Robust SiON rib waveguides for CMOS-compatible surface plasmon resonance sensing of aqueous analytes

1 Introduction

Silicon oxynitride (SiON) and silicon nitride (Si₃N₄) have emerged as highly promising material platforms for compact, integrated waveguide systems. The SiON platform is particularly advantageous because it enables continuous refractive index tuning between the values of SiO₂ (n ∼ 1.45) and Si₃N₄ (n ∼ 2.0) allowing waveguide parameters to be precisely optimized (Ye et al., 2024; Piccoli et al., 2022; Podlucký et al., 2021). By adjusting the oxygen to nitrogen ratio during fabrication, the optical confinement can be precisely tailored through the refractive index contrast between the core and cladding. Compared to high-index platforms like silicon-on-insulator (SOI), the SiON platform offers relatively large mode field diameters, leading to lower propagation losses and more efficient coupling to standard optical fibers, while remaining fully compatible with conventional CMOS manufacturing (Podlucký et al., 2021; Feiler et al., 2023; Sudas et al., 2022; Lockwood and Pavesi, 2011). Plasma-enhanced chemical vapor deposition (PECVD) is widely used for deposition of SiON layers due to its low-temperature requirements and high deposition rates, facilitating the growth of thick, low-stress material layers (Podlucký et al., 2021; Snyders et al., 2023; Su et al., 2020). Due to their low intrinsic optical losses and wide transparency windows, SiON and Si₃N₄ are also utilized for label-free optical biosensing in the visible (VIS) and near infrared (NIR) spectral ranges. Effective biosensing relies heavily on maximizing the interaction between the guided optical mode and the surrounding environment. While this light-matter interaction can be achieved through various waveguiding configurations, it can be significantly enhanced by leveraging plasmonic effects.

The plasmonic sensors cover wide range of applications (Bing et al., 2024; Cennamo et al., 2017; Almeida et al., 2022; Mahmud et al., 2023; Yesudasu et al., 2021), providing real-time detection with high sensitivity, low cost, immunity to electromagnetic interference and label-free analytical biosensing (Feiler et al., 2023). Surface Plasmon Resonance (SPR) is a highly effective method for characterizing the optical properties of various materials and biomaterials. This phenomenon is based on the resonant coupling of electromagnetic waves to collective charge-density oscillations at metal–dielectric interfaces, commonly referred to as surface plasmon polaritons (SPPs). SPR-based waveguide structures are typically realized using configurations such as metal-insulator-metal (MIM)(Butt, 2023), metal-dielectric-metal (MDM)(Zhang et al., 2022), SOI(Kumari et al., 2022), or optical fibers (Wei et al., 2017). These structural approaches significantly influence sensor performance, which is determined by both material properties and waveguide geometry (Divya et al., 2022). Consequently, the performance of these integrated sensors becomes heavily dependent on the evanescent field interaction at the interface between the target analyte and a deposited thin metal film. While gold (Au)(Almeida et al., 2022; Islam et al., 2021; Zhang et al., 2017), silver (Ag)(Cennamo et al., 2017) and copper (Cu)(Mishra et al., 2016) are the most common plasmonic metals, gold is frequently preferred due to its superior chemical stability in aqueous environments and its inherent biocompatibility. SPR sensors can be implemented through various geometrical platforms, including optical fiber configurations (Islam et al., 2021; Hoseinian and Agha Bolorizadeh, 2019), prism based techniques (Hossain et al., 2022; Fang et al., 2023; Shibayama et al., 2021) and integrated waveguide structures (Cennamo et al., 2017; Suryanarayana et al., 2022). In the category of integrated waveguides, specific topologies such as strip/channel (Feiler et al., 2023; Mishra et al., 2016), slab (Cennamo et al., 2017; Walter et al., 2020), slot (Heydari et al., 2020), rib (Almeida et al., 2022) and photonic crystal (Jain et al., 2022; Srivastava et al., 2011) waveguides are utilized to optimize light-matter interaction.

