Wet-Chemical Fabrication of Functional Humidity Sensors on a TiO₂-Coated Glass Substrate via UV Photodeposition

Abstract

This work demonstrates a completely wet-chemical procedure for the fabrication of a functional impedimetric humidity-sensing device on a titania (TiO₂) surface. Optically transparent anatase TiO₂ thin films were deposited on a glass substrate via dip-coating from a titanium tetraisopropoxide (TTIP)–acetylacetonate (AA)-based sol and surface-functionalized with a nickel oxide (NiO_x) layer by ultraviolet (UV) photodeposition. Photodeposition was employed to form the interdigitated electrode pattern on the TiO₂ surface as well through activation with a silver catalyst to promote electroless copper deposition. The relative humidity (RH) response of the pristine TiO₂- and NiO_x/TiO₂-functionalized sensors was studied by impedance (Z) measurements in the 15%–90% RH range. It was found that while NiO_x functionalization significantly dampens the RH–Z functional dependence, it improves its overall linearity and may successfully be employed for the purposeful design of titania-based sensing devices.

1. Introduction

Titanium dioxide (TiO₂) is a wide-bandgap (E_g = 3.2 eV) n-type semiconductor known for its photocatalytic properties, non-toxicity, affordability, and high chemical stability. These characteristics have established TiO₂ as one of the most studied heterogeneous photocatalysts [1], employed in more than 55% of the literature studies on photocatalysis [2], where it finds numerous applications in various “green” technologies, such as the remediation of polluted water and air, self-cleaning and anti-bacterial surfaces, and the conversion of solar energy to value-added chemicals [3].

Apart from its extensive use as a heterogeneous photocatalyst, TiO₂ is attracting attention as a suitable material for applications in microelectronics due to its electrical properties, such as a high κ dielectric (κ ~ 40 for anatase and 100 for rutile phase) and insulating properties (resistivity > 10¹³ Ω·cm) [4], combined with an excellent optical properties (refractive index for anatase TiO₂ > 2.4) [5,6]. Functional TiO₂ layers have thus been used as a basis for the fabrication of various microelectronic devices, including metal–insulator–metal (MIM) memory devices [7,8], photodetectors [9,10,11], and transparent conductive oxide layers [5]. TiO₂ is also suitable for constructing sensors for environmental monitoring [12], such as chemoresistive hydrogen and volatile organic compounds (VOC) sensors and relative humidity (RH) sensors [13,14,15,16,17,18,19]. Especially for applications in RH-sensing devices, TiO₂ is close to being an ideal material due to its unique combination of optical and electric properties, paired with a high chemical stability, which has been demonstrated in optical and surface plasmon resonance (SPR) [20], MEMS [21], and resistive (impedimetric) sensors [16,19], with the last type having an advantage in simplicity due to the ease of readout of the RH-modulated electrical signal.

Impedimetric TiO₂-based sensing devices, however, may suffer from poor humidity response linearity and limited site selectivity. These issues can be mitigated through its surface functionalization with another transition metal oxide, ideally a p-type semiconductor like nickel oxide (NiO_x), a common material in chemical sensor fabrication [11,18,22,23]. This strategy improves charge transport through the formation of pn junctions and simultaneously is shown to increase the number of chemisorption sites available to adsorbate molecules [24], thus enhancing the overall sensing performance of the TiO₂ layer.

The deposition and surface functionalization of TiO₂ layers for sensing and microelectronic devices typically involves advanced physical or chemical vapor deposition techniques such as magnetron sputtering [13,15], e-beam evaporation [17], atomic-layer deposition [4,23], or pulsed laser deposition [16]. These methods, however, require a costly investment in equipment and specialized operator skills and expertise. By comparison, most studies in heterogeneous photocatalysis employ cost-effective wet-chemical methods for TiO₂ layer deposition, such as sol-gel-based spin-coating [25], dip-coating [26], spray-coating [27], and hydrothermal deposition [11,22]. These methods can also be used for NiO_x functionalization, which is achievable via hydrothermal methods [18,22] or wet-chemical impregnation [11]. While wet-chemical approaches are also applicable for forming functional TiO₂-based sensing devices [9,19], advanced vacuum-based deposition techniques are still required to deposit the electrically conductive patterns to form the electrodes, converting the sensing layer response into an electrically measurable readout [9,11,13,19,23,24].

An alternative wet-chemical approach for the metallization of dielectric surfaces, applicable to forming conductive layers on TiO₂-based microelectronic devices, is the electroless deposition (ELD) method. Widely used in the microelectronics industry for plating conductive metal or metallic alloy coatings, the most widely used ELD variant is the copper electroless deposition (Cu-ELD), which is considered an industrial standard for forming Cu interconnects and vias in printed circuit board (PCB) fabrication [28].

Cu-ELD plating is typically achieved through pre-activating the substrate with a noble metal catalyst (usually Pd-based [29]) and immersion in the Cu-ELD plating solution, which contains at least three components: a source of Cu²⁺ ions, a complexing and stabilizing agent (typically EDTA), and a weak reducing agent (formaldehyde—HCHO in classical formulations). When brought to a highly alkaline pH (>12), the reducing agent reacts with the Cu²⁺ ions on the ELD catalyst surface, reducing them to metallic copper [30], as shown in the reaction scheme in Equation (1).

$$\mathrm{C}{\mathrm{u}}^{2+}+2\mathrm{H}\mathrm{C}\mathrm{H}\mathrm{O}+4\mathrm{O}{\mathrm{H}}^{-}\to \mathrm{C}{\mathrm{u}}^0+2\mathrm{H}\mathrm{C}\mathrm{O}{\mathrm{O}}^{-}+2{\mathrm{H}}_2\mathrm{O}+{\mathrm{H}}_2$$

(1)

Two major drawbacks of Cu-ELD are as follows: (i) the need for a palladium-based catalyst, which adds to the bulk of the costs; and (ii) the inability for spatial patterning, as the entire surface is activated during the catalyst pre-activation step, achieved through immersion coating. Fortunately, both deficiencies can be effectively addressed by replacing Pd-based catalysts with more cost-effective Ag-based ones [31,32], which may also be patterned during the pre-activation step using techniques such as inkjet printing [33].

