Near-Infrared Plasmonic Characteristics of Sputter-Deposited Nanocomposite Coatings formed by Indium Tin Oxide Nanoparticles in a SiO₂ Matrix

Abstract

This study explores the fabrication of nanocomposite coatings combining a gas aggregation cluster source (GAS) with reactive sputtering to embed plasmonic Indium Tin Oxide (ITO) nanoparticles (NPs) into a SiO₂ matrix. ITO NPs absorb in the infrared while remaining transparent in the visible range, enabling applications across a broad spectral window. By adjusting the relative deposition rates of SiO₂ and ITO NPs, we achieve coatings with varied filling factors. These structures exhibit localized surface plasmon resonance peaks between 2.2 and 2.5 μm. The resulting microstructure, shaped by self-shadowing during ballistic deposition, leads to persistent porosityeven at low ITO loadings (5 vol %). Spectroscopic ellipsometry and cross-sectional analysis reveal void fractions up to 30 vol %, resulting in an effective refractive index as low as 1.2. Postdeposition annealing in N₂ at 350 °C for 65 min enhances plasmonic response, shifting and amplifying the infrared absorption peak. This modulation in the short-wave infrared (SWIR) region is visualized using hyperspectral imaging. The study underscores the importance of deposition dynamics in shaping microstructure and optical behavior, highlighting challenges in producing dense nanocomposites with tailored infrared optical properties.

1. Introduction

Plasmonic nanomaterials have the potential to enhance the efficiency of already existing devices, while also unlocking new avenues for future technological advancements. − Nanostructures with a metal-dielectric interface spatially confine plasmons, resulting in wavelength-selective light absorption - a phenomenon known as localized surface plasmon resonance (LSPR). Historically, noble metals (Au, Ag, Cu) have been the most studied and utilized plasmonic materials due to their strong light coupling in the visible (vis) regime and a plethora of mature fabrication techniques.

Technological advancements in semiconductor technology have made the near-infrared (NIR, defined here as 0.75 μm - 1.0 μm) and short-wave infrared (SWIR, defined here as 1.0 to 3.0 μm) spectral regions readily accessible for imaging, driving the exploration of new materials with tunable LSPR within these regions of interest. The rapid advancement of commercial SWIR imaging technologies, driven by growing demand in scientific and industrial applications, has highlighted the importance of plasmonic materials with tunable responses in the SWIR spectral range. Nowadays, plasmonic materials with a tunable LSPR in the NIR/SWIR find their use in a handful of applications, including smart coatings for architectural glass, machine vision, bioimaging and surgery guiding to name a few.

Indium tin oxide (ITO) is a well-known transparent conductive oxide (TCO) often used in low emissivity (low-e) coatings due to its high luminous transmission (up to 90%) in the visible (vis, 0.36 μm – 0.75 μm) range, while remaining reflective in the NIR. , When its size is reduced down to a zero-dimensional form (i.e., nanoparticles) ITO retains its transparency within the vis spectral region while absorbing the light in the NIR/SWIR regimes through localized plasmons. − On the one hand, ITO has a limited number of free electrons and charge carrier mobility as compared to noble metals, which results in weaker LSPR response and lower conductivity. On the other hand, ITO possesses lower optical losses which is beneficial for the efficiency of plasmonic devices and can provide a solution to one of the key challenges in modern plasmonics - the significant and inherent optical losses of metallic components.

Akin to noble metal nanoparticles (NPs), the plasmonic response of ITO NPs can be tuned by adjusting their size and morphology, as well as the dielectric function of the surrounding medium. For example, it has been shown that the LSPR wavelength of ITO NPs can be readily controlled from 858 to 1758 nm simply by adjusting their size. Similarly, Gu et al. reported on the synthesis of branched colloidal ITO nanocrystals with a shape-dependent LSPR which manifests itself via the appearance of a new low-energy absorption peak as the particle morphology becomes more anisotropic. Finally, the semiconductor nature of ITO allows for an additional degree of freedom when it comes to tailoring its optoelectronic properties, through controlling the Sn content and/or oxygen vacancies. This has enabled the modulation of the absorption and reflection spectra of ITO NP-based coatings within the vis range. To place the present work in context, Table summarizes representative literature reports on the localized surface plasmon resonance of ITO nanoparticles, highlighting both the achievable spectral range and the principal tuning mechanisms. Across different synthesis routes and material compositions, reported LSPR peak positions span from the near-infrared to deep short-wave infrared, extending from approximately 850 nm up to several micrometers. This wide tunability is most commonly achieved through control of the free carrier concentration via Sn doping or oxygen vacancy engineering, while additional spectral shifts can be induced through nanoparticle size, morphology, or postdeposition thermal treatments.

1. Comparison of LSPR Wavelength of ITO Nanoparticles.

Author	LSPR (peak absorption)	Tunability
Wei	850 −1900 nm	annealing
Gu	1600–2150 nm	shape
Kanehara	1600 −1850 nm	Sn doping
Blemker	2250 nm
Shelemin	1600 nm
Lee	1800 nm
Caetano-Zeballos	2000 nm
Hong	2500 nm
Ma	2600–9000 nm	Sn doping
Maho	2200 nm
Lee	1500–2200 nm	Sn doping
Ma	1800–2000 nm	Sn doping
Li	2200–3000 nm	Sn doping

Despite the progress in exploring the SWIR/NIR characteristics of ITO NPs, most studies focus on standalone ITO NPs or colloidal solutions thereof, leaving a knowledge gap regarding solid-state nanocomposite coatings, typically applied in the glass/optical industries, where ITO NPs are embedded within a dielectric medium, underlying growth mechanisms in such systems and their optical characteristics.

In our previous work, we developed a single-step plasma-based synthesis method for plasmonic ITO NPs using a gas aggregation cluster source (GAS). This nonchemical approach with limited environmental impact has encouraged further interest to fabricate ITO NP-assembled conductive films.

