Insight into the Oxygen-Sensing Mechanisms of TiO₂–CeO₂ Mixed Oxides Treated in a High-Energy Ball Mill: An XPS Analysis

Abstract

This study explored the oxygen-sensing mechanism of CeO₂ modified with TiO₂ via high-energy ball milling at different speeds. Different characterization techniques were employed to investigate the obtained materials. Quantitative surface analysis by X-ray photoelectron spectroscopy was conducted to elucidate their sensitivity mechanisms and assess the impact of the introduction of TiO₂. A comparable concentration of oxygen vacancies was found in the samples milled at 350 and 450 rpm. Electrical measurements conducted at temperatures lower than required for semiconductor gas sensors revealed the higher sensitivity of these two samples in comparison to pure CeO₂ at an oxygen concentration above 10%. In contrast, the samples derived from precursors milled at the highest speed exhibited the lowest sensitivity. This may be linked to a slight decrease in the vacancy concentration and the presence of a differentially charged carbon-containing phase. Eventually, the C 1s line provided significant insight into the surface characteristics of the materials. The uniform and non-uniform charging found for pure TiO₂ and CeO₂, respectively, along with the high charging of CeO₂, suggest that TiO₂ promotes the contact between the sensing layer and the overlayer. Sensor testing showed the significantly lower resistance of mixed oxides in comparison to CeO₂, which increases the utility of metal oxide-based sensors.

1. Introduction

The increase in CO₂ emissions dates back to the Industrial Revolution and continues to this day [1]. The main contribution comes from human activities, primarily the burning of fossil fuels for electricity, heat, and transport. An effective way to reduce the carbon footprint is to control the internal combustion conditions of engines within the power and heating sectors. To achieve this, a reliable and fast response by the sensor used for monitoring the oxygen concentration is necessary [2]. Accurate oxygen measurement is also of great importance in a variety of industrial environments to predict hazards in real time when oxygen levels drop significantly due to oxidation or combustion [3]. Among the many types of oxygen sensors developed over time [4], semiconductor-based oxygen sensors have emerged as the most suitable for a wide range of industrial applications. Their simplicity, cost-effectiveness, and low power consumption make them ideal for integration into distributed sensor networks intended for monitoring the combustion processes and oxygen concentrations [4,5]. Due to its exceptional oxygen storage capacity, cerium dioxide (CeO₂) is one of the most commonly used materials for oxygen-sensing applications. CeO₂ efficiently absorbs and releases oxygen in response to changes in its concentration due to the reversible and fast Ce⁴⁺–Ce³⁺ redox transition. CeO_2−x is a non-stoichiometric form of the oxide with a certain amount of Ce³⁺, accompanied by the creation of nearby oxygen vacancies that act as electron donors to chemisorbed oxygen [4,6,7,8,9]. CeO₂ exhibits the highest oxygen sensitivity at a temperature of approximately 600–700 °C [4]. However, its poor thermostability leads to sintering at elevated temperatures, which significantly reduces its oxygen storage [10] and long-term response stability [4,6]. At lower temperatures, electrical measurements of metal oxide-based gas sensors often encounter difficulties due to the inherently high resistance of oxides [11]. A widely studied method for enhancing the electronic, catalytic, and sensing properties of metal oxide-based sensors is the introduction of a second oxide [6,12,13]. Over the past two decades, research has identified TiO₂ as a particularly promising dopant for this purpose [14,15,16,17]. Various methods have been used for the preparation of TiO₂–CeO₂ materials, such as hydrothermal, co-precipitation, sol–gel, microemulsion, and incipient wetness impregnation [18]. In response to the growing need for sustainable methods with minimal environmental impact, this study employed high-energy ball milling. This type of mechanochemical synthesis offers significant advantages over conventional methods, including reduced energy consumption, elimination of harmful waste, and avoidance of toxic precursors and multistep procedures. Large quantities of the powder catalysts, alloys and ceramics can be obtained under ambient conditions in a short time span, making mechanochemistry a suitable tool for industrial applications [19,20,21,22,23,24,25,26,27,28,29]. During the milling process, energy is released through collision, shear, and friction between the balls and the reactants, and it is adsorbed by the solid reactants, stimulating chemical reactions. The intense mechanical stress and induced bond breakage lead to structural changes that increase the surface reactivity, decrease the particle and crystallite size, possibly increasing the specific surface area, and introduce dislocations and oxygen vacancies [21,26,27]. Among the numerous parameters that influence the milling process [24], the frequency (rotational speed) and time have been identified as critical [26].

Despite extensive research, the understanding of the sensing behavior of mixed oxide sensors remains insufficient. Since the initial step in the sensing process involves the atomic-level interactions between the oxygen and the top layer of the material [4,11], describing the surface characteristics is essential for elucidating the underlying sensing mechanisms. To this end, we conducted a comprehensive XPS analysis of TiO₂-modified CeO₂ materials (for simplicity, also referred to as TiO₂–CeO₂) that were high-energy ball-milled at three different rotational speeds: 350, 450 and 550 rpm. Thick sensing films were subsequently fabricated from the powder precursors. The resulting samples were labeled TiCe-1, TiCe-2, and TiCe-3, corresponding to milling speeds of 350, 450, and 550 rpm, respectively.

Our study focused on several key aspects: the efficiency of creating oxygen vacancies through high-energy ball milling, the effect of TiO₂ on these defects, and the characterization of the overall surface chemistry of the resulting materials, including their dependence on the milling speed. Calculations derived via XPS were used to model the outermost surface layer. The structural and morphological properties were analyzed using the particle size distribution (PSD), X-ray diffraction (XRD), 2D profilometry, scanning electron microscopy (SEM), and Raman spectroscopy. By correlating the results of the characterization methods with the sensor response, we aimed to understand how materials’ structural features govern sensing performance. Given the limited research on the sensing properties of mechanochemically treated, TiO₂-modified CeO₂, this study also contributes to expanding the field of application of this treatment in the sensor fabrication industry.

2. Results

2.1. Particle Size Distribution (PSD) Analysis of Powders

Considering the crucial role of the particle size in affecting the functionality of semiconductor sensors [4,12], it is important to examine the granularity of the obtained ball-milled powders. Figure 1 shows the influence of different milling speeds on the particle size distribution (PSD) of TiO₂–CeO₂. The following characteristic diameters are listed in

Table 1. Characteristic diameters (d₁₀, d₅₀, and d₉₀) and span values of the particle size distributions.