The most recognized conventional methods for SPP excitation are the Kretschmann and Otto configurations utilizing prism coupling (Shibayama et al., 2021; Chlpík et al., 2023). In waveguide-based SPR sensors the dielectric waveguide acts as the light-delivery medium. The guided modes within the dielectric core generate an evanescent wave that penetrates the thin metal layer deposited on the waveguide surface. This mechanism mimics the Kretschmann configuration, where the phase-matching condition between the guided light and the surface plasmons is achieved at a fixed effective incidence angle determined by the waveguide’s propagation constant (Homola, 2008). Since the resulting SPR signal is highly sensitive to the input and output coupling efficiencies, ensuring precise and stable alignment is critical for achieving reproducible and accurate spectral characterization. To confine and guide light within these SPR-based architectures, several coupling techniques commonly used for dielectric waveguides can be employed, including edge coupling (Mishra et al., 2016; Walter et al., 2020), grating coupling (Cheng et al., 2017) and prism coupling (Saravani and Mohammadkhani, 2022; Bonneville et al., 2021) configurations. Among these, edge coupling (butt-end coupling) provides a direct and robust approach for physical characterization, and is the specific methodology implemented in this study to evaluate the fabricated sensor structures.

There is significant ongoing interest in developing SPR-based waveguide sensors using alternative material platforms to expand their practical application potential. Recent studies have investigated optimal SPP excitation for single-mode SiON rib waveguides (Almeida et al., 2022), alongside the impact of gold-layer configuration for photonic crystal fiber sensors (Jain et al., 2022) and Kretschmann configurations (Chlpík et al., 2023). The proposed waveguide sensor builds upon the design principles described by Feiler et al. (2023), which explored multimode SiON structures with various refractive-index configurations and gold-layer parameters using numerical modeling. In multimode structures, the SPR-induced transmission dip is typically broader than in single-mode counterparts due to the convolution of multiple propagating modes, each associated with a distinct incidence angle at the metal interface (Arcadio et al., 2023).

In this paper, we present the optimized design and experimental validation of a robust SiON rib waveguide SPR sensor. By combining experimental characterization with numerical FEM study, various aspects of proposed SiON sensor platform are evaluated and verified. The unique adjustability of the SiON material platform with rib structures provided suitable basis for determining the optimal metal layer thickness for plasmonic phenomena to occur. The subsequent study conducted on the rib structures pointed to the high degree of sensor robustness and suitability for lab-on-a-chip applications in the realm of waveguide SPR sensors.

2 Plasmonic properties of rib waveguide structures and FEM simulations

Surface plasmon resonance (SPR) based waveguide sensors are focused on detecting variations in spectra by using various aqueous analytes with different refractive indices. The operating principle of SPR sensors is dependent on the resonant coupling between the waveguide guided modes and SPR mode. The adjustability of effective refractive index of the SiON material platform offers highly tailored sensing for aqueous analytes in both the visible and near infrared wavelength ranges (Piccoli et al., 2022; Podlucký et al., 2021).

A numerical calculation, to obtain values of initial SPP excitation wavelength ranges, was done for SiON core layer (n _*SiON* = 1.54), using various gold layer thicknesses and different aqueous analytes, which defined the effective refractive index between the analyte and the gold layer. This calculation utilized the refractive index value of waveguide core material along with the effective refractive index value of the SPP as used by Almeida et al. (2022), which was obtained using the following equation:where ε _*m* is the complex permittivity of metal and ε _*d* is the complex permittivity of analyte. The results of this analytical approach provided baseline estimation of the simulation bounds for subsequent numerical FEM analysis.

