Optically Transparent TiO₂ and ZnO Photocatalytic Thin Films via Salicylate-Based Sol Formulations

Abstract

Sol compositions for transparent TiO₂ and ZnO photocatalytic thin film deposition are of interest for the wet-chemical fabrication of self-cleaning coatings. The choice of stabilizing agent is crucial for the sol film-forming properties, with acetylacetone and monoethanolamine conventionally employed for TiO₂ and ZnO deposition sols, respectively. Salicylic acid (SA), capable of chelating both Ti(IV) and Zn(II) precursors, remains underexplored. This study presents novel SA-based sol formulations for the deposition of both TiO₂ and ZnO films, based on titanium tetraisopropoxide (TTIP) and zinc acetate dihydrate (ZAD) precursors, in a fixed 1:3 (TTIP:SA) and 1:2 (ZAD:SA) ratio, and isopropanol solvent, varied across the 1:10 to 1:20 precursor-to-solvent ratio range. Fourier-Transform Infrared Spectroscopy analysis and Density Functional Theory computations confirmed the formation of H₂Ti[SA]₃ and Zn[SA]₂·2H₂O complexes. Scanning Electron Microscopy, X-ray diffraction, and Ultraviolet-Visible spectroscopy were employed to study the structural and optical properties of the dip-coated films, revealing dense TiO₂ (86–205 nm) and ZnO (35–90 nm) layers of thickness proportional to the salicylate concentration and transmittance in the 70–90% range. Liquid-phase Methylene blue (MB) photooxidation experiments revealed that all films exhibit photocatalytic activity, with ZnO films being superior to TiO₂, with 2.288 vs. 0.366 nm h⁻¹ cm⁻² MB removal rates.

1. Introduction

Since its discovery in the late 1970s by Honda and Fujishima [1], semiconductor heterogeneous photocatalysis has become pivotal in “green” environmental technologies, such as water and air pollution abatement [2,3,4]; photoelectrochemical water splitting and CO₂ reduction for energy conversion [5,6,7]; microbial disinfection [8]; and others [9]. Photocatalytic activity is attributed to a variety of wide-bandgap transition metal oxides; however, titanium dioxide (TiO₂) and zinc oxide (ZnO) are the most widely studied photocatalysts [4,10,11]. Absorbing light quanta of energy exceeding their respective bandgaps, E_g ≈ 3.2 eV (λ ≤ 387 nm) for TiO₂ and E_g ≈ 3.37 eV (λ ≤ 367 nm) for ZnO [7,12], these materials generate electron-hole pairs, also known as excitons. While most of the photogenerated excitons recombine, a small fraction (typ. < 1%) manage to migrate to the photocatalyst surface, interacting with surface-adsorbed species and facilitating their photocatalytic degradation through photooxidation and photoreduction reactions [12].

While TiO₂ and ZnO photocatalysts are frequently used as nanoparticle suspensions in research [4,13] for practical applications and catalyst reusability, the development of photocatalytic coatings holds greater significance [11]. Typically, highly efficient TiO₂ and ZnO coatings are in the form of thick mesoporous layers to maximize their specific surface area [14]. However, these mesoporous coatings may not be practical for applications requiring functional photocatalytic layers with intrinsic optical transparency, such as self-cleaning and antifogging coatings on windowpanes or photovoltaics [15,16,17].

Transparent TiO₂ and ZnO layers can be fabricated through a variety of deposition methods. Vacuum-based physical vapor deposition (PVD) techniques, such as reactive magnetron sputtering [15,17,18,19], are commonly favored, however, they can be cost-prohibitive and require specialized equipment and expertise, making them less suitable for small-scale applications. As a result, wet-chemical methods, particularly the sol-gel-based techniques, are preferred, especially in research-scale applications, due to their simplicity, cost-effectiveness, and ability to deposit films without the need for high-vacuum conditions [16,20,21]. In sol-gel deposition, the coating is initially formed from a liquid sol formulation, applied to the substrate using techniques like dip-coating or spin coating, and subsequently converted to the desired metal oxide via post-deposition heat-treatment step [22].

The sol formulation contains at least three main components: a metal-ion-containing precursor, an organic solvent, and a stabilizing agent to limit undesired hydrolysis and improve sol stability. In TiO₂ sol formulations, titanium tetrachloride (TiCl₄) [13] and titanium (IV) tetraisopropoxide (TTIP, Ti(O-iPr)₄) [14,17,21,23] are commonly used as precursors, with the latter preferred due to its lower volatility and absence of toxic HCl released during hydrolysis. Isopropyl alcohol (IPA) is generally used as a solvent in the TTIP case. As for ZnO deposition sols, while examples based on zinc alkoxides have been demonstrated [23], ionic Zn(II) compounds like zinc nitrate [17,24] and, in the predominant case, zinc acetate dihydrate (ZAD, Zn(CH₃COO)₂·2H₂O), are more commonly employed in the literature [13,25,26,27]. Formulating ZAD-based ZnO deposition sols presents challenges due to ZAD’s poor solubility in organic solvents and proneness for self-hydrolysis, even in anhydrous media, due to the hydration water molecules [28]. These factors affect both the ease of preparation and long-term stability of ZAD-based formulations, necessitating the use of specialized, harmful solvents like 2-methoxyethanol (2-ME) [24,25,26,27].

The choice of stabilizing agent in both TiO₂ and ZnO sol formulations significantly impacts the quality and structure of the resulting thin films. Acetylacetone (AcAc) is typically used to stabilize TTIP-based sols [13,21,29], while monoethanolamine (MEA) or other substituted amines are employed in the ZAD case [20,26]. However, some studies have shown the interchangeability of these chelating agents between the two cases [13,30]. For instance, Verma et al. compared the effects of AcAc and diethanolamine (DEA) on TiO₂ films, finding that AcAc-stabilized sols produced films with increased transparency, while DEA led to improved electrochemical performance in terms of ion storage capacity, attributed to enhanced porosity [30]. Vajargah et al. studied the effects of different substituted amines on ZAD-based ZnO deposition sols and found that the stabilizer significantly affects sol stability, with mono- and di-substituted amines having the strongest effect [20]. Vilà et al. compared the influence of MEA and its substituted derivatives to aromatic amino alcohols and observed that aromatic-based stabilizers resulted in films with increased transparency [26].

To enhance sol stability and shelf-life and improve the chances of forming transparent coatings, ligands that form stronger chelates with Ti (IV) and Zn (II) could be advantageous stabilizing agents for TiO₂ and ZnO deposition sols. Sato and Nishide demonstrated that sols based on Ti[EDTA] complexes yield TiO₂ thin films with excellent optical transparency and defined this technique as the molecular precursor method (MPM) [31,32]. Similarly, a more recent work by Amakali et al. showed the successful application of this approach for ZnO thin films [25]. While EDTA is the default chelating agent in chemistry practice, there are more environmentally friendly alternatives available. One such example is salicylic acid (SA), a member of the catechol class of molecules, which has been shown to form stable complexes with both Ti [33] and Zn [34], and the stability of the Ti[SA] complex has been demonstrated to surpass that of Ti[EDTA] [35].

Thus, SA shows promise as a stabilizing agent for TiO₂ and ZnO deposition sols. However, this premise remains underexplored in the literature. The SA film-forming properties have been investigated only on one occasion of ZnO deposition sols by Kuznetsova et al. [36], revealing that salicylate sols yield optically transparent ZnO films with improved electrical properties. This study employed a zinc precursor-to-SA ratio of 1:1, deviating from the typical zinc salicylate complex stoichiometry of 1:2 [34]. Additionally, the photocatalytic activity of the ZnO layers was not explored. In the case of TiO₂, the sole available demonstration of SA-stabilized deposition sol is found in the previous work of this paper’s author [37], where a SA-based TTIP sol with a 1:3 TTIP:SA molar ratio proved suitable for a highly-conformal TiO₂-functionalization coating on a high-aspect ratio anodic alumina substrate [36], and the selection of SA as the stabilizing agent stemmed from its affinity to the alumina surface. Hence, in the present study, the deposition of photocatalytic thin films of TiO₂ and ZnO from novel SA-based sol formulations, governed by Ti[SA] and Zn[SA] complexation stoichiometry of 1:3 and 1:2, respectively, is thoroughly investigated. Fourier Transform Infrared (FTIR) spectroscopy, along with Density Functional Theory (DFT) simulations, are employed to confirm the formation of salicylate complexes in the deposition sols. The effects of salicylate concentration vs. IPA solvent (1:10, 1:10, and 1:20 molar ratio) and their impact on the morphology and optical properties of the final coatings are investigated. These findings are further correlated to the photocatalytic activity in terms of liquid-phase Methylene blue (MB) dye photooxidation rates obtained at varied (UV) illumination intensity. The results may facilitate advancements in the field of photocatalytic thin film deposition, opening new possibilities for environmentally friendly and optically transparent photocatalytic coatings from SA-based formulations.

2. Materials and Methods

2.1. Reagents

All chemicals were of reagent grade or higher: titanium (IV) tetraisopropoxide (TTIP, ≥97%, Alfa Aesar, Haverhill, MA, USA); zinc acetate dihydrate (ZAD, Zn(CH₃COO)₂·2H₂O, ≥98%, Alfa Aesar); Salicylic acid (SA, ≥99%, Alfa Aesar); isopropyl alcohol (IPA, HPLC grade, ≥97%, Macron Fine Chemicals, Shanghai China); Methylene blue (MB, biological stain grade, Alfa Aesar). Distilled water (conductivity > 18 MΩ cm⁻¹) was used for all aqueous solution preparations.

2.2. TiO₂ and ZnO Sol Preparation and Thin Film Deposition

The workflow for the preparation of SA-based TiO₂ and ZnO sols and photocatalytic thin film deposition is illustrated schematically in Figure 1. The steps include synthesis and aging of the deposition sol solutions and dip-coating deposition on glass substrates and are described in greater detail in the following two subsections.