Sample	d10 [µm]	d50 [µm]	d90 [µm]	Span
TiO2	1.25	3.70	10.61	2.53
CeO2	3.80	22.58	55.81	2.30
TiCe-1	0.67	4.46	21.06	4.57
TiCe-2	0.75	2.64	8.53	2.94
TiCe-3	0.90	3.375	12.01	3.29

: d₁₀ (10% of particles have a diameter smaller than this value), d₅₀ (median value, sometimes taken as a measure of the average particle size), and d₉₀ (90% of particles are smaller than this value). The width of the distribution is quantified by calculating the span as (d₉₀ − d₁₀)/d₅₀ [30].

The fragmentation of agglomerates during the milling process is noticeable in the extended particle size distribution curve of the TiO₂–CeO₂ powders [31,32]. The bimodal and extensive distribution of the TiCe-1 powder indicates the lowest particle uniformity among the measured samples. The span of the particle size distributions (2.94) and the disappearance of the bimodal shape suggest the highest uniformity of TiCe-2 among the powders. With a further increase in the rotation speed, the second peak emerges again, and the distribution curve broadens. While the d₁₀ value shows a minimal variation with the milling speed, a significantly greater variation is observed in the larger particle size dimensions, particularly in the d₅₀ and, most notably, in the d₉₀ values. The smallest median particle size is observed for the powder TiCe-2 milled at the second speed, 450 rpm (2.604 µm). The cause of the non-linear change in the particle size is attributed to the impact coefficient and impact frequency of the milling balls, both of which are amplified with the speed [33]. This leads to higher surface energy, and eventually, to particle agglomeration as a mechanism to reduce the energy under intense forces. Notably, the d₁₀ increases linearly with the milling speed, indicating the coalescence of smaller particles and supporting this interpretation of the PSD measurements.

2.2. X-Ray Powder Diffraction (XRPD)

The diffractograms of the isolated and mixed oxide powders are presented in Figure 2. Characteristic reflections of the cubic form of CeO₂ were identified in all the samples. The presence of TiO₂ was not detected, possibly due to its low weight percentage compared to CeO₂ or its small crystallite size [29]. No evidence of a chemical reaction between TiO₂ and CeO₂ was found, due to the absence of reflections corresponding to new crystal structures in the TiO₂–CeO₂ diffractograms [34]. This observation indicates a physical mixture of the two oxide phases rather than a chemical interaction.

The data obtained from the analysis of the most intense CeO₂ (111) peak in the PANalytical X’pert software (version 4.9) are given in

Table 2. CeO₂ (111) peak position and interplanar, or d-spacing, in the pure and TiO₂-modified oxides.

Sample	Peak Position [°]	d-Spacing [Å]
CeO2	28.55	3.124
TiCe-1	28.54	3.125
TiCe-2	28.53	3.126
TiCe-3	28.61	3.117

. Apart from the effect on the particle and crystal size, the synergy of the impact and frictional forces created a certain level of lattice distortion, also known as microstrain [35,36]. The localized atomic displacements induced by the ball milling or chemical dissimilarities showed up as dislocations, vacancies, and stacking faults [37,38]. The change in the particle size is thus a rather complex process influenced by both the surface-to-volume ratio and the lattice distortion [35].

The average crystallite size and microstrain were estimated using the Williamson–Hall plot. The obtained graphs are shown in the Supplementary Materials and numerical data are given in [39]. The results are presented in Figure 3. The analysis of the CeO₂ revealed a crystallite size of 78.3 nm and a microstrain of 1.3∙10⁻⁴. The gradual weakening of the diffraction peaks indicated a reduction in the degree of crystallinity with an increasing rotation speed [40].

The influence of the mechanical action on the crystallite size was the most pronounced for the sample obtained at the highest milling speed (TiCe-3), whilst the largest lattice strain was found for the sample milled at the medium speed (TiCe-2), with the smallest average particles. The reduction in the lattice strain at the highest speed could be explained as follows: the effect of frictional forces and collisions during the milling led to the creation of so-called hotspots, characterized by a short-term sharp temperature surge in the ball–ball or ball–wall contact region [25,41]. As the milling speed increased, both the friction coefficient and the collision frequency amplified [41,42]. Sufficient localized heating facilitated the recovery and recrystallization processes, thereby minimizing the crystal lattice deformation and promoting grain growth [25,38,43].

The slight up-shift in the CeO₂ reflections (Table 2) observed in the TiCe-3 diffractogram indicated a modification of the d-spacing [44,45]. Namely, the shift in the XRD peaks of the mixed oxide system suggested a change in the lattice parameters due to the incorporation of dopants into the host lattice [44]. In the TiO₂–CeO₂ system, the ionic radius of Ti⁴⁺ (0.068 nm) was significantly smaller than that of Ce⁴⁺ (0.093 nm) or Ce³⁺ (0.103 nm), which facilitated the substitution of cerium ions by titanium ions [46]. This resulted in the formation of a fluorite-type crystal structure with the formula Ce_1−xTi_xO₂. Such solid solutions of TiO₂ and CeO₂ have already been well documented in the literature [46,47,48]. Furthermore, the incorporation of Ti⁴⁺ into the fluorite-like structure of other tetravalent oxides has also been documented [49]. This may account for the absence of TiO₂ peaks in the TiCe-3 diffractogram.

2.3. Two-Dimensional Profilometry of Sensing Layers

The roughness and thickness of the TiO₂–CeO₂ films was determined using 2D profilometric measurement. The surface roughness can affect the gas sensitivity of the film, as higher roughness may lead to an increased contact area with the gaseous species [50]. The thickness of the sensing material governs the sensor response by affecting the morphology, porosity, and finally, roughness. However, there is inconsistency among studies regarding the observed changes in the sensing properties with increasing thickness [51]. The average surface roughness is given in the form of the arithmetic average height (Ra), which is defined as the arithmetic mean of the absolute values of the vertical deviation from the mean line passing through the profile of a film. As a parameter that is more sensitive to larger deviation than the Ra, the root mean square roughness (Rq) is also used for describing the surface [52]. The values of the surface roughness parameter and thickness of the mixed oxides are shown in

Table 3. Surface roughness parameters and thicknesses of the mixed oxides obtained from the 2D profilometry data.

Sample	Ra [nm]	Rq [nm]	Thickness [μm]
TiCe-1	660	820	≈7.7
TiCe-2	543	752	≈4.4
TiCe-3	562	711	≈4.7

. Numerical data and calculation of their values are given in [39]. Considering that the effect of screen printing and the particle size on the roughness and thickness of the films is not the focus of this research, a discussion on these properties, as well as 2D profiles of a surface, can be found in the Supplementary Materials.