n e f f s p p = ε m × ε d ε m + ε d

To optimize the sensor design, a two-dimensional (2D) cross sectional Finite Element method (FEM) study was performed on the proposed SiON rib waveguide structure featuring a thin gold layer for SPP excitation. The cross-section detail of the sensor is shown in Figure 1. The structure contains a thermal SiO₂ buffer layer (∼2 µm) on a silicon substrate and SiON rib waveguide with a slab thickness H _*slab* = 5.6 µm, a total rib height H _*rib* = 8 µm (etch depth H _*rib* - H _*slab* = 2.4 µm), rib width W = 10 µm. A thin gold layer (20–40 nm) is placed on top of the waveguiding SiON layer. Specifically, the refractive index of the SiON waveguide core (n _*SiON* = 1.543, λ = 633 nm) was obtained using a Cauchy equation model based on post-fabrication ellipsometry measurements. Similarly, the gold layer was simulated using a complex refractive index n = 0.183 + 3.433i at λ = 633 nm. The gold layer was modeled using tabulated optical constants from Johnson and Christy (1972). Since the permittivity of thin polycrystalline gold films deviates from bulk tabulated values and depends on film thickness and morphology below approximately 50 nm (Yakubovsky et al., 2017) the simulated gold thickness was treated as an effective fitting parameter rather than a fixed nominal value. To enhance computational efficiency while maintaining accuracy, the fully cladded input and output section of the rib structure were excluded from the simulation domain. The presented model focuses exclusively on the etched sensing window where the gold is exposed, as the SPP excitation is confined strictly to the interface between the gold layer and analyte. The rest of materials used in the simulations were set to correspond to the thermal SiO₂ (n = 1.45) buffer and analytes matching refractive indices of common aqueous analytes (n = 1.33–1.47). The FEM simulations were done for both polarizations of the fundamental (zero-order) mode. While higher-order modes were also present in the simulation results, they primarily propagate within the slab part of the rib waveguide structure and do not significantly contribute to the SPP excitation. Therefore, the spectral analysis focuses exclusively on the fundamental mode results.

The primary parameter calculated from the simulation of guided modes was the complex effective refractive index, over a free-space wavelength range of 550 nm–915 nm. For the SPP excitation, the imaginary part of effective refractive index (Im _*neff*) is critically important as it represents the absorption of the TM polarized light coupled into the surface plasmon polariton. These imaginary components of the effective refractive index for various gold layer thicknesses and dimethyl sulfoxide (DMSO) as analyte are shown in Figure 2A) for TE and Figure 2B) for TM polarization.

Since the imaginary part of effective refractive index (Im _*neff*) is not directly measurable, the simulation results must be converted into total transmission, which can be measured at the chip output. The obtained simulation values for TM and TE modes were used to calculate the absorption coefficient α for each polarization by using the formula:where Im _*neff* represents the imaginary part of the effective refractive index obtained from the simulation results, and λ is the wavelength. The total transmission loss over a sensor length l = 1000 µm for one mode is determined using exponential power loss equation:

α = 4 π × Im n eff λ

T = e − α l

However, since the experimental setup used broadband light source without a polarizer, capturing both polarizations simultaneously, the simulated TE and TM spectral contributions were combined and averaged (50:50) to provide a more accurate representation of the measured data. Consequently, an accurate correlation with the unpolarized experimental data necessitates an average of both modal contributions, even though the TM response is numerically dominant in defining the resonance depth and the spectral position of SPP. The combined calculated transmission results for 30 nm gold layer thickness are shown in Figure 3A. In the simulation, the sharp resonance dip occurs exclusively for the TM mode at the SPP resonant wavelength of 724 nm. The corresponding electric field distributions are shown in Figure 3B for TE mode outside of SPP resonant wavelength (600 nm) and Figure 3C for the TM mode at the SPP resonant wavelength (724 nm).

The combined transmission losses provided initial gold layer thickness parameters for experimental sample fabrication and detailed the spectral behavior of the proposed waveguide structure in the presence of DMSO as analyte.