Combining the aforementioned wet-chemical deposition, functionalization, and metallization techniques, it is possible to achieve a fully wet-chemical technological procedure for the fabrication of functional TiO₂-based devices. An effective shortcut in fabricating patterned TiO₂-based sensing structures that are rarely explored in the literature is based on employing the intrinsic photocatalytic activity of TiO₂. Upon illumination with ultraviolet (UV) light of energy larger than its bandgap (E_g ≈ 3.2 eV, λ ≤ 387 nm) [1], TiO₂ generates electron-hole (e⁻/h⁺) charge pairs capable of driving redox reactions with surface-adsorbed species. This process can be used for the surface functionalization of titania through photodeposition of metal ions (Meⁿ⁺) to metallic products via photoreduction (Equation (2)) or metal oxides (MeO_x) via photooxidation and reaction with water molecules (Equation (3)) [34,35,36], depending on their redox potential. Typically, noble metal ions undergo the former, and transition metal ions undergo the latter route.

$$\mathrm{M}{\mathrm{e}}^{\mathrm{n}+}+\mathrm{n}{\mathrm{e}}^{-}\to \mathrm{M}{\mathrm{e}}^0$$

(2)

$$\mathrm{M}{\mathrm{e}}^{\mathrm{n}+}+\mathrm{x}{\mathrm{h}}^{+}+\mathrm{x}{\mathrm{H}}_2\mathrm{O}\to \mathrm{M}\mathrm{e}{\mathrm{O}}_{\mathrm{x}}+\mathrm{x}2{\mathrm{H}}^{+}$$

(3)

One of the main benefits of the photodeposition TiO₂ functionalization approach is that it relies on energy supplied by external UV illumination to drive the photooxidation and photoreduction reactions. Hence, it allows for stepwise selective patterning of a photocatalyst surface through photolithography, enabling additive decoration of the surface with more than one photodeposited chemical species, which is conceptually illustrated in Figure 1.

While photodeposition-based approaches have been demonstrated separately in the literature for both functionalizing TiO₂-based sensing devices [37] and forming conductive electrode patterns on TiO₂ surfaces [38], including in previous works by this author [39], there are no examples in the literature on combining these two approaches for the complete photodeposition wet-chemical fabrication of functional TiO₂-based devices.

Hence, this work aims to demonstrate for the first time the full wet-chemical photodeposition-based fabrication of a TiO₂-based sensing device, including the sensing layer and readout electrodes. An impedimetric RH sensor design was opted for due to the simplicity of fabrication and response readout. A base TiO₂ layer is formed via sol-gel dip-coating on soda-lime glass substrates and subsequently functionalized with NiO_x via photodeposition from an aqueous Ni²⁺ solution. A conductive pattern of the interdigitated electrode (IDE) is then formed onto the pristine and functionalized TiO₂ surface via selective surface activation with an Ag-based catalyst via Ag⁺ photoreduction and copper plated via Cu-ELD. The as-prepared devices are then tested as functional impedimetric humidity sensors, and the effects of Ni functionalization versus the pristine TiO₂ layer are discussed.

2. Materials and Methods

2.1. Reagents

All of the chemicals used in the study were of reagent grade or higher: titanium(IV) tetraisopropoxide (TTIP, 97+%) acetylacetone (AA, 99%), nickel(II) nitrate hexahydrate (Ni(NO₃)₂.6H₂O, 98%), potassium iodate (KIO₃, ≥99.4%), silver nitrate (AgNO₃, ≥99.9% metals basis), and ethylenediaminetetraacetic acid (EDTA, 99%) were supplied from Alfa Aesar (now Thermo Scientific, Waltham MA, USA). Isopropyl alcohol (IPA, HPLC grade ≥ 99.8%) was supplied by Macron Fine Chemicals (VWR, Radnor, PA, USA). Copper sulfate pentahydrate (CuSO₄·5H₂O, ≥99.0%) and formaldehyde (HCHO, 37% aq. sol.) were supplied by Valerus (Sofia, Bulgaria).

2.2. Preparation of TiO₂-Coated Glass Substrates

The TiO₂-coated glass substrates were prepared via sol-gel dip-coating, as depicted in Figure 2. The TiO₂-sol comprised titanium tetraisopropoxide (TTIP), acetylacetone (AA), and isopropanol (IPA) in a 1:3:15 molar ratio. TTIP was added dropwise to the AA/IPA solution, prepared beforehand, and magnetically stirred for 2 h at room temperature (RT) to obtain a clear yellow sol, aged for one week in darkness prior to use.

Standard microscope soda-lime glass slides (75 × 25 × 1 mm, Deltalab, Rubí Barcelona, Spain) were used as substrates. Prior to dip-coating, the glass slides were cleaned in three steps: (i) mechanical cleaning with a mild detergent, (ii) distilled water rinse followed by hot air drying, and (iii) ultrasonic cleaning for 3 min in acetone.

Dip-coating was performed using a homemade apparatus equipped with a speed-controlled stepper-motor-actuated linear slider. Each single dip-coating cycle was conducted at a slide withdrawal rate of 0.5 mm s⁻¹ (2.5 mm s⁻¹ immersion rate), followed by a heat-treatment procedure for 15 min at 350 °C and 60 min at 500 °C (2 °C min⁻¹ ramp rate). The substrates were weighed on a 0.1 mg precision analytical balance prior to and after each deposition cycle to monitor the resulting mass-loading change, which was normalized by the coated area of each slide (≈61 mm coated length S ≅ 32 cm² coated geometric area). A total of five subsequent dip-coating cycles were performed on a set of 12 substrates used in the following experiments.

2.3. Photodeposition Functionalization and Wet-Chemical Fabrication of Conductive Cu Tracks

The TiO₂-coated substrates were cut into 20 mm squares and used without further modification for NiO_x functionalization and deposition of conductive interdigitated electrode (IDE) arrays to form the humidity-sensing devices. The procedure is schematically represented in Figure 3 and described in detail in the following subsections.

