Dataset and publication "Thermally-induced microstructural evolution in nanoparticle-based CuO, WO3 and CuO-WO3 thin films for hydrogen gas sensing"

Contents lists available at ScienceDirect Applied Surface Science Advances journal homepage: www.sciencedirect.com/journal/applied-surface-science-advances Full length article Thermally-induced microstructural evolution in nanoparticle-based CuO, WO3 and CuO–WO3 thin films for hydrogen gas sensing Kalyani Shajia,1, Stanislav Haviara,1, Petr Zemana,1, Michal Procházkab,1, Radomír Čerstvýa,1, Nirmal Kumara, Jiří Čapeka,∗,1 aDepartment of Physics and NTIS - European Centre of Excellence, University of West Bohemia in Pilsen, Univerzitní 8, Plzeň, 301 00, Czech Republic bNew Technologies–Research Centre, University of West Bohemia in Pilsen, Univerzitní 8, Plzeň, 301 00, Czech Republic A R T I C L E I N F O Keywords: CuO, WO3, CuO–WO3 Microstructural evolution Thermal annealing Nanoparticle-based thin films Gas aggregation source Conductometric gas sensors Hydrogen gas sensing A B S T R A C T This study systematically investigates the microstructural evolution of nanoparticle-based CuO, WO3, and composite ‘CuO–WO3’ thin films induced by their post-deposition annealing. The films were reactively deposited using a magnetron-based gas aggregation technique, with the composite films consisting of alternating monolayers of CuO and WO3 nanoparticles. After deposition, the films were annealed in synthetic air at temperatures ranging from 200 to 400◦C and characterized using scanning electron microscopy, Xray diffraction, Raman spectroscopy, and X-ray photoelectron spectroscopy. Annealing of the CuO films led to the most pronounced changes associated with a gradual enhancement of crystallinity accompanied by significant particle growth with increasing annealing temperature, while the WO3 and CuO–WO3 films were more thermally stable to crystallization and particle growth. Notably, at 400◦C, the CuO–WO3 films crystallized into a novel 𝛾-CuWO4 phase. The annealed films were further evaluated for their gas-sensing performance upon H2 exposure and the obtained results were analyzed in relation to film properties and the microstructural evolution induced by annealing. 1. Introduction Hydrogen has gained significant interest as a clean and renewable energy carrier, which makes it a promising candidate for various applications, including power generation, large-scale energy storage, and chemical manufacturing [1,2]. However, its widespread applicability is hindered by safety concerns due to its low flammability limit (4% in air) and extremely low ignition energy (<0.02mJ), which present substantial explosion risks [3]. Furthermore, delicate devices such as fuel cells require precise control of low hydrogen flow rates to ensure optimal performance. As a result, the development of reliable hydrogen sensing technologies is essential for monitoring and regulating hydrogen concentrations across various applications, ensuring both safe and efficient operations. Among gas-sensing technologies, conductometric sensors have attracted considerable interest for their high sensitivity, fast response, and cost-effectiveness [4,5]. These sensors operate on the principle that the electrical conductivity of the sensing material changes in response to the target gases (analytes), which enables their detection and quantification. In particular, metal oxide semiconductors (MOS) ∗Corresponding author. E-mail address: [email protected] (J. Čapek). 1Researcher. are renowned for their superior performance as conductometric sensors [6,7]. In this case, gas species from the ambient atmosphere adsorb and desorb on the semiconductor surface, which alters the carrier concentration near the surface (known as the depletion layer (DL) for n-type MOS) and, in turn, the electrical properties of the material. Pure MOS sensors (not enhanced by noble-metal catalysts) are typically operated at elevated temperatures (200–400◦C) to facilitate the dissociation of H2 molecules on the surface, which is a critical process for the sensing mechanism. One of the key factors in the gas sensing performance is the microstructure of the sensing material, as it directly influences interactions between the sensing material and the analyte [8]. From this point of view, nanostructural materials are of particular interest. First, their high surface area significantly increases the number of active sites for gas adsorption and subsequent reactions. Second, optimizing the dimensions and spatial arrangement of the individual building blocks (e.g., nanoparticles (NPs), nanotubes) within the nanostructural materials significantly enhances the ability (known as transducer function) to convert chemical interactions with the analyte gases into a measurable electrical signal [9,10]. Third, so-called ‘necks’ can be formed at the https://doi.org/10.1016/j.apsadv.2025.100768 Received 3 March 2025; Received in revised form 2 May 2025; Accepted 10 May 2025 Applied Surface Science Advances 28 (2025) 100768 Available online 4 June 2025 2666-5239/© 2025 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ ). K. Shaji et al. junctions between the individual building blocks of the material, for example, when the material is exposed to an elevated temperature. These ‘necks’ significantly affect the changes in the measured conductivity of the percolation network in the material depending on whether the material is exposed to the analyte [11]. The reason is that smaller ‘necks’ can be ‘closed’ or ‘open’ for the electrical current, depending on the dimensions of the DL formed. Fourth, combining building blocks of different metal oxides may lead to the formation of heterojunctions at their interfaces. These heterojunctions are typically highly sensitive to the analyte and can lead to the formation of even deeper DL. This, in turn, amplifies changes in the electrical conductivity of the percolation network, since even larger ‘necks’ can be ‘closed’ or ‘open’ during the sensing process [12,13]. Among MOS materials, both WO3 and CuO have demonstrated excellent gas-sensing properties [14–20]. Several studies [21–24] have reported that