3 Fabrication process of SiON rib waveguide structures

The SiON rib waveguide platform was selected based on numerical optimizations and its suitability for in-house cleanroom processing, facilitating potential integration into lab-on-a-chip applications. The target refractive index for SiON core layer (n _*SiON* = 1.54) was achieved by precisely varying the relative flow rates of precursor gases during PECVD (Plasmalab 80plus, Oxford Instruments) process. Specifically, the flow rates were set to 200 sccm for silane (SiH₄), 500 sccm for nitrous oxide (N₂O), and 600 sccm for ammonia (NH₃).

The SiON rib waveguides were fabricated on a standard 100 mm polished silicon wafer (thickness of 525 µm), featuring a 2 µm thick thermal SiO₂ buffer layer. A nominally 8 µm thick SiON core layer was deposited by PECVD method. This specific layer thickness was based on prior numerical optimization by Feiler et al. (2023), to maximize mode matching for efficient light coupling with a standard optical fiber. Additionally, this thickness supports the desired robust profile of the waveguide, which enhances the overall light throughput and signal detection capabilities of the structure. To define the SiON waveguide structure, a 150 nm copper hard mask was deposited using physical vapor deposition (PVD, Kurt J. Lesker PVD75+). AZ 701 MIR positive photoresist was spin-coated followed by UV exposure through the photolithography mask and development in AZ 400K developer. The rib waveguides were formed by etching the SiON layer by approximately 30% (2.4 µm) of its initial thickness using an inductively coupled plasma/reactive ion etching (ICP-RIE, Oxford Plasmalab 100, Oxford Instruments). The waveguide structure was set to width of 10 µm. This specific etching profile was selected based on previous numerical optimizations by Feiler et al. (2023). FEM modal analysis shows that higher-order modes propagate primarily within the slab region, leaving the fundamental TM mode as the primary cause for SPP excitation at the rib interface.

Following the fabrication process, the SiON rib waveguide structures were passivated with a 2 µm SiO₂ cladding layer. To enable SPR signal measurement, sensing windows were fabricated by etching through the passivation layer to expose the waveguide core for metal deposition. Gold was selected as the plasmonic material due to its robust chemical stability, inherent biocompatibility, and suitability for aqueous analytes measurements. Although silver can theoretically provide sharper resonance peaks, it is highly prone to oxidation and corrosion in aqueous sensing environments. The thin gold sensing layer was deposited onto the exposed sensing window by PVD method (Kurt J. Lesker PVD75+). Finally, the input and output facets of the chip were polished using diamond lapping papers, which enables direct coupling and outcoupling of light to/from optical fibers.

4 Characterization of SiON rib waveguide structures

Following the fabrication of SiON rib waveguide structures, a dedicated measurement workstation was constructed for direct edge coupling characterization. Precise fiber to sample alignment was achieved by incorporating a top-view camera combined with three axis piezoelectric stages. The optical fiber, coupling light from a broadband deuterium-halogen source (Avantes AvaLight-DH-S), was aligned with the polished edge of the waveguide core. For each measurement, proper light propagation was verified to ensure the signal coupled strictly through the waveguide structure, eliminating undesired paths or cladding crosstalk. The outcoupled signal was collected by multimode optical fiber connected to a spectrometer (Ocean Optics USB4000) and the spectral data were recorded using a SpectraSuite software. A simplified schematic of this experimental measurement configuration is shown in Figure 4.

The experimental investigation focused on validating the variation of the SPR spectral dip position, which depends on thickness modification of thin gold layer. Based on previous experimental observations (Chlpík et al., 2023) and results from numerical models, the thin metal layer thickness directly shifts the resonance wavelength within the transmission spectrum. To validate these results, experimental measurements were performed using samples with nominal gold thicknesses of 20 nm, 30 nm and 40 nm (nominal values are deposited by PVD and regulated by Quartz crystal monitor during deposition) in presence of DMSO(n = 1.47) as analyte. These experimental results were subsequently cross-referenced with the results of 2D FEM numerical simulations, as shown in Figure 5A.