2.3.1. NiO_x Functionalization

The TiO₂-coated glass substrates were photodeposition NiO_x-functionalized using a homemade UV exposure setup with an ultraviolet LED source (KTDS-3234UV365B, Kingbright Electronic Co., Ltd., Taipei, Taiwan) with emission λ = 365 ± 5 nm and 620 mW luminous flux. The UV source was positioned approximately 30 mm below a petri dish containing the TiO₂-coated substrate and the photodeposition electrolyte, thus illuminating it from beneath. The UV intensity at the substrate surface level was measured with a calibrated Thorlabs PM400 optical power meter equipped with an S175C thermopile head (Thorlabs, Newton, NJ, USA) and found to be 11–12 mW cm⁻². This was correlated to the signal of a photodiode-based UV intensity sensor (ML8511, Lapis Semiconductor, Kanagawa, Japan) positioned between the UV LED and the illuminated surface, used as feedback to determine the equivalent UV dose ($D^{UV}$, J cm⁻²) during exposure.

The NiO_x photodeposition electrolyte was aqueous 5 mM Ni(NO₃)₂, with an equimolar amount of KIO_3, added as an electron scavenger. The UV illumination duration was approximately 900 s for an equivalent $D^{UV}$ of 10 J cm⁻² and 1800 s for $D^{UV}$ of 20 J cm⁻². Finally, the NiO_x/TiO₂ layers were thoroughly rinsed with distilled water, dried for 15 min at 110 °C, and further heat-treated for 60 min at 350 °C to ensure the conversion of photodeposition products.

2.3.2. Selective Ag-Catalyst Patterning and Cu-ELD Conductive Pattern Formation

Electroless copper deposition (Cu-ELD) was utilized to create conductive IDE patterns and electrical connection pads.

The NiO_x/TiO₂ surfaces were activated with a silver (Ag) catalyst prior to Cu-ELD. Ag patterning was achieved through photolithography: a negative dry-film photoresist (40 µm) was applied to both sides of the TiO₂-coated substrate, exposed with a positive photomask at $D^{UV}$ of 0.25 J cm⁻², and developed for 90 s in 2 wt.% aqueous Na₂CO₃ at 30 °C. After a DI water rinse, the developed pattern was UV-exposed in aqueous 100 mM AgNO₃ at 2.5 J cm⁻². The Ag-activated substrate was rinsed in DI water, and the photoresist was stripped with acetone. The Ag-activated areas were apparent as darker zones, indicating the reduction of Ag⁺ on the TiO₂ surface.

The Cu-ELD plating was carried out at room temperature (RT) for 5 min in a bath consisting of 18 g L⁻¹ CuSO₄.5H₂O, 48 g L⁻¹ EDTA, 57 mg L⁻¹ K₄[Fe(CN)₆], 1 mL L⁻¹ conc. HCl, and 12.5 mL L⁻¹ 37% HCHO [38]. The bath pH was adjusted to 12.3 with 5 M NaOH.

Three different photomasks, shown in Figure 4, were employed: “short-circuit” and “open-circuit” control patterns (Figure 4a,b) and an IDE pattern of 18 interdigitated pairs of 200 µm electrode width, 400 µm finger pitch, and 7 mm finger length (Figure 4c).

2.4. Determination of Relative Humidity Impedimetric Response

The relative humidity (RH)–impedance (Z) curves were generated using the experimental setup depicted graphically in Figure 5.

A pair of peristaltic pumps were employed to supply RH-adjusted airflow. Each pump’s output was directed through separate gas-bubbler flasks: one containing distilled water, thermostated at 30 °C in a water bath to provide humid airflow, and the other containing blue silica gel desiccant to supply dry air. The humid and dry airflows were merged and passed through a glass cell housing a reference RH sensor (HIH-4000-003, Honeywell Inc., Charlotte, NC, USA). The pump rates were differentially controlled via pulse-width modulation (PWM) by an Arduino board. The PWM setpoints were calibrated to ensure a combined airflow of 50 mL min⁻¹ and an effective RH range of 15% ÷ 90% RH. All experiments were conducted at an ambient temperature of 20 ± 1 °C.

The RH-adjusted airflow was directed through a custom-made aluminum flow cell equipped with two window holders. One holder contained a transparent glass slide (fabricated from the same glass as the substrates) for sample observation and illumination with UV light (λ = 365 ± 5 nm, incident intensity at the sample’s surface 1 mW cm⁻², including losses through the top pane), while the other served as a sensor position holder. Electrical connections were established through a custom-made printed circuit board equipped with a pair of copper contact fingers that firmly pressed onto the sensors’ contact pads when the flow cell was assembled.

Impedance measurements were conducted using an RCL meter (PM6304, Fluke Philips, Everett, WA, USA) at test frequencies of 100 Hz, 1 kHz, 10 kHz, and 100 kHz. The signal level was set at 1 V for the IDE pattern, 50 mV for the “short-circuit” pattern, and 2 V for the “open circuit” control patterns.

2.5. Characterization Methods and Equipment

The volumetric density of the TiO₂ deposition sols was determined gravimetrically using a calibrated liquid pycnometer (10.0 mL vol.). An SNB-2T rotary viscometer (Dongguan Lonroy Equipment Co., Ltd., Dongguan, China), equipped with a stainless-steel 0# L rotor, was used for viscosity measurements. The sol’s surface tension was determined via stalagmometry (drop-counting method), where the number of drops resulting from releasing 2.0 mL of sol from a dedicated stalagmometer was counted versus the IPA solvent, and the surface tension was calculated as $\sigma =({\sigma }_{IPA}.\rho .n_{IPA})/({\rho }_{IPA}.n)$, where $n$ and $\rho$ are the number of drops and the gravimetric density of the sol, respectively, $n_{IPA}$ is the number of drops for the reference liquid (isopropanol), and ${\rho }_{IPA}$ and ${\sigma }_{IPA}$ are the density and surface tension of isopropanol (${\rho }_{IPA}$ = 0.785 g cm⁻³ and ${\sigma }_{IPA}$ = 20.89 mN m⁻¹ at 293K [40]).

