon impurities. The incorporation of MWCNTs introduced new peaks corresponding to various functional groups present in the MWCNT structure, such as C=C (sp²),⁵⁵ C–C (sp³),⁵⁵ C–O,⁵⁵ C=O,⁵⁷ and O–C=O.⁵⁵ The presence of these functional groups confirmed the successful integration of the MWCNTs. The SiO₂ content did not considerably affect peak intensities or binding energy positions, indicating its minimal impact on the chemical environment of the C bonds formed in the TiO₂–MWCNT–SiO₂ composites. As a minor change, the peak attributed to C–O appeared sharp in some samples of TiO₂–MWCNT, TiO₂–MWCNT–SiO₂-20, -30, and -80, where the bandgaps determined from the Tauc plot are narrow, except for TiO₂–MWCNT–SiO₂-80. This narrow bandgap is possibly owing to new midgap states formed via the C–O(−Ti) bond.^12,58

The characteristic peaks observed in the Si 2p region are attributed to the Si 2p_3/2 and Si 2p_1/2 signals, indicating the presence of Si⁴⁺ (Figure 8c). In the TiO₂–MWCNT sample, the Si signal primarily originates from the glass of the ITO substrate. The addition of SiO₂ remarkably amplified the Si signals, while the introduction of MWCNTs increased the Si–O binding energy by ∼0.25 eV, indicating strong interactions between MWCNTs and SiO₂.

The high-resolution O 1s spectra (Figure 8d) exhibited two main peaks for pristine TiO₂, attributed to O₂^– and hydroxyl groups (−OH) resulting from oxygen vacancies on the TiO₂ surface. The introduction of SiO₂ led to the formation of Ti–O–Si bonds,^18,19,46,56 suggesting an interaction between Si⁴⁺ and Ti–O, shortening the bond length and increasing the binding energy.⁵⁶ Notably, Ti–O–Si peaks were dominant in TiO₂–MWCNT–SiO₂-20 and -30, suggesting the formation of new midgap states that narrow bandgaps determined from the Tauc plot.

3.7. UPS Spectra

UPS analysis was employed to investigate the ionization energy, which is the energy difference between the VBM and the vacuum level of the TiO₂-based samples. The UPS spectra for each sample are shown in Figure 9a,b. The ionization potential (I.P.) for each sample, determined using eq 1, is shown in Table 1.

1

where hν is the energy of the irradiated UV light (21.22 eV for He I), and W is the energy width of the spectrum (E_min – VBM position). E_min is the high-binding-energy cutoff (secondary-electron cutoff).

Figure 9.

Enlarged view of (a) high-binding-energy cutoff and (b) VBM edge in the UPS spectra of TiO₂, TiO₂–SiO₂-30, TiO₂–MWCNT (TiO₂–MWCNT–SiO₂-0), and TiO₂–MWCNT–SiO₂-x, where x is the added volume of SiO₂ sol. (c) Summary of electronic structures examined using UPS, UV–vis, and MS analyses.

Table 1. Parameters for the Band Structure of Each TiO₂-Based Sample.

	Bandgap/eV	I.P./eV	W.F. (UPS)/eV	W.F. (MS)/eV
TiO2	3.25	8.54	5.52	4.15
TiO2–MWCNT	3.26	7.99	5.27	3.54
TiO2–SiO2-30	3.42	9.28	4.35	3.97
TiO2–MWCNT–SiO2-10	3.26	8.26	5.29	3.73
TiO2–MWCNT–SiO2-20	3.26	8.35	5.47	4.57
TiO2–MWCNT–SiO2-30	3.26	8.46	5.41	4.65
TiO2–MWCNT–SiO2-40	3.42	8.56	5.41	4.55
TiO2–MWCNT–SiO2-50	3.42	8.38	4.39	4.79
TiO2–MWCNT–SiO2-60	3.42	8.37	5.27	5.07
TiO2–MWCNT–SiO2-70	3.42	8.22	5.15	5.27
TiO2–MWCNT–SiO2-80	3.42	8.34	5.45	5.42