composites of CuO and WO3 exhibit enhanced sensing performance toward gases such as xylene and H2S gas. However, the response of CuO–WO3 composites to H2 gas has not been explored yet. In addition, most of these studies have focused on wet-chemical synthesis routes, such as hydrothermal methods, which can compromise the purity of the resulting NPs. In contrast, physical synthesis techniques offer the advantage of producing high-purity NPs [25]. Among these, gas aggregation source (GAS), used in this work, additionally offers enhanced ways of control over NP size and composition, both critical factors influencing gas-sensing performance. In our recent works [26–28], we demonstrated that the H2 sensing performance of a WO3 thin film can be further enhanced by the synergistic combination with CuO NPs due to the formation of nano-sized p-n heterojunctions between n-type WO3 and p-type CuO. Building on these results, we have optimized our deposition technique, which uses a magnetron-based GAS, for the one-step controlled synthesis of the composite thin films consisting of alternating monolayers of CuO and WO3 NPs [29]. This physical deposition technique, leveraging lowtemperature discharge plasma, allows us to prepare high-purity NPs with tunable size, composition, and morphology, and further explore the effect of a composite of CuO and WO3 NPs on hydrogen gas sensing. Given that the microstructure of nanoparticle-based (NP-based) thin films is of key importance in determining their gas-sensing performance, we conducted a comprehensive multi-scale structural study (from atomic to morphological scale), which is complemented with hydrogen-sensing measurements. In particular, we systematically investigated the effects of post-deposition thermal annealing in air at temperatures between 200–400◦C (mimicking also operation of the MOS sensors at elevated temperatures) on the microstructure evolution of NP-based CuO, WO3, and CuO–WO3 thin films with the aim to further optimize the properties of these films and achieve a deeper understanding of the thermally-induced mechanisms that govern their gas-sensing behavior. 2. Experimental details 2.1. Film deposition and post-deposition annealing The thin films investigated in this study were prepared using a custom-built deposition system consisting of a GAS source mounted on a DN200ISO-K six-way cross vacuum chamber equipped with a rotating substrate holder and load-lock system. The GAS (Nanogen-Trio, Mantis deposition) consisted of three 1’’ magnetrons embedded into an axially movable holder of magnetrons installed within a grounded cylindrical aggregation chamber terminated by a domed ending with an exit orifice of 4mm in diameter. In this work, just two magnetrons equipped with high-purity (at least 99.95%) 3.2mm thick Cu and W targets were used. Each magnetron was driven by a DC power supply operated at a constant discharge power of 25W. A quartz crystal microbalance (QCM) mounted on a motorized shutter was used to measure the mass flux of NPs before and during the deposition process. Our in-house-built LabVIEW-based software effectively orchestrated the power supplies, flow controllers, and motorized QCM, ensuring reliable process repeatability. More details about the deposition process can be found in Ref. [29]. For this work, individual NP-based thin films composed exclusively of CuO and WO3, as well as complex composite thin films with alternating monolayers of CuO and WO3 NPs, were prepared. Note that from here on, we denote the individual NP-based thin films of the respective material as CuO or WO3, and the composite films of alternating CuO and WO3 monolayers as CuO–WO3. To explore the effects of annealing of the films, six samples each of CuO, WO3, and CuO–WO3 composite were prepared. The films were deposited onto polished and ultrasonically-cleaned 5×15mm2 Si (100) substrates. To ensure uniformity of the film across the entire substrate area, the depositions were divided into two sets, with three substrates per deposition. One sample each from CuO, WO3, and CuO–WO3 was kept aside as the as-deposited, while the other five were separately annealed to temperatures 200, 250, 300, 350, and 400◦C which corresponds to the typical operating temperatures of most MOS-based conductometric sensors. The films were annealed in a furnace (Clasic 1800) under synthetic air conditions at atmospheric pressure for 6h, followed by comprehensive characterization. First XRD, Raman spectroscopy, and SEM top-view analyses were carried out. Subsequently, each sample was pre-scratched and carefully broken into two pieces, each measuring 5×7.5mm2, with one part subjected to XPS study and the other to SEM cross-section analysis. 2.2. Film characterization The scanning electron microscope (SEM) (Hitachi SU 70) was used to make top-view and cross-sectional images of the as-deposited and annealed samples and investigate the effect of heating on their microstructure. A primary energy of 10keV and 15 keV was used for cross-sectional and top-view imaging, respectively. Top-view SEM images were analyzed using the ImageJ software to determine NP sizes. Approximately 50 particles, where the boundaries were reasonably visible, were selected and their diameters were measured manually with the imaging tools. The size distribution was then plotted as a histogram and fitted with log-normal distribution to estimate the average particle diameter. From the cross-sectional micrographs, the thickness of the as-deposited films was estimated to be 300±10nm. The crystallographic structure of the prepared films was examined by glancing incidence X-ray diffraction utilizing a diffractometer (X’Pert PRO MPD, PANalytical), operating at an accelerating voltage of 40kV and a tube current of 40 mA, using CuK𝛼 (𝜆 = 0.154187nm) at a glancing angle of 0.6◦. Diffraction patterns were obtained by continuous scanning in the 2𝜃 range of 10–80◦ with a step size of 0.05◦ and time per step of 12s. Data analysis was conducted using the PANalytical software package, HighScore Plus. The phase composition was also studied with Raman spectroscopy (LabRAM HR Evolution, Horiba Jobin Yvon) using a 532nm laser. X-ray