Attenuated Total Reflection Fourier-Transform Infrared (ATR-FTIR) spectra of the sol were obtained using a Cary 630 spectrometer equipped with a Diamond-ATR accessory (Agilent Technologies Inc., Santa Clara, CA, USA). UV–visible transmittance spectra of TiO₂-coated slides were obtained using an SP-V1100 spectrometer (DLAB Scientific Co., Ltd., Beijing, China). Raman spectroscopy was performed with an Eddu TO-ERS-532 Raman system (Thunder Optics S.A.S., Montpellier, France) equipped with a 532 nm laser source and a 20× objective lens. Raman spectra were acquired at 50 mW laser power and 2500 ms integration time. X-ray Diffraction (XRD) patterns were obtained using a D8 Advance diffractometer (Bruker AXS GmbH, Karlsruhe, Germany) with a CuKα source and LynxEye PSD detector. Scanning Electron Microscopy (SEM) and Energy Dispersive X-ray (EDX) analysis were performed on a TESCAN Lyra I XMU microscope (TESCAN Group, Brno, Czech Republic) equipped with a Quantax 200 EDX detector (Bruker Nano GmbH, Berlin, Germany). Sheet resistance measurements on the Cu-ELD layers were conducted using a Veeco FFP-5000 four-point probe setup (Veeco Instruments Inc., Plainview, NY, USA).

2.6. Data Processing and Software

Data processing and visualization of all the results presented in this work were exclusively performed using custom scripts in R version 4.4.0 [41]. Image analysis, including the determination of geometric areas from calibrated photography images for mass–area loading assessments, was performed in Fiji (version 2.3.0/1.53q) [42]. Artwork and illustrations were created in Inkscape v.1.3.2 [43].

3. Results and Discussion

3.1. TiO₂ Sol-Gel Dip-Coating of Glass Substrates

3.1.1. FTIR Analysis of the TTIP:AA:IPA Sol

Figure 6 depicts a set of ATR-FTIR spectra obtained for the pure AA stabilizing agent, its IPA solution, and the final Ti:AA:IPA (1:3:15 molar ratio) sol, taken after aging.

The FTIR data confirms the purity of the AA complex-forming agent, exhibiting typical peaks at 1728 cm⁻¹ and 1708 cm⁻¹, associated with the asymmetric and symmetric ν(C=O) modes of the AA keto form, respectively. A peak at 1604 cm⁻¹ is associated with the ν(C=C-OH) vibration of the enolic form [44]. The peaks at 1416 cm⁻¹ and 1359 cm⁻¹ relate to the asymmetric and symmetric δ(CH₃) bands, respectively. The 1245 cm⁻¹ and 1177 cm⁻¹ peaks are associated with the δ(OH) and vinyl δ(C=C-H) modes of the enolic form of AA, respectively, followed by the 1156 cm⁻¹ band, associated with a ρ(CH₃) rocking mode. The peaks at 914 cm⁻¹ and 777 cm⁻¹ are assigned to the ν(C-CH₃) stretching and the olefinic γ(C=C-H) out-of-plane bending modes, respectively. In the AA:IPA solution (1:5 molar ratio), all of the AA bands are preserved, though diminished in intensity. The IPA band appearing includes the following: 3337 cm⁻¹ assigned to the ν(OH) vibration; a triplet at 2972, 2934, and 2886 cm⁻¹ related to the ν(CH) band; 1465 cm⁻¹ and ≈1370 cm⁻¹ asymmetric and symmetric δ(CH₃) bands; and a triplet at ≈1130 cm⁻¹, corresponding to various ν(C-C) and ρ(CH₃) vibrations.

After adding the TTIP and aging, the most drastic change in the TTIP:AA:IPA sol’s ATR-FTIR spectra is observed in the carbonyl area, where the 1728/1708 cm⁻¹ ν(C=O) AA keto form doublet completely disappears, and a pair of new bands form at 1584 and 1530 cm⁻¹, consistent with AA-chelated Ti(IV) [45]. The peak at 1604 cm⁻¹, related to the enolic ν(C=C-OH) AA band, is still discernible. New peaks appear in the low-frequency area: 1014, 990, and 850 cm⁻¹, associated with the isopropoxy group, and a peak at ~666 cm⁻¹, which could be associated with Ti-O-C bands [46]. These observations suggest that at 1:3 TTIP:AA molar ratio, some of the titanium complex is in the form of titanium diisopropoxide bis-acetylacetonate (Ti[(AA)₂(OiPr)₂]). Nevertheless, according to other authors, the stabilization of Ti-acetylacetonate sols is highest at TTIP:AA molar ratios larger than 1:2.86 [45]; however, molar ratios higher than 1:3 have no influence on the quality of the dip-coated thin films [27]. The sol used in this study remained stable without the appearance of turbidity or sedimentation for at least eight months after its preparation.

3.1.2. Physicochemical Properties of the TTIP:AA:IPA Sol and Resulting TiO₂ Films Mass Loading

The physicochemical properties of the TTIP:AA:IPA (1:3:15 molar ratio) sol were experimentally determined prior to deposition at 20 °C (293 K). The properties were as follows: density ($\rho$) = 0.853 ± 0.003 g cm⁻³, viscosity ($\eta$) = 2.602 ± 0.007 mPa·s, and surface tension ($\sigma$) = 24.54 ± 1.2 mN m⁻¹. These parameters allow for predicting the expected mass loading of the dip-coated film using an adapted version of the Landau–Levich equation [47], which can predict the dip-coated sol layer thickness. The thickness of the liquid layer was converted into the expected TiO₂ mass loading per geometric area, as shown in Equation (4):

$$\Delta m^{L-L}=\left(\frac{0.94{\left(\eta U\right)}^{\frac{2}{3}}}{{\sigma }^{\frac{1}{6}}\sqrt{\rho g}}\right)\mathsf{\rho }{w}_{\mathrm{T}\mathrm{T}\mathrm{I}\mathrm{P}}\frac{M_w^{Ti{O}_2}}{M_w^{TTIP}},$$

(4)

where $\eta$, $\sigma$, and $\rho$ are the sol’s dynamic viscosity, surface tension, and density, respectively, $U$ is the dip-coating withdrawal rate (0.5 mm s⁻¹), $w_{\mathrm{T}\mathrm{T}\mathrm{I}\mathrm{P}}$ is the mass fraction of TTIP in the sol, and $M_w^{Ti{O}_2}$ and $M_w^{TTIP}$ denote the molecular weights of TiO₂ and TTIP, respectively. The predicted $\Delta m^{L-L}$ TiO₂ mass loading was 10.4 μg cm⁻².