The pristine TiO₂ showed an I.P. of 8.54 eV, indicative of its inherent wide bandgap and strong UV absorption capabilities, which are larger than the 7.46⁵⁹ and 7.52 eV⁶⁰ values reported in the literature, partially owing to the incorporation of impurities such as carbon. The introduction of MWCNTs reduced the I.P. to 7.97 eV, suggesting an increased absorption of visible light. While a reduction in I.P. by hybridization with oxidized MWCNTs has been observed in other studies, the I.P. value for TiO₂–MWCNT in this study is larger than the 5.51 eV reported in the literature.⁵⁹ This reduction in the I.P. is expected to decrease the energy barrier for electron transfer to reductive species, potentially enhancing the photocatalytic performance for the water-splitting reaction under solar irradiation. The incorporation of SiO₂ further modified the VBM position, with TiO₂–SiO₂ exhibiting an increased I.P. of 9.26 eV, likely owing to alterations in the surface chemistry of TiO₂ or surface dipole orientation contributed by SiO₂, thus affecting the W.F. measurement. As the volume of SiO₂ sol increased, the I.P. or VBM position of TiO₂–MWCNT–SiO₂-x initially decreased to 40 μL and then increased above 40 μL (Figure 9c). These shifts in the VBM position correlate with changes in the bandgap and visible light absorption, which is roughly consistent with those obtained using UV spectra. Each conduction band minimum (CBM) position, calculated by adding the bandgap value to the VBM value, is presented in Figure 9c. The volume-dependent changes may reflect the interplay between the band structure modifications of TiO₂ owing to SiO₂ and MWCNTs. Namely, the upward shift of the VBM above 40 μL is mainly due to the formation of TiO₂–MWCNT heterojunctions that introduce midgap states, as suggested by the UV, Raman, and XPS spectra, while the downward shift below 40 μL is attributed to the formation of Ti–O–Si bonds, as seen in a peak in XPS spectra (Figure 8c), and/or dipole modification by surface SiO₂. This interpretation is consistent with the XRD results, indicating that above 40 μL, large SiO₂ particles reduce the ability of SiO₂ to suppress the anatase-to-rutile transformation of TiO₂, indicating weak chemical interactions between SiO₂ and TiO₂. Another interpretation is that the increased anatase-to-rutile ratio at 40 μL corresponds to the downshifts of the VBM and CBM based on density functional theory, indicating that the VBM and CBM of anatase TiO₂ are at deeper positions than those of rutile TiO₂.⁶¹ As an exception, the VBM and CBM positions of TiO₂–SiO₂ with the smallest anatase-to-rutile ratio were closer to the vacuum level than those of the other samples. Herein, the anatase-to-rutile ratio is not the only factor influencing the positions. Further theoretical analysis is required to elucidate the impact of chemical and microstructural changes on the electronic properties, which will be addressed in a separate study.

The W.F. (φ), defined as the energy difference between the Fermi level and the vacuum level, was determined by subtracting the cutoff energy from the photon energy (21.22 eV).⁵⁹ The W.F. values for TiO₂, TiO₂–SiO₂-30, and TiO₂–MWCNT were 5.52, 4.35, and 5.27 eV, respectively, and those for TiO₂–MWCNT–SiO₂-x are shown in Table 1. These values range from 4.3 to 5.6 eV and align with the literature values for CNTs (4.353–5.6 eV)^62,63 and TiO₂ (4–5.1 eV).^58,64,65 The relative positions of these W.F.s correspond to the Fermi level positions determined from the flat-band potential (Figure 9c). As mentioned in the review, W.F. may have different values depending on the involved measurement method and environment,⁶¹ possibly because of changes in surface charge with varying measurement conditions and the effect of dipole moments on the UPS results (Section 3.9).

3.8. MS Plots

The flat-band potential and donor density of each sample are estimated via MS analysis using eq 2:

2

where C is the capacitance of the space charge region and ε (=ε₀ ε_r), A, N, V, and V_fb are the permittivity, area, donor (or acceptor) density, applied potential, and flat-band potential of the semiconductor, respectively. Furthermore, e is the electric charge, K_B is Boltzmann’s constant, and T is the absolute temperature. Here, ε₀ is the permittivity of free space, and ε_r is the dielectric constant of the semiconductor film.