2.4. Scanning Electron Spectroscopy (SEM) and Energy Dispersive Spectroscopy (EDS) of Sensing Layer

SEM inspection of the CeO₂ film surface (Figure 4 revealed a randomly organized structure of larger flake-like particles and smaller ones that appeared spherical. Many agglomerates allocated beneath the smaller grains formed a discontinuous structure with visible cracks. The surface of the mixtures appeared to be slightly different from that of the pure CeO₂, featuring fewer agglomerates and better cross-linking of the particles (Figure 4). All the mixtures exhibited similar surface characteristics, displaying a lack of particle ordering. The morphology of both TiO₂–CeO₂ and CeO₂ arises from the wide range of particle size distributions present in the powders and the chosen method of preparation.

The EDS analysis was performed on a surface of 6.5 × 5 μm and the results are summarized in

Table 4. Results of the EDS elemental analysis of the TiCe-1, TiCe-2 and TiCe-3.

Sample	Weight (%)					Ti:Ce Weight Ratio Expected/Calculated
	Ti	Ce	O	C	Al
TiCe-1	11.00	61.93	20.79	5.99	0.29	0.25/0.18
TiCe-2	11.23	64.71	17.77	5.96	0.33	0.25/0.17
TiCe-3	10.44	59.83	21.28	8.12	0.33	0.25/0.17

(images and graphs can be found in the Supplementary Materials). The ratio of Ti and Ce on the surface of the mixtures did not correspond to the initial ratio of the oxides (1:4). We attribute this discrepancy to random experimental errors that may have been introduced during the multiple steps of the paste preparation, to the retention of powder on the walls of the milling vessel, but mainly to the non-homogeneous distribution of the two oxides across the film depth. Moreover, given the poor sensitivity of the EDS technique for the detection of elements with a low atomic number, it is likely that the actual presence of oxygen is underestimated, which may have affected the accuracy of the quantitative analysis.

The significant amount of carbon detected in all the samples probably belongs to the decomposition products of the organic matter formed during the heating in argon flow during the sensor fabrication. Namely, the heat treatment of organic compounds in an oxygen-free environment, a process known as pyrolysis, yields carbon-containing products when using CeO₂ as a catalytic material [53]. The highest percentage of carbon was found in the sample milled at the highest speed, TiCe-3. Lastly, the small amount of aluminum identified in all the mixtures suggested that the surface of the alumina substrate was not completely or uniformly covered by the sensing layer.

2.5. Raman Spectroscopy of Sensing Layers

Figure 5 shows the Raman spectra of the film of the starting oxides and mixtures. In agreement with the literature data, the starting TiO₂ shows distinctive Raman bands of anatase and rutile phases, while the CeO₂ exists in the fluorite cubic form [54,55]. The most intense TiO₂ band is assigned to the E_1g mode [56], while the single CeO₂ band corresponds to the F_2g mode [57]. No compounds between TiO₂ and CeO₂ are detected.

The feature at 642 cm⁻¹, present only in the TiCe-1 spectra, is assigned to the vibrations of a moderately distorted anatase lattice (E_3g mode) [56,58]. A reason for its absence in the spectra of the TiO₂ is masking by the intense rutile band A_1g (A_1gr in Figure 5). The origin of the band at 177 cm⁻¹ observed for all the mixed oxides (marked with an asterisk) remains unclear. According to the literature, the similar features at 181 cm⁻¹ [59] and 170 cm⁻¹ [60] are correlated with the TiO₂ II phase, which is irreversibly formed from anatase or rutile at high pressures. Earlier studies have documented the phase transformation from anatase to TiO₂ II through high-energy ball milling [60,61].

The relative intensities of the E_1g anatase to the A_1g rutile mode (Ia–Ir) in the mixtures differ from those in the spectrum of the pure TiO₂ (

Table 5. Intensity ratio of the anatase-to-rutile phase.

Spectrum	Ia–Ir
TiO2	1.12
TiCe-1	5.49
TiCe-2	1.46
TiCe-3	1.12

). A significant rise in the anatase–rutile intensity ratio in the spectrum of TiCe-1 implies that the presence of CeO₂ impedes the anatase-to-rutile transformation [10,62]. The Raman spectra of TiCe-2 and TiCe-3 reveal an exactly opposite effect: the weakening of the anatase band and a decrease in the Ia:Ir ratio, which is indicative of the promotion of the anatase-to-rutile transition. The phase transformation is presumably driven by additional impact energy and heat generated upon a rise in the rotation speed [63,64]. Namely, it has been found that the Ce atom effectively stabilizes the anatase crystals, but only up to a critical temperature point [65]. Another possible explanation is that ball milling induces the amorphization of the anatase phase. Notably, at the medium rotation speed, a significant loss of anatase is observed, yet TiO₂ II remains present. As the rotation speed increases, TiO₂ II also diminishes, ultimately resulting in a predominantly rutile phase.

The composition of the TiO₂ phase can be of great importance from the aspect of gas detection. Materials composed of CeO₂ and rutile–TiO₂ have been demonstrated to possess a higher content of Ce³⁺ compared to materials involving CeO₂ with other TiO₂ phases [66]. The broad and asymmetric nature of the bands, as observed for F_2g in mixed oxides’ spectra, is commonly attributed to the grain size reduction, the inhomogeneous strain, and the presence of defects [67,68,69]. The blue shift of the anatase band, E_1g, observed in the spectra of the mixtures reaches about 7 cm⁻¹ for the TiCe-3 sample. This may be attributed to the change in the materials’ crystal cell [44,70,71,72,73], although it also can be due to the change in the background slope.