2.2.1. TiO₂ and ZnO Sol Synthesis

Both TiO₂ and ZnO deposition sols were based on the same general recipe, with SA as a stabilizing agent and IPA solvent, differing in several aspects: (1) the precursor—TTIP and ZAD for TiO₂ and ZnO deposition sols, respectively; (2) the precursor-to-stabilizer molar ratio, which was 1:3 for TTIP:SA and 1:2 for the ZAD:SA sol, as determined by the stoichiometry of the most-stable salicylate complex of the respective metal ion [33,34]; and (3) the use of reflux heating in the ZnO case, applied to accelerate the poor solubility of ZAD in IPA at room temperature (RT).

To investigate the impact of salicylate concentration on photocatalytic coatings, three formulations with different precursor-to-solvent molar ratios were prepared. The syntheses began with a fixed volume of 35 mL IPA. The appropriate amount of SA was added under magnetic stirring for 30 min, followed by the precursor. The specific synthesis procedures were as follows:

• TiO₂ Sol: TTIP was added dropwise into the SA/IPA mixture. During this step, a light-yellow fluffy precipitate was observed, especially in the Ti:10 and Ti:15 mixtures, but it redissolved under magnetic stirring. The mixture was continuously stirred for 3 h at RT, resulting in a clear, intensively red-orange colored sol. The three sols are denoted as Ti:10, Ti:15, and Ti:20, with 1:3:10, 1:3:15, and 1:3:20 TTIP:SA:IPA molar ratios, respectively.
• ZnO Sol: The calculated amount of ZAD was added to the SA/IPA solution at RT and magnetic stirring. The mixture was then brought up to 60 °C under reflux and stirred for an additional 3 h until the ZAD was completely dissolved. Finally, a clear sol with a faint pink color was obtained. The three sols are denoted as Zn:10, Zn:15, and Zn:20, with 1:2:10, 1:2:15, and 1:2:20 ZAD:SA:IPA molar ratios, respectively.

After synthesis, the sols were transferred to sealed polypropylene containers and aged for 1 week before deposition.

2.2.2. Dip-Coating of TiO₂ and ZnO Thin Films

The thin films were deposited via dip-coating on standard microscopy (soda-lime) glass slides (76 mm × 26 mm × 1.1 mm, Deltalab, Barcelona, Spain) serving as substrates. The substrates were mechanically cleaned with a mild detergent, thoroughly rinsed with distilled water, and dried before a 3-min ultrasonic cleaning in acetone. The initial weight of each slide was measured on an analytical balance (0.1 mg precision) to determine the deposition mass-loading.

Dip-coating was performed using a custom-made setup with a stepper-motor-driven linear actuator. The substrate immersion rate was set at 2.5 mm s⁻¹, followed by a 5-s dwell time in the sol, and a withdrawal at a 0.25 mm s⁻¹ rate. After coating, the samples underwent a two-step heat-treatment process: 15 min at 300 °C and 60 min at 450 °C, with a 4 °C min⁻¹ temperature ramp rate, and allowed to cool naturally in the furnace. Two coating cycles were applied to all samples used in characterization and photocatalytic experiments.

2.3. Characterization Methods

2.3.1. Physico-Chemical Characteristics of the Sols

The sol specific gravity ($\rho$) was determined gravimetrically in a 10 mL pycnometer flask. Surface tension was estimated by the stalagmometric (drop-counting) method, employing a capillary stalagmometer with a 2.5 mL capacity. The number of drops for each sol ($n_{sol}$) was counted, and the surface tension (${\sigma }_{Sol}$) calculated, according to the formula ${\sigma }_{Sol}={\sigma }_{IPA}({\rho }_{Sol}{n}_{IPA}/{\rho }_{IPA}{n}_{Sol})$, where ${\rho }_{Sol}$ is the specific gravity of the sol, $n_{IPA}$ is the number of drops for pure IPA, and ${\rho }_{IPA}$ and $n_{IPA}$ are the density (0.781 g cm⁻¹) and surface tension (20.9 mN m⁻¹) of IPA according to the literature values [38,39]. The kinematic viscosity (ν) was measured using a calibrated Ubbelohde viscometer (viscometer constant $\kappa$ = 0.010022 mm² s⁻¹), based on the average flow time (τ) from three separate measurements, and converted to dynamic viscosity (η) using the respective sol density, according to Equation (1):

$$\eta \left[\mathrm{P}\mathrm{a}{\mathrm{s}}^{-1}\right]=\mathsf{\tau }\left[\mathrm{s}\right]\times \kappa \left[\mathrm{m}{\mathrm{m}}^2{\mathrm{s}}^{-1}\right]\times {10}^{-6}\times \rho \left[\mathrm{k}\mathrm{g}{\mathrm{m}}^{-3}\right].$$

(1)

2.3.2. Instrumental Methods

Attenuated Total Reflection Fourier-Transform Infrared (ATR-FTIR) spectra of the sols were acquired using an Carry 630 spectrometer (Agilent Technologies Inc., Santa Clara, CA, USA) equipped with a Diamond-ATR sampling module. Transmittance Ultraviolet-Visible (UV-Vis) spectra of the thin films were measured in the 300–1000 nm range on an Evolution 300 spectrometer equipped with a DRA-EV-300 Diffuse Reflectance Accessory. Scanning Electron Microscopy (SEM, Thermo Fischer Scientific Inc., Waltham, MA, USA) was performed on a LYRA I XMU microscope (TESCAN GROUP, Brno, Czech Republic), and X-ray Diffraction (XRD) patterns were obtained on D8 Advance diffractometer (Bruker AXS GmbH, Karlsruhe, Germany) with CuKα X-Ray source and LynxEye PSD detector.

2.4. Photocatalytic Activity Determination

The photocatalytic activity of the TiO₂ and ZnO thin films was assessed by liquid-phase photooxidation (PCO) experiments of Methylene blue (MB) dye—a widely-employed model pollutant and the by-default standard for photocatalyst coatings activity screening, as featured in the ISO 10678:2010 specification. The MB PCO removal mechanism is relatively well-understood and thoroughly investigated by Houas et al. [40]. In the aqueous liquid-phase, it is known to be governed mainly through the generation of reactive species, such as hydroxyl radicals ($H{O}^{•}$), as shown in Equations (2)–(5), which in turn react with the MB dye molecules, resulting in its decolorization and mineralization.

$$Ti{O}_2,ZnO\stackrel{h\upsilon }{\to }{h}^{+}+e^{-}$$

(2)

$$H_{2}O+h^{+}\to H{O}^{•}+H^{+}$$

(3)

$$O_2+e^{-}\to O_2^{•-}$$

(4)

$$2O_2^{•-}+{2H}^{+}\to 2H{O}_2^{•}\to H_2{O}_2+O_2\to H{O}^{•}+O{H}^{-}$$

(5)

For the MB PCO screening experiments, the as-deposited TiO₂ and ZnO photocatalyst-coated slides were cut to 15 mm × 26 mm rectangles, corresponding to a geometric area of 3.9 cm², noting that approx. 5 mm from the bottom and top of the microscope slides were discarded to exclude dip-coating artifacts.

The experiments were conducted in a custom-made automated setup, whose development has been previously detailed in a separate publication [41]. A cross-sectional view of the photocatalytic reactor is depicted in Figure 2a. In brief, a 5-cm-path optical cuvette (50 mm × 18 mm × 36 mm inside dimensions), serving as the reaction cell, contains the sample, reaction medium, and a 10-mm stirring bar to ensure its agitation.

Illumination was provided by a 3W UV (λ = 365 ± 5 nm) LED (GP-3WUVA-G45, GMKJ LED, Shenzen, China), positioned above the sample, behind a collimating lens, and a UV intensity sensor (ML8511, Lapis Semiconductor, Kanagawa, Japan) to provide optical power feedback, which was correlated to the UV illumination intensity (I^UV, mW cm⁻²) at the sample surface by calibration with a PM400 optical power meter, equipped with a S175C thermopile sensor head (Thorlabs Inc., Newton, NJ, USA).

MB concentration is measured in situ by a laser-diode-based (λ = 650 nm, 5 mW optical power) colorimeter, whose emission wavelength is a close match with the MB absorbance peak at 662 nm, as shown by the MB UV-Vis absorbance spectra available in Figure S1a in the Supplementary materials.

The laser intensity is monitored by photodiode detectors before (I₀) and after (I₁) the liquid reaction cell, allowing the calculation of the MB absorbance ($A^{MB}$) according to Equation (6):

$$A^{MB}=-\mathrm{l}\mathrm{o}\mathrm{g}(I_1/I_0)$$

(6)

MB concentration ($C^{MB}$) is proportional to $A^{MB}$ and can be calculated for each data point using Equation (7), where $A_0^{MB}$ and $C_0^{MB}$ represent the initial absorbance and concentration of MB, respectively. A set of $A^{MB}$ vs. $C^{MB}$ calibration plots are supplemented in Figure S1b–d in the Supplementary Materials.

$$C^{MB}=C_0^{MB}\times (A^{MB}/A_0^{MB})$$

(7)

In experiments, 15 mL of aqueous MB solution ($C_0^{MB}$ = 1 ppm) was added to the reaction cell. Initially, a 30 min phase ensures for adsorption-desorption equilibrium is reached, followed by 60 min of UV illumination. $C^{MB}$ is measured every 5 min throughout the entire sequence of the experiment to obtain a concentration profile, as shown in Figure 2b. Two quantitative descriptors were extracted from each such dataset:

• The MB saturation concentration ($C_{sat}^{MB}$) was estimated by fitting the non-dissociative Langmuir adsorption model (Equation (8)).

  $$(1-C^{MB})=C_{sat}^{MB}\times \left(1-e^{-K_{Ads}^{MB}t}\right)$$

  (8)
  where $K_{Ads}^{MB}$ is the MB adsorption rate constant, and $t$ is the time coordinate. $C_{sat}^{MB}$ was then converted to molar saturation coverage per geometric area (nmol cm⁻²), as shown in Equation (9), where $V^{Cell}$ represents the reaction volume (15 mL), $M_W^{MB}$ is the molecular weight of MB dye (319.85 g mol⁻¹), and $S^{Film}$ is the geometric area of the sample (3.9 cm²).

