Boron-Modified Anodization of Preferentially Oriented TiO₂ Nanotubes for Photoelectrochemical Applications

Abstract

This study investigates the synthesis and characterization of boron-modified nanotubular titania (NTO) arrays fabricated via a single-step anodizing process with varying concentrations of boric acid (BA). Following anodization, a reductive heat treatment was applied to facilitate the crystallization of the anatase phase in the boron-modified NTO. The effect of the BA concentration on the structural, morphological, and photoelectrochemical (PEC) properties of the NTOs was systematically explored through scanning electron microscopy (SEM), X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), luminescence, and UV-Vis spectrometry. The introduction of boron during anodization facilitated the formation of sub-bandgap states, thereby enhancing the light absorption and electron mobility. This study revealed the optimal BA concentration that yielded a 3.3-fold enhancement of the PEC performance, attributed to a reduction in the bandgap energy. Notably, the highest incident photon-to-current conversion efficiency (IPCE) was observed for NTO samples anodized at a 0.10 M BA concentration. These findings underscore the promise of boron-modified NTOs for advanced photocatalytic applications, particularly in solar-driven water-splitting processes.

1. Introduction

Nowadays, green energy technologies such as photolysis and photocatalysis play a crucial role in everyday life. Recent studies have increasingly focused on enhancing the properties of existing materials, particularly semiconductors, which are critical in the surface heterogeneous processes of photocatalytic water splitting (PCWS). Titanium dioxide is one of the most advantageous and widely used materials in this field. The improvement of its photocatalytic properties is typically achieved through well-known techniques such as cationic [1,2,3,4] and anionic [3,5,6] doping, as well as nanoparticle decoration [7,8,9,10]. Anionic doping of titania by boron is frequently highlighted due to its positive effects on the photocatalytic properties of titanium dioxide [3,5,11,12,13,14]. However, the evaluation of the doped boron state and composition in such materials is quite challenging, mostly due to the complexity of its detection. Boron-modified titania is currently studied and produced in various forms, including powders, films, coatings, nanoparticles, and nanotubes, using methods such as sol-gel synthesis [15], chemical vapor deposition (CVD) [16], hydrothermal synthesis [17], and anodizing [18,19,20]. This study focuses on anodizing as the most suitable method, offering the advantage of a one-step in situ modification of oxide films or coatings with minimal energy and time costs.

Several studies have suggested the boron occupancy of a wide range of positions within the crystal structure of titanium dioxide, potentially shifting the bandgap energy toward the visible spectrum and enhancing the conductivity and electron mobility of the semiconductor [21,22,23,24,25,26,27,28,29,30]. These changes affect the photocatalytic properties of titania. However, despite extensive theoretical modeling, the impact of boron doping on the photoelectrochemical properties remains unclear and does not always correlate with the experimental results [31,32]. The trends in the effect of boron doping on the properties of nanotubular titanium oxide were previously reported. Szkoda et al. demonstrated that boron doping can reduce the bandgap of nanotubular titania from 3.09 eV to 2.91 eV [33]. Further, Georgieva et al. noted that boron can occupy interstitial positions within the titania structure [18]. Other researchers [16,34] have reported that boron doping can enhance the maximum photoconversion efficiency by a factor of 2 to 2.25. Additionally, it has been found that boron doping slows the formation and growth of anatase grains [35]. When boron is introduced into the anodizing electrolyte, it can increase the decomposition rate of methylene blue up to 97.3% within 1 h and 100% within 3 h [19,34], and of acidic yellow 1 (AY1) by 93% within 2 h [36], compared to undoped titania. Furthermore, the photocurrent has also been observed to increase by a factor of 3.1 to 3.5 [19,34]. Boron doping in nanotubular titania enhances the incident photon-to-current conversion efficiency (IPCE) from 25% at a wavelength of 530 nm to 34% at 500 nm in a DSSC cell sensitized by [RuL₂(NCS)₂]TBA₂ [20], shifts the flatband potential cathodically from −0.26 V to −0.37 V vs. Ag/AgCl, and slightly decreases the open-circuit potential of the photovoltaic cell from 0.70 to 0.66 V in pure versus doped titania [20]. Simultaneously, boron doping could enhance the photocatalytic efficiency, accompanied by reductions in the geometry of nanotubular arrays compared to undoped nanostructured TiO₂ (NTO) [19]. Moreover, the surface active area increases due to doping, which is considered advantageous [37,38,39]. The extended electron lifetime resulting from doping leads to a decreased recombination rate between the doped NTO and the DSSC cell electrolyte [20]. The various concentration of boron additives was also studied in [40], where an improvement in the photocatalytic performance in terms of rhodamine B decomposition was shown. The qualitative composition of boron was determined according to XPS, along with its movements to TiO₂ surfaces, with its interstitial position according to the calculation of Heffner et al. [41]. The exfoliation of boron also leads to an increase in the photocatalytic decomposition of rhodamine B [42], where exfoliation could be connected to boron migration to outer surfaces.

Boric acid (BA) is recognized as a favorable dopant due to its widespread availability, low cost, and versatility. For instance, it has been utilized in various methods, such as non-anodizing electrochemical treatment in an aqueous H₃BO₃ solution [18,33] or via single-step anodizing with the addition of BA to the anodizing electrolyte [43,44]. Momeni et al. have partially investigated the properties of NTO doped in situ with low BA concentrations of up to 0.025 M [43]. Their work pointed out that the photocurrent increased with boric acid concentrations up to 0.01 M, followed by a subsequent decrease.

Moreover, several studies [45,46,47,48] have reported that heat treatment in reductive environments (e.g., hydrogen) increases defects in the crystal structure [27,49]. However, only a few studies have explored such treatment of boron-modified anodized titania with enhanced photoelectrochemical properties and unchanged morphology [50,51]. Defect formation was suggested to facilitate boron incorporation into the titania structure [52].