photoelectron spectroscopy (XPS) measurements were performed in an ultrahigh vacuum chamber with a base pressure of ≤3×10−8Pa, using a non-monochromatic X-ray source XR 50 operated with Mg K𝛼 line (h𝜈 = 1253.6eV) and the hemispherical analyzer (Phoibos 150, SPECS Surface Nano Analysis GmbH) with a multichannel CMOS detector. The survey spectra were obtained by 5 scans with an energy step size of 0.5eV and pass energy of 50eV. Core-level spectra were measured by 20 scans with an energy step size of 0.05eV and pass energy of 30eV. Samples were mounted on Ti or Mo sample holders using conducting silver paste. For analyzing the measured spectra, the KolXPD software was used. All spectra have been charge corrected according to adventitious C 1s spectral component (C–C, C– H) with a binding energy (BE) of 284.8 eV [30]. The peaks were fitted using Shirley background and Voigt and Voigt doublet functions for deconvolution. Applied Surface Science Advances 28 (2025) 100768 2 K. Shaji et al. 2.3. Measurement of gas sensing properties Gas sensing measurements were performed on 200nm thick films deposited onto 9×9mm2 quartz glass substrates with square 2×2mm2 electrodes in the corners comprising sputter-deposited 50nm Cr layer followed by 100nm Pt layer. A custom-built sensor testing station was developed to assess the H2 gas sensing performance of the annealed samples using the fourpoint probe Van der Pauw method. Stable electrical contacts were established using platinum-coated contact pointy clamps arranged in a square pattern with a spacing of 8mm, which were pressed onto the sample electrodes. This whole setup was inside a brass chamber with a total volume of 3cm3. The sample was heated using a ceramic hot plate (Bach Resistor Ceramics GmbH, Germany), and a thin thermocouple (0.3mm in diameter) was attached to the top surface of the sample to monitor its surface temperature. More details can be found in Ref. [31]. The testing atmosphere was precisely regulated using three mass flow controllers (Alicat Scientific Ltd.) for the N2, O2, and H2 gases. To evaluate the H2 response, the sample was exposed to alternating 5min pulses of 1% H2 by vol. in synthetic air while maintaining the total flow rate at 100sccm. Resistance measurements were performed using a precise current source (Keithley 6220), and two voltmeters (Keithley 6514). Each sample was subjected to the following testing protocol. Prior to each measurement, the samples were pre-heated to 150◦C in a constant flow of synthetic air and maintained at this temperature for one hour to stabilize the baseline resistance and remove surface-adsorbed moisture from the sample and measuring chamber, thereby ensuring that the sensing measurements were conducted under stable, dry (non-humid) conditions. Subsequently, the gas sensing response was recorded for selected annealed samples at an operating temperature of 200◦C. 3. Results 3.1. SEM The top-view SEM micrographs, which illustrate the structural evolution of NP-based thin films of CuO, WO3 and their CuO–WO3 composite as a function of the annealing temperature, are shown in Fig. 1. The corresponding cross-sectional views are presented as insets. The films in their as-deposited state have already been thoroughly examined in our previous work [29]. The films consist of spherical NPs with a mean diameter of approximately 9nm for both materials with corresponding Full Width at Half Maximum (FWHM) values of 1.5 and 3nm for CuO and WO3 NPs, respectively. These values were determined by analyzing top-view SEM micrographs of isolated NPs dispersed on the substrate. However, this approach could not be applied to the annealed films in the present study, as the focus here is on the interactions between touching NPs during the annealing. For this reason, we utilized the ImageJ software package to estimate the dimensions of NPs from the SEM micrographs of the annealed films, as described in detail in Section 2.2. Although individual NPs with clearly distinguishable boundaries were selected manually from the SEM micrographs, the limitation in the resolution of objects at the nanoscale does not prevent the possibility of selecting agglomerates (loosely bonded NPs) and/or aggregates (fused or strongly bonded NPs forming a new particle) instead of individual NPs. Consequently, the measured size may be larger than their real dimensions. Nevertheless, calculating the mean diameter still provides a reasonable approximation of the overall particle size. Considering the microstructure of the as-deposited films, the WO3 film exhibits a less porous microstructure compared to the CuO film, despite the fact that NPs are of the same size. As expected, the porosity of the composite film appears to be between that of the WO3 film and the CuO film, but the microstructure is more similar to that of the WO3 film. The microstructure of the CuO films remains unchanged up to 250◦C when annealed. A noticeable increase in particle size begins at 300◦C, with significantly larger NPs observed at 350 and 400◦C. Based on our analysis of the SEM micrographs, the mean diameter of the NPs was derived from the fitted log-normal distributions, yielding values of 11, 19, and 27nm for the films annealed at 300, 350, and 400◦C, respectively, with corresponding FWHM values of 8, 8, and 15nm. For the WO3 films, the microstructure remains stable up to 300◦C. A slight increase in the diameter of the NPs is observed at 350◦C. This increase becomes more pronounced at 400◦C, where the mean diameter of NPs is 12nm (FWHM=7 nm). These findings align with previous studies, which report a more pronounced growth of WO3 NPs at temperatures exceeding 600◦C [32]. Finally, the microstructure of the CuO–WO3 composite film remains stable up to 350◦C, which suggests that the presence of both constituents in the film contributes to its structural stability. This stabilization is notable, as the microstructures of the individual constituents are stable only to lower temperatures (250◦C for the CuO NPs and 300◦C for the WO3 NPs). Annealing at 400 ◦C results in a slight increase in particle size with a mean diameter of approximately 11nm (FWHM=6nm). Furthermore, it is noteworthy that while the thicknesses of the WO3 and CuO–WO3 films remain relatively unchanged across all annealing temperatures, the CuO films exhibit a noticeable reduction in thickness, with an almost 25% decrease observed between the asdeposited and the 400◦C annealed samples. This trend indicates that CuO films become more compact at higher annealing temperatures, whereas the microstructure of WO3 and CuO–WO3 composite films exhibits enhanced thermal stability. 