The purpose of the present work is to advance the field of novel materials by producing plasmonic composite coatings featuring ITO NPs in a SiO₂ matrix. Such coatings maintain high luminous transparency in the vis while offering tunable transparency in the SWIR range. With the aim of going beyond the state of the art and addressing the knowledge gap in the growth mechanisms of nanocomposite coatings produced by combination of GAS with reactive magnetron sputtering (R-MS), we develop a multiscale ballistic deposition model to simulate coating growth, predict the coating microstructure, and correlate it with the optical response of the ITO/SiO₂ nanocomposites. We also demonstrate that the intensity and position of the ITO absorption band can be modulated within the SWIR spectral region by hyperspectral imaging. Finally, we investigate the ITO NPs’ optoelectronic properties to elucidate the mechanisms behind the observed tunable optical response.

2. Experimental Section

2.1. Deposition System

The ITO/SiO₂ composite coatings were fabricated using a deposition system schematically shown in Figure a. It consisted of a custom-built gas aggregation cluster source (GAS) mounted onto a separate vacuum chamber (hereinafter referred to as the processing chamber or PC) equipped with an additional magnetron. Both the GAS and the PC magnetrons were equipped with independently actuated shutters for the deposition process control.

1.

(a) Schematic of the deposition setup. (b) HRTEM micrograph of ITO NPs (a lower-magnification image used to determine the size distribution is provided in Figure S1). (c) Histogram and log-normal fit of the ITO NP size distribution, obtained by analyzing the TEM micrographs.

ITO NPs were synthesized in the GAS section using magnetron sputtering of an ITO target (2-in. diameter, 0.250-in. thickness, In₂O₃/SnO₂ 90/10 wt %, 99.99% purity, Kurt J. Lesker, USA). The synthesis process, described in more detail in a previous paper, was conducted at an elevated buffer gas pressure of 863 mTorr, while introducing 15 sccm of the argon (Ar, 5N purity, Air Liquide, QC, Canada) via an automatic mass flow controller (MKS Instruments, USA). The magnetron, equipped with the ITO target, was powered by a DC power supply (Pinnacle, Advanced Energy, CO, USA) operating in a constant-current mode at 500 mA. The orifice diameter was 2 mm.

The SiO₂ matrix was produced in the PC chamber by reactive sputtering of a silicon target (2-in. diameter, 0.250-in. thickness, 99.99% purity, Kurt J. Lesker, USA) at 15 mTorr in an Ar/O₂ atmosphere. The working gas atmosphere was formed by mixing 15 sccm of Ar, supplied through the GAS orifice, with 5 sccm of oxygen (O₂, 5N purity, Air Liquide, QC, Canada) injected directly into the PC via a separate gas feed line. The Si target was sputtered using a 150 W radiofrequency (RF, 13.56 MHz) plasma, powered by an RF power supply (AG0613, T&C Power Conversion Inc., USA). The PC magnetron was positioned 15 cm from the substrate holder at a 45-degree angle relative to the holder’s normal. All depositions were carried out without intentional substrate heating, i.e., at nominal room temperature. The substrate temperature was not actively monitored during deposition; while a mild temperature increase during SiO₂ sputtering cannot be excluded, thermal effects were not a controlled parameter and are not expected to significantly affect the microstructure during growth.

The experimental conditions for the NP and matrix depositions were optimized based on real-time mass flux data monitored by a retractable quartz crystal microbalance (QCM) prioritizing long-term process stability over the highest deposition rates.

The composite coatings were prepared by depositing ITO NPs and the SiO₂ matrix in a sequential layer-by-layer manner using the shutter in front of cluster source. The total deposition time for the ITO NPs was fixed at 30 min for all experiments, while the amount of the SiO₂ matrix was varied by doubling its deposition time in successive steps: 0, 30, 60, 120, and 240 min (see Table for the details). The deposition process alternated between two modes. In the first mode, both the ITO NPs and the matrix were deposited simultaneously for 3 min. In the second mode, only the matrix was sputtered. In total, the deposition modes were alternated 10 times. The duration of the matrix-only deposition step was adjusted to ensure that the deposited SiO₂ amount to be equivalent to 0, 30, 60, 120, and 240 min of pure SiO₂ sputtering. Such an approach allows one to track the relative amounts of ITO and SiO₂ within the composites. To ensure a more uniform deposition, the substrate holder was rotated at 1 Hz. Note, a reference (Sample 1) containing standalone ITO NPs was produced using GAS with the PC magnetron off.

2. Total Deposition Times of the ITO NPs and SiO ₂ Matrix Layers Used in the Experiments.

Sample ID	ITO NP deposition time [min]	SiO2 deposition time [min]
1	30	0
2		30
3		60
4		120
5		240

2.2. Sample Characterization

The morphology of the produced NPs were characterized using bright-field (BF) imaging of a submonolayer of NPs deposited onto a 30 nm-thick Si₃N₄ membrane grid (Norcada, AB, Canada). The imaging was performed with a transmission electron microscope (TEM, JEM 2100F, Jeol USA Inc., MA, USA). The obtained BF micrographs were analyzed statistically using the ImageJ software package to determine the NP size distribution and flux, defined as the number of NPs deposited per unit area per unit time (see Figure c).

The NP crystallinity was further analyzed with grazing incidence X-ray diffraction (GIXRD) using a diffractometer (D8, Bruker, MA, USA) equipped with a Cu target (Kα source, X-ray wavelength of 0.154 nm).

The microstructure of the prepared nanocomposite ITO/SiO₂ coatings was studied by performing cross-sectional secondary electron (SE) imaging with a scanning electron microscope (Quattro S, Thermo Fischer Scientific, US). For this purpose, the coatings were deposited onto precleaved Si wafers.

The specular transmission (T _spec) and reflection (R _spec) spectra of the composites were measured using a variable-angle UV–vis-NIR spectrophotometer (Cary 7000, Agilent Technologies Inc., CA, USA) equipped with a universal measurement accessory (UMA). Total transmission (T _total) and total reflection (R _total) spectra were obtained with a UV–vis-NIR spectrophotometer (Lambda 1050, PerkinElmer Inc., MA, USA) equipped with a 150 mm diameter integrating sphere.