The experimental mass loading was determined for a set of 12 samples. The mass change reflects each of the five deposition cycles, including the thermal annealing step, and the resulting gravimetric mass change (vs. the mass of the bare substrate) was normalized by the coated geometric area to obtain $\Delta m$. The experimental data are presented in Figure 7, along with the theoretical $\Delta m^{L-L}$ according to Equation (4).

As shown in Figure 7, there is a constant linear growth in $\Delta m$ as a function of the coating cycles, indicating the stability of the previously deposited TiO₂ layers in the TTIP:AA:IPA sol. A linear fit of the experimental data revealed a deposition rate of 12.1 μg cm⁻² per coating cycle, which is close to the $\Delta m^{L-L}$ prediction of 10.4 μg cm⁻². The final mass–area loading was found to be 60.2 ± 6.5 μg cm⁻².

3.1.3. Phase Composition of the TiO₂ Thin Films

Figure 8a presents a set of Raman spectra obtained from the TiO₂ thin films after each of the five dip-coating deposition and heat-treatment cycles. The formation of the anatase TiO₂ phase is evident immediately after the first coating cycle, as indicated by the appearance of its characteristic bands: E_g (146 cm⁻¹, 640 cm⁻¹), B_1g (398 cm⁻¹), and A_1g (519 cm⁻¹) [48]. The intensities of all of the characteristic anatase TiO₂ Raman bands increased as a function of dip-coating cycles, concomitantly with the glass substrate mass loading increase, signifying the increase in thickness and improved crystallinity of the anatase TiO₂ layer being formed. No significant shifts in position or full width at half-maximum (FWHM) were observed in the most insensitive E_g band at 146 cm⁻¹, confirming a stable growth of the TiO₂ layer.

To gain further insight and confirm the phase composition of the anatase TiO₂ layers, XRD analysis was conducted, with a representative diffractogram of the final TiO₂-coated substrates shown in Figure 8b. It confirms the formation of the anatase TiO₂ phase, as suggested by the Raman analysis. The characteristic diffraction peaks associated with the (101) and (200) crystallographic planes are observed at 2θ of 25.3° and 48.1°, respectively (JCPDS card No. 21-1272). Scherrer analysis of the (101) peak revealed a mean crystallite size of 26 nm, which is consistent with the sizes reported for TiO₂ thin films deposited by the TTIP-acetylacetonate sols in other literature studies [27].

3.1.4. Optical and Optoelectronic Properties of the TiO₂ Thin Films

The UV–Vis transmittance of the TiO₂-coated slides, coated with five dip-coating cycles, is depicted in Figure 9a. The samples exhibit suitable optical transparency with 60%–80% transmittance within the visible region, modulated by interference fringes typical for thin optical coatings. Swanepoel’s method [49] was employed to estimate the thin films’ thickness and refractive index, utilizing the position of the interference fringes. Firstly, a set of polynomial functions was applied around the interference pattern, as shown in Figure 8a.

The refractive index of the TiO₂ film was obtained from the enveloping functions, according to Equation (5).

$$n\left(\mathsf{\lambda }\right)={\left[N\left(\mathsf{\lambda }\right)+\sqrt{{N\left(\mathsf{\lambda }\right)}^2-{s\left(\mathsf{\lambda }\right)}^2}\right]}^{1/2},$$

(5)

where $N\left(\mathsf{\lambda }\right)$ is expressed as follows:

$$N\left(\mathsf{\lambda }\right)=2s\frac{T_{max}\left(\mathsf{\lambda }\right)-T_{min}\left(\mathsf{\lambda }\right)}{T_{max}\left(\mathsf{\lambda }\right)T_{min}\left(\mathsf{\lambda }\right)}+\frac{{s\left(\mathsf{\lambda }\right)}^2+1}{2},$$

(6)

where $T_{max}\left(\mathsf{\lambda }\right)$ and $T_{min}\left(\mathsf{\lambda }\right)$ are the numerical values of the enveloping polynomials, while $s\left(\mathsf{\lambda }\right)$ is the refractive index of the substrate, which can be approximated from its optical transmittance ($T_{sub}$), viz.:

$$s\left(\lambda \right)=T_{sub}{\left(\mathsf{\lambda }\right)}^{-1}+\sqrt{{T_{sub}\left(\mathsf{\lambda }\right)}^{-2}-1}.$$

(7)

The thickness of the TiO₂ films can be estimated via Equation (8):

$$d=\frac{{\mathsf{\lambda }}_{\mathrm{m}}{\mathsf{\lambda }}_{\mathrm{m}+1}}{2({\mathsf{\lambda }}_{\mathrm{m}+1}n\left({\mathsf{\lambda }}_{\mathrm{m}}\right)-{\mathsf{\lambda }}_{\mathrm{m}}n\left({\mathsf{\lambda }}_{\mathrm{m}+1}\right))},$$

(8)

where ${\mathsf{\lambda }}_{\mathrm{m}}$ and ${\mathsf{\lambda }}_{\mathrm{m}+1}$ are the positions of any two subsequent interference maxima or minima in the UV–Vis transmittance spectra.