Figure 10a–k shows the MS plot, and Figure 10l compares the flat-band potential for each TiO₂-based sample, determined from the intercept of the linear range of the MS plot with the potential axis. Figure 10m shows the donor (or acceptor) density of each sample calculated from the slope of the linear range of the MS plot. According to literature, the dielectric constants for anatase TiO₂ and anatase TiO₂ on SiO₂ were 55⁶¹ and 76,⁶⁶ respectively. Herein, 60 was used to calculate the N for all samples. As shown in Figure 10a, pristine TiO₂ exhibits a negative y-intercept (−0.49 eV) and a positive slope. The incorporation of MWCNTs into TiO₂ results in a negative shift of the y-intercept (−0.67 eV) and a smaller positive slope compared to that of the pristine TiO₂, indicating increases in the dielectric constant of the film and donor density because of an increased loaded amount of TiO₂ (Section 3.1, 3.3, and 3.4) on the MWCNT with a large surface area. However, the calculated density may not be directly compared because surface heterogeneity from microstructural changes can lead to miscalculations. The addition of SiO₂ to TiO₂ leads to a negative shift in the y-intercept (−1.78 eV) and a slope much larger than those corresponding to pristine TiO₂ and TiO₂–MWCNT, suggesting a notable decrease in the donor densities or dielectric constants owing to the incorporation of SiO₂. For the TiO₂–MWCNT–SiO₂ samples with varying volumes of added SiO₂ sol, the flat-band potential and donor (or acceptor) density dramatically increased up to 30 μL, remained almost constant between 30 and 50 μL and then slightly increased up to 80 μL, as shown in Figure 10l,m. In contrast, the TiO₂/SiO₂ molar ratio (Figure 2a) remained almost constant below 30 μL and dramatically decreased above 40 μL. This means that the SiO₂ content, with a much lower dielectric constant (<5)⁶⁷ compared to anatase TiO₂ (>55),⁶¹ which may change the dielectric constant of TiO₂–MWCNT–SiO₂-x, is not the only factor changing the y-axis value in Figure 10m. Adsorption of anions and cations at the sample–solution interface can shift the flat-band potential to baser and nobler values than that before the adsorption, similar to changes observed with pH variations.⁶⁸ For TiO₂–MWCNT and TiO₂–MWCNT–SiO₂-10, the flat-band potential has a large negative value, possibly due to the negative charge effects of the MWCNTs or BA on the MWCNTs. In TiO₂–SiO₂-30 and TiO₂–MWCNT–SiO₂ samples prepared with SiO₂ sol volumes between 20 and 40 μL, where the flat-band potential is still more negative than that above 50 μL, the strong chemical interaction between SiO₂ and TiO₂, attributed to the large surface area provided by a large amount of small SiO₂ crystallites, may form Ti–O–Si bonds (Figure 8c), providing a Lewis acidic hydroxyl group¹⁷ with negative charges of −O^– on the surface in 0.1 M Na₂SO₄ at pH ∼ 6. Substantial changes in the flat-band potential are further discussed in Section 3.9. Here, the slope of the MS plot changed from positive (n-type semiconductor behavior) to negative (p-type semiconductor behavior) with increasing SiO₂ sol volume above 30 μL, while the flat-band potential changed from a baser to a nobler potential than the set potential for photoresponse measurements. This is a phenomenon observed at an interface exposed to stimulated sunlight in Na₂SO₄ (Section 3.9). Therefore, the observed p-type behavior does not indicate that the TiO₂–MWCNT–SiO₂ film wholly changed from n-type to p-type with the addition of the SiO₂ sol (Figure 9c). The band structures of the sample–solution interface of TiO₂ prepared with MWCNT and SiO₂ sol volumes less than 20 μL, at ∼30 μL, and above 40 μL are illustrated in Figure 11a–c, assuming the flat-band potential only reflecting the varying Fermi level position (E_F) depending on the SiO₂ sol volume. Therefore, an accumulation/depletion layer for holes was formed at the semiconductor–solution interface below or above 30 μL. Below 30 μL, despite hole accumulation at the interface, the high position of the VBM results in a low oxidation rate using photoexcited holes. Above 30 μL, band bending in TiO₂ at the TiO₂–solution interface forms a hole-depleted layer, increasing the energy barrier for charge transfer and retarding the oxidation reaction, although the VBM for accepting electrons is sufficiently low. In fact, the charge transfer resistance (R_ct) acquired from EIS analysis (Section 3.10) decreased up to 30 μL and then increased with increasing SiO₂ sol content. Thus, increasing or decreasing SiO₂ sol volume from 30 μL can retard the oxidation reaction owing to either a wide hole-depleted layer or an upward shift in the VBM, respectively. In accordance with the change in the crystal structure and chemical bindings confirmed using XRD, XPS, and Raman spectra, the transition from the n-type to p-type behavior observed in samples with high SiO₂ and MWCNT contents may be attributed to changes in Fermi levels in the solution under the irradiation owing to the formation of new energy levels and their density caused by the addition of MWCNT and SiO₂ above a certain amount. The reduced crystallite size of TiO₂ in the TiO₂–MWCNT–SiO₂ ternary composites prepared via dual sol–gel EPD increased their grain boundary density, resulting in a high density of new energy levels formed at the boundary, thereby substantially changing the band structure. Additionally, variations in the microstructure can influence the electronic properties of the semiconductors. However, the calculated width of the depletion layer (W) at 0.245 V vs Ag/AgCl (saturated KCl) (right axis in Figure 10m) using eq 3(61) exceeds the size of anatase crystallites for all the samples. Hence, each band bending is possibly smaller than the expected width of the depletion layer, which may have an advantageous effect on the efficiency of photoelectrochemical reactions using holes.