2.6. X-Ray Photoelectron Spectroscopy (XPS) of Sensing Layers

The XPS measurements of the TiO₂, CeO₂, and mixed oxides were characterized by sample charging caused by their reduced conductivity, particularly at room temperatures. The most obvious consequence of the sample charging during the XPS measurements was the shift of the photoelectron and Auger peaks to higher binding energy. However, the effects of non-uniform charging, where different regions of the analyzed sample showed different electrical potentials, required a more detailed analysis. Non-uniform (or differential) charging contributes to the broadening of all the photoelectron lines and in some cases deforms the line shapes, making it impossible to follow even modest changes in the binding energy attributed to different chemical bonds. In the most severe cases of non-uniform charging, the same photoelectron line may appear at two positions in the spectrum. This phenomenon arises from sudden changes in the electrical potential between two surface regions, potentially representing separate chemical phases with different electron (or hole) transport properties. If the actual binding energy of a particular phase is known, the measured binding energy can be used to calculate the electric potential of a sample, or of a part of the sample containing that phase. If two photoelectron lines have distinctively different binding energy shifts, they obviously originate from different sample regions and consequently cannot belong to the same chemical phase. Similar ideas have been recently used to separate the chemical phases from samples with strong non-uniform charging. Since the exact shape of a photoelectron line carries information about the spatial distribution of a certain element, lines of the same chemical phase should have the same shape [74]. By applying these considerations to the XPS analysis within this study, it appeared that the seemingly least significant photoelectron lines in our samples, that is, the C 1s lines from adventitious carbon, carried potentially very important information about the overall surface structure of the samples. Therefore, it is in our interest to properly describe and quantify the overlayer(s) of the hydrocarbons that are inevitable contaminants in air-exposed samples.

2.6.1. TiO₂ Sample

Detailed spectra taken from the reference TiO₂ sample are shown in Figure 6a,c,d. The fitting of the Ti 2p line was performed with the spin–orbit pair of only one contribution, which corresponded to the TiO₂ phase (Figure 6a). Contributions attributed to Ti₂O₃ or TiO, which would appear at lower binding energies, were not observed. If the Ti 2p_3/2 line originated from TiO₂, its position should be at 458.6 eV [75]. Accordingly, the atoms of the TiO₂ phase should be at the potential of 1.6 V. The dominant peak of the C 1s line (Figure 6c), corresponding to adventitious carbon, was situated at about 1.8 eV relative to its expected position [75], implying rather uniform surface charging. Three contributions were observed in the spectrum of the O1s line (Figure 6d), which can be addressed (from right to left) as bulk metal oxide O1, always accompanied by defective/hydroxylated surface oxygen O2, and oxygen of organic compounds O3 [76]. We were applying the same peak model for the O 1s line for all the samples. As the sample was charged, all the peaks appeared to shift about 1.6 eV toward the higher binding energies (Figure 6d). C–O bonds could be also identified in the C 1s line (Figure 6c).

2.6.2. CeO₂ Sample

The overall shape of the Ce 3d photoelectron line (Figure 6b), taken from the CeO₂ sample, was successfully reconstructed using the fitting model proposed by Burroughs et al. [77]. The small discrepancies between the fitting results and the experiment can be attributed to the background calculation and carbon contamination that preferentially screens the Ce 3d signal on the high binding energy side. The line was fitted to eight peaks, four for each value of the angular momentum (j = 5/2, 3/2). The sets of “v” and “u” peaks belonged to the Ce 3d_5/2 and Ce 3d_3/2 photoelectron line, respectively. The peak v/u was the main contribution of the CeO₂ phase, v′/u′ corresponded to the Ce³⁺ oxidation state and the rest of the peaks represented satellites attributed to the CeO₂ phase. By comparing the measured position of the main Ce 3d_5/2 (the peak v) with that expected in the non-charged sample [78], the potential of CeO₂ was found to be about 7.2 V. The amount of Ce atoms in the +3 oxidation state evaluated from the sum of the v′ and u′ peak areas relative to the overall line intensity was determined to about 18.4%. If all the Ce³⁺ were placed at the surface, implying their sole origin from oxygen vacancies, and given that the introduction of one vacancy results in the formation of two Ce³⁺ ions [79], we may roughly estimate the upper limit of these defects to be about 9% relative to the total Ce surface concentration. The hydrocarbon layer with C–O bonds (Figure 6c) and the metallic oxide phase (Figure 6a,b) were on different potentials, which is an example of non-uniform charging, contrary to the TiO₂ sample. We presume the reason for this is the lack of ohmic contact between the hydrocarbon layer and the Ce–oxide, which would otherwise enable equal transport and distribution of electrons through the two distinct phases.

2.6.3. TiCe-1 Sample

Both the Ce 3d_5/2 and Ti 2p_3/2 photoelectron lines were shifted for about 6.1 eV relative to their reference positions (882.3 eV and 458.6 eV, respectively), implying uniform charging of the two oxide phases. A hydrocarbon layer on top of this sample was found to follow the potential of the metallic oxides. By comparing with the results obtained for the single oxide samples, it appeared that the TiO₂ grains established a galvanic contact between the CeO₂ and the hydrocarbon layer.

The expected position of the O1 line in the CeO₂ phase, calculated from its relative position with respect to the Ce 3d_5/2 line [76], was 529.5 eV, practically coinciding with that of the O1 peak of the TiO₂ phase (529.7 eV [75]). Therefore, the O 1s line was successfully fitted to three contributions, as in the case of the single oxide phase samples. The same peak fitting model was then applied for the samples TiCe-2 and TiCe-3.

2.6.4. TiCe-2 Sample

The metallic oxide phases of the TiCe-2 sample appear to be on a rather uniform electric potential of about 6.7 V during the XPS measurements (Figure 6a,b). Yet, the C 1s line shows up in the form of two distinct peaks separated for 5 eV, which indicates that TiCe-2 is non-uniformly charged (Figure 6c). The structure at the higher binding energy CH1 has a contribution of adventitious carbon and an additional one that we attribute to C–O bonds. Its position corresponds to that of the metallic oxides. The signal at the lower energies CH2 also has two peaks that we attribute to adventitious carbon from two differentially charged regions. This interpretation is supported by the fitting of the O 1s line, consisting of only three contributions being on the same potential as the oxide phases (Figure 6d). After correcting for charging (6.7 V), the O3 is found at 532.6 eV, further confirming that the CH1 structure lies at the same potential as the metallic oxides and that the region from which CH2 originates does not contain oxygen.

2.6.5. TiCe-3 Sample

Strong non-uniform charging is also observed while measuring the TiCe-3 sample, as seen from the detailed spectra of the C 1s line (Figure 6c). The potential of the structure CH1 is equal to the common potential of metal oxides, which is for this sample approximately 5.8 V. The peak of the highest intensity belongs to a differentially charged structure, CH2, with an electric potential of 0.8 V. The O 1s line is fitted to four peaks. Apart from the bulk and surface metallic oxide contributions O1 and O2, two contributions for organic oxygen have to be added: one for each of the differentially charged hydrocarbon regions CH1 (O3) and CH2 (O4). The relative position of the O1 peak with respect to the metallic photoelectron lines for TiCe-3 is O1-Ti 2p_3/2 = 71.1 eV and Ce 3d_5/2-O1 = 352.9 eV, being identical, in the frame of the experimental uncertainty, with the other samples. After correcting the binding energies for the potential of the structures CH1 and CH2, the peaks O3 and O4 are both situated at 532.5 eV. Their position practically coincides with the binding energies of organic oxygen in the previous spectra. From the relative positions of the metal and O 1s photoelectron lines, it appears that the pure and mixed oxides have the same chemical bonds: Ti is present as TiO₂, while Ce is present dominantly as CeO₂, with a certain amount of Ce³⁺.