To evaluate light extinction caused by absorption (A) and scattering (S) both T _spec and R _spec as well as T _total and R _total were measured. The A was calculated using the following formula:

$$A=100-T_{total}-R_{total}$$ 1

while the directional (forward as well as backward) scattering was calculated as follows:

$$S=T_{total}-T_{spec}+R_{total}-R_{spec}$$ 2

To gain insight into the optoelectronic properties of the ITO and the impact of the encapsulating medium on them, the T _spec and R _spec spectra were fitted using a custom script in MATLAB implementing a Bruggeman Effective Medium Approximation (B-EMA) model and the Transfer Matrix Method (TMM). The model consisted of three components: ITO, SiO₂, and voids (representing pores within the coatings to better reflect the present scenario). The fit was restricted to a range between 600 and 2500 nm as the model did not account for scattering effects, which were shown to be more prominent at shorter wavelengths.

Indium tin oxide (ITO) is predominantly an n-type degenerate semiconductor, where free charge carriers arise from both extrinsic doping by Sn and intrinsic doping via oxygen vacancies. Unlike ITO thin films, which reflect infrared light due to their metallic nature, ITO NPs absorb light within the SWIR/NIR regions through LSPR. The dielectric permittivity of TCOs such as ITO films can be described by the Drude-Sommerfeld theory:

$$\epsilon (\omega )={\epsilon }_{ITO,\infty }-\frac{q^{2}N}{{\epsilon }_0{m}_e^{*}}\left(\frac{1}{{\omega }^2+i\gamma \omega }\right)$$ 3

where m _e = 0.45m _e is the effective electron mass, ε_ITO,∞ = 3.74 is the high-frequency permittivity, q is the elementary charge, and ε ₀ is the vacuum permittivity. The N parameter represents the free charge carrier concentration in ITO, while γ is a damping constant.

To describe the plasmonic behavior of ITO NPs within a dielectric matrix, we use effective medium theory (EMA). Metallic inclusions in nanocomposites are often modeled with the Maxwell-Garnett approximation, yet we opted for the Bruggeman model − due to the broad range of filling factors in our samples and given by the following mixing formula:.

$$f_{Void}\frac{{\epsilon }_{Void}-{\epsilon }_{eff}}{{\epsilon }_{Void}+2{\epsilon }_{eff}}+f_{SiO2}\frac{{\epsilon }_{SiO2}-{\epsilon }_{eff}}{{\epsilon }_{SiO2}+2{\epsilon }_{eff}}+f_{ITO}\frac{{\epsilon }_{ITO}-{\epsilon }_{eff}}{{\epsilon }_{ITO}+2{\epsilon }_{eff}}=0$$ 4

where f _ITO, f _SiO2, f _Void are the filling factor and ε _ITO, ε _SiO2, ε _Void are the dielectric functions of each component.

The T and R spectra were fitted using a three-component model, comprised of ITO, SiO₂, and void, where the ITO component contribution was described by the Drude model with the free charge carrier concentration and the damping constant as fitting parameters. The refractive indices for the SiO₂ and the void refractive indices were assumed constant and set at 1.45 and 1.00 respectively across the entire spectral range (600–2500 nm) of interest. In all cases, the roughness of the composites was modeled as a 100 nm layer using the Bruggeman effective medium approximation, consisting of a 50:50 mixture of the composite’s effective dielectric function (calculated from eq ) and void.

The optical and structural properties of the nanocomposites at macroscopic scale were further analyzed using variable angle spectroscopic ellipsometry (VASE, RC2-XI, J.A. Woollam, NE, USA). Measurements were conducted over the same wavelength range of 600–2500 nm at four angles of incidence (45°, 55°, 65°, and 75°). They were performed with the use of focusing probes and a motorized stage enabling mapping across the sample. Postdeposition annealing (350 °C, N₂) experiments were carried out using a heat-cell accessory installed on the RC2-XI (Linkam temperature stage, UK). The ellipsometry data (ψ and Δ) were fitted using the CompleteEASE software (J.A. Woollam) using a B-EMA model configured similarly to the one employed for the T and R spectra fitting. The fitting of ellipsometry was performed in accordance with the fitting of T and R, obtaining comparable results.

Imaging of the samples was performed using a digital visible light camera and a hyperspectral SWIR imaging setup L-EOS equipped with ZephIR SWIR camera (Photon etc., QC, Canada). To estimate the transmittance of the samples within the SWIR range, we use the average transmission which is calculated as follows:

$$T_{av}=\frac{{\int }_{0.9}^{2.5}\eta (\lambda )·T(\lambda )·d\lambda }{{\int }_{0.9}^{2.5}\eta (\lambda )·d\lambda }$$ 5

where η is the quantum efficiency of a SWIR photodetector, and T is the measured specular transmission for the wavelength range between 0.9 and 2.5 μm. We use the quantum efficiency of a ZephIR 2.5 SWIR camera PD as our sensitivity reference function.

The relative deposition rates of ITO nanoparticles, the SiO₂ matrix, and the corresponding nanocomposites were analyzed using a combination of QCM and ellipsometry. See section 3.3 for more details

3. Results and Discussion

3.1. ITO NP Morphology and Crystallinity

In the first part of the study, we analyze the morphology and crystallinity of ITO NPs. The ITO NPs produced via GAS exhibit a variety of shapes, including spherical, octahedra and truncated dodecahedrons (Figure b), with a mean size of 21 ± 13 nm (Figure c). Multiple NP structures are observed, including single crystals, poly crystalline and amorphous. The crystalline domains are assigned to the cubic Ia̅3 space group as confirmed by the X-ray diffraction measurements described below. The variety of obtained shapes and the presence of an amorphous phase both indicate complex growth dynamics and rather fast thermalization resulting in the formation of out-of-equilibrium shapes.