In this study, NTOs were synthesized with varying contents of boric acid in the anodizing electrolyte through a single-step anodizing process. Reductive heat treatment was further conducted to facilitate boron modification of the titania. The NTO samples were comprehensively characterized using scanning electron microscopy (SEM), energy-dispersive X-Ray spectroscopy (EDX), X-ray diffraction analysis (XRD), X-ray photoelectron spectroscopy (XPS), ultraviolet–visible light (UV-Vis) spectrometry, photoluminescence, and electrochemical and photoelectrochemical (PEC) methods. This research investigates the relationship between boron modification, the structural disordering of the preferentially oriented doped nanotubular titania, and its PEC efficiency. These analyses provide critical insights into the role of boron in enhancing the photoelectrochemical properties of modified nanotubular titania.

2. Materials and Methods

2.1. NTO Synthesis

Titanium plates (99.9%, VSMPO-AVISMA, Yekaterinburg, Russia) with dimensions of 2 × 2 × 0.1 cm were prepared for anodizing by electrochemical polishing using a solution of 80%_vol ethylene glycol (99.9%, Ekos, Moscow, Russia) and 20%_vol isopropanol (99.8 wt.%, Ekos, Moscow, Russia), with the addition of 1 M NaCl (99.9%, Chemreactivsnab, Yekaterinburg, Russia). The modified polishing process is described elsewhere [53,54,55]. The titanium substrates were electropolished at room temperature, maintaining a DC voltage of 20 V for 50 min. The NTOs were grown on titanium substrates by anodization in ethylene glycol with the addition of an ammonium fluoride solution (99.9%, 40% aqueous solution, Sigma-Tek, Moscow, Russia) of 2.3%_vol and 1.5%_mol, respectively, with various concentrations of boric acid (99.8%, Ormet, Yekaterinburg, Russia). Anodizing was carried out in electrolytes with the following concentrations of boric acid: 0, 0.01, 0.05, 0.10, 0.25, 0.50 and 1 M. The samples are designated as NTO-x/y, where “x” represents the BA concentration in the anodizing electrolyte and “y” indicates the anodization time. The anodization process was performed using the titanium plates as both the cathode and anode, under a constant voltage of 30 V for 40, 60 and 120 min, using a power supply (AKIP-1105, PriST, Moscow, Russia). After anodization, the nanostructured coatings on the substrates were annealed at 400 °C for one hour in air, followed by annealing in a hydrogen atmosphere for an additional 1 h, with heating and cooling rates of 1 °C/min.

2.2. Phase and Chemical Characterization

The phase composition of the obtained nanostructured coatings was analyzed using the XRD method using a Stadi automated diffractometer (Stadi, Darmstadt, Germany) equipped with a CuKα X-ray source. The diffraction patterns were analyzed, and the unit cell parameters were calculated using the Le Bail method. The thickness, microstructure, and morphology of the NTO arrays were examined using SEM with a JEOL 6390 (JEOL, Tokyo, Japan) and a Vega Compact (TESCAN, Brno, Czech Republic). Compositional analysis was conducted through EDX using a JEOL JED 2300 (JEOL, Tokyo, Japan), and XPS using a VG ESCALAB MKII electron spectrometer (VG Scientific Ltd., East Grinstead, UK) with a Mg Kα (E = 1253.6 eV) X-ray source.

2.3. Optical Properties Research

An FS5 spectrofluorometer (Edinburgh Instruments, Livingston, UK) was used to measure the diffuse reflectance spectra at the UV and visible wavelengths. A Teflon cuvette was used as a white reference. The luminescence and excitation photoluminescence spectra were recorded using the Cary Eclipse Varian (Agilent, Santa Clara, CA, USA) spectrofluorometer, using fluorescence mode (registration threshold of the emission process by time was 2 μs). Additionally, the luminescence decay curves of the TiO₂ samples were measured.

2.4. Photoelectrochemical Properties Studies

Illuminated open-circuit potential measurements (i-OCP) and j-V analysis of the material in a three-electrode cell were carried out using a P-20X potentiostat–galvanostat (Electrochemical Instruments, Moscow, Russia) with a light-emitting diode (LED) with the wavelength of 365 nm (full width at half maximum ~15 nm) and light power of 170 ± 10 mW/cm². The j-V curves were obtained with a scanning rate of 10 mV/s. The working electrode (WE)—NTO samples in the form of Ti/TiO₂ plates, counter electrode (CE)—Pt, and the reference electrode (RE)—Ag/AgCl and KNO₃, 0.1 M. The light intensity at 365 nm LED on NTO surfaces was determined by the 11XLP12-3S-H2 (Standa, Vilnius, Lithuania) light power detector.

The incident photon-to-electron conversion efficiency (IPCE) was estimated using the electrochemical cell setup mentioned above. The SCS10-PEC (Zolix Instruments Co., Beijing, China) spectrometer irradiated light at a selected wavelength ranging from 300 to 480 nm with a step of 2.5 nm. A potential of 800 mV between the WE and RE electrodes was applied by the SCS10-PEC system equipped with Xe-arc lamp XBO 500W/H OFR (Osram, Munich, Germany) and incorporated CHI600e potentiostat–galvanostat (CHInstruments, Austin, TX, USA). The PEC activity was calculated according to the IPCE parameter using the following equation [56,57]:

$$\mathrm{I}\mathrm{P}\mathrm{C}\mathrm{E}=1240·{\displaystyle \frac{Ip\left(\lambda \right)}{P\left(\lambda \right)·\lambda }}$$

(1)

where I_p(λ) is the photocurrent density at a selected wavelength (A·m⁻²), P(λ) is the power of the light flux at a given wavelength (W·m⁻²), and λ is the incident light wavelength (nm). The proposed PEC measurements are similar to the techniques described elsewhere [58,59,60,61]. The most detailed description of the i-OCP, j-V analysis, and IPCE is presented in [61].

3. Results

3.1. Synthesis and Surface Morphology of NTO Arrays

SEM surface images of both unmodified and boron-modified NTOs are shown in Figure 1.

Nanotubular arrays were successfully formed in the electrolytes with BA concentrations up to 0.10 M, as confirmed by the cross-sectional images shown in Supplementary Figure S1. However, when the BA concentration exceeded 0.10 M, the NTO morphology shifted from a nanotubular to a nanoporous structure. At concentrations of 0.50 M or higher, the anodizing process formed a non-porous barrier layer. Detailed morphological parameters, such as the pore diameter and wall thickness, are presented in

Table 1. Structural parameters and morphology types of NTO coatings with varied BA concentrations during a 120 min anodizing process.