  $${\theta }^{MB}[\mathrm{n}\mathrm{m}\mathrm{o}\mathrm{l}\mathrm{c}{\mathrm{m}}^{-2}]=\frac{{C_{sat}^{MB}\left[\mathrm{p}\mathrm{p}\mathrm{m}\right]\times V^{Cell}\left[\mathrm{m}\mathrm{L}\right]\times 10}^3}{M_W^{MB}\left[\mathrm{g}\mathrm{m}\mathrm{o}{\mathrm{l}}^{-1}\right]\times S^{Film}\left[\mathrm{c}{\mathrm{m}}^2\right]}$$

  (9)
  The ${\theta }^{MB}$ parameter is thus indicative of the differences in the specific surface area between the photocatalytic samples.
• The MB PCO reaction rate (${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$) and reaction rate constant (${\mathrm{K}}_{PCO}^{\mathrm{M}\mathrm{B}}$) were determined from the decrease in $C^{MB}$ during UV illumination. ${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$ was obtained by linear regression, as shown in Equation (10), where $C_{0-PCO}^{MB}$ represents the initial $C^{MB}$ after the dark adsorption phase.

  $$C^{MB}\left[\mathrm{p}\mathrm{p}\mathrm{m}\right]=C_{0-PCO}^{MB}\left[\mathrm{p}\mathrm{p}\mathrm{m}\right]-{\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}\left[\mathrm{p}\mathrm{p}\mathrm{m}{\mathrm{h}}^{-1}\right]\times t\left[\mathrm{h}\right]$$

  (10)
  ${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$ was converted to molar MB PCO removal rate per geometric area (nmol h⁻¹ cm⁻²) using a similar approach as that shown in Equation (9). Additionally, ${\mathrm{K}}_{PCO}^{\mathrm{M}\mathrm{B}}$ rate constants (h⁻¹) were calculated using the pseudo-first-order kinetic model, commonly employed to describe MB PCO removal rates, as shown in Equation (11).

  $$-\mathrm{l}\mathrm{n}(C^{MB}/C_{0-PCO}^{MB})={\mathrm{K}}_{PCO}^{\mathrm{M}\mathrm{B}}\left[{\mathrm{h}}^{-1}\right]\times t\left[\mathrm{h}\right]$$

  (11)

To provide context for the PCO values, a reference coating of a commercial photocatalyst was evaluated alongside the SA-based TiO₂ and ZnO thin films. It was prepared by drop-casting an aqueous suspension of 0.5 mg mL⁻¹ Degussa P25 (now AEROXIDE^® TiO₂ P25, Evonik Industries AG, Essen, Germany). The suspension was sonicated for 15 min, and a 2 mL aliquot was applied to the surface of a clean 15 mm × 26 mm glass substrate and allowed to dry naturally at RT, followed by heat treatment at 450 °C for 1 h.

Additionally, to compensate for the contributions of MB adsorption on the reaction cell walls and stirring bar during, as well as photolysis during, the PCO phase, a set of blank experiments was conveyed using uncoated glass slides of the same 15 mm × 26 mm dimension. ${\theta }^{MB}$ was determined to be 0.146 ± 0.05 nmol cm⁻² and ${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$ was found to be 0.109 ± 0.029 nmol h⁻¹ cm⁻².

2.5. Computational Methods

2.5.1. Density Functional Theory (DFT) Computations

Density Functional Theory (DFT) computations were employed to predict the vibrational frequencies of the expected Ti[SA] and Zn[SA] complexes. The structures and input files were created using Avogadro ver. 1.2.0 [42], and the GAMESS 2021-R1 [43] package was utilized for the computations. The B3LYP DFT functional [44,45] with Grimme’s D3 dispersion corrections [46] was employed. Hydrogen, oxygen, and carbon atoms were described with 6-311G(d,p) triple zeta basis sets [47], while Ti and Zn atoms were treated with the LANL2DZ-ECP basis with effective core potentials [48,49]. The basis set parameters were obtained from the Basis Set Exchange software [50]. All geometry optimizations and IR-frequency predictions were carried out with simulated IPA solvent with the Polarizable Continuum Model (PCM) implemented in GAMESS.

2.5.2. Data Treatment and Model Fitting

All data treatment, plotting, and model fitting were performed in R version 4.2.2 [51]. Non-linear multivariate fitting was carried out using the Hooke and Jeeves method, as implemented in the {optimr} R package [52]. Additionally, the {pracma} package was used for its complex implementation of the digamma and cotangent functions [53], which were required for some of the optical data model fitting.

3. Results and Discussion

3.1. ATR-FTIR & DFT Investigation of the TiO₂ and ZnO Salicylate Sols

ATR-FTIR spectra of the salicylate TiO₂ and ZnO deposition sols are shown in Figure 3b,c, alongside a reference spectrum of the SA ligand in IPA depicted in Figure 3a. The spectra encompass the fingerprint region (1800–650 cm⁻¹), which will be comprehensively discussed in the subsequent paragraphs. The ν(C-H) and ν(O-H) group frequency range (3500–2800 cm⁻¹) is also presented, although no significant changes are apparent within this interval. It features a broad band around 3337 cm⁻¹ and a triplet at 2972, 2934, and 2886 cm⁻¹, correlated with ν(O-H) and ν(C-H) modes, respectively, and dominated by the signature of the IPA solvent. The full-scale, uncropped, experimental spectra are included in the Supplementary Materials (Figure S2).

In the free SA ligand spectrum (Figure 3a), the carboxylic ν(C=O) stretching band is evident at 1670 cm⁻¹, followed by a peak at 1610 cm⁻¹ with a 1588 cm⁻¹ shoulder, corresponding to the ν_as(COO⁻) asymmetric vibration and ν(C=C) aromatic ring modes [54,55,56], implying partial dissociation of SA in IPA. IR absorbance bands at 1487 cm⁻¹ and 1465 cm⁻¹ are linked to ν(C=C) vibrational modes of the aromatic ring [56,57]. The ν_s(COO⁻) symmetric band is anticipated around 1380 cm⁻¹ [55,56] and apparently is obscured by the IPA solvent background. A broad peak at 1301 cm⁻¹ is associated with the carboxylic δ(OH) bending mode, while the doublet at 1245 cm⁻¹ and 1219 cm⁻¹ corresponds to phenolic ν(C-OH) and δ(OH) modes, respectively [54]. Relatively low-intensity features at 1032 cm⁻¹, 850 cm⁻¹, 757 cm⁻¹, 701 cm⁻¹, and 657 cm⁻¹ arise from δ(C-H) modes of the benzene ring [55]. Other discernible features in the free SA spectra include a pronounced IPA solvent contribution, evident from its standalone FTIR spectra overlayed as a grey infill in the Figure 3a plot for reference.

In the Ti[SA] set (Figure 3b), the ν(C=O) band is reduced in intensity while remaining discernible at 1670 cm⁻¹, alongside the ν_as(COO⁻), ν(C=C) doublet, subtly shifted to 1604 cm⁻¹ and 1583 cm⁻¹. Within the ring vibration area, 1500–1400 cm⁻¹, a 1460 cm⁻¹ band emerges, growing more prominent with the Ti[SA] concentration. The carboxylic δ(OH) peak at 1301 cm⁻¹ broadens and becomes upshifted to 1310 cm⁻¹. Concomitantly, a decrease in absorbance is observed in the phenolic δ(OH) shoulder, implying an interaction between the Ti atom and oxygens in both the -COOH and -OH groups of SA. New peaks manifest at 887 cm⁻¹ and 671 cm⁻¹, potentially associated with Ti-O-C modes.

By comparison, in the Zn[SA] set (Figure 3c), the ν(C=O) band is suppressed, indicative of a robust interaction between the Zn atom and the carboxyl oxygens. The emergence of a new band at 1713 cm⁻¹ corresponds to the ν(C=O) of CH₃COOH released by the dissolution of ZAD. The ν_as(COO) and ν(C=C) doublet transforms into a triplet, centered at 1599 cm⁻¹, whereas the 1610 cm⁻¹ and 1588 cm⁻¹ bands shift in opposite directions: to 1630 cm⁻¹ and 1569 cm⁻¹. This effect, akin to observations in other bivalent metal salicylates, such as Cu²⁺ and Fe²⁺ [56], underscores interaction with the -COOH group. Significantly, the ν_s(COO⁻) region, now centered at 1376 cm⁻¹, becomes more prominent, while the phenolic ν(CO) and δ(OH) peaks broaden, rendering the carboxylic δ(OH) peak indistinguishable. Analogous to the Ti[SA] case, the ν(C=C) feature at 1460 cm⁻¹ becomes more pronounced, accompanied by the emergence of new bands at 866 cm⁻¹ and 671 cm⁻¹, potentially assignable to Zn-O-C modes.

In both instances, the emerging spectral features attributed to the salicylate formation exhibit an increase proportional to that of the precursor concentration. The evidence of bidentate bridging configuration, reliant on both the carboxylic and phenolic moieties of SA for the Ti[SA] complex and exclusively on the carboxyl group for the Zn[SA] one aligns with the existing literature regarding the preferred complexation modes of these ions with catechol molecules [34,58,59]. Drawing from this insight, it is plausible to infer that the Ti[SA] structure aligns with the H₂Ti[SA]₃ configuration elucidated by Gigant et al. [33]. Similarly, the Zn[SA] complex matches with the Zn[SA]₂·2H₂O structure proposed by Klug et al. [60], accommodating the hydration water molecules released by the ZAD precursor, as well.

DFT Modelling of Ti[SA] and Zn[SA] Complexes

To match the experimental ATR-FTIR data with the expected H₂Ti[SA]₃ and Zn[SA]₂·2H₂O complexes structure, their IR spectra were predicted computationally. The geometries of both complexes, depicted in Figure 4a, were based on the aforementioned studies by Gigant and Klug [33,60] and optimized at the B3LYP-D3 LANL2DZ-ECP/6-311(d,p) level of theory. The XYZ coordinates of the optimized geometries are provided in the Supplementary Materials (Tables S1 and S2).