To further characterize NP crystallinity, GIXRD was used as a complementary technique. The diffraction pattern of the ITO NPs (shown in Figure S2) closely matches both the XRD pattern of the original target material from which the NPs were sputtered and the cubic In₂O₃ reference pattern from the database (ICSD 50847, comparable to JCPDS card no. 06–0416).

The assignment of crystalline, polycrystalline, and amorphous nanoparticles is based on diffraction evidence reported in our previous work, where selected-area electron diffraction (SAED) combined with HRTEM was used to explicitly distinguish nanoparticles exhibiting clear diffraction features from those showing no diffraction contrast. The present study builds on that work and focuses on nanoparticle–matrix growth and optical response; therefore, TEM is used here primarily to document particle morphology and size distribution, while the structural interpretation relies on the previously established SAED/HRTEM analysis in combination with XRD measurements performed on the samples presented here.

3.2. Deposition Rates of ITO NPs and SiO₂

Before performing a detailed characterization of the ITO/SiO₂ composite coatings, we first estimate the intended ratio between the ITO content and the SiO₂ matrix. As discussed, the deposition time for the nanoparticles was kept constant across all samples, while the deposition time of the matrix was varied. Estimating the amount of matrix was straightforward: SiO₂ films deposited under the same conditions as the composites were prepared, and their thickness determined by fitting ellipsometry data with a simple model (Table ).

3. Estimated (Design) Thicknesses of ITO NPs and of the SiO₂ Matrix for Different Deposition Conditions.

Sample ID	SiO2 matrix equivalent thickness [nm]	ITO NP equivalent thickness [nm]	Total equivalent thickness (SiO2 + ITO NPs) [nm]
1	0	150	150
2	250		400
3	500		650
4	1000		1150
5	2000		2150

In contrast, estimating the amount of deposited ITO is more challenging. The nanoparticle coatings form a highly porous structure, where the total thickness does not directly correspond to the material quantity. To address this, a calibration combining quartz crystal microbalance (QCM) measurements and variable-angle spectroscopic ellipsometry (VASE) was performed. As the QCM measures mass flux, by depositing both the matrix and nanoparticles individually at comparable QCM frequency shifts, we ensured that the coatings contained a similar mass of material.

Using the bulk densities of ITO (7.14 g/cm³) and SiO₂ (2.13 g/cm³), along with the ellipsometry-measured thickness of the SiO₂ coating, we estimate that the equivalent thickness of a compact ITO thin film would be 150 nm. However, ellipsometry measurements of the ITO nanoparticle coatings using the B-EMA model indicate a thickness of approximately 450 nm. This suggests that the volumetric filling factor of the pure nanoparticle film is about 0.33, which aligns well with the 0.36 filling factor obtained from the B-EMA model (Section 3.6). To account for the porosity component in the coatings, we introduce the filling factor (FF) and define it as follows:

$${FF}_x=\frac{V_x}{V_{ITO}+V_{{SiO}_2}+V_{VOID}}=\frac{t_x}{t_{ITO}+t_{{SiO}_2}+t_{VOID}}$$ 6

The volume of the component V _x corresponds to the equivalent thickness t _x , as the deposited surface area remains the same on a macroscopic scale. This provides a straightforward approach to quantitively compare deposition rates, composite thicknesses, and filling factors obtained from optical models, which all are expressed in the volumetric form. For all of the samples, we assume the same amount of nanoparticles corresponding to Sample 1 which contains no matrix material.

It should be noted that the void fractions estimated from eq are subject to uncertainty arising from the different methods used to determine the individual thickness contributions, as well as from the inherently indirect nature of this approach, which may be susceptible to sample-to-sample variations. The SiO₂ matrix thickness is obtained from ellipsometric calibration coatings and is therefore relatively well-defined. In contrast, the equivalent thickness of the ITO nanoparticles is inferred indirectly from QCM-based deposition rate calibration performed in a separate experiment and may be affected by run-to-run variations in deposition rate even under nominally identical conditions.

3.3. Microstructure and Thickness of ITO/SiO₂ Nanocomposite Coatings

The microstructure and thickness of the deposited composites were analyzed using SEM imaging in the cross-section orientation. The resulting SEM micrographs are shown in Figure a-e. The measured thicknesses of the composites were 300 nm for films composed only of ITO NPs and 1100, 1500, 2000, and 3000 nm for the ITO/SiO₂ composites, corresponding to a higher matrix concentration (see Table ). The total composite thickness measured from cross-section SEM carries additional uncertainty due to sample cleaving, local thickness variations, and imaging geometry. Because the void fraction is derived from a combination of independently estimated effective thickness components, the absolute values of the void fraction should therefore be regarded as approximate. To assess the impact of measurement uncertainty, uncertainty bounds were estimated based on the individual methods: ± 25% for the effective ITO nanoparticle thickness inferred from QCM calibration, ± 5% for the SiO₂ thickness obtained by ellipsometry, and ± 10% for the total thickness measured by cross-sectional SEM. While these uncertainties affect the absolute void fraction values, the observed trend of decreasing porosity with increasing SiO₂ matrix content remains robust within these bounds.

2.

(a–e) Cross-section SEM micrographs of ITO/SiO₂, thickness of the composites corresponding to the values in Table shown as blue columns. (f) Filling factors obtained from SEM analysis; (g and h) SEM micrographs of higher magnification showing the microstructure of the composite coatings (Sample 1 and Sample 5, respectively). The SEM micrograph in (h) is shown in fake colors using Green-Red channels that correspond to signals from SE/BSE detectors, respectively.

We note that the measured thicknesses (see Table ) are significantly higher than the sum of the matrix thickness and the equivalent thickness of the ITO NPs predicted by the design (see Table ). This indicates the presence of porosity in all the deposited coatings.