3.2. XRD Fig. 2 shows the XRD patterns of the as-deposited films and the films annealed up to 400◦C. From Fig. 2a, which corresponds to the XRD pattern of CuO films, it is evident that the annealing process has an influence on the intensity and the width of diffraction peaks. The as-deposited film shows a broad peak indicating a low crystallinity of the film. The peak position overlaps with the positions of the powder diffraction standards of monoclinic CuO (m-CuO; PDF Card No. 00-048-1548) and cubic Cu2O (c-Cu2O; PDF Card No. 00-005-0667), indicating the presence of both phases in the as-deposited film. When the film is annealed to 200◦C, two distinct diffraction peaks emerge at 2𝜃 values of about 35.5◦ and 38.7◦. These peaks can be assigned exclusively to m-CuO. The first peak corresponds to (002) and (11 1) diffraction doublet, while the second corresponds to (111) and (200) doublet. With increasing annealing temperature, these peaks become narrower and more intense, which suggests the growth of crystallites and improvement in the crystallinity of the film, respectively. The size of coherently diffracting domains was calculated from the FWHM of the (111) peak using the Scherrer equation, yielding values of 5, 7, 12, 18, and 23nm for the films annealed at temperatures of 200, 250, 300, 350 and 400◦C, respectively. Let us note that these values are in a good agreement with the size of the CuO NPs determined from the SEM micrographs (see Section 3.1). The XRD patterns of the WO3 films in Fig. 2b exhibit very broad diffraction peaks from the as-deposited state up to an annealing temperature of 250◦C. Such peaks are characteristic of a disordered structure. At these relatively low temperatures, the thermal energy is insufficient to promote a significant crystallization or crystallite growth [32,33]. At 300◦C, the XRD pattern contains distinct diffraction peaks at 2𝜃 values of 23.25, 23.75, and 24.35◦. These peaks correspond to the (002), (020), and (200) lattice planes of monoclinic WO3 (m-WO3; PDF Card No. 01-083-0950), respectively. At higher temperatures, the intensity of the (200) peak increases, which is attributed to improving crystallinity. However, the changes in FWHM are too small to draw definitive conclusions about the crystallite growth. Applied Surface Science Advances 28 (2025) 100768 3 K. Shaji et al. Fig. 1. SEM micrographs of top-views of the NP-based CuO, WO3, and CuO–WO3 films for the as-deposited state and after annealing at different temperatures. The insets show the cross-sectional micrographs of the respective samples with the same magnification as for the top-views. The as-deposited CuO–WO3 composite film, consisting of alternating CuO and WO3 monolayers, exhibits an XRD pattern (Fig. 2c) characterized by broad peaks similar to those of the individual CuO and WO3 films. Its disordered state is preserved up to 350◦C, which is a temperature higher than the crystallization thresholds of both binary films. Notably, further annealing to 400◦C leads to the emergence of a novel phase, which we refer to as 𝛾-CuWO4. Its XRD pattern does not show conclusively peaks corresponding to CuO, WO3, or any of their compounds recorded in the PDF database (e.g., triclinic CuWO4 (aCuWO4); PDF Card No. 04-009-6293 or triclinic Cu2WO4 (a-Cu2WO4); PDF Card No. 00-041-0948). Instead, it closely resembles to the known triclinic 𝛾-CuMoO4 phase (PDF Card No. 04-009-2227). In our parallel Applied Surface Science Advances 28 (2025) 100768 4 K. Shaji et al. Fig. 2. XRD patterns of NP-based CuO (a), WO3 (b), and CuO–WO3 (c) films for the as-deposited state and after annealing at different temperatures. The diffraction peaks labeled with an asterisk ’*’ in the XRD patterns can be attributed to the silicon substrate. study, we used ab initio simulations [34] and we confirm the existence of 𝛾-CuWO4 as a metastable form that might be stabilized by its lower density and/or Cu-rich composition. 3.3. Raman spectroscopy Fig. 3 presents Raman spectra of the as-deposited and annealed films. The Raman spectrum of the Si (100) substrate is shown at the bottom for reference. From the Raman spectra of CuO films (Fig. 3a), a peak overlapping with the Si vibrational mode is visible at 296cm−1, which corresponds to the A𝑔 mode of m-CuO, linked to phase rotations [35]. With heat treatment, there is an increase in the relative intensity of this peak to the intensity of the Si peak at around 513cm−1, which confirms that the changes of the peak at 296cm−1 can be associated with CuO. A second, weak peak is observed at 340cm−1, which gradually intensifies with increasing temperature. Additionally, a peak at 625cm−1 may indicate the presence of another vibrational mode of CuO due to its asymmetrical nature. Both peaks correspond to the B𝑔 modes of m-CuO, with the 340cm−1 peak attributed to Cu–O bending and the 625 cm−1 peak associated with symmetric oxygen stretching [35]. The sharper and more intense Raman peaks observed at higher temperatures can be attributed to the substantial increase in particle size of CuO, as discussed in Section 3.1 [35,36]. The Raman spectra of the WO3 films (Fig. 3b) in both the asdeposited state and after annealing up to 250◦C show a broad, asymmetric peak in the range of 700–800cm−1. This broad feature may indicate a low crystallinity of the film [37,38], which is consistent with the XRD results (Fig. 2b) that show a disordered state below 300◦C. Upon annealing to 300◦C, a sharp peak appears at 798 cm−1 along with weak peaks at 265 and 704cm−1. A very weak peak at about 320cm−1 also emerges at 300◦C. These bands (at 265, 320, 704, and 798cm−1) closely align with the wavenumbers of the four strongest modes of m-WO3 [38]. The bands at 265 and 320cm−1 correspond to the O–W–O bending modes of the bridging oxygen, while the bands at 704 and 798cm−1 correspond to the stretching modes [38,39]. With further heat treatment, the peaks become sharper and more intense. Unlike CuO, these changes are not linked to particle growth, as the size of WO3 NPs remains nearly unchanged after annealing (see Fig. 1c), but they are attributed to enhanced crystallization. In the CuO–WO3 composite films, the Raman spectra from the asdeposited state up to 350 ◦C are characterized by broad peaks, while distinct peaks appear only after annealing at 400◦C (Fig. 3c). This observation aligns well with the XRD findings (Fig. 2c). Although the peak identification cannot be fully conclusive, the peaks can be attributed to the CuO, WO3 and CuWO4 phases. The weak signal at 261cm−1 and the broad band observed around 750cm−1(likely deconvoluted into two prominent peaks at 704 and 798 cm−1) correspond to m-WO3, while the weak peak at 345cm−1 can be attributed to m-CuO. Other Raman peaks detected at 400cm−1, 825 cm−1 and 881cm−1 appear to be associated with CuWO4. The peak at 400cm−1 is commonly observed in the stable triclinic CuWO4 phase (referred to here as 𝛼-CuWO4) [40]. The peaks at 825cm−1 and 881 cm−1 closely match the characteristic positions of 𝛾-CuMoO4 nanostructures studied by Ali et al. [41]. Specifically, the Raman spectra of 𝛾-CuMoO4 show peaks at 811 and 837cm−1 along with a prominent peak at 880cm−1. Since no Mo was detected in the composite film, and the novel 𝛾-CuWO4 phase was revealed by XRD and further confirmed by ab initio simulations as a structural prototype of 𝛾-CuMoO4 [34], we suggest that the peak at 825cm−1 in the composite film may result from a combination of the 811 and 837cm−1 modes, while the peak at 881cm−1 is consistent with the peak at 880 cm−1. This alignment indicates that these peaks likely represent vibrational modes equivalent to the 𝛾-CuMoO4 phase. 3.4. XPS XPS analysis was conducted to investigate the evolution of the bonding states in the films with increasing annealing temperature. Note that the surface of the composite film has been topped by a layer of the CuO NPs with 50% coverage of the underlying WO3 monolayer, instead of the complete coverage by the CuO monolayer (10nm). This approach ensured that comprehensive information is obtained via XPS which has a characteristic penetration depth below 10nm. The analyzed core-level spectra of O, Cu, and W are summarized in Figs. 4–6. In the Cu2p spectrum of the as-deposited CuO film (Fig. 4a), it is possible to identify a component at 932.89 eV, likely corresponding to a mixture of Cu1+ and metallic Cu0 states, [42,43]. The second peak at 934.33eV corresponds to Cu2+ [42,43]. The position of the Cu2+ is slightly shifted compared to most references, but the split of Cu2+ and Cu1+ states is in the reasonable range (here 1.43 eV). The remaining features in the spectrum include typical satellite peaks of CuO and Cu(OH)𝑥. The O 1s spectrum (Fig. 4b) consists of three components: the Applied Surface Science Advances 28 (2025) 100768 5 K. Shaji et al. Fig. 3. Raman spectra of NP-based CuO (a), WO3 (b), and CuO–WO3 (c) films for the as-deposited state and after annealing at different temperatures. The spectrum corresponding to the bare Si substrate is shown as well as a reference. one at 529.77eV represents both copper oxides, Cu2O and CuO. The other two components at 531.23eV and 532.07eV correspond to surface contamination such as Cu(OH)𝑥, other hydroxides, or carbonates. With annealing, we can observe that the intensity of the contaminationrelated peaks decrease due to desorption while that of the copper oxide peaks increase. This trend is evident in both Cu2p and O 1s spectra. Additionally, there is an increase in the relative abundance of the Cu2+ state at the expense of Cu1+, which indicates oxygen saturation and the formation of the CuO phase. At the same time, while it is not possible to confirm or deny a metallic component Cu0, from the strong diffraction peaks of m-CuO at elevated temperatures (Fig. 2a), one can assume that any metallic state present in the as-deposited film likely vanishes during annealing, if it existed at all. The XPS spectra of WO3 films are shown in Fig. 5. In the W4f spectra (Fig. 5a), a major component at 35.84eV can be identified as W6+, which is consistent with literature values of 35.7 eV [44] and 35.8eV [43]. A minor component at 34.38 eV is attributed to an unspecified substoichiometric oxide state [43,45]. With increasing annealing temperature, a small but continuous decrease in intensity of this peak is observed, indicating further oxidation. Fig. 5b shows the evolution of the O1s peak in the WO3 films. Similar to the W 4f spectrum, the main peak at 530.82eV corresponds to the tungsten oxides [43,44]. The other two components at higher binding energies are attributed to surface contamination, likely originating from hydroxides, carbonates, or other surface-bound species. With increasing temperature, in addition to oxidation toward fully stoichiometric WO3, the W6+ peak shifts to lower binding energies above the temperature of 300◦C. This shift may be due to a microstructural evolution and crystallization of WO3, as clearly evident from XRD pattern for 300◦C in Fig. 2b. A similar phenomenon was reported by Liu et al. [46] after TiO2 crystallization during annealing at 400◦C. Fig. 6 shows XPS results for the CuO–WO3 composite film. Similar to the CuO film, three components can be identified: Cu1+(+Cu0) at 932.86eV, Cu2+ at 934.86 eV, and Cu(OH)𝑥 at 936.30eV. In contrast to the CuO film, the composite films exhibit a larger peak for the less oxidized copper states compared to the Cu2+ state. This can be explained by a preferential binding of the available oxygen to W, as can be seen in the W4f spectra in Fig. 6b, where only a single component W6+ is identified at 35.47eV. A sudden change in the Cu2p component ratio is visible at 400◦C. Here, CuWO4 is formed, as evident in the XRD pattern in Fig. 2c. Studies in Refs. [47,48] supports the potential formation of CuWO4 with the Cu2+ state appearing at the same position as in CuO. Therefore, the peak at 934.86eV likely represents both CuO and CuWO4. The same applies to the W4f spectra, where the only spectra component represents both WO3 and CuWO4. Fig. 6c shows O 1s spectra of the composite film. The main peak at 530.76 eV represents the states of all the oxides mentioned above. The two peaks at higher binding energies, 533.14eV and 531.99 eV, are attributed to surface contamination such as organic compounds and hydroxides [42,43]. These compounds desorb with increasing annealing temperature, as indicated by decreasing intensity of the corresponding peaks. All peaks (Cu2p, W 4f, O1s) in Fig. 6 exhibit subtle but consistent shifts toward lower binding energies with increasing annealing temperature. This behavior can be related to the crystallization of the material at an elevated temperature. 