3

Figure 10.

(a–k) MS plot for TiO₂, TiO₂–SiO₂-30, TiO₂–MWCNT (TiO₂–MWCNT–SiO₂-0), and TiO₂–MWCNT–SiO₂-x, where x is the added volume of SiO₂ sol, measured in a 1 M Na₂SO₄ solution. m indicates the slope [(μF)⁻²/V]. (l) Comparison of flat-band potentials and (m) donor (or acceptor) densities and width of depletion layer at 0.245 V vs Ag/AgCl (saturated KCl) for each TiO₂-based sample, provided that ε_r = 60. Donor (or acceptor) densities were calculated by using a geometric area of 0.5 cm², corresponding to the substrate area for TiO₂ thin films or each TiO₂ composite thin film.

Figure 11.

Band structures of TiO₂ surfaces prepared with MWCNTs and SiO₂ sol before and after contacting with a 1 M Na₂SO₄ solution based on MS analysis when the added volume of SiO₂ sol is (a) <20 μL, (b) ∼30 μL, and (c) >40 μL. An accumulation or depletion layer for holes is formed at the semiconductor–solution interface below or above 30 μL, respectively, while band bending is negligible in the solution at 30 μL.

3.9. Band Structure Based on Various Analysis

The coherence among UPS, UV–vis, and MS analyses provides a comprehensive understanding of surface and bulk electronic modifications. Figure 9c shows a summary of the electronic structures, that is, VBM, CBM, and E_F positions, examined using these analytical methods. The Fermi levels, calculated separately from the W.F. via UPS results and the flat-band potential via the MS plot, are shown in Figure 9c. The pristine TiO₂ exhibits a wide bandgap typical of the anatase phase, limiting its responsiveness to UV light. The incorporation of MWCNTs shifts the VBM and CBM positions upward by the formation of TiO₂–MWCNT heterojunctions that introduce midgap states, enhancing solar absorption and electron–hole separation. Conversely, the addition of SiO₂ to pristine TiO₂ shifts the VBM and CBM downward and widens the bandgap. However, this effect is not monotonic for TiO₂–MWCNT–SiO₂-x; as the SiO₂ sol volume increases, the VBM and CBM shift downward up to 40 μL and then shift upward above 40 μL with the bandgap widening.