Figure 4b presents the predicted IR vibrational frequencies for both complexes, juxtaposed with the experimental ATR-FTIR spectra of the Ti:10 and Zn:10 sols. A synthetic IR spectrum is generated by augmenting them with Gaussian functions (7 cm⁻¹ RMS width). The experimental IPA spectrum is included as a grey overlay in Figure 4b, alongside its sum with the Gaussian-broadened DFT-predicted spectra. Qualitatively, the DFT predictions match well the principal features observed in the experimental spectra.

Table 1. A list of vibrational frequencies predicted by DFT modeling of the free SA ligand and H₂Ti[SA]₃ and Zn[SA]₂·2H₂O complexes, assigned to experimentally observed features in their respective ATR-FTIR spectra. All of the predicted frequencies were scaled by a factor of 0.98.

Vibrational Modes *	SA		H2Ti[SA]3		Zn[SA]2·2H2O
	DFT, cm−1	Observed, cm−1	DFT, cm−1	Observed, cm−1	DFT, cm−1	Observed, cm−1
ν(C=O)	1731	1670	1657	1670	-	-
νas(COO−) + ν(C=C)	1623 1577	1610 1588	1616, 1608 1595	1604 1583	1627 1584 1540	1630 1599 1569
ν(C=C)	1483 1470	1487 1465	1466 1451, 1438	1487 1460	1482 1460	1487 1464
νs(COO−)	1380	1379	1327	1338	1397, 1371	1397, 1376
δ(OH) + ν(COO−)	1324, 1312	1301	1404 1296	1400 1310	-	-
ν(Ph-OH)	1240	1245	1269, 1242	1245, 1224	1247	1248
δ(Ph-OH)	1236	1219	-	-	1227	1223
δ(CH)	1181 1155 1100	1159 1126 1107	1182 1150 1123, 1090	1159 1126 1107	1162 1143	1159 1126 1107
δ(C=C)	1023 842 631	1032 850 657	1034 845 -	1032 850 -	1029 815 -	1032 - -
δ(C=C), δ(CH) + δ(COO−) out of plane	753 695	757 701	770 712	757 701	754 708	757 704
ν(Me-O)	-	-	883 680	887 671	875 674	866 671

* Notation: ν—stretching (ν_as—asymmetric, ν_s—symmetric), δ—bending, C=O—carboxylic carbonyl, C=C—aromatic ring, COO⁻—carboxylic group, OH—carboxylic hydroxyl, Ph-OH—phenolic hydroxyl, CH—aromatic hydrogens.

provides a compilation of significant IR vibrational frequencies predicted for both complexes, as well as the free SA ligand. The frequencies are attributed to distinct vibrational modes based on the associated atom displacements. The table also correlates these frequencies with the most prominent ATR-FTIR spectral characteristics that were experimentally detected within corresponding spectral ranges.

For the free SA ligand, the calculated frequencies are in line with previous combined FTIR-DFT investigations [55]. The predicted ν(C=O) frequency (1731 cm⁻¹) is higher compared to the observed one (1650 cm⁻¹). However, in organic solvents, SA tends to form dimeric structures [54], while in this case, free SA was modeled as a monomer. For the H₂Ti[SA]₃ and Zn[SA]₂·2H₂O complexes, the predicted frequencies account well for the emergence of new bands in the lower-frequency range, including Ti-O-C and Zn-O-C stretching modes, anticipated at 883 cm⁻¹ and 875 cm⁻¹, respectively, consistent with experimental observations.

The primary distinctions between the two complexes spectra stem from the different interactions between the Ti and Zn central atom and the -COOH group. In the H₂Ti[SA]_3, this interaction exhibits asymmetry with respect to the two carboxyl oxygens, which is further accentuated by the presence of SA ligands with different protonation states (Ti[SA²⁻][HSA⁻]₂), in accordance with the structure proposed by Gigant et al. [33]. Notably, the appearance of a ν(C=O) band is solely caused by the doubly-deprotonated SA ligand, whereas the other two contribute to carboxylate FTIR region through their ν_as(COO⁻) modes, collectively intensifying the ~1600 cm⁻¹ area. The ν_s(COO⁻) mode is suppressed and red-shifted from 1380 cm⁻¹ (in free SA) to 1327 cm⁻¹, resulting in a broad feature, combined with the carboxylic δ(OH) mode and phenolic ν(Ph-OH) C-O stretch, predicted at 1296 cm⁻¹ and 1269 cm⁻¹.

In the Zn[SA]₂·2H₂O complex, the Zn center engages symmetrically with both carbonyl oxygens, resulting in a subdued ν(C=O) band. The ν_as(COO⁻) disperses across several IR frequencies in the 1630–1540 cm⁻¹ range, with a concurrent reduction in intensity. In contrast, the ν_s(COO⁻) band, predicted around 1397–1371 cm⁻¹, displays enhanced intensity, in agreement with the observations from the experimental Zn:10 spectra.

An interesting aspect illuminated by the calculated IR spectra is the subtle difference between the two complexes, observed in the in-plane δ(CH) bending region (1160–1100 cm⁻¹) of the experimental ATR-FTIR spectra. For Zn[SA]₂·2H₂O, the predicted intensity of the in-plane δ(CH) bending modes is notably attenuated when compared to the H₂Ti[SA]₃ case. In the standalone IPA spectrum, this region is characterized by a distinct triplet at 1159 cm⁻¹, 1126 cm⁻¹, and 1107 cm⁻¹, a pattern maintained in the Zn:10 case, as shown in Figure 4b. However, for H₂Ti[SA]₃, the additional absorbance in this area contributes to enhancing the 1159 cm⁻¹ and 1126 cm⁻¹ peaks, causing the 1107 cm⁻¹ band to appear as a shoulder, in line with the experimental Ti:10 result.

In summary, the DFT calculations provide strong confirmation of the presence of the H₂Ti[SA]₃ and Zn[SA]₂·2H₂O complexes, offering a qualitative match with the experimental ATR-FTIR observations. These results underscore the formation of these complexes within the TiO₂ and ZnO deposition sols.

3.2. Effect of Precursor Concentration on the Physicochemical Parameters of TiO₂ and ZnO Deposition Sols and Resulting Thin Film Mass-Loading

Elucidating the formation of Ti and Zn salicylates in the sols, the next aim is to investigate the effects of their concentration on the physicochemical properties of the deposition solutions. The sol-specific gravity ($\rho$), surface tension ($\sigma$), and viscosity ($\eta$) collectively impact the thickness of the dip-coated layers [22].

Table 2. Summary of physicochemical parameters of the TiO₂ and ZnO deposition sols at the precursor to IPA ratios of 1:10, 1:15, and 1:20. Values include specific gravity ($\rho$), surface tension ($\sigma$), and dynamic viscosity ($\eta$). The theoretically predicted area-normalized mass loadings, according to the Landau–Levich model ($∆m^{theor}$), are also listed, as well as the experimentally obtained ones for the first ($∆m^I$) and second ($∆m^{II}$) dip-coated layers, and their cumulative value ($∆m^{tot}$).

Composition	$\mathit{\rho }$ * g cm−3	$\mathit{\sigma }$ * mN m−2	$\mathit{\eta }$ * mPa s−1	$∆{\mathit{m}}^{\mathit{t}\mathit{h}\mathit{e}\mathit{o}\mathit{r}}$  µg cm−2	$∆{\mathit{m}}^{\mathit{I}}$  µg cm−2	$∆{\mathit{m}}^{\mathit{I}\mathit{I}}$  µg cm−2	$∆{\mathit{m}}^{\mathit{t}\mathit{o}\mathit{t}}$  µg cm−2
Ti:10	0.913 ± 0.005	26.80 ± 1.03	8.38 ± 0.01	17.31	21.49 ± 1.42	33.63 ± 1.70	55.12 ± 2.16
Ti:15	0.878 ± 0.004	24.26 ± 0.73	6.07 ± 0.03	11.32	13.16 ± 1.70	13.64 ± 1.54	26.80 ± 2.38
Ti:20	0.856 ± 0.001	22.19 ± 0.54	4.77 ± 0.01	8.15	6.86 ± 1.51	10.80 ± 3.10	17.66 ± 3.32
Zn:10	0.960 ± 0.001	25.62 ± 0.53	6.83 ± 0.01	18.21	25.00 ± 4.58	18.70 ± 2.80	43.71 ± 4.42
Zn:15	0.912 ± 0.001	24.34 ± 0.14	4.46 ± 0.01	10.59	16.50 ± 2.96	6.89 ± 1.55	23.40 ± 2.30
Zn:20	0.876 ± 0.001	23.09 ± 0.81	3.53 ± 0.01	7.37	11.17 ± 3.15	8.49 ± 3.14	19.65 ± 1.40

* Values were obtained at room temperature (25 ± 1 °C) immediately prior to the dip-coating of the respective samples for validity during the dip-coating process.

lists the experimentally obtained $\rho$, $\sigma$, and $\eta$ values, measured immediately after aging and prior dip-coating, for all six sols as a function of their precursor-to-solvent ratios (1:20, 1:15, and 1:10).

Using the values of $\rho$, $\sigma$, and $\eta$, the anticipated dip-coated layer thickness ($h^{L-L}$) can be predicted by the Landau–Levich equation [22,61]:

$$h^{L-L}=0.94\frac{{\left(\eta U\right)}^{2/3}}{{\sigma }^{1/6}\sqrt{\rho g}},$$

(12)

where $U$ is the withdrawal velocity (0.25 mm s⁻¹) and $g$ signifies the gravitational acceleration (9.8 m s⁻²). Thus obtained, $h^{L-L}$ reflects the thickness of the liquid sol layer drawn along the moving substrate during the dip-coating process. However, it can be readily converted to the anticipated post-calcination mass-per-area metal oxide loading, using Equation (13) to enable a comparison with experimental gravimetric analysis values:

$$∆m^{theor}=h^{L-L}\rho w_{prec}\frac{M_w^{prec}}{M{w}^{oxide}},$$

(13)

where $\rho$ denotes the sol density, $w_{prec}$ is the precursor weight fraction, and $M_w^{prec}$ and $M{w}^{oxide}$ are the molar masses of the precursor (220 g mol⁻¹ for ZAD and 284 g mol⁻¹ for TTIP), and the oxide (81 g mol⁻¹ for ZnO and 78 g mol⁻¹ for TiO₂), respectively.