3.5. Gas sensing measurements The CuO, WO3, and CuO–WO3 films annealed at temperatures of 200, 300, and 400◦C were exposed to 1% by volume of H2 gas in synthetic air, and their dynamic response characteristics (Fig. 7) were studied at a fixed operating temperature of 200◦C. All annealed CuO films exhibit p-type behavior, as indicated in Fig. 7a by an increase in resistance upon H2 exposure. This increase is caused by the interaction of H2 with oxygen species (O− 2, O−, O2−) adsorbed from air onto the surface, which causes previously trapped electrons to be released back into the hole accumulation layer of p-type CuO, thereby reducing the hole concentration [12]. The film annealed at 300◦C demonstrates a slightly enhanced response compared to those annealed at 200◦C and 400◦C. In addition, the response and recovery are also faster for this sample as can be seen from the estimated response, Tres, and recovery, Trec, times indicated in Fig. 7. This enhanced performance arises from an interplay between the particle size and the crystallinity. Although the film annealed at 200◦C has a larger surface area, its low crystallinity limits the sensitivity, whereas the highly crystalline film annealed at 400 ◦C exhibits larger NPs and thus a reduced surface area [49]. The effect of annealing on the sensing response of the WO3 films is particularly interesting and distinctly different from that of conventional n-type WO3, which exhibits a decrease in resistance upon H2 exposure as electrons previously trapped by adsorbed oxygen species are released back into the electron depletion zone of WO3. The most pronounced anomalous behavior can be observed in the WO3 film with the disordered structure annealed at 200◦C. In particular, the resistance initially drops upon exposure to H2, then gradually increases, and eventually stabilizes. When H2 supply is stopped, the resistance slowly decreases, which also mimics the p-type behavior. In the film annealed at 300◦C, where crystallization already occurs, this anomalous response persists but is less pronounced. It becomes even less significant in the film annealed at 400◦C. In these films, the resistance initially Applied Surface Science Advances 28 (2025) 100768 6 K. Shaji et al. Fig. 4. XPS core-level spectra of Cu 2p (a) and O 1s (b) of NP-based CuO film for the as-deposited state and after annealing at different temperatures. The dotted lines indicate the respective peak positions in the as-deposited state. Fig. 5. XPS core-level spectra of W 4f (a) and O 1s (b) of NP-based WO3 film for the as-deposited state and after annealing at different temperatures. The dotted lines indicate the respective peak positions in the as-deposited state. Fig. 6. XPS core-level spectra of Cu 2p (a), W 4f (b), and O 1s (c) of NP-based CuO–WO3 film for the as-deposited state and after annealing at different temperatures. The dotted lines indicate the respective peak positions in the as-deposited state. Applied Surface Science Advances 28 (2025) 100768 7 K. Shaji et al. Fig. 7. Dynamic responses to 1 vol.% H2 gas at a fixed operating temperature of 200◦C for NP-based CuO (a), WO3 (b), and CuO–WO3 (c) films for the as-deposited state and after annealing at different temperatures. The sensing response is defined as R𝑔/R𝑎, where R𝑔 is the resistance in synthetic air containing H2, and R𝑎 is the resistance in synthetic air. The response values of CuO is shown at its original scale, whereas that of WO3 and CuO–WO3 are magnified by the factors indicated on each curve. The response time, Tres , and recovery time, Trec , defined as the times required to reach 90% of the total resistance change upon exposure to and removal of H2, respectively, are indicated in the figure for the respective annealing condition for both CuO and CuO–WO3. Note that an accurate estimation of response and recovery times was not feasible in the case of WO3 due to its anomalous behavior. drops upon exposure to H2, but then slowly increases instead of the decreasing trend typical for the n-type semiconductors. That is, the crystallinity seems to be a crucial factor controlling the p-type vs. ntype sensing response of the NP-based WO3 films. Let us note that the estimation of Tres and Trec were not feasible in this case due to the anomalous sensing behavior. Finally, the CuO–WO3 composite films exhibit an n-type sensing response at all annealing temperatures, as evidenced by their resistance decreasing upon H2 exposure. The response is very similar for the disordered films annealed at 200◦C and 300◦C, while the crystalline film with the 𝛾-CuWO4 phase, annealed at 400◦C, shows a slightly reduced response to H2. Tres and Trec exhibit a similar trend, with the 400◦C annealed film showing the slowest dynamics. While the composite films display slightly faster Tres and Trec at all annealing temperatures compared to the corresponding values for CuO films, none of the composite films demonstrate a significant enhancement in H2 sensing compared to the individual binary films, contrary to our initial expectations. 