4. Thickness and Experimentally Determined Filling Factors by SEM Analysis for Different Deposition Conditions.

Sample ID	FFITO (SEM)	FFSiO2 (SEM)	FFVoid (SEM)	Thickness (SEM) [nm]
1	0.38 ± 0.1	0.00 ± 0	0.62 ± 0.1	300
2	0.14 ± 0.04	0.23 ± 0.03	0.63 ± 0.05	1100
3	0.10 ± 0.03	0.33 ± 0.04	0.57 + 0.05	1500
4	0.07 ± 0.02	0.50 ± 0.06	0.43 ± 0.07	2000
5	0.05 ± 0.01	0.67 ± 0.08	0.28 ± 0.08	3000

While the presence of porosity in the coatings comprised of only NPs (Figure f) is expected due to its ballistic growth regime, , one can anticipate a higher densification of the structure as the amount of SiO₂ matrix increases. However, even for a coating with the highest deposition time of SiO₂ matrix, the resulting structure contains nearly 30% of voids. This estimation is based on the expected thickness of the matrix (2000 nm), the equivalent thickness of ITO nanoparticles (150 nm), and the total thickness measured by SEM (3000 nm) (see eq ).

To investigate the composite coatings’ microstructure in more detail, we superimpose the secondary electron (SE) and the backscattered electrons (BSE) signals obtained during SEM imaging. The BSE provides atomic contrast (Z-contrast) as heavier elements (In and Sn), because of their bigger nuclei, can deflect incident electrons more strongly than lighter elements (Si), while SE conveys information on the specimen’s morphology. Here, the green and red channels (appearing as yellow) represent SE and BSE signals, respectively.

Figure g shows the resulting SEM micrograph obtained by mixing SE/BSE signals. Within the coating structure, the ITO NPs are covered on top by SiO₂ plumes, suggesting that the ITO NPs shadow their immediate surroundings during predominantly line-of-sight deposition of the matrix material. We hypothesize that this mechanism plays a major role in generating pores within the coating structure. We propose a straightforward growth model to explain the formation of voids and dome-shaped deposits on the ITO NPs, which is discussed in-details in the next section.

3.4. Nanocomposite Growth Model

To study the phenomenon of shadowing by NPs, we developed a multiscale ballistic deposition (BD) growth model combining two classical ballistic models. Such an approach offers a qualitative description of the observed characteristics of the deposited coatings.

The growth rule of the 2D model is as follows: start with a substrate line of length L. Particles (points) are released from a parallel line source of the same length L, positioned far above the substrate. These particles move perpendicularly toward the substrate and adhere upon first contact with either the growing aggregate or the substrate. After N particles, where N ranges from 10⁵ to 10⁶ for different filling factors, a particle representing an ITO NP with a radius of 200 points is released according to the same rules. This process is repeated in cycles until 200 ITO NPs have landed onto the growing film. The filling factor in simulations is adjusted by doubling the number of matrix atoms, replicating the real deposition process, where the deposition time of the matrix was also doubled between experiments. The filling factor in simulations can be estimated in a similar manner to SEM cross-section image analysis. Since we know the thickness of the simulated films without nanoparticles (from additional simulations) and the thickness of the simulations including nanoparticles, while keeping the amount of nanoparticles constant across simulations, we can apply the same approach. By making a similar assumption as in the SEM analysis-that pure nanoparticle films have a porosity of approximately 60% - the filling factors can be estimated accordingly.

Compared to other BD models, we simplified our approach by making the small particles (atoms) the same size as the grid cells. The insets in Figure illustrate the structures originating from the interaction between these BD processes. The resulting process produces branched, tree-like structures growing from the NPs. At the NP scale, the internal structure of the matrix becomes irrelevant, and only the effective density of the material remains significant.

3.

(a) Filling factors obtained from MBM. (b) Simulated structures; Insets show the fine details of the simulated structures. (c) Raw BSE SEM micrograph of Sample 5 for comparison of the microstructure.

Figure b illustrates a series of simulations with varying amounts of matrix material. At lower relative deposition rates of matrix atoms in comparison with NPs, the matrix material forms plumes that cover the nanoparticles. These plumes enlarge the effective top cross-section of the NPs, increasing their size in both lateral and longitudinal dimensions. In this scenario, the growth mechanism closely resembles the pure NP BD process, except that the deposited nanoparticles with plumes are effectively larger.

With an increasing amount of matrix material, the growth of the matrix begins to dominate, homogenizing the structure by inhibiting propagation of the topographical features during growth. The regions directly around the NPs remain more porous due to shadowing effects. Since the initial growth mechanism remains the same - formation of plumes on the nanoparticles - the matrix grows in a columnar structure, with pores occurring between these columnar or tree-like structures. This closely resembles the structures shown in the cross-section SEM (BSE signal only) image (see Figure c). Even if the NP filling factor is reduced, the shadowing effects still prevent a complete homogenization of the film. The presence of larger grains leads to amplification of shadowing effects, hindering the homogenization even further.

We attempted to quantitatively compare the BD model simulations with the structures analyzed by SEM. In accordance with the experimental conditions, we kept the amount of NPs constant across all the simulations while doubling the amount of the matrix. We then computed the filling factors using the previously developed methodology. The results are presented in Table .

5. Comparison of Estimated (Design) Component Ratios with Filling Factors Calculated with a Multiscale Ballistic Depostion Model (MBM) for Different Deposition Conditions.

Sample ID	FFITO (MBM)	FFSiO2 (MBM)	FFVoid (MBM)
1	0.60	0.00	0.40
2	0.26	0.43	0.31
3	0.16	0.55	0.29
4	0.10	0.66	0.24
5	0.06	0.78	0.16

Although the simulations yield different numerical values compared to the other methods, the overall trends remain similar. It is important to note that the model described here represents a highly simplified two-dimensional structure, whereas the actual composites are three-dimensional. Additionally, the simulated thickness is significantly smaller than that of the experimental composites due to simulation time limitations.