Equation 4 is used to calculate the E_F from the flat-band potential:^61,69

4

As shown in Figure 9c, for TiO₂–MWCNT–SiO₂-x, the E_F derived from MS analysis at 100 kHz decreased with an increasing SiO₂ sol volume, although it remained almost constant between 20 and 40 μL. Figure S5a–k shows MS plots; these plots were prepared based on the capacitance values obtained by a conventional fitting of the semicircular part of EIS spectra using parallel circuits of R and CPE. Figure S5l,m shows the V_fb and N of each sample calculated based on the MS plots. As shown in Figure S5l, V_fb shifts to a nobler potential as the voume of added SiO₂ sol increases, which is similar to the shift observed in V_fb derived from MS analysis at 100 kHz (Figure 10l). As shown in Figure S5a–k, the p-type behavior was not observed, consistent with the single-point measurement results at frequencies below 50 kHz. Slow processes such as ion movement may vary the potential of the Helmholtz layer at low frequencies, resulting in the varying pseudocapacity at each potential. In contrast, the change in the Fermi level based on the UPS results was much smaller than that based on the flat-band potential. E_F values vary depending on the measurement method, prominently for TiO₂, TiO₂–MWCNT, and TiO₂–MWCNT–SiO₂-10, 20, 30, and 40: the MS-derived values are closer to the vacuum level than those from UPS spectra, indicating that the flat-band potential is more negative than other samples. Generally, the flat-band potential shifts when Helmholtz potential changes owing to ion adsorption, surface chemical bond modifications (electric dipole moments), and surface-level electron and hole capture of a semiconductor. Thus, the flat-band potential may indicate a more negative value, even if the Fermi level under a vacuum is not close to the vacuum level, when a larger amount of negative charge, such as ions or electrons, is attached to the sample surface compared to the other samples. For TiO₂–MWCNT and TiO₂–MWCNT–SiO₂-10, the E_F values are much closer to the vacuum level, possibly owing to the negative charge effects of MWCNTs or BA on MWCNTs. In TiO₂–MWCNT–SiO₂ samples prepared with SiO₂ sol volumes below 40 μL, where E_F is still closer to the vacuum level than that derived from the UPS spectra, the strong chemical interaction between SiO₂ and TiO₂, attributed to the large surface area provided by a large number of small SiO₂ crystallites, may result in a Lewis acidic surface with a high density of negative O^– charges on the sample surface. The Fermi level obtained from MS analysis is even above the CBM for several samples—a feature of the quasimetallic state degenerated from n-type semiconductors. This may be due to different measurement methods, i.e., CBM and E_F are derived from UPS analysis in vacuum and MS analysis in the solution, respectively. As discussed above, adsorption of negative ions in the solution may shift V_fb to a nobler potential, resulting in the position of E_f being closer to the vacuum level. The quasimetallic state may be degenerated by the ITO substrate, as the thickness of TiO₂–MWCNT–SiO₂ film immobilized on the ITO is small below 30 μL (Figure S2). E_f based on the UPS results varies as the SiO₂ sol volume changes were much smaller than E_f based on V_fb. This is partially due to the varying dipole moments of sample surfaces that affect the W.F. measurement by UPS. A noticeable shift in the Fermi level derived from the UPS close to the vacuum level was also observed in TiO₂–SiO₂-30.

At all frequencies between 10 Hz and 100 kHz, E_f based on V_fb shifted to a deep position with the addition of either or both SiO₂ and MWCNT. The shift of E_f based on the V_fb to a deep position with increasing SiO₂ sol volume was evident in the MS analysis results obtained at 100 kHz. The varying band structure reflected in the V_fb shift may be attributed to the exclusion of the effects of several slow physical changes, including charge movement, caused by the increasing frequency.⁷⁰⁻⁷³ However, using the density functional theory in the future can clarify the correlation of physicochemical properties and varying band structure, including V_fb.

3.10. EIS

Figure 12a presents the Nyquist plots for the pristine TiO₂, TiO₂–MWCNT, TiO₂–SiO₂-30, and TiO₂–MWCNT–SiO₂-x samples with different volumes of SiO₂ sol. As shown in Figure 12a, a semicircle or part of a semicircle was observed in all samples, indicating an electron transfer reaction. The Nyquist plots were fitted using an equivalent circuit, as shown in the inset, which includes a parallel circuit of R_ct and a constant phase element (CPE), which is connected in series with a resistance (R_s). The CPE denotes the pseudocapacitance corresponding to R_ct. The impedance of CPE (Z_CPE) is calculated as follows:

5

where T is the capacitance [Ω^–1 s^p], ω is the radial frequency, and p is a correction constant of 0–1. For all samples, R_s values remained almost constant between 60 and 80 Ω. Figure 12b–d shows the R_ct and CPE parameters (T and p) acquired by fitting the plots with the equivalent circuit. The pristine TiO₂ sample exhibited a considerably large R_ct, suggesting a low charge transfer efficiency. The integration of MWCNTs led to a notable decrease in R_ct and R_s (data not shown), suggesting that MWCNTs serve as a superior medium for electron transfer. The Nyquist plot of TiO₂–SiO₂-30 shows a parallel rise along the imaginary axis, characteristic of a dielectric material. Consequently, the R_ct for this sample is considerably larger than that of the pristine TiO₂ because of the insulation properties of SiO₂. For TiO₂–MWCNT–SiO₂-x, R_ct decreased gradually up to 30 μL of the SiO₂ sol and then increased with further additions. TiO₂–MWCNT–SiO₂-30 showed the lowest R_ct value, suggesting optimal photoelectrochemical activity under an applied potential of 0.245 V vs Ag/AgCl (saturated KCl). The lowest I_d/I_g ratio for TiO₂–MWCNT–SiO₂-30 (Figure 7i) indicates high conductivity owing to the lowest defect density in the graphene structure of MWCNTs, which corresponds with the lowest R_s value, resulting in improved charge transfer efficiency, corresponding to reduced R_ct. The increased R_ct with excessive SiO₂ sol above 40 μL is owing to the combination of a decreased TiO₂ to SiO₂ ratio in the film, an increased bandgap, a decreased anatase/rutile ratio, and a high defect percentage in the graphene structure of the MWCNT (>50 μL), all of which decrease the effective carrier density for oxidation as discussed in previous sections, and owing to shallow E_F position and formation of a hole-depleted layer, reducing charge transfer efficiency.

Figure 12.

(a) Nyquist plots of TiO₂, TiO₂–SiO₂-30, TiO₂–MWCNT (TiO₂–MWCNT–SiO₂-0), and TiO₂–MWCNT–SiO₂-x, where x is the added volume of SiO₂ sol, measured in a 0.1 M Na₂SO₄ solution at 0.245 V vs Ag/AgCl (saturated KCl). (b) R_ct, (c) CPE-T, and (d) CPE-p for each TiO₂-based sample, acquired by fitting the plots with the equivalent circuit shown in the inset of (a).

The CPE-T values, which relate to double-layer capacitance, increase with the integration of MWCNT and SiO₂, which is owing to a combination of the expanded surface area and the high dielectric constant from the newly formed hydroxyl group,¹⁷ respectively. The large CPE-T for TiO₂–MWCNT–SiO₂-30 suggests that its large surface area is consistent with the large film thickness (Section 3.1). Meanwhile, CPE-p values close to 1 (ideal capacitor) for all samples suggest a homogeneous electrochemical performance. The EDS mapping results confirm the homogeneous distribution of TiO₂, SiO₂, and MWCNTs in TiO₂–MWCNT–SiO₂-x (data not shown.).

3.11. Photocurrent Measurement

Figure 13a shows the photocurrent responses of the pristine TiO₂, TiO₂–SiO₂-30, TiO₂–MWCNT, and TiO₂–MWCNT–SiO₂ samples at different volumes of the SiO₂ sol. Figure 12b depicts a bar graph comparing these photocurrents. The pristine TiO₂ sample exhibits a photocurrent density of 1.46 μA/cm². Incorporating SiO₂ into TiO₂ (TiO₂–SiO₂-30) results in a slight photocurrent density increase to 1.47 μA/cm², likely owing to increased hydrophilicity and a deep valence band, whereas the film resistance is large. For TiO₂–MWCNT, the photocurrent density increases to 1.82 μA/cm² owing to the abundant electron/hole carriers from the high TiO₂ loading on MWCNT and efficient electron transport channels provided by MWCNTs,^12,45 as evidenced by decreased R_ct and R_s values from EIS analysis. The addition of MWCNTs enhances charge separation and reduces electron–hole recombination. However, the increase in photocurrent density is smaller compared to our previous study¹² due to differences in EPD time and annealing temperature, resulting in different film structures, including chemical composition and film morphology. Increasing the SiO₂ sol volume in TiO₂–MWCNT–SiO₂-x raises the photocurrent density to 30 μL and then gradually decreases beyond this point, suggesting an optimal SiO₂ volume for maximum photoelectrochemical performance. The highest photocurrent density of 10.9 μA/cm² for TiO₂–MWCNT–SiO₂-30 (10 min EPD) is ∼7.4 times higher than that for TiO₂–MWCNT prepared under the same conditions except for the addition of the SiO₂ sol. This value surpasses our previous highest results of 7.26 μA/cm² for TiO₂–MWCNT prepared under the same annealing temperature with sol–gel EPD (5 min) from a mixed solvent of acetone and 2-propanol, with an optimal volume ratio of 1:9,⁴⁵ where longer deposition times than 5 min resulted in film peeling from the substrate during EPD. However, TiO₂–MWCNT–SiO₂-30 and TiO₂–MWCNT–SiO₂-40 showed lower stabilities against on–off cycles than in other conditions; stability decreased to <70% after 10 cycles owing to unknown reasons (Figure S6). Hence, improving the stability will be a future challenge. High photocurrent value and stability were achieved for TiO₂–MWCNT–SiO₂-20.