Preliminary evaluation indicates a strong spectral agreement between the experimental data and the numerical model, specifically for the nominal 30 nm gold structure. However, to account for minor geometric variations and fabrication intolerances in the remaining curves, a thickness trimmed analysis was performed. As shown in Figure 5B, adjusting the simulated gold thickness to 25 nm and 35 nm yielded a significantly improved spectral match with the measurement curves. This adjustment confirms that a slight variation in local metal thin film morphology during the deposition accounts for the minor variations observed in the nominal 20 nm and 40 nm experimental curves.

Following the thickness study, the sensor refractometric response was evaluated for measurements of various refractive indexes in the range of 1.33–1.38 in six steps, which are typical for biological and chemical analytes. For this characterization, a gold layer thickness of 40 nm was selected to maintain a stable plasmonic response across the entire target refractive index sweep. The liquid analytes consisted of volumetric mixtures of demineralized water (n = 1.331) and isopropyl alcohol (IPA, n = 1.376). The corresponding simulated spectral curves are shown in Figure 6A, while the experimental measurements are presented in Figure 6B. Experimentally, the resonance wavelength shifted from approximately 619.2 nm for pure demineralized water up to 627.1 nm for pure IPA. To quantify the refractometric capability of the device, the sensor sensitivity S was determined using the following equation:where Δλ represents the shift in resonant wavelength and Δn is the change in the refractive index of used analyte. Based on the simulation results obtained within the refractive index range of water-IPA (1.33–1.38), the sensor demonstrated average sensitivity of 158 nm/RIU ranging from 80(water) up to 350 (IPA) nm/RIU increasing with the analyte refractive index as shown in Figure 7.

S = Δ λ Δ n n m / R I U

The analysis of resonant wavelength peaks obtained from measured data did not yield to reliable information. The highest uncertainty of measured dip positions was ±1.75 nm and results in LOD 33 × 10^-3RIU. The whole computation process can be found in Supplementary Material.

The acquired experimental transmission spectra exhibit spectral broadening and a slightly lower resonance depth compared to the idealized numerical models. This behavior is likely attributed to scattering losses induced by surface roughness at the SiON/gold interface, as well as non-ideal fiber to chip coupling facets. Furthermore, variations in the local morphology of the thin gold film can dampen the SPP effect, leading to broader spectral results (Treebupachatsakul et al., 2021).

To mitigate the high frequency experimental noise while preserving the true resonance minima position and shapes, all acquired transmission spectra were post processed using 50-point (10 nm) Savitzky-Golay smoothing filter with second polynomial order. All experimental measurements were performed under stable laboratory conditions at a room temperature of approximately 21 °C within the 550–915 nm spectral range.

The gold thickness study was conducted by repeated deposition and stripping of gold layers with differing thicknesses (20, 30 and 40 nm), using DMSO as the analyte to induce resonance. The gold thickness interrogation resulted in reproducible attenuation of total transmission across the spectrum. The subsequent study using water-IPA solutions, was done on another rib waveguide sample to characterize the sensor’s response to refractive index change. The gold thickness used for this study was 40 nm. Both studies used the same spectrometer settings in terms of integration time (20s) and averaging (10 times).

To achieve a direct quantitative comparison between experimental results and numerical models, the resulting transmission curves were normalized from counts to relative units, to the correct influence of light source and waveguide transmission.

The close alignment between the observed experimental dip positions and the simulated plasmonic trends strongly suggests the successful excitation of surface plasmon modes within the robust SiON rib waveguide structures. Using only TM polarized source light should theoretically isolate highly responsive TM components, which should lead to sharper dips. Unfortunately, implementation of such source in this specific test setup introduced severe insertion losses that drastically degraded the outcoupled signal to noise ratio.