It should be emphasized that the ballistic deposition model presented here is not intended to be fully predictive but rather serves as a semiquantitative framework to rationalize the experimentally observed growth behavior of the present nanoparticle–matrix system. The model relies on several simplifying assumptions, including idealized ballistic particle trajectories, unit sticking probability upon first contact, and a simplified treatment of void formation without allowing for surface diffusion, restructuring, or coalescence effects. In addition, the simulations are two-dimensional, whereas the experimental coatings are inherently three-dimensional, and the simulated film thickness is limited by computational constraints. Despite these simplifications, the model captures the essential shadowing-driven growth mechanisms and reproduces the observed trends in porosity and filling factor as a function of matrix deposition, providing order-of-magnitude agreement with experimentally inferred values rather than exact quantitative predictions.

3.5. Optical Properties

The optical spectra (Figure a1–a5, c1–c5) reveal a strong absorption feature in the NIR/SWIR region, with the LSPR maximum located at the upper limit of the measured spectral range (∼2500 nm). Despite efforts to maintain consistent NP concentration across all samples as per design, Samples 1 and 5 display less intense plasmon peaks, which could potentially indicate a slightly lower NP content. Nevertheless, for simplicity, we assume, in the following analysis, that there is a similar NP concentration in all samples.

4.

UV–vis–NIR spectra recorded on the ITO/SiO₂ composite coatings: (a1–a5) Transmission (T) spectra; (b1–b5) Reflection (R) spectra; (c1–c5) Absorption (A) and Scattering (S) calculated from T and R data. Black solid lines correspond to the specular T and R, red to their fit obtained using a B-EMA model. Green solid lines represent Absorption, and blue solid lines represent Total Scattering (Scattering in T + Scattering in R).

The fitted transmission and reflection spectra with the B-EMA model are shown in Figure as red lines. The charge carrier concentration and damping constant were extracted from the fit. The estimated free charge carrier concentration ranged from N = 3.5 × 10²⁰ cm^–3 to 4.8 × 10²⁰ cm^–3. Reported values for ITO thin films fall within 1 × 10²⁰ cm^–3 and 2 × 10²¹ cm^–3, − indicating that the synthesized NPs exhibit carrier concentrations comparable to their thin films counterparts. The damping constant was found to be around 3.8 × 10¹⁴ rad/s which is higher than typical values for ITO thin films (∼1.0 × 10¹⁴ rad/s). This increase is attributed to the size-dependent effects of nanoparticles on the dielectric function, which is standardly described as γ = γ₀ + A/R, where γ₀ is the damping constant of the bulk material, A is an experimentally estimated constant, and R is the radius of the nanoparticle.

In addition to material properties of the ITO NPs, the optical model provides additional insights regarding the composites, including the effective refractive index (n) and extinction coefficient (k) (Figure a), along with the calculated filling factors (Figure b). The refractive index of the samples increases with the matrix content which aligns well with prior SEM and VASE observations indicating more dense coatings. For the most dense coating (Sample 5), the refractive index is close to that of the matrix with a value of 1.45 at 550 nm. Despite this, the presence of a high surface roughness (100 nm) still renders the sample antireflective (Figure ). At the same time, the extinction coefficient of the samples with a smaller filling factor is reduced as the NPs become more homogeneously dispersed in the volume of the composite. The obtained filling factor trends (Figure b) are consistent with the filling factors obtained from the SEM analysis.

5.

a) Refractive index (n) and extinction coefficient (k) of the composites computed by fitting T and R data with the optical model. Sample 3 is not shown to improve readability, as its slightly blue-shifted LSPR leads to significant overlap with the other curves. Version of panel (a) including Sample 3 is provided in the Supporting Information in Figure S3. (b) Filling factors for each component obtained from the optical model.

The studied UV–vis-NIR spectra reveal one additional important property of the matrix-rich composites, which was not taken into consideration in the optical model: the presence of increased scattering with a higher concentration of matrix. As the amount of matrix increases, the scattering also increases, as demonstrated in Figure c1–c5). In Sample 5, the haze level reaches approximately 20% at 400 nm. We attribute this enhancement of scattering to growth of the SiO₂ plumes on top of the ITO NPs as shown in Figure and Figure . The plumes grow larger with increasing amount of SiO₂ deposited, reaching dimensions comparable to visible light wavelengths.

Finally, optical modeling also allows one to estimate the thickness. Since the T and R spectra measurements were taken on glass samples placed next to silicon wafers with an approximately 1 cm offset in the axis of rotation, the thickness estimates obtained by fitting the spectroscopic data are slightly lower and generally comparable to the VASE measurement at ± 1 cm. Despite these slight differences, both techniques show very consistent results (see Table ).

6. Comparison of the Filling Factors As Determined by Fitting VASE Data and Fitting T and R Spectra for Different Deposition Conditions.

Sample ID	FFITO (VASE)	FFSiO2 (VASE)	FFVoid (VASE)	FFITO (vis–NIR)	FFSiO2 (vis–NIR)	FFVoid (vis–NIR)
1	0.36	0.00	0.64	0.39	0.00	0.61
2	0.13	0.30	0.57	0.13	0.35	0.52
3	0.09	0.40	0.51	0.11	0.59	0.30
4	0.06	0.61	0.33	0.07	0.71	0.22
5	0.03	0.77	0.20	0.03	0.85	0.12

The VASE linear profiling across the area of interest was performed with a spatial step resolution of 2 mm. The spectroscopic data (ψ and Δ) were collected at each point and fitted with the same B-EMA model.

For Sample 1 consisting of only ITO NPs, the thickness profile resembles a pair of partially overlapped bell-shaped peaks. This could be caused by an offset of the sample rotation axis with respect to the nanoparticle beam, and/or a gas flow effect previously observed for deposition of Cu nanoparticles with GAS. The bell-shaped profile is clearly visible for all the samples.

The entire NP deposition pattern has a diameter of approximately 4 cm. The deposition of the SiO₂ film without NPs is much more uniform, where the difference in thicknesses between the center and the edge reaches 10% as a maximum. The edge of the measured area corresponds to the film without nanoparticles. Therefore, the points at 6 cm in Figure a show similar or slightly lower than the designed thicknesses of the matrix.

6.

(a) Thickness profiles measured with VASE in comparison with thicknesses obtained via SEM and fitting vis–NIR. (b) Filling factors of the same model from fitting VASE data.