Boric Acid [M]	0	0.01	0.05	0.10	0.25	0.50	1.00
Array Type	Nanotubular				Nanoporous	Barrier Layer
Pore Diameter [nm]	51.5 ± 4.1	52.1 ± 3.1	54.6 ± 4.7	55.2 ± 5.1	23.0 ± 2.1	Not Detectable
Wall thickness [nm]	11.3 ± 0.8	10.4 ± 0.8	9.9 ± 0.6	17.1 ± 1.0	-

.

The pore diameter of nanotubes or nanopores increases slightly from 51.5 ± 4.1 nm to 55.2 ± 5.1 nm as the concentration of BA in the anodizing electrolyte increases from 0 to 0.10 M. This increase in the BA concentration also leads to a reduction in the self-organization of the nanotubular arrays. In addition, when the BA concentration exceeds 0.10 M, the excessive formation of titania “nano-grass” on top of the nanotubular array is observed. At a BA concentration of 0.25 M, weakly self-organized nanoporous titanium dioxide forms instead of the nanotubular structure. Further increases in the BA concentration result in the development of a non-porous barrier layer.

The thickness of the nanotubular NTO coatings, measured from cross-sectional images, is detailed in

Table 2. Thicknesses of NTO coatings versus anodizing conditions.

BA Concentration [M]	0	0.01	0.05	0.10
Anodizing Time [min]	Thickness [µm]
40	2.8 ± 0.1	2.3 ± 0.3	4.1 ± 0.5	5.8 ± 0.3
60	3.0 ± 0.2	2.9 ± 0.3	3.9 ± 0.6	6.7 ± 0.5
120	4.0 ± 0.2	3.7 ± 0.6	6.6 ± 1.2	9.9 ± 0.7

. The thickness of the NTO coatings increased with both the time and BA concentration in the electrolyte, ranging from 2.3 ± 0.3 μm to 9.9 ± 0.7 μm

The current–time curves during anodization are depicted in Figure 2. The behavior of the curves presented in the image sheds light on the morphological features and the thickness of the synthesized NTO.

In the absence of BA, titanium anodization in the electrolyte follows a typical sequence: the initial formation of a barrier layer is accompanied by a decrease in current (short dot arrow 1 in Figure 2). The subsequent current growth leads to the formation of a pore (dashed arrow 2 in Figure 2). Thereafter, the process transitions into a steady-state growth phase, characterized by a slight decline in current. At BA concentrations of 0.01 and 0.05 M, a noticeable delay in the transition between pore nucleation and nanotube growth is observed from the position of the current peak that belongs to the pore/tube formation process. When the BA concentration exceeds 0.10 M, the current–time curves do not display this transition, which may indicate a lack of nanotubes or nanoporosity, as confirmed by SEM images. It can be concluded that the morphological changes in the synthesized NTOs are strongly affected by the concentration of BA, which mostly acts as a modifying additive. This modification is mainly associated with the incorporation of fluoride ions into the anodizing electrolyte. The increased concentration of BA leads to the extraction of fluoride ions from titania complexes [TiF_x]^4−x and their bonding to boron fluoride complexes [BF_x]^3−x. This consequently leads to the decreased solubility of titania during anodizing, which increases the density of the titanium oxide layer, ultimately leading to the formation of barrier-type coatings. Such coatings are commonly observed in the anodization of titania in electrolytes that poorly dissolve the formed titania.

3.2. Phase and Chemical Composition

The X-ray diffraction patterns of the synthesized NTOs annealed in air, followed by hydrogen annealing, are presented in Figure 3 and Supplementary Figure S2a,b.

The XRD pattern revealed two main phases: anatase phase (A) with I4₁/amd symmetry, which was created during the processing, and the phase of titanium, consistent with previous reports [59,62]. However, only faint traces of rutile and anatase phases were detected in the NTO-0.50/y and NTO-1.00/y samples, as displayed in Figure S2a,b. The diffraction peaks corresponding to the (100), (002), (101), (102), (110) and (103) planes are attributed to the titanium substrate (COD database card no. 00-044-1294). For the NTO-0/y, NTO-0.01/y, NTO-0.05/y and NTO-0.10/y, anatase crystal planes (101), (103), (004), (200), (105) and (116) were observed (COD database card no. 01-073-1764). The intensities of the anatase peaks (101), (103), (004) and (200) increased with the increase of the BA content in the electrolyte up to 0.10 M. At a BA concentration of 0.25 M, a transition coating with nanopores was formed, where only the anatase (101) plane was observed. When the BA concentration exceeded 0.25 M, only traces of rutile and anatase phases were detected in the XRD patterns. This absence of titania peaks in the XRD patterns confirms the exceptionally low thickness of titania coatings at high BA concentrations, validating the formation of thin barrier-layer titania coatings in non-solvent electrolytes. Notably, the anatase (101) and (004) are slightly left-shifted with the increase of the BA concentration up to 0.25 M, as shown in Figure 3b. According to [63], such a shift is accompanied by an increase in crystal cell volume and may be connected with boron doping or modification of NTO.

The unit cell parameters for anatase formed under the selected anodizing conditions were calculated from the XRD patterns and are summarized in Table S1.

Based on a slight increase in the crystal cell parameters in the a, b, and c directions with longer anodizing times and an insignificant impact of the BA concentration in the anodizing electrolyte on the crystalline structure, it is challenging to reliably determine the effect of the BA concentration on the crystalline lattice of anatase.