The $∆m^{theor}$ values, predicted via Equations (12) and (13), for all sol compositions are listed in Table 2, along with experimentally obtained values. The latter were averaged from sets of six thin films, deposited in two dip-coating cycles, from each sol composition, with a full 450 °C calcination after each cycle. The experimental values reflect the mass change after the first ($∆m^I$) and second ($∆m^{II}$) cycle, as well as the cumulative loading ($∆m^{tot}$).

The data in Table 2 clearly demonstrates that precursor concentration results in a proportional rise in $\rho$, $\sigma$, and $\eta$ in both the TiO₂ and ZnO deposition sols, which, in accordance with the Landau–Levich model, consequently, leads to an increased mass loading, respectively thickness, of the dip-coated layers. A qualitative comparison between $∆m^{theor}$ and experimental ∆m values reveal a good agreement. The Landau–Levich model accurately predicts a marginally higher mass loading for ZnO films in all cases. However, the predicted values tend to underestimate the observed values by approximately 20% for the Ti-sols and 35% for the Zn-sols.

This discrepancy can be attributed to the effects of solvent evaporation during dip-coating, which are not accounted for by the Landau–Levich equation and are generally thought to increase the solute concentration in the deposited liquid layer, as demonstrated by Kuznetsov and Xiong [62], which in turn effectively increases the resulting mass-loading.

A notable difference in the behavior of the TiO₂ and ZnO deposition sols, reflected in the application of subsequently dip-coated layers, is illustrated in Figure 5. Figure 5a plots the effect of precursor concentrations against the experimentally observed mass-change ($∆m$) for the initial- ($∆m^I$ in Table 2) and subsequently dip-coated layer ($∆m^I+∆m^{II}$, or $∆m^{tot}$ in Table 2). To aid interpretation, the salicylate content was converted to molar concentration, applicable both for the TTIP and ZAD precursors, as well as the anticipated H₂Ti[SA]₃ and Zn[SA]₂·2H₂O complexes. Notably, the linear increase in mass-loading vs. precursor content, discussed above, is observed in both cases only for the initial layer. However, for the overcoated layer, $∆m$ is systematically lower in the ZnO case.

This discrepancy may be attributed to differing interactions between the salicylate solution and the two surfaces. Catechols, such SA, possess a strong affinity towards MeO_x surfaces, including TiO₂, with the possibility of forming bridging structures [59] even with the available -COOH groups when in the H₂Ti[SA]₃ complex. Consequently, when overcoated on the previously deposited TiO₂ surface, the Ti-sol may have a lower surface tension, facilitating a highly conformal coating, akin to the observations for alumina surfaces using a salicylate-based Ti-sol in our previous publication [37].

Conversely, ZnO is known for low stability in acidic environments and in the presence of strong ligands [63]. Hence, potential dissolution of previously deposited ZnO cannot be discounted. To test this hypothesis, in a separate experiment, TiO₂ and ZnO films were deposited with up to five consecutive layers, employing the Ti:15 and Zn:20 sols, which yield similar mass loading per layer, of 11 μg cm⁻². As shown in Figure 5b, a clear linear increase in $∆m$ vs. the number of dip-coating cycles is obtained for the Ti:15 sol, while a plateau in the $∆m$ values after the second dip-coating cycle is notably observed in the Zn:20 case, indicating dissolution of the previously deposited layers.

In summary, the results demonstrate that precursor concentration affects the physicochemical parameters of the sol deposition solutions, which in hand have a direct effect on the resulting thin film thicknesses. The experimental data also suggest greater applicability of the Ti-salicylate system on the more chemically stable TiO₂ surface for achieving tunable film thickness through multiple dip-coating cycles, as compared to the Zn-salicylate sol, due to ZnO dissolution.

3.3. Structure & Morphology of Salicylate-Based TiO₂ and ZnO Thin Films

SEM micrographs revealing the morphology of films deposited through two dip-coating cycles from each of the six sol compositions are shown in Figure 6.

The TiO₂ films (Figure 6a–c) exhibit dense, uniform surfaces with fine grains, consistent with typical sol-gel titania films [29,30]. No apparent morphological changes attributed to precursor concentration effects are discernible. Cross-sectional micrographs (Figure 6a) reveal a thickness of approx. 200 nm for the Ti:10-deposited layers, alongside a compact internal structure.

The ZnO films (Figure 6d–f) exhibit a comparable morphology, featuring slightly more granular structures. Notably, this morphology differs from traditional sol-gel films obtained from ZAD:MEA:2-ME sols, often defined as “wrinkled,” ganglia-like patterns [26], and is more consistent with the one obtained by the EDTA-based MPM method [25]. Nevertheless, the solvent influence cannot be ignored, as evidenced by the work of Vajargah et al. [20], where uniform, finely granular morphologies were achieved using ethanolamine-stabilized sols based on 1-propanol. The thickness of the Zn:10-deposited ZnO film is approximately 90 nm (Figure 6d).

Comparing the Ti:10 and Zn:10 cases, the observed thickness contrasts with their similar $∆m^{tot}$ values of 55 ± 2 µg cm⁻² for TiO₂ and 44 ± 4 µg cm₋₂ for ZnO, which, alongside bulk densities of 4.2 g cm⁻³ for TiO₂ and 5.6 g cm⁻³ for ZnO, can be used to roughly estimate an expected thickness of 131 ± 5 nm and 79 ± 7 nm, assuming dense films. This discrepancy suggests that the deposited films are of lower density due to intrinsic porosity.

XRD diffraction patterns for all TiO₂ and ZnO films are shown in Figure 7 and, given their low thickness, exhibit a strong amorphous halo caused by the glass substrate.

Nevertheless, in the TiO₂ films’ XRD patterns (Figure 7a) the formation of anatase TiO₂ can be suggested by the (101) reflection observed in the Ti:10-deposited film. Scherrer analysis estimates an average crystallite size of 14 nm for this case.

Similarly, the ZnO films from the Zn:10-sol (Figure 7b) exhibit (100), (002), and (101) reflections, consistent with the zincite structure. Noticeably, the comparable intensity of the (100) reflection to that of (101) implies an a-axis preferential orientation, characteristic of ZnO sols with precursor concentrations > 0.2 mol L⁻¹, as noted by Znaidi [27]. Scherrer analysis of the (101) peak indicates a mean crystallite size of 22 nm for the Zn:10 deposited film.

Quantitative analysis of crystallite size was unfeasible for other cases due to lower XRD peak intensity, potentially attributed to film thinness or lower crystallinity, resulting from salicylate compositions with precursor-to-solvent ratios below 1:10.

3.4. Optical Properties of the TiO₂ and ZnO Thin Films

As evident by the photographic images and UV-Vis transmittance spectra, shown in Figure 8, all of the TiO₂ (Figure 8a,c) and ZnO films (Figure 8b,c), exhibited optical transparency in the visible region, regardless of the deposition sol concentration. The transmittance spectrum of the bare glass substrate is also shown in Figure 8b,d for reference.

The TiO₂ films exhibited a slight color tint, also reflected as interference fringes in their transmittance spectra (Figure 8b), with a noticeable red-shift in proportion to deposition sol precursor concentration, arising from TiO₂‘s higher refractive index (2.52 for bulk anatase) compared to ZnO (2 for bulk zincite), combined with and greater film thickness of the titania films, as discussed in the previous sections.

The ZnO films exhibited no significant features in their transparency range (400 nm < λ < 1000 nm), with transmittance values approaching that of the bare substrate. Below 400 nm, the TiO₂ films show a sharp transmittance drop due to bandgap excitation. Instead, ZnO films are defined by a distinct dip-and-plateau feature around 359 nm, attributed to excitonic absorption [26], which becomes more prominent in the films deposited from higher precursor concentration sols and apparently is related to their thickness.

The optical bandgaps ($E_g$) of the TiO₂ and ZnO thin films were determined using the Tauc relation, ${\left(\alpha E\right)}^n=B(E-E_g)$, where $\alpha$ represents the absorption coefficient obtained from transmittance spectra, $B$ is a constant, and $E$ is the photon energy in eV. For indirect bandgap semiconductors like TiO₂, $n=½$ [18], while for direct-bandgap semiconductors like ZnO, $n=2$ [19,26]. Analysis of ${\left(\alpha E\right)}^n$ vs. ($E$) plots for Ti:10 deposited TiO₂ (Figure 9a), and Zn:10 deposited ZnO films (Figure 9b) revealed $E_g$ values of 3.34 eV and 3.3 eV, respectively.

Tauc plots for films obtained from the other sol solutions are provided in Supplementary Materials (Figure S3). Reducing the Ti[SA] complex-to-solvent ratio from 1:10 (Ti:10) to 1:20 (Ti:20) led to an $E_g$ increase to 3.43 eV, attributed to poorer crystallinity and reduced thickness. Conversely, in ZnO films, $E_g$ remained stable, with a marginal redshift from 3.30 to 3.28 eV for the lower Zn[SA] precursor concentration sols, values consistent with that of sol-gel deposited ZnO thin films [25].

3.4.1. Transfer Matrix Method Modelling of TiO₂ and ZnO Thin Film Transmittance Spectra

To gain further insight into the thin films’ optical properties, their UV-Vis transmittance spectra were mathematically modeled, employing the transfer matrix method (TMM) [64]. As depicted in Figure 10, the dip-coated samples were modeled as a three-layer configuration.

Each layer is defined by its thickness ($d$) and complex refractive index ($\tilde{n}$), defined as $\tilde{n}=n+ik$, where $n$ represents its ordinary index of refraction and $k$—the optical extinction coefficient. The thin films layers (TiO₂ or ZnO) are symmetrically deposited on both sides of the substrate ($d_2=d_{sub}$, ${\tilde{n}}_2={\tilde{n}}_{sub}$), and assumed to be of the same thickness ($d_1=d_3=d_{film}$) and refractive index (${\tilde{n}}_1={\tilde{n}}_3={\tilde{n}}_{film}$). The stack is enveloped by air (${\tilde{n}}_0={\tilde{n}}_4=1+i0$).