2.6.6. XPS Quantitative Analysis

Employing the same approach as for CeO₂, the Ce³⁺ percentage in TiCe-1, TiCe-2 and TiCe-3 is calculated to be 19.1%, 17.8%, and 14.9%, respectively. Considering the various sources of uncertainty, we can say that the amount of Ce³⁺ in the mixtures and pure CeO₂ (18.4%) is constant except in the last sample, where a certain decrease in the amount of Ce³⁺ ions is evident. The binding energies of all the contributions in the Ti 2p, Ce 3d, C 1s and O 1s lines are listed in the

Table 6. Binding energies [eV] of the Ti 2p, Ce 3d, C 1s, and O 1s core-level peaks. Only one component of the spin–orbit splitting for Ti 2p and Ce 3d is shown.

	Ti 2p	Ce 3d				C 1s		O 1s
	3/2	v	v′	v″	v‴	C-C(H)	C-O	O1	O2	O3	O4
TiO2	460.1	/	/	/	/	286.6	288.1	531.0	532.3	533.9	/
CeO2	/	889.5	891.9	896.3	905.5	285.5	287.0	536.7	539.2	532.8	/
TiCe-1	464.7	888.6	890.6	895.4	904.6	290.8	291.8	535.8	537.1	538.6	/
TiCe-2	465.3	889.2	891.2	896.0	905.2	291.5 (CH1) 286.4 (CH2 I) 285.0 (CH2 II)	292.4 (CH1)	536.5	537.8	539.3	/
TiCe-3	464.4	888.4	890.6	895.2	907.0	290.7 (CH1) 285.5 (CH2)	292.8 (CH1) 287.5 (CH2)	535.4	536.7	538.3	533.0

. In contrast to the reference TiO₂, CeO₂, and TiCe-1, the spectra of TiCe-2 and TiCe-3 reveal two distinct hydrocarbon contributions (Figure 6c). This observation suggests a lack of electrical conductivity between the CH1 and CH2 components, with only one of these acting as the top layer of the film.

Through bond identification, we have determined that the covering layer of TiCe-1 shares the same potential as the bulk of the oxides. Consequently, we propose that the CH1 structure in the C1s spectra of TiCe-2 and TiCe-3 originates from the hydrocarbon layer atop both the TiO₂ and CeO₂ in the respective samples. This can be confirmed by calculating its thickness. The fraction of the total carbon layer covering TiO₂ is denoted as θ_Ti, while the part covering the CeO₂ patches is θ_Ce. The details of the quantitative analysis, the relevant literature sources [23,80,81,82,83,84,85,86], as well as the modeling of the TiO₂-modified CeO₂ film, are given in the Supplementary Materials. The numerical data and calculation within the quantitative analysis can be found in [39]. The results derived from the analysis based on the hypothesized model are listed in

Table 7. The surface coverage values (θ_Ti and θ_Ce), hydrocarbon layer thickness (d), and concentration ratio (O:Ti–O:Ce).

Sample	θTi:θCe [%]	d [nm]	O:Ti–O:Ce
TiO2		1.15	O:Ti = 1.98
CeO2		2.73	O:Ce = 2.52
TiCe-1	43:57	1.51	O:Ce = 2.65
TiCe-2	31.7:68.3	2.42	O:Ce = 2.72
TiCe-3	34.5:65.5	2.15	O:Ce = 2.70

. According to the data, the thickness of the carbon-containing overlayer above the mixed oxides is significantly smaller than that on the pure CeO₂. The thinnest overlayer is found for TiCe-1 sample, which contains the highest surface coverage of TiO₂ patches (θ_Ti) among the mixed oxides.

2.7. Results of Electrical Measurements

The measurements of the TiO₂–CeO₂ sensor (hereinafter referred to as the TCO sensor) presented in this study were performed at 310 °C. This operating temperature was selected because it represents the minimum temperature at which the resistance of the sensor over the entire O₂ concentration range is low enough to be measured by most available instruments, while the stability of the response is satisfactory. To investigate the sensing behavior of the obtained TiO₂–CeO₂ materials, we measured the current–voltage curves of the TCO sensors at the chosen working temperature, using the Keithley 2450 Source Meter Unit (SMU) (Solon, OH, USA). It was found that all the mixtures exhibited ohmic contact characteristics, as can be seen in the example given in the Supplementary Materials.

The resistance values were recorded for 15 min at each O₂ concentration. As an example, the result of the time-dependent electrical measurements for TiCe-3 is given in the Supplementary Materials. The resistance value for each O₂ percentage as well as the standard deviation were calculated from the final 200 s of exposure. The dependence of the DCR values of the sensor on the oxygen percentage is presented in Figure 7. The data plotted with a linear scale are shown in the Supplementary Materials. To evaluate the reproducibility of the sensing layer fabrication process, two mixtures were screen-printed twice and their response to O₂ was measured. Plots of the same color in Figure 7 belong to TCO sensors manufactured from identical mixtures.

Variations in parameters such as the squeegee pressure and amount of paste during the manual application of each layer may affect the repeatability of a process, which can be reflected in the performance of the final sensing material. The responses of two TiCe-1 sensors based on identical mixtures showed slight differences, although the resistance values at the same O₂ exposure tended to cluster (Figure 7). This discrepancy may also arise from random errors in the paste preparation and the wide particle size distribution of the milled powders. Indeed, the sensor responses of the two TiCe-2 samples, made from powder with a narrower span in the particle size distribution (Table 1), appeared to match. However, the small inconsistency in the sensor response did not affect the application of high-energy ball milling for fabrication of the sensing layer: each sensor will be calibrated before use.