4. Discussion NP growth, as observed in all investigated CuO, WO3, and CuO– WO3 films, is primarily driven by the tendency of the system to minimize the overall surface energy. Thermal annealing provides sufficient thermal energy to overcome the activation energy for bond breaking and subsequent atomic migration [50–52]. Two primary mechanisms are commonly associated with the growth of NPs during annealing: coalescence and Ostwald ripening [53]. In dense NP-based materials like those in our study, the predominant growth mode is coalescence. In this process, the growth occurs at relatively low energies via surface migration of atoms and grain boundary diffusion between NPs in contact. With sufficient energy, this may result in the formation of a single, larger NP. In contrast, Ostwald ripening typically involves atoms detaching from smaller NPs and migrating to larger ones, which leads to a redistribution of material without direct particle contact. This process causes smaller NPs to shrink, while larger ones grow as a result of differences in surface energy. However, Ostwald ripening is more common in well-separated systems, where higher temperatures than those used in this study are required to facilitate atom evaporation and reattachment [53,54]. During coalescence, when two NPs come into contact, high-energy, unstable surface atoms (with few neighbors) rapidly diffuse to the more stable interface or point of contact (with more neighbors) [55]. The formation of new bonds at the interface and the annihilation of free surfaces produce heat sufficient to temporarily ‘melt’ the interface and form a neck; ‘melting’ here refers to thermally induced atomic disordering and bond softening that allow atoms to rearrange into a lower-energy configuration [56]. In amorphous NPs, this triggers a ’crystallization wave’ propagating through them as heat dissipates, which results in full crystallization [55,57]. Thus, coalescence may promote the crystallization of amorphous particles, and the entire process is accelerated at higher temperatures due to enhanced atomic migration [52,58,59]. When fully crystalline NPs come into contact, their crystallographic orientations are often misaligned due to random positioning. At lower temperatures, NPs reorient to maximize the contact area and thus enhance interfacial stability [56]. Initial contact leads to free surface annihilation via surface diffusion and the formation of a neck between particles. Disordered interface atoms gain energy from heat released during bond formation, which enables rearrangement to align with the crystallographic orientation of one NP [57]. In the case of incomplete alignment, grain boundaries form at the interface due to mismatched crystal planes [52,57,60]. Neck broadening, driven by surface and grain boundary diffusion, stabilizes the interface, often resulting in a dumbbell-shaped structure with a grain boundary at the interface. At elevated temperatures, enhanced atomic mobility driven by surface and grain boundary diffusion can facilitate complete coalescence through extensive atomic rearrangements and epitaxial alignment. However, when the misalignment between interacting NPs is substantial, the process may be hindered by the formation of stable grain boundaries due to kinetic and thermodynamic constraints that limit the extent of atomic reorganization. As a result, metastable configurations often emerge at the grain boundaries, which remain energetically stable over extended periods [56,57]. Our study reveals distinct crystallization behaviors for the CuO, WO3, and CuO–WO3 films upon thermal annealing. The CuO films exhibit crystallization at relatively low annealing temperatures (200◦C), with significant NP growth evident in the SEM micrographs, particularly at 300◦C. In contrast, the onset of crystallization in the WO3 films is delayed, and the increase in particle size is less pronounced compared to CuO. In the composite films, the XRD patterns show broad peaks Applied Surface Science Advances 28 (2025) 100768 8 K. Shaji et al. from the as-deposited state up to 350◦C, with the transition to a novel crystalline 𝛾-CuWO4 phase at 400◦C. This transition is accompanied by minimal changes in particle size. The ease in crystallization of CuO at relatively low temperatures can be attributed to its comparatively low enthalpy of atomization (7.71eV/atom). This low enthalpy facilitates bond breaking and enables sufficient atomic migration, even at lower annealing temperatures. At 200◦C, the available thermal energy is sufficient to break atomic bonds, thus promoting surface diffusion and the rearrangement of atoms into lower-energy configurations. As a result, early crystallization occurs, which is followed by the growth of NPs. The improving crystallinity with increasing annealing temperature in the CuO films is reflected in the intensifying diffraction peaks corresponding to mCuO. This observation is further supported by the vibrational modes in the Raman spectrum, which correspond to m-CuO. Furthermore, the decrease in FWHM of the diffraction peaks with annealing suggests an increase in crystallite size due to an enhanced coalescence. Notably, the negligible difference between the crystallite size calculated from FWHM of the (111) diffraction peak and the particle size estimated from SEM micrographs (at 300, 350, and 400◦C) implies that the particles are predominantly single-crystalline [52]. In addition, XPS analysis reveals that the fraction of Cu2+ states increases at the expense of Cu1+ with annealing, which is related to an enhanced oxygen incorporation and the progressive formation of stoichiometric CuO. Compared to the CuO films, the growth of NPs in the WO3 films is significantly limited to the growth observed after annealing at 400◦C, as evidenced by SEM micrographs. This slower growth can be attributed to the much higher enthalpy of atomization of WO3 (25.29eV/atom), which is three times higher than that of CuO. Consequently, more energy is required for atoms to break their bonds, diffuse and promote neck formation, atomic rearrangement and crystallization. The XRD patterns of the WO3 films annealed up to 300◦C show broad peaks, which indicate their small crystallite sizes and significant structural disorder. At 300 ◦C, the energy supplied is already sufficient for atoms to overcome bonding constraints resulting in the crystallization of WO3 to monoclinic phase. Further annealing does not significantly