As discussed above, the thickness was also estimated from the cross-section SEM images at the center of the samples. The SEM thickness measurements are in good agreement with data obtained from VASE (see the Figure a), as well as the obtained filling factors (see the Figure b).

The NP embedding results in the formation of pores which are propagated to the macroscale, at least within the NP deposition area. This limitation could be potentially overcome by employing less directional deposition techniques for the matrix fabrication, such as plasma-enhanced chemical vapor deposition (PECVD).

3.6. SWIR Transparency of ITO/SiO₂

In this section, we demonstrate that the tunable optoelectronic properties of ITO NPs can be used to fine-tune transparency in the SWIR range while maintaining visible transparency. This underscores the application potential of the produced nanocomposites.

In general, the absorption level of the coating can be tuned by adjusting the NP concentration, either through variations in coating thickness or NP loading. For example, Chang et al. report on self-assembled colloidal ITO NP coatings that achieve spectrally selective IR absorption (2–5 μm) by varying the number of deposited layers. Chen et al., showed that absorption of the nanocomposite films can be varied with different concentrations of the NPs. To further illustrate the possibilities discussed above, we use the B-EMA optical model introduced previously representing an ITO/SiO₂ composite on a glass substrate. The effective optical constants of ITO and SiO₂ matrix - namely, the refractive index (n) and extinction coefficient (k) - are extracted by fitting ellipsometric and spectrophotometric data. Figure S4 presents simulated transmission spectra of ITO/SiO₂ nanocomposite coatings with varying thickness (Figure S4a) and NP filling factor (Figure S4b) using the obtained effective optical constants for the mixture. Both increased coating’s thickness, and higher filling factors lead to enhanced plasmonic absorption due to greater optical density.

The optoelectronic properties of ITO thin films can be modified through postdeposition annealing. The plasma frequency of ITO is strongly influenced by the concentration of oxygen vacancies and the films’ microstructure. An improvement of the film conductivity and an increase of free charge carrier concentration can be attributed to improved lower electron scattering and an increase in oxygen vacancies, which results from the release of oxygen atoms from the film surface. As such, annealing in vacuum has been shown to help prevent additional oxygen incorporation into the films. The calculations performed for ITO/SiO₂ nanocomposite featuring ITO with various free charge carrier concentrations, show that carrier density not only amplifies the absorption peak but also shifts its position toward shorter wavelengths, reflecting the direct dependence of the plasma frequency on the density of free carriers (Figure S4c).

In nanoparticle-assembled films, however, the damping constant should be regarded as an effective parameter that reflects not only intrinsic charge carrier mobility within individual nanoparticles, but also microstructural factors such as interparticle contact resistance, percolation pathways, and void morphology. Annealing can promote partial densification of the nanoparticle network, improve electrical connectivity between neighboring particles, and modify the distribution of voids within the composite, all of which can reduce electron scattering and contribute to the observed decrease in γ. While the free charge carrier concentration remains the dominant parameter governing the LSPR spectral position and absorption strength, the reduction of the damping constant upon annealing is attributed to a combined effect of improved electronic transport across the nanoparticle network and microstructural reorganization, rather than to grain growth alone.

To demonstrate this experimentally, ITO/SiO₂ nanocomposite (Sample 3) coating was subjected to annealing in a N₂ atmosphere while in situ ellipsometry measurements were performed. Then, the obtained ellipsometric data were fitted with the previously described three-component B-EMA model. For each temperature step, the model film thickness and filling factors were kept constant as significant structural changes were not expected. The free charge carrier concentration N and the damping constant γ of the ITO Drude model were the only two parameters allowed to be fitted. The resulting dependencies of N and γ_on as a function of temperature are shown in Figure a.

7.

(a) Free charge carrier concentration and damping constant as functions of annealing temperature, measured by VASE and modeled using a B-EMA model for the ITO NP/SiO₂ composite (Sample 3). (b) Transmission spectra of a bare glass substrate and glass coated with (i) SiO₂-passivated ITO film and (ii) ITO/SiO₂ composite (Sample 3) before and after annealing at 350 °C in N₂ for 65 min. The shaded area indicates the quantum efficiency curve of the ZephIR SWIR camera sensor (Photon Etc., QC, Canada), adapted from ref . (c, d) Photographs of the same samples: captured with a standard visible-light camera (c) and a hyperspectral L-EOS imaging system with a ZephIR SWIR camera (d). Sample coding: bare glass (1/blue), SiO₂-passivated ITO film (2/magenta), ITO/SiO₂ before (3/orange) and after (4/red) annealing.

The observed increase in free charge carrier concentration with temperature up to 400 °C is consistent with previous studies on ITO annealing, , with free charge carrier concentration values ranging between N = 1 × 10²⁰ cm^–3 - 1 × 10²¹ cm^–3. Oxygen vacancies play a dominant role in the ITO electrical conductivity, each contributing two free electrons, whereas each incorporated Sn⁺ ion donates one free electron. We attribute the increase in charge carrier concentration to the formation of oxygen deficiencies and the activation of Sn⁺ dopants during the annealing process, which is also consistent with the literature. ,

The observed decrease in the damping constant is attributed to an improved mobility of the charge carriers. Previous studies have shown that annealing of the ITO films results in an increase of the grain size, and larger grains are associated with higher charge carrier mobility. For example, Lee et al. report that postdeposition annealing increases the grain size from 2.5 to 6.5 nm - comparable to the electron mean free path ranging from 4.1 to 6.7 nm. Although the charge carrier concentration remains unaffected, annealed films show reduced resistivity by about a factor of 3.

Annealing of the films also increase Hall mobility from ∼4.0 to ∼11.3 cm²/V·s. The improvements of the electron transport can be attributed to reduced grain-boundary scattering, as fewer grain boundaries per unit volume allow for less electron scattering. In ITO NP-assembled coatings, grain boundary scattering events can have even more significant effects on the electron transport. Gondorf et al. report a stark contrast between the electron mobility within individual NPs (22–35 cm²/V·s) and the macroscopic Hall mobility (0.12–0.7 cm²/V·s). This indicates that while electrons move efficiently within the crystalline domains, interparticle transport limits the conductivity, reducing it by roughly one to 2 orders of magnitude.