The highest change in the sensors’ resistance was in the range of low oxygen concentrations (0–10%), approaching saturation thereafter. The high resistance of the CeO₂ film, reaching several hundreds of MΩ, posed a challenge for practical applications due to the difficulties in measuring the resistances of this order. Even at the low O₂ percentages, its resistance exceeded 100 MΩ. In contrast, the sensor based on the TiO₂–CeO₂ films showed significantly lower resistance over the entire O₂ concentration range, whilst preserving a similar relative change in the resistance. The resistance of the mixed oxide films decreased with increasing milling. It was also observed that TiCe-2 and TiCe-3 showed a similarity in the DCR value at identical O₂ concentrations, while the mixture milled at the lowest speed (TiCe-1) exhibited greater resistance by about an order of magnitude. In this study, the sensitivity was calculated as the slope of the DCR-O₂ dependence curve divided by the resistance value at 0% O₂, labeled as DCR_Ar [87]:

S = ∆DCR⁄∆O₂∙1⁄DCR_Ar

(1)

The sensitivity, presented in Figure 8, is a function of the oxygen partial pressure due to the non-linear response of the sensors (Figure 7). The sensitivity strongly depended on the O₂ concentration, with a maximum observed in the 0–10% region, followed by a sharp decrease to a saturated value. At lower percentages of O₂ (0–10%), the best oxygen sensitivity was observed for the pure CeO₂, which was about 0.06 above that of TiCe-1. At higher partial pressures of O₂, TiCe-1 and TiCe-2 were more sensitive than CeO₂, while the TiCe-3 film (obtained by the highest rotation speed) had the lowest sensitivity throughout the whole concentration range.

3. Discussion

At the grain surface, oxygen molecules extract electrons from the conduction band, trapping them at the surface in the form of ions, which induces upward band bending. This causes the formation of an electron-depleted region (space charge layer within the grain), which represents a potential barrier that hinders the movement of electrons between the two grains (Figure 9). Consequently, the electron mobility, as well as the number of electrons free to act as a carrier, is reduced, resulting in a decrease in the electrical conductivity of the n-type semiconductor as the oxygen partial pressure increases, following the Equation (2):

σ = A∙exp∙(−E_A∙(k∙T))∙p_O^1/N

(2)

Here, σ represents the conductivity, A is a constant, E_A is the activation energy for conductivity, k is the Boltzmann constant, T stands for the operating temperature, and p_O is the oxygen partial pressure. N is a value that depends on the type and number of the defects, and for the case of n-type semiconductors, it is negative [4]. The increase in the DC resistance with the rising O₂ concentration suggests that TiO₂-modified CeO₂ acts like a single n-type oxide over the entire partial pressure range. The trend observed in Figure 8 arises from the increasing probability of vacancy filling during the exposure to higher O₂ concentrations. At sufficiently elevated O₂ levels, saturation occurs: vacancies become fully occupied, rendering further oxygen exposure inert. Consequently, the resistance stabilizes as it reaches its maximum value [4,11].

Figure 9. Schematic illustration of the homo-contact between the grains of the same semiconductor (left) and the hetero-contact between two types of semiconductors (right) in contact after reaching equilibrium; “SCL” stands for the width of the space charge layer, “BB” for the banding of the valence and conduction band, “CP” for the contact (hetero) potential and “∆BD” for the band edge difference. Two different colors (blue and gray) represent different types of semiconductors.

Figure 9. Schematic illustration of the homo-contact between the grains of the same semiconductor (left) and the hetero-contact between two types of semiconductors (right) in contact after reaching equilibrium; “SCL” stands for the width of the space charge layer, “BB” for the banding of the valence and conduction band, “CP” for the contact (hetero) potential and “∆BD” for the band edge difference. Two different colors (blue and gray) represent different types of semiconductors.

Increasing the number of oxygen vacancies is expected to enhance the sensitivity of the detection process [4,6]. However, the X-ray photoelectron spectroscopy (XPS) analysis reveals that the relative amount of Ce³⁺ remains consistent in both the CeO₂ and the mixtures (approximately 18%), although there is a slight reduction noted for the TiCe-3 (14.9%). These findings suggest that increasing the milling speed within the tested range does not significantly affect the formation of oxygen vacancies until a certain threshold (TiCe-2), beyond which it appears to act as an inhibitor. The presence of TiO₂ also does not affect the formation of vacancies in TiO₂–CeO₂ mixtures. According to the Ce 3d and Ti 2p detailed spectra, oxygen vacancies are generated exclusively in the CeO₂ domains. The XPS did not detect Ti³⁺ ions, indicative of the presence of oxygen vacancies at room temperature. The difference in the amount of defects will persist under elevated temperatures and oxygen-deficient conditions, due to the substantial differences in stability between the two oxides.

The effects of the so-called “poisoning” of a surface by carbon-containing species can lead to the occupation of the active sites and the reduction in the concentration of oxygen species adsorbed [88]. Two different hydrocarbon phases (CH1 and CH2) are identified on the surfaces of TiCe-2 and TiCe-3. In the case of TiCe-3, both are partially oxidized, while for the TiCe-2, they appear to be just differentially charged adventitious carbons, with no evidence of oxidation. The two carbon-containing phases are galvanically isolated, indicating the absence of physical contact between them. Otherwise, we would expect to observe some deformation of the C 1s line, rather than a complete separation into two distinct regions. Therefore, only one of these phases can be considered an overlayer. We propose that it is the one with the same potential as the oxide phases, CH1, found in all three mixtures. All the evidence suggests that CH2 is present on the samples’ surface but is physically displaced by metal oxides. It is possible that this phase resides at the boundaries between two identical or different grains. Grain boundaries, whether a homo- or hetero-interface, represent a type of microstructural defect characterized by an electronic property that differs from that of the surrounding surface [17]. This distinctive electronic structure may cause a different charging of the CH2 phase, which could partially or completely coat the grain surface, somehow alternating the electron transport or oxygen diffusion. Since it is not observed for the TiCe-1 mixture, such carbon-containing species could have been introduced during the milling at a rotational speed above 350 rpm. Nevertheless, the common potential of the metallic lines suggests that CH2 does not inhibit the contact between TiO₂ and CeO₂.