Assigning distinct transmittance coefficients ($T_m^{\prime }$) to each of the three layers, the overall transmittance of the structure ($T$) can be defined as their product:

$$T={\prod }_{m=1}^3{T}_m^{\prime },$$

(14)

where the values of $T_m^{\prime }$ are determined from each layer’s individual system transfer matrices $S_m$:

$$S_m=\left[\begin{array}{cc}{S}_{11}& S_{12}\\ S_{21}& S_{22}\end{array}\right]=I_{m-1,m}{P}_m{I}_{m,m+1}$$

(15)

where $I_{m-1,m}$ and $I_{m,m+1}$ denote the interface matrices between the m-layer and its neighboring media, and $P_m$ is the propagation matrix. The interface and propagation matrices for the optically coherent thin film layers (m = 1, 3 in Figure 10) are defined in Equations (16) and (17):

$$\begin{gathered} I_{m,n}=\frac{1}{t_{m,n}}\left[\begin{array}{cc}1& -r_{n,m}\\ r_{m,n}& \left(t_{m,n}{t}_{n,m}-r_{m,n}{r}_{n,m}\right)\end{array}\right] \\ \mathrm{w}\mathrm{h}\mathrm{e}\mathrm{r}\mathrm{e}{t}_{m,n}=\frac{2{\tilde{n}}_m}{{\tilde{n}}_m+{\tilde{n}}_n}\mathrm{a}\mathrm{n}\mathrm{d}{r}_{m,n}=\frac{{\tilde{n}}_m-n}{{\tilde{n}}_m+{\tilde{n}}_n} \end{gathered}$$

(16)

$$P_m=\left[\begin{array}{cc}{e}^{-i{\delta }_m}& 0\\ 0& e^{i{\delta }_m}\end{array}\right],\mathrm{w}\mathrm{h}\mathrm{e}\mathrm{r}\mathrm{e}{\delta }_m=\frac{2\pi {\tilde{n}}_m{d}_m}{\lambda }$$

(17)

and $\lambda$ is the transmitted light wavelength. Thus, for the optically coherent thin films $T_m^{\prime }$ is obtained by applying Equation (18) on the $S_{11}$ element, returned by the system transfer matrix in Equation (15).

$$T_m^{\prime }={\left|\frac{1}{S_{11}}\right|}^2$$

(18)

Applying the abovementioned approach to the substrate (m = 2 in Figure 10), whose thickness is three orders of magnitude greater than the transmitted wavelengths, results in high-frequencies oscillations in the resulting transmittance spectra, which cannot be resolved by instrumental UV-Vis spectroscopy. Thus, optical incoherence is introduced for the substrate by converting its transfer matrix to an intensity matrix, employing the square magnitude of its elements, by modifying Equations (16) and (17) into the form shown in Equation (19) [65].

$$I_{m,n}=\frac{1}{{\left|t_{m,n}\right|}^2}\left[\begin{array}{cc}1& -{\left|r_{n,m}\right|}^2\\ {\left|r_{m,n}\right|}^2& \left({\left|t_{m,n}-t_{n,m}\right|}^2-{\left|r_{m,n}-r_{n,m}\right|}^2\right)\end{array}\right],\mathrm{a}\mathrm{n}\mathrm{d}{P}_m=\left[\begin{array}{cc}{\left|{\mathrm{e}}^{-i{\mathsf{\delta }}_m}\right|}^2& 0\\ 0& {\left|{\mathrm{e}}^{i{\mathsf{\delta }}_m}\right|}^2\end{array}\right]$$

(19)

Using this approach, Equation (18) is simplified to:

$$T_m^{\prime }=\frac{1}{S_{11}}$$

(20)

The division of the overall transfer matrix into separate ones allows for separate treatment of the thin films and the intermediate substrate layer as optically coherent and incoherent components, respectively. Consequently, the overall transmittance of the dip-coated layer stack becomes a functional relationship of the transmitted wavelength, the complex refractive index dispersions of the substrate, ${\tilde{n}}_{sub}\left(\lambda \right)$, with constant, and the unknown thickness, $d_{film}$, and complex refractive index dispersion, ${\tilde{n}}_{film}\left(\lambda \right)$, of the dip-coated layers, that can be treated as fitting parameters.

The refractive index dispersion of the substrate, ${\tilde{n}}_{sub}\left(\lambda \right)$, is modeled using the wavelength-dependent refractive index, $n\left(\lambda \right)$, and optical extinction constants, $k\left(\lambda \right)$, derived from UV-Vis transmittance and reflectance spectra of the bare substrate, following the procedure of González-Leal [66]. These values were fitted to parametrize a Cauchy absorbent model [67], expressed in Equation (21).

$$\left\{\begin{array}{c}n\left(\lambda \right)=A+\frac{B\times {10}^4}{{\lambda }^2}+\frac{C\times {10}^9}{{\lambda }^4}\\ k\left(\lambda \right)=D\times {10}^{-5}+\frac{E\times {10}^4}{{\lambda }^2}+\frac{F\times {10}^9}{{\lambda }^4}\end{array}\right.$$

(21)

where A through F are the empiric constants. Consequently, the complex refractive index of the substrate becomes: ${\tilde{n}}_{sub}\left(\lambda \right)=n\left(A,B,C,\lambda \right)+ik(D,E,F,\lambda )$.

The refractive indices of the thin films are described with appropriate dispersion models of their complex dielectric function $\tilde{\epsilon }(\lambda$), instead, and subsequently transformed into a complex refractive index through the relation $\tilde{n}\left(\lambda \right)=\sqrt{\tilde{\epsilon }\left(\mathsf{\lambda }\right)}$.

In the case of TiO₂, a single Tauc–Lorentz oscillator model was employed, which is typically used for titania thin films [68]. Its functional form is: ${\tilde{\epsilon }}_{T-L}={\epsilon }_{T-L}^{Re}\left(E_g,{\epsilon }_{\infty },E_0,A,C,E\right)+i{\epsilon }_{T-L}^{Im}(E_g,{\epsilon }_{\infty },E_0,A,C,E)$, where $E_g$ signifies the optical bandgap, ${\epsilon }_{\infty }$ represents the dielectric function value at high frequencies, $E_0$ stands the oscillator energy, and $A$ and $C$ denote the amplitude and broadening coefficient of the oscillator peak, respectively. $E$ is the photon energy ($E=hc/\lambda$). The complete analytical expressions for both ${\epsilon }_{T-L}^{Re}$ and ${\epsilon }_{T-L}^{Im}$, as proposed by Jellison and Modine [69], were directly employed in the TiO₂-coated samples TMM models without modification.

For ZnO films, features in the transmittance spectra emerge only below its $E_g$ and are not covered by the Tauc–Lorentz dispersion. Instead, the Tanguy dispersion model, based on Wannier excitons [70], was applied and shown to yield a good description of ZnO’s excitonic features [19,71]. In this context, the complex dielectric function is directly given by Tanguy’s dispersion: ${\tilde{\epsilon }}_{Tanguy}(E_g,A,\mathsf{\Gamma },R)$. Here $E_g$ signifies the optical bandgap, $A$ represents the amplitude pre-factor, $\mathsf{\Gamma }$ is a broadening term of the energy levels, and $R$ denotes the exciton binding energy. The analytical solution given by Tanguy is applied directly [72] with the addition of the Sellmeier term, as shown in Equation (22), to improve the description of the transparency region.

$$\tilde{\epsilon }={\epsilon }_{\infty }+\frac{B}{E^2-E_S^2}+{\tilde{\epsilon }}_{Tanguy}(E_g,A,\mathsf{\Gamma },R)$$

(22)

The entire mathematical apparatus described above, including the transfer matrix equations, Cauchy, Tauc–Lorentz, and Tanguy dispersion functions, were implemented in R-scripts, which are provided in the Supplementary Materials (Appendix S1).

3.4.2. Results from Modelling of TiO₂ and ZnO Thin Film Transmittance Spectra

The TMM model fits the transmittance data presented in Figure 11, and the extracted parameters for the TiO₂ and ZnO films, obtained from all of the salicylate sol compositions, are listed in

Table 3. A summary of parameters obtained from optical modeling of the UV-Vis spectra of TiO₂ and ZnO thin films. The optical bandgap from Tauc analysis, denoted as ${\mathrm{E}}_{\mathrm{g}}^{Tauc}$, is provided for reference. ${\chi }^2$ is the quality of fit parameter.

TiO2 Sample	${\mathbf{E}}_{\mathbf{g}}^{\mathit{T}\mathit{a}\mathit{u}\mathit{c}}$(eV)	Tauc—Lorentz model fitting parameters							${\mathit{\chi }}^2$	${\mathit{d}}_{\mathit{f}\mathit{i}\mathit{l}\mathit{m}}$(nm)	$\mathit{n}$ *2	$\mathit{P}$ *3 (%)
		${\mathit{\epsilon }}_{\mathbf{\infty }}$		${\mathbf{E}}_{\mathbf{g}}$(eV)	$\mathit{A}$(eV)		${\mathit{E}}_0$(eV)	$\mathit{C}$(eV)
Ti:10	3.34	1.88		3.39	111.25		4.36	1.203	0.49	205	2.02	21
Ti:15	3.36	1.77		3.44	111.49		4.53	1.161	0.50	137	1.99	22
Ti:20	3.43	2.28		3.47	114.26		4.35	0.897	0.51	86	1.91	27
ZnO Sample	${\mathbf{E}}_{\mathbf{g}}^{\mathit{T}\mathit{a}\mathit{u}\mathit{c}}$ (eV)	Tanguy model fitting parameters							χ2	d (nm)	$\mathit{n}$ *2	$\mathit{P}$ *3 (%)
		${\mathit{\epsilon }}_{\mathbf{\infty }}$	${\mathbf{E}}_{\mathbf{g}}$(eV)	$\mathit{A}$(eV3/2)	$\mathit{R}$(meV)	$\mathsf{\Gamma }$(meV)	$\mathit{B}$(eV2)	${\mathit{E}}_{\mathit{s}}$(eV)
Zn:10	3.30	1 *1	3.45	8.22	60 *1	48	26 *1	7.1 *1	0.88	72	1.57	34
Zn:15	3.29		3.46	7.68		60			0.57	52	1.55	36
Zn:20	3.28		3.45	5.69		68			0.44	35	1.48	43

*¹ Values fixed during model fitting. *² Calculated via the respective model at λ = 450 nm. *³ Calculated via Equation (19), against $n_B$ = 2 for ZnO and 2.52 for TiO₂.