Nevertheless, the intensity of the peaks corresponding to the (101) and (004) planes of anatase varies, which can be attributed to the orientation variation of these planes. These orientation changes can be quantified using the texture coefficient, which reflects the predominant crystallization of material grains. The texture coefficient is calculated according to the Ariosa method (2) [64].

$${\mathrm{T}\mathrm{C}}_{\mathrm{x}\mathrm{y}\mathrm{z}}={\displaystyle \frac{{\mathrm{I}}_{\mathrm{x}\mathrm{y}\mathrm{z}}/{\mathrm{I}}_{\mathrm{x}\mathrm{y}\mathrm{z}}^0}{\left(\frac{1}{\mathrm{N}}\right)\sum {\mathrm{I}}_{\mathrm{h}\mathrm{k}\mathrm{l}}/{\mathrm{I}}_{\mathrm{h}\mathrm{k}\mathrm{l}}^0}}$$

(2)

where I_hkl—X-ray intensity of the hkl reflection, I⁰_hkl—X-ray intensity of TiO₂ powder with randomly oriented grains (card no. 01-073-1764), and N—the number of considered Bragg reflections (six and four in our calculations for the nanotubular and nanoporous arrays, respectively). The texture coefficients for the (101) and (004) planes are presented in

Table 3. The texture coefficient values of the anatase phase in the NTO as obtained by sequential annealing in air and hydrogen at 400 °C.

Anodizing Time [min]	BA Concentration [M]	0	0.01	0.05	0.10	0.25	0.50	1
	Texture Coefficient
40	TC (101)	0.10	0.09	0.13	0.16	0.77	Undetectable
	TC (004)	3.49	3.56	3.54	2.47	1.98
60	TC (101)	0.10	0.22	0.19	0.25	0.90
	TC (004)	3.47	2.94	2.71	2.26	1.81
120	TC (101)	0.09	0.12	0.08	0.13	1.41
	TC (004)	3.73	3.94	4.13	2.86	1.55

.

The preferred orientation of the (004) plane was observed at BA concentrations up to 0.25 M. This orientation is advantageous for surface energy [64,65,66,67,68], and the properties of similarly oriented titania have been studied in recent works [4,56,64,65,66,67,68]. The suppressed preferred orientation of TiO₂ along with the (004) plane is likely caused by the sorption of boron species from the anodizing electrolyte. Previous studies [69,70] suggest the absorption of boron species on the four-, five-, and six-fold coordinated titanium sites in the (101) and (004) planes of TiO₂. In this work, reductive thermal treatment in a hydrogen atmosphere is proposed to form four- and five-fold coordinated Ti on the nanotubular coatings’ surface, facilitating the modification of it by boron species. These interactions ultimately impact the morphology of NTO and the preferred orientation of anatase, leading to lattice distortion and consequently affecting the optical and photocatalytic properties of NTOs.

The XPS spectra, including the Ti 2p, O 1s, C 1s, B 1s regions and survey spectra of the NTO-0.10/120, NTO-0/120, NTO-0.01/120 and NTO-0.05/120 coatings, are shown in Figure 4a–d. The spectra referring to other studied materials are shown in Supplementary Figures S3–S5.

Figure 4a presents the Ti 2p region within the binding energy range of 468 to 455 eV, displaying two distinct peaks at 458.3 eV and 464.1 eV, corresponding to Ti 2p3/2 and Ti 2p1/2 in the 4⁺ oxidation state [56,71,72], respectively, with no evidence of lower oxidation states of titanium. However, the presence of Ti³⁺ at concentrations below the detection limit is approximately 1 at.% cannot be entirely excluded. The O 1s region (537–526 eV), depicted in Figure 4b, reveals three peaks at 529.5 eV, 531.5 eV and 532.7 eV, corresponding to Ti–O, Ti–O–H and C–O–H bonds [56,71,72], respectively. In the C 1s region (291–281 eV) shown in Figure 4c, the spectrum can be deconvoluted into three peaks at 284.5 eV (comprising peaks at 284.1 eV and 285.2 eV), 286.2 eV and 288.1 eV, which are attributed to C–C (sp² and sp³ carbon), C–O and C=O bonds [71,73,74], respectively. The formation of adsorbed oxygen impurities, primarily OH⁻ groups, as well as complex carbon species, and their impact on the photoelectrochemical properties of TiO₂, have been previously explored in studies on oxygen [56,72,75,76,77] and carbon [3,21,56,71,73,78,79,80,81] species. These species are likely connected to surface defect states and oxygen vacancies, particularly in the four- and five-fold coordinated Ti in the (101) and (004) planes of the TiO₂ structure in co-modified NTO.

The B 1s region, spanning binding energies of 194–184 eV, is illustrated in Figure 4d. The detection of boron at low concentrations was challenging; however, two peaks were revealed at 191 ± 1 eV and 187 ± 1 eV. The first is associated with Ti–O–B and B–O bonds, while the second corresponds to Ti–B bonds. It is worth noting that the typical detection limit for boron by XPS, using commonly employed emission sources such as Al Kα and Mg Kα, is approximately 1–3 at.% in Ti and O matrices [82,83,84], which already corresponds to the extreme level of doping for titania. This was confirmed with an experimental detection limit of boron above 2 at.% according to [85]. Therefore, the low signal-to-noise ratio in the B 1s region prevents an accurate quantitative compositional analysis. Thus, the observed signal in the boron region could be mainly attributed to a low concentration of surface-adsorbed boron species, such as Ti–OB(OH)O–Ti and Ti-OB(OH)OB(OH)–Ti adsorbates, as considered in [86], which is in agreement with the migration of boron to the outer surfaces of TiO₂ according to Heffner et al. in [41]. The role of the surface enrichment of TiO₂ with boron was additionally highlighted in [85]. As well as the possible partial presence of boron in interstitial [87,88,89,90] or substitutional [91,92,93] positions within the TiO₂ structure, which impacts the PEC properties of the boron-modified nanotubular TiO₂.