The fabricated structures were inspected by using various surface characterization methods. In Figure 8 are shown scanning electron microscopy (SEM) images of A) the sensing region, and B) the polished cross-section with the measured dimensions. The measured, etch depth of ∼2.4 µm matches the nominal value, whereas the slab thickness of ∼5.38 µm and the rib width of ∼11.3 µm deviate from the design values of 5.6 µm and 10 μm, corresponding to a total SiON thickness of ∼7.8 µm instead of the nominal 8 µm. Owing to the large cross-section, the effective index of the fundamental mode is only weakly affected by sub-micrometer-scale geometric deviations, and the resonance condition is governed primarily by the core refractive index and the gold–analyte interface, so these deviations are not expected to shift the resonance position significantly. The images also revealed a sidewall residue of SiO₂ resulting from an incomplete etching of the cladding layer prior to gold layer deposition. To evaluate the impact of this residue, the 2D FEM simulation model was updated to incorporate this non-ideal sidewall geometry. The updated numerical analysis indicated that the presence of this residue had a negligible effect on both the overall transmission results and the SPP excitation efficiency, because the primary TM polarized mode features electric field vectors that oscillate vertically, and the light interaction occurs at the interface between top gold layer and analyte. Thus, the results and the model are not depicted. This geometric tolerance highlights the structural resilience of the sensor design under standard fabrication conditions.

5 Discussion

The numerical and experimental results demonstrate the viability of the silicon oxynitride (SiON) platform for integrated surface plasmon resonance waveguiding. Two-dimensional FEM modal analysis shows that the fundamental mode acts as the primary mode for light-matter interaction within the rib waveguide structure. While the TM polarization component is mostly responsible for generating SPP response, the total optical throughput is combined for both polarizations due to the broadband, unpolarized light source used in the experimental setup. Higher-order modes were found to propagate predominantly in the slab region rather than in the rib. Owing to their weak lateral confinement, their field is spread across the wide slab, so the local field intensity at the gold–analyte interface is low and their coupling to the SPP is inefficient; the same delocalization also makes these modes more susceptible to scattering losses. Consequently, they contribute a broadband background to the transmitted signal but do not take part in the plasmonic sensing response.

The transmission dips that were observed for different nominal gold layer thicknesses (20,30 and 40 nm) show that thin metal layer thickness changes the resonant wavelength spectral positions as shown in Figure 9A. Slight discrepancy between expected numerical results and measured SPR dip positions might be caused by deviations in deposited thickness of gold layers. The thickness fitting of numerical model shows agreement with trimmed values of gold layer thickness for 25 nm and 35 nm obtained from experimental data shown in Figure 9B. The gold thickness in the numerical model was subsequently adjusted to obtain the best spectral match with the experimental curves. This adjustment should not be interpreted as a correction of the deposited film thickness alone. For polycrystalline gold films thinner than ∼50 nm, the imaginary part of the dielectric function increases markedly with decreasing thickness, and the optical constants depend on the film thickness and morphology even for a nominally identical deposition process (Yakubovsky et al., 2017). The tabulated bulk-like values used in the simulation therefore underestimate the losses of the thinnest films. The fitted thickness thus acts as an effective parameter that jointly compensates for this thickness-dependent permittivity and for the deposition tolerance, bringing the model into closer agreement with the physical behavior of the fabricated films.

A key advantage of the SiON platform lies in its refractive index tunability, which allows the core index to be matched to the target analyte range. This is directly reflected in the measurements: the resonance contrast and dip depth were significantly stronger for the DMSO analyte (n = 1.47) than for the water and IPA analytes, which showed shallower dips in normalized transmission. This is attributed to the higher refractive index of DMSO providing a closer match to the SiON core (n = 1.54), resulting in more efficient evanescent coupling and a stronger SPR response. The same trend holds across the water-IPA mixtures range (n = 1.331–1.376). However, the averaged sensitivity of 158 nm/RIU was calculated from simulation results over range (1.33–1.38). The sensitivity is increased with higher analyte refractive index, reaching from 80 (water) up to 350 (IPA)nm/RIU as shown in Figure 7. This is consistent with initial premise that bringing the analyte refractive index closer to the one of core strengthens the plasmonic interaction and increases sensitivity. To further verify the effectiveness of core matching, average sensitivity of the rib sensor with varying core RI was investigated using FEM simulations shown in Figure 10.