. The quality of the fits is assessed using the ${\chi }^2$ parameter. While the mathematical model fits do not perfectly match the experimental transmittance profiles, particularly in the extreme transparency regions dominated by free-carrier absorption, the results qualitatively capture the interference pattern of the TiO₂ films and the position and absorbance of the excitonic band in ZnO. These fits offer a satisfactory basis for extracting additional analytical insights into the impact of sol concentration on the resulting film properties from the transmittance spectra.

For the TiO₂ films, the TMM results estimate a film thickness ($d_{film}$) of 205 nm in the Ti:10 case, consistent with SEM observations. At lower precursor concentrations in sols Ti:15 and Ti:20, $d_{film}$ decreases to 137 nm and 86 nm, respectively, aligning well with gravimetric analysis data and the observed interference peaks (Figure 11a). Concerning the Tauc–Lorentz fitting parameters, the obtained $E_g$ values qualitatively match those from the Tauc plots, revealing a blueshift from 3.39 eV for Ti:10 to 3.47 eV for Ti:20. Further fitting parameters are detailed in Table 3 and show good agreement with the literature values for sol-gel TiO₂ thin films [68]. The refractive index of TiO₂ films falls within the range of 2.02 to 1.91, decreasing with precursor concentration. Employing these values, the film porosity is estimated at 21% to 27%, determined by Equation (23), derived from the Lorentz-Lorenz effective medium model:

$$\mathrm{P}=\left(1-\frac{\left({\mathrm{n}}^2-1\right)\left(n_B^2+2\right)}{\left(n^2+2\right)\left(n_B^2-1\right)}\right)\times 100[\%]$$

(23)

where $n_B$ represents, in this case, the refractive index of bulk anatase TiO₂ (2.52).

For the ZnO films model fits, the number of free variables was limited, considering the limited features in their UV-Vis spectra. The eight parameters encompassed by Equation (22) to describe Tanguy’s dispersion underwent a reduction by setting ${\epsilon }_{\mathrm{\infty }}=1$ for all fits; fixing the exciton energy $R=60 \mathrm{m}\mathrm{e}\mathrm{V}$, typical for ZnO films [71]; and optimizing Sellmeier term’s $B$ and $E_s$ independently. These are then fixed to the values specified in Table 3 across all three fits. Consequently, the description of ZnO thin films was streamlined to four parameters: $E_g$, $A$, and $\mathsf{\Gamma }$, along the film thickness $d_{film}$.

The resulting fits, depicted in Figure 11b, generally exhibit lower ${\chi }^2$ values and comparatively lower fitting quality compared to the TiO₂ fits. However, they offer some valuable insights. Notably, the TMM fitting indicates $d_{film}$ to span the range of 72 to 35 nm, exhibiting a decreasing trend with precursor concentration—congruent with expectations and gravimetric analysis data. $E_g$ values in the range of 3.45 to 3.46 eV are notably higher than those obtained from Tauc plots. The decrease in film thickness correlates with a systematic reduction in the amplitude ($A$) parameter, concurrently accompanied by an increase in the broadening ($\mathsf{\Gamma }$) term. Such trends align with the anticipated behavior for ZnO films of lower crystallinity [71,73].

Discernible from the UV-Vis spectra and TMM-predicted, the refractive indices within the ZnO films lie between 1.57 and 1.48, diminishing with increasing precursor concentration. Assuming a bulk refractive index of 2.0 for ZnO, these values translate to porosities spanning 34% to 43%.

3.5. Photocatalytic Activity of the TiO₂ and ZnO Thin Films

The effects of Ti[SA] and Zn[SA] precursor concentration in the deposition sols on the photocatalytic activity of the resulting coatings were investigated in liquid-phase MB photocatalytic oxidation (PCO) experiments. Each data point, discussed in this section is obtained from three separate runs of three parallel experiments and, hence, is based on an average of nine experimental values. As discussed in Section 2.3, each experiment provided two parameters: the saturation coverage of the model pollutant (${\theta }^{MB}$) and the rate of pollutant removal (${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$, ${\mathrm{K}}_{PCO}^{\mathrm{M}\mathrm{B}}$).

Beginning with the effects on the deposition sol composition on ${\theta }^{MB}$, the results are depicted in Figure 12a. Notably, in TiO₂ films, the deposition sol concentration yields a linear increase in ${\theta }^{MB}$, from 1.4 to 2.4 nmol cm⁻². This corresponds to approximately 0.87 × 10¹⁵ to 1.47 × 10¹⁵ molecules per square centimeter, while the established surface density of Ti-sites, relevant for monolayer formation on a planar titania surface, is documented at 0.52 × 10¹⁵ [74]. However, it should be noted that MB, being a bulky molecule, would probably not form monolayers with a single molecule per active site. The observed linear correlation between film thickness and ${\theta }^{MB}$ for TiO₂, alongside SEM analysis (Figure 6a–c), suggests that MB adsorption extends beyond the film surface, with dye molecules adsorbed inside the porous structure. Conversely, the ${\theta }^{MB}$ values for ZnO films were relatively stagnant, ranging from 1 nmol cm⁻² to 1.2 nmol cm⁻², regardless the concentration of the deposition sol or its corresponding impact on film thickness. This observation implies that any internal porosity within the ZnO films seems to have limited accessibility to MB dye adsorption.

The MB removal rate (${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$) across the Figure 12a sample set is shown in Figure 12b. Intriguingly, the increased MB adsorption accessible surface area does not yield a concomitant correlation with a high intrinsic photocatalytic activity. Specifically, films derived from the Ti:20 sol exhibited an ${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$ value of 0.149 nmol h⁻¹ cm⁻², increasing to 0.377 nmol h⁻¹ cm⁻² for Ti:15 and stagnating in the Ti:10 case, while the benchmark P25 reference notably outperformed these figures with an ${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$ value of 1.733 nmol h⁻¹ cm⁻².

In contrast, the ZnO films exhibit a clear linear increase in ${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$ with sol concentration, with rates increasing from 0.485 nmol h⁻¹ cm⁻² to 2.288 nmol h⁻¹ cm⁻² along Zn:20 to Zn:10 cases, surpassing the activity of the commercial P25 photocatalyst. This observation could be attributed to several factors. Firstly, the XRD patterns (Figure 7) indicate better crystallinity in the ZnO films, particularly evident in the Zn:10 and Zn:15 cases. Secondly, it is noteworthy that the soda-lime glass microscope slide substrates may allow diffusion of Na⁺ ions into the TiO₂ layer during heat treatment, thereby potentially lowering its activity [75]. This effect is less pronounced in the P25 sample, as it is not reliant on crystallization during this step. In contrast, for ZnO, Na⁺ incorporation has been demonstrated as a form of doping without any detrimental effects [76].

All ${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$ rates shown in Figure 12 were obtained under a similar UV irradiation intensity, $I^{UV}$= 0.88 ± 0.09 mW cm⁻². However, varying $I^{UV}$ during PCO experiments and analyzing the resulting dependence of ${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$ on this parameter can offer insights into the stability and quantum efficiency of a given photocatalyst. Mills et al. [77] established an interpretation that the photocatalytic reaction rate dependency on $I^{UV}$ follows three functional regimes: (1) low-to-medium $I^{UV}$ results in a linear effect on ${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$; (2) medium-to-high $I^{UV}$ leads to a square-root relationship with ${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$; and (3) at high $I^{UV}$, ${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$ is anticipated to plateau due to saturation. The precise transition values between these regimes hinge on the particular photocatalyst, as determined by its reaction turnover number per provided photon count. Typically, medium $I^{UV}$ values are considered to be above 1 mW cm⁻² for thin films of P25.

Figure 13 presents a comparison of the ${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$ vs. $I^{UV}$ profiles derived for TiO₂ thin films deposited using the Ti:10 sol, ZnO thin films from the Zn:10 sol, and the reference P25 sample.

The reference P25 sample exhibits a positive ${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$ vs. $I^{UV}$ dependence, characterized by the expected square-root increase across the 0.8–4.9 mW cm⁻² $I^{UV}$ range, where ${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$ increases from 1.733 nmol nmol h⁻¹ cm⁻² to 2.878 nmol nmol h⁻¹ cm⁻², beyond which no further ${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$ increases are anticipated. Similarly, the Ti:10 deposited TiO₂ thin film displays an increase in ${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$, from 0.366 nmol nmol h⁻¹ cm⁻² to 0.768 nmol nmol h⁻¹ cm⁻², discernible primarily at higher $I^{UV}$. This could imply that the Ti:10 film is already in the UV saturation region, a notion explored in the preceding paragraphs. Interestingly, the ZnO films manifest a detrimental effect of $I^{UV}$ on ${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$, decreasing by about 25%, from 2.288 nmol nmol h⁻¹ cm⁻² to 1.691 nmol nmol h⁻¹ cm⁻². Notably, akin behavior is evident in all Ti- and Zn-salicylate sol deposited samples, irrespective of precursor concentration, which primarily influences the extent of the reaction rate in the ZnO case. One potential explanation for this phenomenon is the photo-dissolution effect, wherein ZnO is prone to photocorrosion in aqueous solutions, especially under UV illumination. This effect hinges on the interaction of photogenerated holes ($h^{+}$) with nearby oxygen atoms on the ZnO surface, ultimately inducing photooxidative dissolution of the ZnO layer itself, as illustrated in Equation (24):

$$2{\mathrm{Z}\mathrm{n}\mathrm{O}}_{\left(\mathrm{s}\right)}+4h^{+}\to 2Z{n}_{aq.}^{2+}+O_2\uparrow$$

(24)

The influence of $I^{UV}$ on the photodissolution rates was examined by Han et al., revealing a positive dependence [63]. Given that the ZnO layers possess a thickness below 100 nm, photodissolution effects are likely to exert an adverse impact on their activity.