Figure 13.

(a) Photocurrent responses of TiO₂, TiO₂–SiO₂-30, TiO₂–MWCNT (TiO₂–MWCNT–SiO₂-0), and TiO₂–MWCNT–SiO₂-x, where x is the added volume of SiO₂ sol, measured in a 1.0 M Na₂SO₄ solution at 0.245 V vs Ag/AgCl (saturated KCl). (b) The photocurrent of the 10th on/off cycle for each sample. LSV results of each sample in a 0.5 M Na₂SO₄ measured at 5 mV/s (c) with and (d) without simulated sunlight irradiation. (e) Saturated current of each sample observed below 1.0 V for each sample with and without simulated sunlight irradiation.

Figure 13c,d shows the LSV results with and without simulated sunlight irradiation. Figure 13e compares the saturated current observed below 1.0 V for each sample with and without simulated sunlight irradiation. Among the TiO₂–MWCNT–SiO₂-x samples prepared with SiO₂ sol volumes between 10 and 80 μL, TiO₂–MWCNT–SiO₂-30 showed the largest current. Beyond a certain SiO₂ threshold, its negative impact on current mirrored the photocurrent response trends (Figure 13a). This negative impact above 30 μL was particularly pronounced in the LSV tests, as the insulating characteristics of SiO₂ became highly evident in high-voltage applications. Regarding the onset potential of the oxidized current, adding SiO₂ to TiO₂ shifted the onset potential to several hundred millivolts more noble than that of the pristine TiO₂, regardless of its deep VBM position, which is consistent with the observed large R_ct. These shifts in onset potential with changing SiO₂ sol volumes, which coincide with changes in R_ct, suggest that variations in film properties affect the effective carrier density, as discussed in Section 3.10.

The highest saturated photocurrent below 1.0 V was 9.17 μA/cm² for TiO₂–MWCNT–SiO₂-30, which is ∼3.4 times higher than that of TiO₂–MWCNT prepared under the same condition except for the addition of SiO₂ sol. This indicates that the addition of SiO₂ sol drastically increases the photocurrent when the applied potential is minimal, as observed in the photoresponse measurement.

Figure 14 shows the interplay between the electronic structure and surface properties of the TiO₂–MWCNT–SiO₂ sample, optimized for enhanced photoelectrochemical water splitting. TiO₂, with its wide bandgap of ∼3.2 eV, serves as the primary photocatalyst, absorbing UV light to generate electron–hole pairs. The inclusion of MWCNTs, depicted as a cylindrical nanostructure adjacent to TiO₂, increases the TiO₂ surface area and likely facilitates electron transport owing to their excellent conductivity, leading to the highly efficient separation of photogenerated electrons and holes. Here, solar energy excites electrons (e^–), denoted by the yellow arrow, which then move into the conductive network of MWCNTs (green arrow), reducing electron–hole recombination and enhancing photocatalytic performance. The associated energy level of SiO₂ is ∼9.0 eV,⁷⁴ which does not participate directly in charge separation but enhances the overall hydrophilicity of the photocatalyst even with MWCNT, promoting interface contact with water molecules. Additionally, the addition of SiO₂ results in a downward shift of the VBM and CBM for TiO₂–SiO₂-30 and TiO₂–MWCNT–SiO₂-x up to 40 μL of the SiO₂ sol. The deep VBM position of approximately −8.5 eV vs vacuum (3.86 V vs Ag/AgCl (saturated KCl)) may provide sufficient overpotential to oxidize water efficiently. Additionally, the newly formed Ti–O–Si and Ti–O–C bonds in TiO₂–MWCNT–SiO₂-20 and -30 may form new midgap levels functioning as traps for electrons/holes to suppress their recombination and/or directly accept electrons during the water-splitting reaction. Furthermore, the band structure, including the Fermi level position, may have been tuned as band bending and a hole-depleted layer at the sample–solution interface are negligible.