3.3. UV-Vis Spectrometry and Photoluminescence

The anatase structure of TiO₂ is an indirect semiconductor with a bandgap of approximately 3.2 eV [94,95,96,97]. Selected UV-Vis diffuse reflectance spectra of the NTOs are presented in Figure S6. These spectra show the fundamental absorption edge that is distinguished for titanium dioxide. Additionally, a broad absorption peak is observed, which can be attributed to O₂⁻ radicals associated with hole centers localized on 2p orbitals, as well as vacancy defects. The mechanism behind the formation of these vacancies can be attributed to a series of photoinduced electron–hole defect processes occurring on the surface of boron-modified titania, as described by Equations (3)–(7).

$$\mathrm{B}:{\mathrm{TiO}}_2+hv \to \mathrm{B}:{\mathrm{TiO}}_2(h_{vb}^{+}+e_{cb}^{-})$$

(3)

$${\mathrm{B}}^{3+}+e_{cb}^{-} \to {\mathrm{B}}^{2+}$$

(4)

$${\mathrm{B}}^{2+}+{\mathrm{O}}_2 \left(\mathrm{ads}\right) \to {\mathrm{B}}^{3+}+{\mathrm{O}}_2^{•-}$$

(5)

B²⁺ + Ti⁴⁺ → B³⁺ + Ti³⁺

(6)

$${\mathrm{Ti}}^{3+}+{\mathrm{O}}_2 \left(\mathrm{ads}\right) \to {\mathrm{Ti}}^{4+}+{\mathrm{O}}_2^{•-}$$

(7)

where a created ${\mathrm{O}}_2^{•-}$ slows down the formation of an electron–hole pair $h_{vb}^{+}$+${ e}_{cb}^{-}$.

To accurately determine the bandgap, the UV-Vis spectrum was transformed using the Kubelka–Munka (KM) Equation (8) [98]:

$$\mathrm{F}({\mathrm{R}}_{\mathrm{\infty }})={\displaystyle \frac{\mathrm{K}}{\mathrm{S}}}={\displaystyle \frac{{(1-{\mathrm{R}}_{\mathrm{\infty }})}^2}{2{\mathrm{R}}_{\mathrm{\infty }}}}$$

(8)

where R_∞—reflection coefficient with respect to a white reference (PTFE, WS-1); K—absorption coefficient; and S—scattering coefficient. The Tauc plot transformation method was used to determine the bandgap energy [99,100]. Figure 5 presents plots of the F(R_∞) as a function of the photon energy. The obtained samples exhibit intraband gap states, which could be attributed to surface modification by B³⁺ ions and defects within the anatase structure. A comparison of the samples depicted for 120 min of the anodizing revealed the intensity reduction of the bands within the 280–470 nm range, which is associated with Ti⁴⁺ charge transfer (CT) transitions (

Table 4. Attribution of the absorption band components for the NTO-x/y samples, where y represents the anodization duration of 120 min.

Attribution	NTO-0/y	NTO-0.01/y	NTO-0.05/y	NTO-0.10/y
Gaussian Peak [eV]
V0•• + ${\mathrm{O}}_2^{•-}$	2.01	2.05	1.99	1.91
2B2g → 2B1g (Ti3+)	2.26	2.37	2.31	2.33
V0 (anatase)	2.55	2.53	2.58	2.63
	2.79	2.71	2.79	2.87
Χ1a → Γ1b	2.90	2.89	2.94	2.96
Χ2b → Γ1b	3.09	3.05	3.10	3.11
Γ3 → Χ1b		3.14	3.21
Χ1a → Χ1b	3.47	3.49	3.52	3.44
Χ2b → Χ1b	3.66	3.67/3.81 (2-component)	3.85	3.73
Γ5′a → Γ1b	4.02			3.93
Γ5′a + Γ2′ → Γ1b		4.12	4.16
Γ2′ → Γ1b	4.21			4.25
Γ3 → Γ5′b	4.63	4.55	4.69	4.62

). This decrease refers to absorption bands corresponding to dopants or impurities in the 470–800 nm region.

Since the duration and concentration of the reducing agent in the atmosphere during heat treatment were consistent for all the coatings, it can be inferred that the intercalation of boron into the facets or inter-grain borders of anatase crystals enhances the absorption intensity of low energy states. These states modify the band structure of the TiO₂ semiconductor. Consequently, the Tauc method is employed to determine the bandgap energy, as described by Equation (9).

$${\left(\mathsf{\alpha }\mathrm{h}\mathsf{\nu }\right)}^{\mathrm{n}}=\mathrm{B}(\mathrm{h}\mathsf{\nu }-{\mathrm{E}}_{\mathrm{g}})$$

(9)

where α—absorption coefficient; hν—photon energy (eV); E_g—optical bandgap energy (eV); B—constant; and n—power degree, which depends on the transition type: n = 1/2 indirect, and n = 2 direct transition; the equation should be modernized according to what is presented in [101]. The modernization includes eliminating the impact of low energy absorption bands (sub-bandgap states) due to defect states, such as a Ti³⁺ oxygen vacancy and ${\mathrm{O}}_2^{•-}$ ion, by rearranging the baseline for bandgap estimation, and it is presented in detail in Supplementary Materials in Equations (S1)–(S5). The calculated results of the corrected bandgap energies (Eg) are presented in Figure 6 and Figure S7 and are followed by the values listed in

Table 5. Bandgap energy calculations for the NTO-x/y samples, where y corresponds to an anodization duration of 120 min.

Sample	Bandgap Calculation
	Direct Transition		Indirect Transition
	Eg [eV]	(Eg)S [eV]	Eg [eV]	(Eg)S [eV]
NTO-0/y	3.19	3.30	2.58	3.27
NTO-0.01/y	3.15	3.30	2.37	3.27
NTO-0.05/y	2.50	3.26	0.2	3.29
NTO-0.10/y	3.01	3.26	1.93	3.23

.

As illustrated in Figure 6, the bandgap energy (E_g) values slightly decrease within the range of x = 0–0.05, while the (E_g)_S values remain nearly unchanged, except for a more significant slight reduction of 0.06 eV observed in samples with x = 0.05 and 0.10, where (E_g)_S equals 3.23 eV. These insignificant changes can be explained by the size factor influence. The difference of the bandgap values for a bulk and nanosized material might be roughly estimated by the following expression:

$$E_g\left(nano\right)=E_g\left(bulk\right)+{\displaystyle \frac{1}{〈r〉}}·(\mathrm{e}\mathrm{V})$$

(10)

where $〈r〉$—medium size of nanoparticles (in nm). Accordingly, when the wall thickness decreases, in the x = 0–0.05, the bandgap value increases; in the x = 0.05–0.10, with wall thickness growth, E_g is consequently reducing. Charged electron and hole defects create impurity energy levels in the NTO bandgap and do not affect the bandgap value itself.