In summary, the impacts of sol concentration on both TiO₂ and ZnO deposited thin films prove intricate, affecting both MB saturation coverage and its PCO reaction rate contingent upon the inherent activity of the sample. Nonetheless, all photocatalytic thin films demonstrate satisfactory PCO activity when related to their thinness. This attribute, coupled with their intrinsic optical transparency, renders them suitable for applications necessitating this particular combination. All the data discussed in this subsection is tabulated in

Table 4. A summary of parameters obtained from the MB adsorption and PCO experiments for the reference P25 sample and all the TiO₂ and ZnO thin films deposited from the sols with a varied precursor concentration, including MB saturation coverage (${\theta }^{MB}$), UV illumination intensity (I^UV), and MB removal rate (${\mathrm{r}}_{PCO}^{\mathrm{M}\mathrm{B}}$) and corresponding pseudo-first-order rate constant (${\mathrm{K}}_{PCO}^{\mathrm{M}\mathrm{B}}$).

Sample	${\mathit{\theta }}^{\mathit{M}\mathit{B}}$  nmol cm−2	IUV mW cm−2	${\mathbf{r}}_{\mathit{P}\mathit{C}\mathit{O}}^{\mathbf{M}\mathbf{B}}$  nmol h−1 cm−2	${\mathbf{K}}_{\mathit{P}\mathit{C}\mathit{O}}^{\mathbf{M}\mathbf{B}}$  h−1
P25	1.447 ± 0.078	0.87 ± 0.09	1.733 ± 0.096	0.153 ± 0.009
		2.08 ± 0.31	2.495 ± 0.201	0.228 ± 0.020
		4.85 ± 0.33	2.878 ± 0.206	0.266 ± 0.021
Ti:10	2.426 ± 0.213	0.88 ± 0.09	0.366 ± 0.035	0.031 ± 0.003
		2.01 ± 0.26	0.324 ± 0.055	0.027 ± 0.005
		4.13 ± 0.27	0.768 ± 0.318	0.068 ± 0.029
Ti:15	1.756 ± 0.127	0.88 ± 0.09	0.377 ± 0.061	0.032 ± 0.005
		2.04 ± 0.28	0.347 ± 0.057	0.029 ± 0.005
		4.42 ± 0.29	0.475 ± 0.067	0.040 ± 0.006
Ti:20	1.441 ± 0.132	0.88 ± 0.09	0.149 ± 0.037	0.012 ± 0.003
		2.09 ± 0.29	0.228 ± 0.035	0.019 ± 0.003
		4.30 ± 0.24	0.404 ± 0.071	0.034 ± 0.006
Zn:10	1.159 ± 0.103	0.87 ± 0.09	2.288 ± 0.138	0.206 ± 0.014
		2.01 ± 0.26	1.931 ± 0.204	0.173 ± 0.019
		4.46 ± 0.36	1.691 ± 0.176	0.150 ± 0.017
Zn:15	1.169 ± 0.128	0.88 ± 0.09	1.232 ± 0.112	0.107 ± 0.010
		1.95 ± 0.25	1.040 ± 0.164	0.090 ± 0.015
		4.31 ± 0.36	0.750 ± 0.077	0.064 ± 0.007
Zn:20	1.007 ± 0.070	0.87 ± 0.09	0.485 ± 0.043	0.041 ± 0.004
		2.11 ± 0.28	0.508 ± 0.041	0.043 ± 0.004
		4.23 ± 0.25	0.468 ± 0.063	0.040 ± 0.005

for convenient reference. For the purpose of comparison with the existing literature, MB PCO rates are presented as conventional pseudo-first-order rate constants (${\mathrm{K}}_{PCO}^{\mathrm{M}\mathrm{B}}$), as well.

Such comparison with the recent literature data, which is available for similar thin film TiO₂ and ZnO coatings, is presented in

Table 5. Comparison between the apparent MB PCO removal constant of the samples with the highest activity in this article and the literature values for similar transparent TiO₂ and ZnO films.

Film (Thickness)/Substrate	Deposition Method	Reaction Conditions	${\mathbf{K}}_{\mathit{P}\mathit{C}\mathit{O}}^{\mathbf{M}\mathbf{B}}$ h−1	Ref.
TiO2 (198 nm)/Glass	Sol-gel dip-coating TTIP:SA:IPA	15 mL, 1 mg L−1 MB 3 W UV LED (365 nm)	0.068	This work
ZnO (90 nm)/Glass	Sol-gel dip-coating ZAD:SA:IPA		0.206
TiO2 (262 nm)/Glass	Sol-gel dip-coating TTIP:HCl:IPA	3 mL, 5.4 mg L−1 MB 15 W UV lamp (254 nm)	0.406	[78]
TiO2 (171 nm)/SiO2/Glass			0.397
TiO2 (219 nm)/SiO2/Glass			0.498
TiO2 (262 nm)/SiO2/Glass			0.798
TiO2 (203 nm)/Quartz	Metal-Organic CVD TTIP precursor	10 mg L−1 MB Natural sunlight	0.462 *	[79]
TiO2 (192 nm)/Glass	Sol-gel dip-coating Ti(oEt)4:EtOH:HNO3:H2O	35 mL, 2.5×10−5 mol L−1 MB 15 W UV lamp (254 nm)	0.882 *	[80]
TiO2 (199 nm)/Glass	Sol-gel dip-coating TTIP:AcAc:EtOH:HNO3:SDS	50 mL, 5 mg L−1 MB 11 W UV lamp	0.255 *	[81]
TiO2 (132 nm)/Glass	Magnetron Sputtering	10 mL, 10 mg L−1 MB 8 W UV tube	0.145 *	[82]
ZnO (122 nm)/Glass			0.412 *
ZnO (324 nm)/Glass	Spin-coating + hydrothermal ZAD:MEA:EtOH	100 mL, 20 mg L−1 MB 300 W Tungsten-Halogen	0.360 *	[83]
ZnO (47 nm)/Glass	Sol-gel dip-coating ZAD:MEA:IPA	45 mL, 10−5 mol L−1 MB 500 lx UV lamp (365 nm)	0.246 *	[84]
ZnO (130 nm)/Glass			0.276 *
ZnO (250 nm)/Glass			0.294 *

* Recalculated from the reported value in min⁻¹.

. In this case, only the ${\mathrm{K}}_{PCO}^{\mathrm{M}\mathrm{B}}$ values are employed, as they are typically reported in these studies. Clearly, the salicylate deposited TiO₂ and ZnO films, albeit with limited thickness and crystallinity, show a comparable activity with that of those presented in other studies. It should be noted, however, that a direct comparison between the reported apparent pseudo-first-order rate constants is difficult and, in this case, verging on meaningless, as they do not reflect on the variations in reaction conditions (e.g., UV illumination intensity, model pollutant concentration, reaction volume or exposed geometric area of the photocatalyst layer, etc.) and only on the MB removal rate from its relative liquid-phase concentration decrease point of view.

4. Conclusions and Outlook

In summary, the results outlined in this work successfully demonstrated the use of salicylic acid (SA) as a stabilizing agent in sol formulations, suitable for dip-coating deposition of thin films of titanium dioxide (TiO₂) and zinc oxide (ZnO). The use of SA, as a strong chelating agent, allowed the same general sol recipe to be used in both cases, employing isopropyl alcohol (IPA) as a solvent, which is notable for the ZnO case, where the higher-toxicity 2-methoxyethanol solvent was effectively replaced. The successful application of traditionally employed precursors for each of the photocatalytic materials—titanium tetraisopropoxide (TTIP) for TiO₂ and zinc acetate dihydrate (ZAD) for ZnO, demonstrates that the salicylate route is universal and may probably be extended for the deposition of other metal oxide photocatalysts, where similar formulations can be devised, based on the stoichiometry of the most-stable salicylate complex.

Regarding the salicylate complex formation, the ATR-FTIR investigation, paired with DFT computations, underscored the formation of titanium (IV) and zinc (II) salicylate complexes in the deposition sol solutions, ascribed to H₂Ti[SA]₃ and Zn[SA]₂·2H₂O, which are established as stable products of the reaction between the respective metal oxide precursors and the SA ligand, and the results of this investigation may be helpful in guiding other researchers in FTIR vibrational mode assignments of metal salicylate complexes.

The relationship between the concentration of the Ti[SA] and Zn[SA] complexes and the sol physicochemical parameters—viscosity, density, and surface tension—was revealed to have a direct influence on the characteristics of the thin films prepared via dip-coating. Notably, the variations in sol concentration exhibited a pronounced impact on the thicknesses of both the resulting salicylate-based sol TiO₂ and ZnO thin films, as evident from SEM imaging, UV-Vis spectroscopy, and mathematical modeling of the transmittance spectra.

The investigation into the photocatalytic activity of the thin films yielded that sol concentration can be linked to the effective surface area of the films, as judged by their saturation coverage. However, it cannot be directly translated to intrinsic photocatalytic activity. In terms of saturation coverage, TiO₂ films exhibited a linear increase in MB saturation coverage with sol concentration, while the ZnO films remained relatively stagnant. Conversely, in MB PCO rate terms, the ZnO films surprisingly displayed a linear increase with sol concentration, surpassing the activity even of the reference P25 TiO₂ sample. However, the dependence of photooxidation reaction rate on the UV illumination intensity revealed a detrimental effect in the ZnO case, possibly caused by photodissolution, indicating potential challenges in the stability of ZnO films under prolonged UV exposure.

In conclusion, this study offers a comprehensive exploration of salicylate-based sol-gel derived TiO₂ and ZnO thin films. The transparency and photocatalytic properties of the films makes them promising for applications, such as transparent coatings for environmental remediation and advanced photoelectrochemical and photonic devices. Further research could explore the optimization of film properties and address challenges related to photodissolution, thereby unlocking the full potential of these materials for practical applications.