Applying Equations (5)–(8) to the interference pattern shown in Figure 9a, the thickness of the TiO₂ layer, deposited within five dip-coating cycles, was found to be 304 nm, with a refractive index (based on $n\left(\mathsf{\lambda }\right)$, where λ = 450, 639, 520, and 849 nm) of 2.25. The TiO₂ films’ packing density (${\rho }_{film}$) was also calculated according to Pulker’s equation [50]:

$${\rho }_{film}=\frac{(n_{film}^2-1)(n_{bulk}^2+2)}{(n_{film}^2+2)(n_{bulk}^2-1)}$$

(9)

where $n_{film}$ is the refractive index obtained from Swanepoel analysis of the transmittance spectrum and $n_{bulk}$ is the refractive index for bulk anatase, with $n_{bulk}$ = 2.51 being a reasonable value for the investigated λ region. Applying Equation (9) to these values, a ${\rho }_{film}$ of 0.9 is revealed, translating to solid film with a porosity of about 10%.

Finally, the optical bandgap ($E_g$) of the TiO₂ films was determined via Tauc analysis. The absorption coefficient $\alpha \left(\mathsf{\lambda }\right)$ was calculated for the UVA range ($\mathsf{\lambda }$ = 320 ÷ 380 nm), near TiO₂’ absorption edge, as $\alpha =d^{-1}\mathrm{l}\mathrm{n}\left(T^{-1}\right)$, where $d$ is the TiO₂ film thickness obtained from Equation (8). Figure 9b shows a Tauc plot of the functional dependence of $\sqrt{\alpha E}$ vs. $E$ ($E=1240{\mathsf{\lambda }}^{-1}$), based on the expectation that the as-formed anatase TiO₂ has an indirect optical bandgap. The optical bandgap was estimated by fitting the linear part of the Tauc plot and found to be 3.32 eV, a reasonable value for a sol-gel-deposited polycrystalline anatase TiO₂.

3.2. Photodeposition Functionalization, Ag-Catalyst Activation, and Resulting Cu-ELD Patterning of the TiO₂ Surface

NiO_x functionalization of the TiO₂ layers was conducted at two UV photodeposition doses ($D^{UV}$) of 10 and 20 J cm⁻², and then activated with Ag catalyst for the Cu-ELD deposition step. To investigate the effects of NiO_x photodeposition treatment and optimize the Ag activation $D^{UV}$, a set of samples from the pristine TiO₂ and NiO_x/TiO₂ layers were treated with Ag at $D^{UV}$ of 1.25 and 2.5 J cm⁻².

SEM imaging, coupled with EDX analysis, was employed to examine the morphology changes of the pristine TiO₂ layer after NiO_x and Ag photodeposition. Figure 10 presents a representative set of micrographs.

Starting with the pristine TiO₂ surface (Figure 10b), its morphology can be described as uniform, crack-free, and with slight graininess. The NiO_x treatment (Figure 10c,d) did not induce any discernable change in surface morphology, regardless of $D^{UV}$, although some loss of features could be speculated. In all cases, the Ag activation step led to the appearance of additional graininess, which is more pronounced on the pristine TiO₂ surface (Figure 10e) compared to the NiO_x/TiO₂ case (Figure 10f,g). The results for surface coverage with Ni and Ag, obtained from EDX analysis, are listed in

Table 1. A summary of the EDX analysis for the surface content of Ni and Ag, obtained for the TiO₂ samples, functionalized at varied photodeposition $D^{UV}$ doses.

Sample	${\mathit{D}}^{\mathit{U}\mathit{V}}$ Ni, J cm−2	${\mathit{D}}^{\mathit{U}\mathit{V}}$ Ag, J cm−2	Ni, at. %	Ag, at. %
Pristine TiO2	-	1.25	-	1.63
		2.5		1.97
NiOx/TiO2	10	1.25	1.49	0.39
		2.5	1.21	0.43
	20	1.25	1.23	0.57
		2.5	1.85	0.61

.

The NiO_x photodeposition resulted in a detectable Ni surface content in the order of 1–2 at. %; however, the differences between the NiO_x/TiO₂ surface functionalized at $D^{UV}$ of 10 and 20 J cm⁻² are not drastic. This outcome is not surprising, given that in this work, the functionalization is carried out in a thin liquid electrolyte layer with finite Ni²⁺ content. A similar observation can be made for the Ag content, where the $D^{UV}$-dependent variation for the same surface is small and within the margin of error for the EDX measurement. However, it drastically varies across the pristine TiO₂ and NiO_x/TiO₂ surfaces. In the former case, up to 1.97 at. % Ag is accumulated, while, on average, four times lower Ag content is observed in the NiO_x/TiO₂ case.

The effects of $D^{UV}$ on the quantity and quality of the resulting Cu-ELD conductive copper coating were investigated by subjecting Ag-activated 20 × 20 mm TiO₂- and NiO_x/TiO₂ substrates to 5 min Cu plating in the Cu-ELD bath described in Section 2.3.2. After plating, the Cu loading was determined gravimetrically and normalized by the geometric area of the coated surface to yield the copper mass–area loading ($\Delta m^{Cu}$), and sheet resistance measurements were taken to obtain the layers’ electrical resistance. The results are summarized in

Table 2. Cu-ELD mass–area loading and sheet resistance for TiO₂ samples, functionalized at varied photodeposition $D^{UV}$ doses.

Sample	${\mathit{D}}^{\mathit{U}\mathit{V}}$ Ni, J cm−2	${\mathit{D}}^{\mathit{U}\mathit{V}}$ Ag, J cm−2	$\Delta {\mathit{m}}^{\mathit{C}\mathit{u}}$, μg cm−2	RS, Ω sq.−1
Pristine TiO2	-	1.25	107 ± 30	0.926 ± 0.195
		2.5	131 ± 6	0.488 ± 0.065
NiOx/TiO2	10	1.25	37 ± 6	0.411 ± 0.008
		2.5	51 ± 13	0.409 ± 0.041
	20	1.25	87 ± 31	0.491 ± 0.036
		2.5	129 ± 35	0.419 ± 0.040

.