However, this matching cannot be pushed arbitrarily. As the analyte index approaches the core, the resonance dips broaden and shift further along the spectrum, eventually limiting reliable detection. This is precisely where the tunability of SiON becomes advantageous: rather than being fixed, the core refractive index itself can be tailored during fabrication to be optimal to the targeted analytes range, keeping the resonance sharp and within the measurable spectral window. This confirms the suitability of SiON as a material platform and demonstrates its potential for further development of SiON-based photonic integrated circuits for refractive-index sensing.

The proposed SiON rib waveguide sensors are compared against alternative SPR sensor configurations across four critical criteria: spectral sensitivity, fabrication complexity, features sizes and material availability. With the SiON rib waveguide platform presented in this work achieved averaged sensitivity 158 nm/RIU, scaling from 80 (water) up to 350(IPA) nm/RIU increasing with the analyte refractive index. Crucially, this performance is achieved while maintaining full compatibility with conventional optical lithography and standard CMOS processes, such as PECVD.

When compared to the sensors operating within similar micrometer scale geometries, the proposed design exhibits distinct performance advantages. For example, SiON rib waveguide structure outperforms conventional anti-resonant reflecting optical waveguide (ARROW) sensor, which require similar fabrication workflow but yield slightly lower sensitivity of 75 nm/RIU (Asha et al., 2022). Similarly, a comparison with silica based slab waveguide utilizing localized gold sphere arrays reveals that the presented sensor achieved higher performance than the theoretically reported sensitivity of 80 nm/RIU (Gu et al., 2020), without requiring specialized nanoparticle deposition tools.

The structural and manufacturing advantages of the SiON rib structure become more evident when contrasted with sub-micron and nanostructured configurations that yield comparable or slightly elevated sensitivities. Compared to hybrid silicon nitride (Si₃N₄) ridge waveguides integrated with gold nanorods reported a sensitivity of 250 nm/RIU (Chamanzar et al., 2013), while slot waveguide ring resonators has achieved sensitivity of 212 nm/RIU (Barrios et al., 2007). However, both designs require low-throughput, high-cost electron beam lithography (EBL) to define critical features, such as 200 nm thick ridge lines or narrow slot gaps. Additionally, the slot ring resonator demands low-pressure chemical vapor deposition (LPCVD) to achieve single mode tolerance. In contrast, robust SiON rib design utilizes a simple, single step thin film deposition, bypassing advanced nanopatterning, while relaxing fiber coupling precision requirements.

Finally, different material configurations, such as Silicon Carbide (SiC) SPR sensor augmented with 10 graphene layers can achieve high average sensitivity of 2290 nm/RIU (Du et al., 2021), trading off physical scalability and structural robustness. The SiC platform depends on thin 100 nm core layer and the precise, intensive transfer of multiple graphene sheets, which introduces severe fabrication risks, susceptibility to defects and calibration complexities.

In summary, while specialized nanophotonic or multi-layered structures can achieve higher sensitivity, they suffer from strict sub-micron manufacturing processes. Although our comparison states the sensitivity value as the chosen metric, we are aware it does not fully represent the sensor. However, it’s consistently reported within the cited sources. The table with additional information was added as Supplementary Material. The proposed SPR rib waveguide sensor offers balanced approach in strong, dynamically scalable sensitivity while offering simplified fabrication process, material longevity and compatibility with existing CMOS foundry lines for easier lab-on-a-chip integration.