Given that the linearization of the fundamental absorption edge of the TiO₂ semiconductor is more accurate for an indirect transition and the resulting values are less than for n = 2, the values obtained for n = 1/2 should be considered the most representative of the actual bandgap, with the influences of sub-bandgap states considered and considering the nanoparticles’ size.

Figure 5 reveals broad absorption bands in the energy range of 1.5–4.5 eV, which are composed of the fundamental absorption band and an additional band, interpreted as Urbach tails (represented by red, blue, green, and yellow Gaussians) and could be associated with sub-bandgap states. These Urbach tails are associated with structural disordering within the anatase unit cells, caused by the formation of defect levels, as described by Equations (3)–(7). The introduction of B³⁺ ions into [TiO₆] octahedra (Equation (10)) mostly leads to the formation of Ti³⁺ ions and V_O vacancies, acting as electron traps that capture electrons from the conduction band (CB).

$${\mathrm{T}\mathrm{i}}_{\mathrm{T}\mathrm{i}}^{\times }+{\mathrm{O}}_{\mathrm{O}}^{\times }\to {\mathrm{T}\mathrm{i}}_{\mathrm{T}\mathrm{i}}^{\prime }+{\mathrm{B}}_{\mathrm{T}\mathrm{i}}^{\prime }+{\mathrm{V}}_{\mathrm{O}}^{••}+{\displaystyle \frac{1}{2}}{\mathrm{O}}_2$$

(11)

The fundamental absorption band results from the interband optical transitions between sublevels of the titanium 3d-shell in the CB and 2p-levels of O²⁻ ions in the valence band (VB). These transitions can be classified into direct (Χ_i → Χ_j, Γ_i → Γ_j) and indirect (Χ_i → Γ_j, Γ_i → Χ_j) types, as depicted in Figure 7.

Previous studies [102,103] have demonstrated that doping with Ag and Al reduces the bandgap. While TiO₂ is typically an n-type semiconductor, doping with B³⁺ ions imparts some p-type semiconductor characteristics. For the samples described in this study, this factor has minimal impact in the cases of x = 0.05 and x = 0.10. The creation of one V₀ vacancy leads to the formation of two Ti³⁺ ions, producing a pair of “electron defect” and “hole defect”.

The bandgap of undoped TiO₂ is 3.30 eV (Figure 6a), closely aligning with standard anatase values. However, the absorption edge is governed by an indirect transition Γ₃ → Χ_1b, with the lower energy level for the direct transition being approximately 3.5 eV Χ_1a → Χ_1b [104,105] (represented by magenta Gaussians in Figure 5a). The experimentally determined energy levels are lower than these values, most probably due to two main reasons: p-d hybridization between the outer shells of B and Ti, and the formation of Ti³⁺ ions, whose energy levels (²T_2g) are situated below those of tetravalent titanium 3d states.

Anatase is an indirect band semiconductor. The data in Figure 6 confirms that the indirect transition model more accurately describes the photon absorption process by electrons transitioning to states corresponding to the CB bottom. This behavior arises from the overlapping bands corresponding to the Γ₃ → Χ_1b indirect transition for Ti⁴⁺ and the ²T_2g → ²E_2g forbidden direct d-d transition of Ti³⁺.

The absorption edge in the general absorbance of NTO-0.05/120 shifts from 2.58 to 0.2 eV, indicating the formation of numerous energy levels associated with nonequilibrium charge carriers, such as electrons on ${\mathrm{O}}_2^{•-}$ radicals and holes (V₀), which act as traps. It is important to note that energy transitions occur between these levels. The E_g values, obtained without separation of the impurity energy states, which are generated by the electron and hole defects, from the host absorption for the samples are 3.19 eV (Figure 6a), 3.15 eV (Figure 6b), and 2.50 eV (Figure 6c). As shown in Figure 6d, the E_g of NTO-0.10/120 increases to 3.01 eV, attributed to electron–hole recombination, as described by Equation (11), or by incorporation of boron into the interstitial position.

$${\mathrm{h}}_{\mathrm{v}\mathrm{b}}+{\mathrm{T}\mathrm{i}}^{3+}\to {\mathrm{T}\mathrm{i}}^{4+}$$

(12)

It should be noted that the anodizing time significantly impacts the E_g energies, as depicted in Figure S7. The bandgap energy decreases with increasing anodizing time, observed as 3.21 eV for NTO-0.05/40, 3.16 eV for NTO-0.05/60 and 2.50 eV for NTO-0.05/120. Despite this, the corrected (E_g)_S values remain constant at 3.29 eV. This reduction in E_g is attributed to the increased influence of boron modification on the Ti₂O₄ cell structure of TiO₂, as described by Equations (3)–(7). Concurrently, the number of V_O defects rises, leading to the formation of Ti³⁺ sites, as indicated by Equation (12).

$${2\mathrm{Ti}}^{4+}+{\mathrm{O}}^{2-} \to {2\mathrm{Ti}}^{3+}+{\mathrm{V}}_0+{\displaystyle \frac{1}{2}}{\mathrm{O}}_2$$

(13)

The Γ_1b level, located in the 3d orbital of Ti⁴⁺, can be excited through indirect transitions Χ_{1a, 2b} → Γ_1b (with peaks at 3.07 eV in the red Gaussian in Figure 8, and peaks at 2.90 eV in the orange and 3.09 eV in the brown Gaussians in Figure 5a–d) or via direct transition Γ_5′a → Γ_1b (violet Gaussians in Figure 5a–d, with peak energy ranging from 3.67 to 3.85 eV, and a peak at 3.87 eV in the dark-blue Gaussian in Figure 9). These radiative transitions between levels in the d-orbital of Ti and the p-orbital of O allow the observation of emission spectra, which are formed by bandgap transitions from the conduction band (CB) to the valence band (VB) (Figure 7).