On average, for all samples, $\Delta m^{Cu}$ was in the range of 40–130 μg cm⁻², which, assuming a dense copper layer and a bulk density of Cu of 8.94 g cm⁻³, suggests that the thickness of the Cu-ELD layer should lay in the ~100 ± 50 nm range. The sheet resistance (R_S) measurements revealed R_S in the ~400 mΩ sq.⁻¹ range, suggesting a similar thickness range: assuming a bulk resistivity (${\rho }_{bulk}$) for the Cu layer of 1.7 μΩ·cm, the thickness can be estimated via the ${\rho }_{bulk}$/R_S ratio, yielding the expectation for a 40 nm thick layer on average for most samples. It must be noted, however, that typically, Cu-ELD layers have a higher ${\rho }_{bulk}$ compared to bulk copper [30].

The effects of $D^{UV}$ on $\Delta m^{Cu}$ and R_S qualitatively align with the amount of Ag detected in the EDX analysis (Table 1). Systematically, the NiO_x/TiO₂-functionalized surfaces yielded lower $\Delta m^{Cu}$ compared to the pristine TiO₂ surface, suggesting a slower deposition rate. In all cases, the $D^{UV}$ of the Ag-catalyst activation had a beneficial effect in lowering the R_S; hence, a $D^{UV}$ of 2.5 J cm⁻² was opted for Ag activation during the later preparation of the functional sensing devices.

Regarding the morphology and crystallinity of the Cu-ELD layer, Figure 11 presents SEM micrographs of the Cu-ELD-plated surfaces: pristine TiO₂ (Figure 11a) and the two NiO_x-functionalized TiO₂ surfaces, photodeposited at 10 J cm⁻² (Figure 11b) and 20 J cm⁻² (Figure 11c), after Ag activation at 2.5 J cm⁻².

As visible in the SEM micrographs, there is a drastic difference in the Cu layer morphology grown on the pristine TiO₂ surface compared to the NiO_x-functionalized TiO₂. The former case exhibits a rougher morphology with coarser features visible, in the order of 100–200 nm, while a smoother and more conformal layer is observed on the NiO_x/TiO₂ case.

Interestingly, the difference in surface morphology is not reflected in the Cu-ELD layer’s crystallinity. Figure 12 depicts XRD patterns obtained from the pristine TiO₂ and the two photodeposition modifications. The presence of metallic copper is confirmed by the appearance of the characteristic reflections for fcc Cu planes (111), (200), and (220) at 43.3°, 50.5°, and 74.1° 2θ, respectively (JCPDS No. 003-1018).

No effects of the NiO_x photodeposition were reflected on the Cu mean crystallite size, estimated through Scherrer analysis of the (111) reflection peak and found to be 14.8 nm for the TiO₂ surface and 15 nm and 15.1 nm for the NiO_x/TiO₂, functionalized at 10 J cm⁻² and 20 J cm⁻², respectively. Similarly, in our previous work on photocatalytic Cu-ELD activation of TiO₂-coated porous alumina surfaces, no effects from the photodeposition treatment on the crystallinity of the Cu layer were observed, even though a drastically different Cu-ELD morphology was obtained on this much rougher surface [39].

3.3. Humidity Response of Functional Sensing Devices, Fabricated via Photodeposition, and Cu-ELD on a TiO₂ Surface

As the main objective of this work is to demonstrate the wet-chemical fabrication of a functional TiO₂-based sensing device, sets of three devices were fabricated onto the pristine TiO₂- and NiO_x-functionalized TiO₂-coated substrates. In all cases, the devices for RH response testing were fabricated at two $D^{UV}$ for NiO_x photodeposition: 10 and 20 J cm⁻²; the Ag activation step was conducted at $D^{UV}$ of 2.5 J cm⁻², and the Cu-ELD was 5 min. Three Cu pattern geometries were fabricated: two control samples—“short-circuit” pattern to confirm that the conductivity of the Cu layer is RH-independent and “open-circuit” pattern to confirm that the substrate remains non-conductive at larger inter-electrode distances—along with one IDE-based pattern to check the RH response as a function of NiO_x functionalization (viz. Figure 4, Section 2.3.1).

The impedimetric RH response for pristine TiO₂ and the NiO_x/TiO₂ surfaces is summarized in

Table 3. Tabulated results for the impedance (Z) dependence on the relative humidity (RH) for the pristine TiO₂ and the two NiO_x-functionalized TiO₂ surfaces. The data obtained from the “short circuit” and “open circuit” geometries (see Section 2.3.1) are included for reference. The Z-RH functional dependences from the IDE-patterned surface are shown graphically in Figure 13.

Material →		Pristine TiO2			NiOx/TiO2 (10 J cm−2)			NiOx/TiO2 (20 J cm−2)
Geometry	RH	100 Hz	1 kHz	10 kHz	100 Hz	1 kHz	10 kHz	100 Hz	1 kHz	10 kHz
	15%	0.12 Ω	0.82 Ω	8.20 Ω	0.21 Ω	0.84 Ω	8.21 Ω	0.24 Ω	0.85 Ω	8.21 Ω
	90%
	15%	>200 MΩ	38.2 MΩ	3.68 MΩ	>200 MΩ	37.3 MΩ	3.66 MΩ	>200 MΩ	37.4 MΩ	3.64 MΩ
	90%		37.7 MΩ	3.66 MΩ
	15%	143 MΩ	14.1 MΩ	1.44 MΩ	118 MΩ	11.9 MΩ	1.21 MΩ	124 MΩ	12.5 MΩ	1.26 MΩ
	90%	4.55 MΩ	3.58 MΩ	1.18 MΩ	30.3 MΩ	9.55 MΩ	1.13 MΩ	44.1 MΩ	10.6 MΩ	1.19 MΩ
	15% + UV	144 MΩ	14.4 MΩ	1.45 MΩ	119 MΩ	12.0 MΩ	1.21 MΩ	124 MΩ	12.4 MΩ	1.26 MΩ
	90% + UV	3.66 MΩ	3.03 MΩ	1.15 MΩ	29.5 MΩ	9.52 MΩ	1.13 MΩ	44.8 MΩ	10.6 MΩ	1.19 MΩ
S (Z15%/Z90%)		31.4	3.9	1.2	3.9	1.2	1.1	2.8	1.2	1.1
S (Z15%/Z90%) + UV		39.3	4.8	1.3	4	1.3	1.1	2.8	1.2	1.1