The changes in electronic properties are manifested through the optical response as shown in Figure b. Here, we compare the transmission spectra of a bare glass substrate (1) with glass substrates coated with a continuous ITO film (2) (Delta Technologies Ltd., CO, USA) and ITO/SiO₂ nanocomposite, both before (3) and after (4) annealing in N₂ according to the protocol described previously.

The bare glass remains transparent across most of the spectral range, with an absorption onset around 300 nm. The ITO film is transparent in the visible region but exhibits a reduced transmission in the NIR and SWIR due to increased reflectivity at longer wavelengths, consistent with the metallic-like behavior of ITO in the infrared range. In addition, the ITO-coated glass shows broad and uniform absorption in the NIR/SWIR region (see Figure S5) while the as-deposited ITO/SiO₂ nanocomposite displays a LSPR peak centered around 2250 nm. Annealing in N₂ enhances the amplitude and induces a blue shift of the LSPR peak.

To demonstrate the effects of annealing and plasmonic absorption enhancement, we imaged all four samples using both a standard visible-light digital camera and the L-EOS hyperspectral imaging system equipped with the ZephIR 2.5 SWIR camera. In Figure b, the shaded region represents the quantum efficiency curve of the ZephIR 2.5′s HgCdTe sensor, which maintains a spectrally uniform response with peak efficiency up to 85% and operates effectively within a spectral range of 850 to 2500 nm. Note that the imaging was performed in reflection mode, where light from the source passes through the samples and is then reflected back by a Spectralon pad positioned beneath them.

All four samples appear visually transparent when imaged with a standard digital camera, as shown in Figure c. The background pattern remains clearly visible through each coated glass window. However, imaging the same set of samples using the L-EOS hyperspectral SWIR camera at approximately 2000 nm (Figure d) reveals distinct differences: the bare glass retains its transparency (T _av = 91.9%) as its reflectivity and absorption remain constant, as expected; the ITO film exhibits a slight tint (T _av = 78.3%) due to increased reflectivity and absorption at longer wavelengths; the as-deposited ITO/SiO₂ nanocomposite shows moderate absorption (T _av = 64.6%); and the annealed nanocomposite becomes opaque at 2000 nm (T _av = 48.3%). This demonstrate that optical response of the coating can be tuned while using the same total amount of ITO.

Notably, SWIR imaging also uncovers inhomogeneities within the nanocomposite, attributable to variations introduced during the fabrication process (see Section 3.6). The T _av values provided in the parentheses are estimate values for the entire spectral range probed by the SWIR camera.

To illustrate the impact of annealing on transparency, we conducted hyperspectral imaging across the entire SWIR range. The resulting hyperspectral data cube was decomposed into a series of images at various wavelengths and assembled into an animated series (see Figure S6). These images reveal a gradual decrease in transparency for the ITO film, while the nanocomposite exhibits varying degrees of absorption before and after annealing, reflecting the slopes of their respective transmission curves.

4. Conclusions

The ITO/SiO₂ nanocomposites were prepared using a combination of a gas agglomeration cluster source and reactive magnetron sputteringan approach potentially compatible with modern optical coating technologies.

The microstructure and homogeneity of the resulting coatings were examined at both nano- and macroscopic scales using SEM and VASE techniques, respectively. The filling factors were estimated through structural analysis (SEM) and optical methods (VASE and vis–NIR spectroscopy), with both approaches showing strong agreement in trends and absolute values.

The resulting coatings exhibit multiscale porosity. The NPs create a shadowing effect on the underlying material, leading to void formation within the growing structure. While the relative porosity decreases with increasing NP content, even a coating in which ITO NPs constitute only 7% of the total material, the system retains approximately 25% of voids. Increasing the deposited SiO₂ content results in denser and more homogeneous coatings, where ITO NPs are distributed throughout the entire volume.

To describe the porous nanostructure of the film, we developed a multiscale ballistic growth model. While the model correctly captures general trends, it aligns with experimental data only qualitatively. This limitation arises from the fact that the model is two-dimensional and represents significantly thinner films, making it more suitable for understanding the deposition process rather than achieving precise quantitative agreement.

Optical properties of the prepared coatings in the vis and the SWIR/NIR ranges were investigated using vis-NIR spectroscopy. All samples exhibit an LSPR peak centered at 2500 nm while maintaining high luminous transparency and reduced reflectivity. The presence of pores leads to some level of scattering, which is primarily observed in the visible range and must be considered on an application-specific basis. On the other hand, porosity lowers the effective refractive index of the nanocomposite films, thereby reducing their luminous reflectivity.

The charge carrier concentration of ITO NPs was estimated to be 10²⁰ cm^–3, comparable to values reported in the literature for thin films. The damping constant was found to be 3.8 × 10¹⁴ rad/s, exceeding that of thin films due to the size-dependent dielectric response.

Finally, we demonstrate that the nanocomposite transmittance in the SWIR range can be tuned by adjusting the NP optoelectronic properties through annealing. The annealing in N₂ at 350 °C of the ITO NP reduces the average transmittance from 64.6% to 48.3% within the spectral range between 0.9 and 2.5 μm. These results highlight the potential of ITO NPs for applications requiring materials with independently controlled optical response in visible and SWIR regions (e.g., counterfeit protection).

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.5c12463.

• Larger TEM image of ITO nanoparticles, XRD pattern of ITO nanoparticles, refractive index and extinction coefficient of nanocomposites including sample 3, influence of different parameters on ITO LSPR, transmission and reflection spectra of ITO film, hyperspectral animation of ITO Nps and ITO thin film measured with SWIR camera (PDF)

†.

Present address: Department of Mechanics of Materials and Nanostructures, EMPA, Thun, 3600, Switzerland

The authors declare no competing financial interest.