The spectrum in Figure 8, obtained by the excitation of the Γ₃ → Γ_5′b (peak at 4.69 eV in the light-blue Gaussian in Figure 9) and Γ_2′ → Γ_1b (peak at 4.22 eV in the broad purple Gaussian in Figure 9) transitions, reveals peaks with energies of 3.07, 2.94 and 2.92 eV, which are associated with Γ_1b → Χ_2b and Γ_1b → Χ_{1b, 1a} transitions. Additionally, peaks at 2.81, 2.71 and 2.55 eV are observed, corresponding to oxygen vacancies in the anatase structure. These vacancies create defect levels within the bandgap, leading to non-emissive recombination of the h⁺/e⁻ pairs, as described by Equations (13) and (14) [102,104].

$${\mathrm{V}}_0+{\mathrm{e}}_{\mathrm{c}\mathrm{b}}^{-}\to {\mathrm{V}}_0^{-}$$

(14)

$${\mathrm{V}}_0^{-}+{\mathrm{h}}_{\mathrm{v}\mathrm{b}}^{+}\to {\mathrm{V}}_0+\mathrm{h}\mathsf{\upsilon }$$

(15)

The spectra for the NTO-0.05/40 sample are presented in Figure 9.

In addition to the previously mentioned transitions, the spectra reveal peaks with energies of 3.35 eV, 3.23 eV and 3.52 eV, corresponding to the Χ_1a → X_1b, Γ₃ → X_1b and Χ_2b → Χ_1b transitions, respectively, as well as a peak at 3.67 eV associated with the Χ_2b → Χ_1b transition. Additionally, a peak at 5.26 eV is linked to the Μ_5’ → Μ_1,4 transition [106]. The spectrum shown in Figure 9 closely resembles recent observations in [102].

Figure S8 presents the luminescence spectra for the NTO-0/120, NTO-0.01/120, NTO-0.05/120 and NTO-0.10/120 samples. Notably, the sample NTO-0.05/120 spectrum exhibits decreased intensity and an increased Urbach tail intensity (Figure 5c). This suggests a significant presence of electron traps that act as sites for non-emissive recombination, primarily involving oxygen radicals ${\mathrm{O}}_2^{•-}$ and oxygen vacancies V₀ within the TiO₂ structure. The spectrum also indicates a reduction in the luminescence intensity between the NTO-0/120 and NTO-0.05/120 samples. This reduction may be attributed to increased surface-absorbed oxygen, as described in Equations (3)–(7). A similar phenomenon is observed for the NTO-0.05/y sample with increasing anodizing time (Figure S9).

In this context, the absorbed oxygen ions on the surface, along with the electrons and holes generated at these sites, likely reduce the formation of electron–hole pairs within the boron-modified TiO₂ structure. Consequently, this results in decreased emissive recombination and lower luminescence intensity. The peaks at 4.22 eV and 4.69 eV indicate the decomposition of the emission structure into two components: a fast and a slow component, with the fast component predominating. The fast component is attributed to the luminescence of Ti⁴⁺ within the nanotube structure, particularly along the {100}, {101} and {001} facets. The slow component is associated with Ti⁴⁺ ions within TiO₂ nanoparticles. Therefore, luminescence decay curves (Figure S10) should be analyzed using a two-exponential function model (Equation (15)):

$$\mathrm{I}={\mathrm{I}}_1\exp \left(-{\displaystyle \frac{\mathrm{t}}{{\mathsf{\tau }}_1}}\right)+{\mathrm{I}}_2\exp \left(-{\displaystyle \frac{\mathrm{t}}{{\mathsf{\tau }}_2}}\right)$$

(16)

Here, the lifetime was calculated using Equation (16) from [107]:

$$\mathsf{\tau }={\displaystyle \frac{{\mathrm{I}}_1{\mathsf{\tau }}_1^2+{\mathrm{I}}_2{\mathsf{\tau }}_2^2}{{\mathrm{I}}_1{\mathsf{\tau }}_1+{\mathrm{I}}_2{\mathsf{\tau }}_2}}$$

(17)

The lifetimes of emissive recombination for the NTO-x/40, NTO-x/60 and NTO-x/120 samples are 5.63, 5.95 and 6.47 μs, respectively. An increase in the emissive recombination time corresponds to a decrease in the luminescence intensity. Despite this, the carrier diffusion and transport times relevant to photocatalytic water-splitting processes occur on the nano- to microsecond scale [108].

The Raman spectrum of the NTO-0.05/60 sample is shown in Figure S11. This spectrum is characteristic of anatase [104], displaying peaks corresponding to Ti−O vibrations at 142, 193 and 392 cm⁻¹, assigned to the E_1g, E_2g and B_1g modes, respectively, representing translational and deformation vibrations. The vibration observed at 512 cm⁻¹ (A_1g + B_1g) corresponds to the valence vibration of the Ti−O−Ti bond, while the 638 cm⁻¹ (E_g) vibration is attributed to the symmetric valence vibration of the Ti−O bond [109]. Additionally, a weak vibration at 798 cm⁻¹ (B_1g) is associated with the first harmonic in the second order.

3.4. Photoelectrochemical Measurements

The flatband potential (E_fb) and onset potential (E_onset) of the NTO samples were determined using “OCP-illuminated OCP” and j-V measurements, as outlined in Section 2.3 and previous studies [61,110]. Selected experimental results from these measurements are presented in Figure S12. NTO samples subjected to sequential thermal treatment in air and hydrogen exhibit n-type semiconductor behavior, as confirmed by a negative shift in the illuminated open-circuit potential (i-OCP) compared to the dark OCP. The E_fb and E_onset values, both negative relative to the reference electrode (RE), are depicted in Figure 10.

The positions of E_fb and E_onset versus the hydrogen evolution reaction (HER) potentials suggest the feasibility of spontaneous water splitting under the given conditions [61]. The HER potential is at −582 ± 6 mV vs. Ag/AgCl (pH = 6.3 ± 0.1). Consequently, for most of the synthesized titania samples, spontaneous water splitting is not achievable without applied external bias. However, for samples NTO-0/y and NTO-0.10/y, spontaneous water splitting is possible, albeit requiring a small overpotential of approximately 200–300 mV.

The incident photon-to-current efficiency (IPCE) measurements provide further insights into the photoelectrochemical properties of the NTO samples. The IPCE spectra for the NTO-x/y samples are shown in Figure 11 and Figure S13, revealing changes in light absorption and its conversion into energy necessary for the water decomposition reaction.