. Starting with the two control samples, “short-circuit” and “open-circuit”, it can be noted that in all three materials, the Cu-ELD process yielded conductive Cu layers with an impedance of 0.12 Ω at 100 Hz in the case of the pristine TiO₂ sample. For the NiO_x/TiO₂ substrates, a small increase in Z (at 100 Hz) to 0.21 and 0.24 Ω was observed for the 10 and 20 J cm⁻² NiO_x-functionalized TiO₂ layers. No effect of RH on the impedance was observed, suggesting an intimate contact of the Cu grains in the conductive layer.

For the “open-circuit” case, all three materials showed similar behavior, with an impedance out of the 200 MΩ range of the RCL meter at 100 Hz, dropping to 37–38 MΩ at 1 kHz in all three cases, further decreasing by an order of magnitude to 3.6–3.7 MΩ at 10 kHz. No observable RH-Z dependence was noticed, except for a negligible decrease in impedance in the case of pristine TiO₂ at RH > 80%.

The IDE-patterned samples exhibited humidity-sensing behavior in all cases. Figure 13 depicts current–voltage (I–V) curves for the IDE-patterned pristine TiO₂- and NiO_x-functionalized devices in the ±2.5 V range.

At DC condition, all of the samples exhibit some conductivity, yielding a current in the order of 5–10 nA at a low RH (15%), which increases six-fold for the pristine TiO2 case and three-fold for the NiO_x/TiO₂ layers when the RH is increased to 90%. The frequency-dependent impedimetric RH–Z dependence curves are shown in Figure 14.

In all cases, a relatively low RH modulation of the impedance is observed at lower RH (≤40%), followed by a large drop at higher H₂O gas-flow contents. This is consistent with the accepted mechanism of operation for similar TiO₂-based RH sensors, where surface conductivity is formed by proton hopping across surface-bound -OH groups at low surface H₂O coverages, whereas condensed water layers form at further RH increase [21], leading to a concomitant drop in impedance, which may even be caused by surface available alkali ions (Na⁺, K⁺, etc.).

The highest impedance modulation was observed at 100 Hz and fell within the 10–100 MΩ range. The RH effect on Z can be converted to sensitivity (S), considering change in Z, within the studied relative humidity range, or S = Z^RH=15%/Z^RH=90%.

Pristine TiO₂ showed the highest sensitivity with S = 31.4 at 100 Hz; however, its RH–Z curve exhibited the largest non-linearity, expressing a logarithmic dependence, typically observed in similar humidity-sensing devices. NiO_x functionalization improved the RH–Z linearity in both the NiO_x/TiO₂-functionalized surfaces, effectively reducing the modulation response at higher RH levels, however, ultimately decreasing the sensitivity factor by an order of magnitude to 3.9 and 2.8, for the NiO_x/TiO₂ devices functionalized at $D^{UV}$ of 10 J cm⁻² and 20 J cm⁻², respectively.

This may be related to the possible occupation of TiO₂ surface sites by the Ni²⁺ photodeposition products, thereby decreasing its hydrophilicity. It has been demonstrated by Macovei et al. that the addition of Ni content to TiO₂ thin films both increases their water contact angle and limits the number of sites for H₂O molecule adsorption [51]. Nevertheless, both NiO_x-functionalized samples exhibited lower impedance at RH = 15% and better linearity. Increasing the test frequency to 1 kHz, 10 kHz, and 100 kHz led to a reduction in sensitivity in all three cases, as depicted in Figure 13, except for the pristine TiO₂-based device, which exhibited a noticeable RH response at RH > 50% in all cases.

Finally, given that the TiO₂-based RH sensors are still expected to be based on an active photocatalyst, another set of Z-RH measurements was conducted to determine the effects of UV illumination (λ = 365 nm, 1 mW cm⁻² incident UV intensity on the TiO₂ surface) on the response of the impedimetric humidity sensor. These are overlayed in gray for all the Z-RH datasets in Figure 14, and the impedance obtained at the two endpoints of the RH range of measurements is also included in Table 3. Briefly, the UV illumination exhibited an effect only on the pristine TiO₂ surface, where it was reflected by an increase in Z at low-to-medium RH (15%–50%), followed by a more pronounced drop in impedance at RH > 50%. On both NiO_x/TiO₂-based devices, the UV illumination had almost no measurable effect on the RH-Z dependence.

4. Conclusions

A completely wet-chemical procedure for the fabrication of a functional TiO₂ thin film-based impedimetric humidity sensor was successfully demonstrated. While the sensing performance of the device is based on a change in the MΩ range, the procedure is reproducible and is not affected by altering the surface chemistry of the base photocatalytic layer, which is the case of the NiO_x photodeposition; hence, there is room for future improvements, which can be based on improving the resolution of the photomasks and decreasing the IDE finger pitch in order to further bring down the impedance in the operable RH range and make the sensing device applicable for practical applications.

It was also successfully demonstrated that photodeposition MeOx functionalization can be employed to purposefully modulate the TiO₂ layer response, improving its linearity in the 15%–90% RH range and decreasing its impedance at lower RHs. While it must be stated that more work is needed to elucidate the UV photodeposition dose effects on this observation, this initial observation opens possibilities for the preparation of functional TiO₂ layers with tunable sensing responses. Ultimately, while the results of this work, as judged by the sensing performance of the as-fabricated TiO₂-based devices, are not directly applicable to practical RH sensors, it is clearly demonstrated that the possibilities to use photodeposition-based approaches to pattern a photocatalytic surface in a spatially selective manner may be applied for the purposeful design and fabrication of photo-electrochemical devices with an even higher degree of integration than the ones demonstrated herein.