increase the intensity of the diffraction peaks, which may indicate a limited improvement in crystallinity. However, Raman spectroscopy of the WO3 films show that the intensity of vibrational modes associated with the monoclinic phase increases at and above 300◦C, which suggests an enhanced crystallinity. This apparent discrepancy in the results arises from the higher surface sensitivity of Raman spectroscopy compared to XRD [35]. Furthermore, XPS analysis reveals that the WO3 films are nearly fully stoichiometric, with only minor substoichiometry in the asdeposited state and after annealing at 200◦C. Increasing the annealing temperature leads to oxidation toward fully stoichiometric WO3 and improved crystallization, as indicated by the shift in the W6+ peak to lower binding energies above 300◦C. In the CuO–WO3 composite films, a highly disordered atomic arrangement is supposed to occur at the interface between CuO and WO3 NPs, where a grain boundary forms due to the significant difference in the crystallographic orientations of the two types of NPs [57]. This grain mismatch, which only results in incomplete coalescence, may explain the broad diffraction features observed in the films up to 350◦C. Furthermore, the composite film comprises alternating monolayers of CuO and WO3 NPs in equal volumetric proportions, which, however, does not correspond to an inherently equal ratio of Cu and W atoms due to the differences in the atomic volumes of CuO and WO3. CuO with a smaller atomic volume (approximately 21Å3 per metal atom) allows more Cu atoms to be packed into the same unit volume compared to WO3 (approximately 54Å3 per metal atom). This Cu-rich environment may then facilitate the incorporation of Cu atoms into the W sublattice, thereby relieving local lattice strain and promoting the stabilization of the metastable 𝛾-CuWO4 phase at the grain boundary as the annealing temperature rises to 400 ◦C [61,62]. This is supported by XRD findings and further corroborated by the Raman and XPS analyses. The gas-sensing measurements of all three NP-based films clearly reveal that the observed microstructural evolution is a key factor governing their sensing response. The coalescence of NPs, along with their size, crystallinity, and phase composition, directly influences the adsorption and reaction dynamics on the film surfaces upon H2 exposure. In the case of p-type CuO films, while the enhanced crystallinity with increasing annealing temperature is generally beneficial, it is counteracted by the pronounced growth of NPs with a reduction in surface area. As a result, optimal sensing performance is achieved at 300◦C when a delicate balance between the crystallinity and the surface area is maintained. In contrast, the much less pronounced growth of NPs in the WO3 films upon annealing suggests that the crystallinity enhancement is the dominant factor affecting their sensing behavior. The anomalous sensing response, characterized by a combination of p-type and n-type behavior, observed in these films can be attributed to complex surface interactions between the surface of WO3, oxygen species, and hydrogen gas, which may depend on the structural disorder as proposed in [63– 65]. The low crystallinity and thus the higher degree of structural disorder of the films annealed at lower temperatures might favor the adsorption of molecular oxygen (O− 2) rather than the dissociation of O– O bonds and the subsequent adsorption of atomic oxygen (O−). When H2 reacts with O− 2, it depletes electrons in WO3 forming hydroxide ions (OH−) and the resistance increases. On the other hand, the reaction of H2 with O− generates H2O molecules while releasing electrons, thereby reducing resistance. As the crystallinity improves with increasing annealing temperature, the latter reaction becomes dominant, causing the sensing response to gradually shift toward the typical n-type behavior of WO3. The n-type CuO–WO3 composite films showed the highest thermal stability of the microstructure among all the investigated films, with no noticeable NP growth or crystallization up to 350◦C. We attribute this behavior to the effect of the alternating CuO and WO3 NPs monolayers, where the coalescence of NPs and their growth is limited, as the CuO NPs are adjacent to the WO3 NPs from above and bottom and vice versa. Consequently, their sensing response remains largely unchanged and no significant enhancement is observed. At 400◦C, the crystallization into the novel 𝛾-CuWO4 phase occurs, but even this transformation does not result in enhanced sensing performance. The lack of the sensing enhancement compared to the individual binary films may be due to the nearly identical volumetric ratio of CuO and WO3 NPs within the composite films, which could restrict the synergistic effects typically observed in heterojunction-based sensors [13,28]. Our very recent results indicate that optimizing the CuO/WO3 volumetric ratio could yield better results. A more systematic investigation into this optimization, as well as the sensing response of the 𝛾-CuWO4 phase, is currently underway and will be presented in a forthcoming publication. 5. Conclusions NP-based CuO, WO3, and their composite CuO–WO3 films were reactively synthesized using a magnetron-based gas aggregation source controlled by an advanced in-house-built software. The films were subsequently subjected to thermal annealing in air up to 400◦C to systematically investigate the microstructure evolution from the atomic to the morphological scale. The thermally-induced changes in the films, including coalescence of NPs, their size growth, crystallinity, and phase composition were analyzed in relation to their gas-sensing response to H2 exposure. The CuO films exhibit a gradual enhancement of the crystallinity with increasing annealing temperature and a pronounced NP growth occurring at 300◦C and above. The optimal gas-sensing performance of these films, characterized by the p-type behavior, is achieved at 300◦C due to a delicate balance between the particle size (affecting surface area) and the crystallinity. In contrast, WO3 films begin to Applied Surface Science Advances 28 (2025) 100768 9