The fabricated SPR rib waveguide sensor could be improved by reducing the overall roughness of SiON layer, therefore increasing sensitivity. Higher surface roughness lowers the sensitivity of the SPR sensor (Agarwal et al., 2016). A simulation analysis of surface roughness profiles for SPR sensors by Treebupachatsakul et al. (2021) shows that even small surface height changes could affect sensitivity and the SPR dip of the sensor. Optical contrast was lower when the surface roughness height was over 8 nm and the sensitivity increased dramatically with smoothness of simulated layer. Additionally, the gold layer mimics the surface of underlying SiON layer, therefore decreasing the overall surface roughness could lead to more accurate measurement results. The gold layer thickness could be further optimized during the PVD process to be more precise. From the perspective of measurement setup, experimental measurements could be improved by using a stronger light source with polarization controller or passive polarization optics. Furthermore, the chip design could be optimized by integrating polarization-control elements, such as multimode interferometers (MMIs), or efficient polarization-selective coupling schemes, such as grating couplers. However, even with this factor the robust SiON rib waveguide SPR sensors were able to be utilized in spectral sensing of various commonly used liquid analytes.

The numerically calculated data and measurement results show changes in spectral positions, for different gold layer thicknesses. Even though the numerical fitting analysis of gold thin layer showed agreement within the numerical and experimental results, additional characterization is required to prove these differences, which could be attributed to experimental and fabrication-related factors. These include surface roughness, non-ideal fiber to waveguide coupling and outcoupling, general fabrication intolerances and small signal to noise ratio during the measurements.

To summarize, the results of spectral measurements validate the adjustability of the SiON platform by tailoring the core refractive index to be closer to the target analyte. The sensitivity increases as the refractive index of analyte rises, which was experimentally evaluated.

6 Conclusion

This study investigated the design, fabrication and experimental characterization of a silicon oxynitride (SiON) rib waveguide platform optimized for integrated SPR sensing. The SiON platform was chosen specifically for its CMOS compatibility and its refractive index matching capability, which allowed the core refractive index to be tailored to n = 1.54 to optimize light-matter interaction for aqueous analytes.

The presented research validates the sensing mechanism through two scientific approaches. First, numerical FEM simulations provided a robust framework for determining optimal gold layer thicknesses, which was experimentally confirmed through transmission measurements with a dimethyl sulfoxide (DMSO) analyte. Second, the sensor’s refractometric simulated sensitivity was evaluated using water-isopropyl alcohol (IPA) mixtures, resulting in averaged sensitivity 158 nm/RIU, changing from 80 (water) up to 350 (IPA)nm/RIU. The sensor performance is higher for measurements using DMSO analyte over the standard demineralized water and IPA mixtures analytes. This increase in sensor performance might be caused by the strong nonlinearity in sensitivity, which rises with higher refractive index of analyte compared to the refractive index of SiON waveguide core. While minor spectral discrepancies between numerical simulations and experimental results were observed due to fabrication tolerances and local thin-film morphology, these were successfully addressed through a thickness fitting analysis, confirming the reliability of our numerical model.

In summary, the SiON rib waveguide platform offers an effective compromise between high performance and practical implementability. The results confirm the presence of surface plasmon resonance and demonstrate the practical viability of the SiON platform for integrated sensing applications. Moreover, the index tunability and CMOS-compatible fabrication of the SiON platform are assets that extend beyond plasmonic sensing, to integrated photonic sensing applications in general. While specialized nanophotonic structures can achieve higher theoretical sensitivity, the proposed sensor delivers a versatile, tunable, durable, and reproducible solution suitable for integration into lab-on-a-chip systems for various sensing applications. Future work will focus on further optimizing the SiON material properties and waveguide geometry to maximize effective refractive index matching specific biological and chemical targets, thereby expanding the potential of the SiON platform for diverse sensing applications.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/aot.2026.1936319/full#supplementary-material

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Footnotes

Edited by: Daniel Benedikovic, University of Žilina, Slovakia

Reviewed by: Jan Latal, VSB-Technical University of Ostrava, Czechia

Marcin Lelit, Warsaw University of Technology, Poland