The presence of differentially charged oxidized species is not detected in the cases of TiCe-1 and TiCe-2. Both samples demonstrate higher sensitivity than TiCe-3 across the entire range of oxygen concentrations and CeO₂ for the concentrations above 10%. TiCe-1 is also characterized by the absence of a differentially charged CH2 phase, the smallest thickness of the carbon overlayer, and the highest surface roughness. Increasing the powder precursors’ milling speed up to 550 rpm reduces the sensitivity of the layer, potentially due to the introduction of oxygen-containing CH2 species, causing a small decrease in the surface roughness, but most likely due to the lowest amount of oxygen vacancies. It is also worth noting that the EDS results also show that the TiCe-3 sample contains the highest amount of carbon. This observation further supports the hypothesis that the CH2 phase is not a surface contaminant but rather is located in the sub-surface microstructure of the sample. However, the role of the CH2 phase in reducing the sensitivity of TiCe-3 remains inconclusive without XPS experiments conducted at elevated temperatures. This phase, due to the oxygen present in its structure, may potentially inhibit the binding of ambient oxygen that penetrates through defects (gaps) in the overlayer and reaches the grain boundaries. The content of the TiO₂ crystalline phases does not affect the sensitivity in the expected manner, but its incorporation in the CeO₂ layer reduces the amount of surface contamination, i.e., the thickness of the overlayer, d (Table 7). The average surface roughness, Ra, of the mixtures (Table 3) appears to decrease with the increasing thickness of the carbon-containing overlayer CH1, d (Table 7).

The homo-contact established at the boundary between two grains is a decisive factor in the free carrier transport in semiconductor devices. The mechanism of electron transfer from one domain to another in the presence of a potential barrier caused by the existence of a depletion region is described with the double Schottky barrier [6,17,89]. The charge carrier mobility through the barrier occurs at elevated temperatures and governs the conversion of the reaction into an electrical signal [90]. By assembling semiconductors with different physical properties or working functions into one structure, apart from the homo-contacts, hetero-contacts and band edge differences are created. The hetero-contact represents a thin double layer of physically separated positive and negative charges at the interface of the grains of dissimilar semiconductors. The formed potential reduces the mobility of electrons moving in the direction against it, without affecting the electrons coming from other directions toward the positive side of the interface. Yet, their role in shaping the sensing functions is significant only if the structure of the sensing layer is properly designed. In such cases, the response of the sensor to the target gas is doubly affected: by oxygen-binding-induced changes in the surface electron density and by the reduced mobility of electrons passing through the contact, leading to a remarkably higher sensitivity [89,91]. The surface structure of the TiO₂–CeO₂ mixtures consists of irregularly distributed grains of varying sizes forming agglomerates, as observed in the SEM images. N. Yamazoea and K. Shimanoea [89] proposed that mixing the grains of two semiconductor materials almost halves the resistance of the resulting structure, regardless of the presence of the target gas and the combination of materials. This chaotic arrangement allows current to flow through paths of lower resistance, primarily homo-contacts, while hetero-contacts play a minor role in electron transport [88,91]. Indeed, the main difference in the response of the TCO compared to the pure CeO₂ sensor is precisely the markedly lower electrical resistance in the entire range of O₂ concentrations. The latter makes TCO sensors much more suitable, while the sensitivity is slightly lower only for the lowest oxygen concentrations. While the contact potential modulates the drift mobility of the electrons, the band edge difference plays an important role in establishing an exchange current under the unbiased condition [91].

As already said, the common potentials of the Ti 2p and Ce 3d lines suggest the establishment of well-defined contact between TiO₂ and CeO₂, facilitating electronic transport between these two phases. The formation of the well-defined contact is further supported by the Raman spectroscopy, which indicates a noticeable effect of CeO₂ on the phase composition of TiO₂. Given that neither ball-milled nor thermally induced chemical reactions are detected, it is likely that the interactions occurring at the interface between the two oxides are predominantly physical in nature. However, it remains uncertain whether this contact resembles the one depicted in Figure 9 The presence of a rectifying contact within a domain significantly smaller than the overall TiO₂-modified CeO₂ layer cannot be directly observed, as the I–V curves reflect the behavior of the entire sensing material. Even if rectifying properties form at the oxide–oxide interface, they do not significantly affect the sensitivity of the TiO₂–CeO₂ system, that is, not to an observable extent.

4. Materials and Methods

4.1. High-Energy Ball Milling

Three different TiO₂-modified CeO₂ powders were obtained in a high-energy ball mill (Planetari Micro Mill Pulverisette 7, Fritsch, Germany) at three different rotational speeds. TiO₂ (purity > 99%, Alfa Aesar, Kandel, Germany) and CeO₂ (99.9% purity, Johnson Matthey-Alfa, Karlsruhe, Germany) powders served as the starting materials. Milling was carried out under atmospheric conditions in an 80 cm³ silicon nitride vial with 25 silicon nitride balls (10 mm diameter), maintaining a ball to powder mass ratio of 10:1 and a total powder weight of 4.87 g. A CeO₂:TiO₂ mass ratio of 0.8:0.2 was chosen for this study. Milling lasted 160 min, with a 5 min break every 20 min, at speeds of 350, 450 and 550 rpm for each powder. Thick sensing films were made from these powders as described in the following section. The resulting samples were designated as TiCe-1, TiCe-2, and TiCe-3 based on their respective milling speeds: 350, 450 and 550 rpm.

4.2. Paste Preparation and Sensor Fabrication

CeO₂ and mixed oxide pastes were prepared according to the recipe for TiO₂ paste production for dye-sensitized solar cells established by Ito and colleagues [92]. Prior to applying each layer, the pastes were homogenized by stirring on a magnetic stirrer (MR 3001 K, Heidolph, Wood Dale, IL, USA). Six layers were screen-printed on an alumina substrate using custom-made screen equipment. Commercial substrates with interdigitated gold electrodes and a platinum heater on the back side were used (DropSense with interdigitated Gold Electrode/200 μm lines and gaps, IDEAU200-HPT-WB, Metrohm AG, Herisau, Switzerland). Following the printing of each layer, the substrate was dried on a hotplate (PZ28-2, Prazitherm, Harri Gestigkeit GmbH, Dusseldorf, Germany) at 125 °C for 6 min. The samples were then sintered in an oven under an argon flow gradually rising to 800 °C. Prior to this, the samples were briefly exposed to a flow of synthetic air at 400 °C to remove organic components from the paste. By triggering the growth of the so-called neck between the particles, the sintering improved the electromechanical bonding of the particles [93]. The thick film of CeO₂ and the mixtures showed good attachment to the alumina substrate. The resulting structures were mounted between two pairs of copper connector wires, which were custom-made in our facilities.