The relationship between the IPCE and the wavelength enables the evaluation of the material’s conversion bandgap. This parameter can be calculated similarly to the optical bandgap using absorption spectra and a Tauc plot. As per [58], the conversion bandgap (E_ipce) is obtained by extrapolating the linear portion of the IPCE versus the photon energy (hν) plot in the coordinates (IPCE·hν)^1/n − hν, where n = 0.5 for a direct transition. Thus, the E_ipce or IPCE bandgap represents the minimum photon energy required for the PEC process. Figure 12 illustrates the E_ipce values as a function of the boric acid concentration in the anodizing electrolyte.

The data clearly show that E_ipce depends on the BA concentration in anodizing electrolytes. As shown in Figure 12, the E_ipce decreases with an increasing BA concentration from 0 to 0.10 M. However, when the BA concentration reaches 0.25 M, resulting in a nanoporous layer, the E_ipce dramatically increases to 3.40 ± 0.05 eV, which negatively affects the PEC properties of the NTO. Further increases in the BA concentration above 0.25 M, leading to the formation of a non-porous layer, result in E_ipce values exceeding 3.45 eV. A general trend of decreasing E_ipce with increasing anodizing time is observed across the entire range of BA concentrations. Notably, the E_g and E_ipce values both decrease with the anodizing time, as illustrated by the Eg values of 3.21, 3.15 and 2.50 eV for NTO-0.05/40, NTO-0.05/60 and NTO-0.05/120, respectively. Although the absolute values of E_g and E_ipce are not equal, their similar trends highlight their different origins, as previously discussed in [56,58]. Specifically, E_g encompasses all the possible electron–hole excitation and recombination pathways, while E_ipce reflects only those electron–hole pairs with sufficiently long lifetimes to participate in photoelectrochemical processes.

The luminescence spectra of NTO-0.05/y (Figure S8) indicate a slight decrease in intensity and an increased electron–hole recombination rate, which correlates with the higher intensity of the Urbach tails, non-radiative recombination via continuous defect energy levels, and a significantly reduced bandgap (E_g = 0.2 eV). Conversely, the IPCE spectra (Figure 11 and Figure S13) reveal a minor decrease in photoconversion efficiency, which is also linked to these defect energy levels. However, the luminescence intensity across various dopant concentrations remains relatively consistent (Figure S8), suggesting that the observed changes in the PEC efficiency are not solely due to variations in luminescence. While some studies have proposed a direct relationship between luminescence and PEC efficiency [19,33], our findings indicate that the energy levels identified in the luminescence spectra of boron-modified samples are not directly associated with changes in the IPCE spectra or PEC efficiency. Instead, we propose that the introduction of B³⁺ into the system likely suppresses the formation of oxygen vacancies typically associated with Ti³⁺ defects, thereby increasing the number of non-radiative recombination centers. This suppression likely represents the primary influence of in situ boron modification on PEC properties.

Non-porous and nanoporous NTOs exhibit low PEC efficiency due to the insufficient thickness and limited active surface area, whereas nanotubular NTOs demonstrate enhanced PEC properties, with the integral IPCE efficiency increasing up to three-fold at higher BA concentrations. The integral IPCE efficiency (Figure 13) correlates well with the thickness of the NTO coatings (Table 2) for the NTO-0/y, NTO-0.01/y and NTO-0.10/y series, though the NTO-0.05/y series shows an inverse correlation.

This anomaly is likely due to an increased charge carrier pathway, heightened recombination through intra-bandgap states, and structural disordering, as indicated by the increased defect energy levels, non-radiative luminescence, and Urbach tail phenomena.

The IPCE peak position shift in nanotubular regions (Figure 11 and Figure S13) is minimal, not exceeding 10 nm, while nanoporous and non-porous titania coatings show a more significant shift of up to 20 nm. However, the last-mentioned shift toward deeper UV regions is not conducive to effective utilization. The decreased performance of the NTO-0.05/y series is likely due to the ineffective combination of defect energy levels within the bandgap, which are associated with structural or surface disordering of the anatase lattice. This disordering leads to a substantial decrease in E_g and an increase in Urbach tail intensity, promoting the recombination of photoinduced electron–hole pairs rather than their transfer to photoelectrochemical active sites.

Pristine NTO-0/y and low-level doped NTO-0.01/y samples exhibit similar integral IPCE efficiency due to the relatively high E_g and E_ipce values and strong preferential grain orientation along the (004) plane. In contrast, the NTO-0.10/y series demonstrates superior performance across all the anodizing times, attributed to a lower E_g and E_ipce, moderate crystal lattice disordering, and reduced preferential orientation of anatase grains along the (004) plane. These characteristics, likely influenced by boron modification and the appropriate influence of B³⁺ species on the properties of modified NTO, contribute to an advantageous IPCE efficiency in PEC-type PCWS processes.

4. Conclusions

Boron-modified reduced NTO crystalline and energy structures were systematically analyzed using XRD, UV-Vis diffuse reflectance, luminescence, and IPCE spectrometry. The results demonstrate that varying the boric acid content in an ethylene-glycol-based electrolyte containing fluoride ions promotes the synthesis of different NTO coating types, including nanotubular, nanoporous, and non-porous structures. NTO coatings formed under optimal conditions exhibit a strongly oriented crystalline grain structure along with the (004) plane, which is advantageous for PEC applications. In contrast, nanoporous titania coatings are composed of a more conventional texture. Luminescence and UV-Vis analyses identified defects such as Ti³⁺, V₀, surface O^2•− radicals and Urbach tails induced by boron introduction. The primary effect of beneficial boron modification appears to be the suppression of oxygen vacancy formation, reducing the non-radiative recombination, as indicated by the luminescence spectra. The highest observed IPCE efficiency was approximately 60%, with a conversion peak at 355 nm and corresponding E_g, (E_g)_s and E_ipce values of about 1.93, 3.22 and 3.26 eV, respectively. The integral IPCE efficiency of boron-modified NTOs ranged from 6.8% to 22.9% across the 300–420 nm wavelength range, leading to a 3.3-fold increase in PEC efficiency.

This study expands the potential applications of NTOs in photocatalytic processes and highlights additional methods for enhancing PEC performance through the strategic incorporation of dopants.