Figure 14.

Schematic of the interplay between the electronic structure and surface properties of the TiO₂–MWCNT–SiO₂ sample designed for enhanced photoelectrochemical water splitting.

4. Conclusion

TiO₂–MWCNT–SiO₂ composite thin-film photoelectrodes were fabricated by sol–gel EPD with varying amounts of SiO₂ sol. The introduction of both MWCNT and SiO₂ through dual sol–gel EPD enhanced the photocatalytic performance of TiO₂ by reducing the crystallite size range of anatase TiO₂ to between 3.5 and 5 nm. The photocatalytic performance can be further improved by optimizing the volume of the SiO₂ solution added, especially when the applied potential is minimal. The optimal SiO₂ content was determined by the interplay between the increased surface area of SiO₂ and the disadvantage of the decreased conductivity. The optimized size of SiO₂ crystallites with sufficient surface area enhances hydrophilicity and efficiently suppresses anatase-to-rutile phase transformation. Moreover, Ti–O–Si bonds formed with appropriate volumes of SiO₂ sol while maintaining the integrity of the graphene structure of the MWCNT, which is key to high electrical conductivity. UPS and MS analyses confirm that the integration of both MWCNTs and SiO₂ into TiO₂ matrices modulates its electronic structure, altering the VBM and CBM, and midgap states, thereby possibly shifting the Fermi level position. Superior photoelectrochemical activity is achieved with a deep VBM and/or negligible band bending at the TiO₂–solution interface, optimizing the generation and separation of charge carriers under light irradiation, although neither the accurate V_fb for each sample nor the reason for the shift of V_fb toward a nobler potential with increasing SiO₂ sol volume is not entirely clear due to the limitations of MS analysis. In addition, preparing an electrode that exhibits high photoelectrochemical activity and long-term stability than those achieved herein is one of the next challenges.

In conclusion, tailored engineering of TiO₂–MWCNT–SiO₂ composite thin films using dual sol–gel EPD yields a film electrode with an optimized electronic structure, enhancing light harvesting with rapid charge carrier processes and suppressing recombination of electrons and holes, leading to a highly efficient photocatalytic water-splitting reaction under solar irradiation. The rise of oxidation current is baser, and the photoresponse current density at 0.245 V vs Ag/AgCl (saturated KCl) is 7.4 times higher than for TiO₂–MWCNT under simulated sunlight irradiation.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.4c08658.

• EDS results of pristine TiO₂ film on indium tin oxide (ITO); film thickness of each TiO₂ film measured from cross-sectional SEM images; bandgap of each sample calculated from Tauc plots when the data with an absorbance above 3 before subtracting the background are ignored; XRD patterns in the 22–29° range of each sample; results of MS analysis using CPE; and stability of each sample against on–off cycles (PDF)

Author Contributions

^§ M.M. and Y.Y. equally contributed to this work. Conceptualization: M.M. and Y.Y.; methodology: Y.Y.and M.M.; formal analysis and investigation: Y.Y. and M.M.; writing—original draft preparation: Y.Y. and M.M.; writing—review and editing: M.M.; funding acquisition: M.M.; resources: M.M.; supervision: M.M.

Partial financial support was received from the Chuo University Personal Research Grant (2024) and the Chuo University Grant for Special Research (2024). The authors have no relevant financial or nonfinancial interests to disclose.

The authors declare no competing financial interest.