4.3. Characterization of Powders

To determine the effect of the milling speed on the average particle size of the TiO₂–CeO₂ powders, particle size distribution (PSD) analysis was employed. The used instrument was the laser-diffraction-based Mastersizer 2000 (Malvern Panalytical, Malvern, UK) analyzer, covering the particle size range of 0.02–2000 µm. To perform the PSD measurements, the powders were dispersed in ethanol and subjected to ultrasonic treatment at a frequency of 40 kHz and power of 50 W. The crystal structure of the powders after undergoing mechanochemical treatment was investigated by the X-ray powder diffraction (XRPD) technique. Diffraction data were acquired in Bragg–Brentano geometry over the range of the scattering angle from 15° to 99°, with a step size of 2θ = 0.05° and a scan step time of 55.245 s. Measurements were performed using a PANalytical X’Pert PRO MPD diffractometer (Almelo, The Netherlands) equipped with an X’Celerator Ultrafast 1D detector based on Real Time Multiple Strip (RTMS) technology. The CuK_α source with a wavelength of 1.542 Å was used. The peak analysis was performed with PANalytical X’pert (version 4.9) software to identify the phases and determine their position, full width at half maximum (FWHM or β), and d-spacing. The Williamson–Hall plot was constructed based on the obtained results to distinguish the contributions of the crystallite size and microstrain of the milled powders.

4.4. Characterization of Sensing Material

The roughness and thickness of the screen-printed films were determined using a Rank Taylor Hobson Ltd. profilometer (Leicester, UK), with a step setting of 0.04 mm for the thickness and 0.02 mm for the roughness measurements. The roughness and thickness were calculated as the average values of data collected from the three, i.e., five regions on the sample, and shown in Table 3. The surface profile was obtained by applying the roughness filter (0.33 Hz) of the instrument. The results of the 2D profile measurements of the films are given in the Supplementary Materials. Scanning electron microscopy (SEM) and energy dispersive spectroscopy (EDS) (Zeiss Auriga with Oxford XMax 150 EDS, Jena, Germany) were employed to investigate the morphology of the CeO₂ and TiO₂–CeO₂ films on the substrate. The Raman spectra were acquired using a DXR Raman microscope (Thermo Fisher Scientific, Medison, WI, USA), equipped with a research optical microscope and a CCD detector. A frequency-stabilized single-mode diode laser with an excitation wavelength of 532 nm was employed. The laser power was adjusted to 2 mW and the exposure time to 5 s. The XPS measurements were performed on a XPS system with a Class 100 energy analyzer (VSW, Bushey, UK), which was part of an experimental setup Multitécnica assembled for the surface investigation [94]. The survey spectrum was recorded in a fixed transmission mode with a pass energy of 44 eV (FAT 44), an energy step of 0.5 eV and a dwell time of 0.5 s. Due to the overall low signal intensity and charging problems introducing line broadening, detailed spectra of the Ce 3d core-level were also recorded in FAT 44 mode, with an energy step of 0.2 eV and dwell times of 8 s and 12 s per scan. A non-monochromatic Mg Kα line with photon energy of 1253.6 eV was used as an excitation source. Calibration of the energy axis was performed using a single Ag (110) crystal and polycrystalline Au previously cleaned by Ar ion sputtering. The energy was calibrated to the peak positions of the Ag 3d_5/2 and Au 4f_7/2 core-level spectra, with binding energies of 368.22 eV and 83.96 eV, respectively. The samples were in a high vacuum for 24 h prior to the measurements.

4.5. Sensor Response Measurements

Electrical measurements of the obtained sensing structures (Figure 10) were conducted to explore the sensing behavior of the TiO₂-modifed CeO₂. A scheme of the setup for measuring the sensor response in the gas flow regime developed for testing the TCO sensors is given in [29]. The sensors were exposed to varying volumetric concentrations of O₂, ranging from 0 to 100%. To achieve different O₂ concentrations, the flow rates of O₂ (5 N purity) and Ar (5 N purity) were adjusted using two mass flow controllers (MC-200SCCM-D/5M, Alicat Scientific, Tucson, AZ, USA).

The total gas mixture flow remained at 200 SCCM during all the measurements. Monitoring of the O₂ concentration was carried out by measuring the response of the reference sensor (AA428-210 (AO₂) Citicel, Honeywell City Technology, Bangalore, India) with a multimeter (34461A, Keysight, Santa Rosa, CA, USA). The TCO sensor mounted in a copper holder was placed in a closed glass flask with inlets for the gas mixture and electrical contacts, and gas outlets. The sensor was heated by connecting a back-side platinum heater to a direct current power supply (9174B, B&K Precision, Yorba Linda, CA, USA), labeled DCPS. The supply voltage level was set to 7 V, 8 V, and 9 V to obtain different operating temperatures of the TCO sensor (270 °C, 310 °C, and 350 °C). The direct current resistance (DCR) of the sensors was measured using a multimeter (34410A, Agilent, Santa Clara, CA, USA), with a range for measuring the resistance up to 1 GΩ.

5. Conclusions

The surface properties of TiO₂-modified CeO₂ layers, treated in a high-energy ball mill at three different rotational speeds, were examined to elucidate their sensing behavior when exposed to oxygen. The TiCe-1 sample prepared from precursors milled at the lowest speed (350 rpm) exhibited the highest oxygen sensitivity. Both TiCe-1 and TiCe-2 (450 rpm) outperformed pure CeO₂ in terms of the sensitivity at oxygen concentrations above 10%. All the measurements of the sensor response were conducted at operating temperatures lower than those typical for single metal oxide-based sensors. The XPS analysis revealed that incorporating TiO₂ in the CeO₂ layer helped in reducing the surface contamination, i.e., decreasing the thickness of the overlayer of adventitious carbon. At the speeds above 350 rpm, a new carbon-containing phase, CH2, in introduced. This phase is hypothesized to be physically separated from the carbon overlayer and located near the grain boundaries, potentially modifying the electron transport properties of the TiCe-2 (450 rpm) and TiCe-3 (550 rpm) layers. The TiCe-3 sample, treated at the highest milling speed, exhibited the lowest sensitivity among the tested samples. Potential contributing factors to this are the partially oxidized structure of the CH2 phase or a slightly reduced number of oxygen vacancies in comparison to the other samples. Furthermore, the XPS results indicated that oxygen vacancies, the primary active sites for oxygen adsorption, were exclusively formed within the CeO₂ domains. This suggests that while TiO₂ enhances electrical conductivity, its role in oxygen reactions is minimal or indirect. It is worth noting that the addition of TiO₂ also significantly lowered the resistance of the CeO₂ sensor across the entire range of O₂ concentrations, thereby simplifying the practical application of the sensor.