Cascaded Directional Coupler-Based Triplexer Working on Spectroscopically Relevant Wavelengths for Multiple Gas Detection

Abstract

In this article, we present experimental and simulation results of a novel high-performance cascaded directional coupler-based triplexer. The device is designed to combine the wavelengths of 1530 nm, 1653.7 nm, and 2003 nm for use in spectroscopy systems targeting the detection of ammonia, methane, and carbon dioxide gases, respectively. The triplexer’s functions focus on enhancing the coupling efficiency and selectivity, while facilitating the on-chip integration of diode lasers. The experimental results demonstrate that the coupling efficiency is 82%, 73%, and 91% for the respective wavelengths of 1530 nm, 1653.7 nm, and 2003 nm. The results highlight the triplexer’s capability as a multifunctional beam combiner and an adaptable power source, essential for advanced gas sensing techniques and integrated couplers.

1. Introduction

Optical gas sensors can detect greenhouse gases and air pollutants by measuring the unique absorption spectra of target molecules. The near-infrared (NIR) wavelength range is suitable for the detection and quantification of key gas molecules due to their unique absorption features, which can be effectively analyzed using various spectroscopic techniques [1,2]. Various gas analyzers are available in the market, utilizing spectroscopic techniques such as photoacoustic, tunable diode laser spectroscopy (TDLS), cavity ringdown, thermal FTIR, and comb sources [3,4,5,6,7]. Photoacoustic-based gas absorption techniques, such as Quartz Enhanced Photoacoustic Spectroscopy (QEPAS), unlock real-time in situ measurements with a response time > 1 s and offer high gas specificity. They are capable of part-per-quadrillion (ppq) detection sensitivity [8,9]. This technique utilizes a quartz tuning fork, known for its stability and precision, and is independent of wavelength. This enables the detection of unique absorption features of gas molecules at their respective wavelengths across a broad spectrum, ranging from the ultraviolet to the terahertz spectrum [10,11,12]. However, these devices often face limitations in terms of tradeoffs between size, weight, sensitivity, selectivity, sensor lifetime, calibration, and performance in harsh environments. A Vernier effect-based quantum cascade laser (QCL) multi-gas detection system using QEPAS is demonstrated for detecting Carbon Monoxide (CO), Nitrous Oxide (N₂O), and Carbon Dioxide (CO₂), with absorption peaks at 4708 nm, 4520 nm, and 4457 nm, respectively [13]. This system has restricted spectral selectivity and free-space optical configurations are prone to misalignment due to positioning issues with mirrors and lenses. Portability remains a primary challenge, especially for wide-area sensing in agricultural or remote environments. Improving these features with a focus on lightweight, compact, and cost-effective designs for gas-sensing technologies remains crucial. Optical fiber-based solutions on the one hand offer benefits in terms of compactness and lightweight designs. On the other hand, challenges are still faced related to material selection, the need for collimators, dispersion, and nonlinearity, and absorption losses due to interactions with optical phonons and inherent fragility and complicated deployment in harsh environments [14,15,16,17].

Recent advancements in integrated photonics and semiconductor laser technologies facilitate the miniaturization of gas sensors, enabling the integration of lasers and optical components in a miniaturized, chip-scale design. Specifically, silicon nitride (Si₃N₄)-based chip-scale devices offer a wider bandwidth and lower power consumption, paving the way for more efficient, multispectral, and cost-effective gas sensing solutions. QEPAS can benefit from the use of on-chip laser-integrated triplexers, enabling compact, lightweight multi-gas sensing. This technology can be integrated into unmanned aerial vehicles for wide-area gas sensing, such as monitoring ammonia, methane, and carbon dioxide in agricultural and industrial applications. The use of on-chip triplexers can potentially provide cost advantages over fiber-based multiplexors as hybrid integration of lasers will ultimately avoid the need for pigtailing.

While various couplers are available for telecommunications applications, they often lack the bandwidth required for gas absorption wavelengths. Several passive optical components, including Directional Coupler (DCs) [18,19,20], Mach–Zehnder interferometers (MZIs) [21], arrayed waveguide gratings (AWGs) [22], and multimode interference (MMI) couplers [23], enable multiplexing functionalities. However, in MMI-based structures, wavelength multiplexing requires finding a common self-imaging point that suits the three different wavelengths, where implementing angled multimode interference (AMMI) leads to significant losses caused by asymmetry in the multimode waveguide [24,25]. Among these options, the DC stands out as particularly promising due to its compact footprint, high reliability, and minimal insertion loss [19,26]. It handles multiple channels and is compatible with other multiplexing devices.

This work presents a novel, compact, high coupling efficiency (CE) and selective Si₃N₄-based triplexer suitable for sensing systems for the detection of ammonia (NH₃), methane (CH₄), and carbon dioxide (CO₂) in the NIR spectral region. This approach exploits the distinct absorption lines of the selected gases at specific wavelengths: λ_NH3 = 1530 nm, λ_CH4 = 1653.7 nm, and λ_CO2 = 2003 nm. The device is based on a simple coupled mode mechanism that employs DCs for wavelength combining. The numerical investigation focuses on maximizing CE for all three wavelengths, which is achieved through a detailed assessment of the design parameters, ensuring that the lengths of the three channels are a common multiple of the coupling lengths for the three wavelengths. The integration of laser diodes via flip-chip or transfer-printing techniques will enable the development of a robust, self-contained chip capable of generating multiple wavelengths, thereby enhancing gas sensing performance [27,28].

2. Methods and Designs

When two or more waveguides are placed in close proximity (a few hundreds of nanometers), as shown in Figure 1a, they couple through their evanescent fields. This allows power to be exchanged between waveguides as the guided modes propagate along them. Using coupled-mode theory, the relationship between coupling length and optical transmittance can be determined, assuming the modes in each waveguide remain approximately unchanged in the absence of the other [29]. The coupling only affects the amplitude of these modes, without altering their transverse spatial distribution or propagation constant (β) [30]. The coupling length (L_c) can be calculated as follows:

$$L_c={\displaystyle \frac{{\lambda }_0}{2\Delta n_{eff}}}$$

(1)

where L_c is the distance at which the power is fully transferred to the adjacent waveguide. This process depends on the wavelength of light λ₀, and the change in effective refractive index (Δn_eff) of the coupled waveguides, which is further influenced by the gap (g) between the waveguides. The waveguide dimensions are designed to effectively confine the modes of micro-on-chip diode lasers; furthermore, the fundamental mode of each waveguide is considered in the simulation.

Multiplexer device configurations are realized based on the Si₃N₄ on silicon dioxide (SiO₂) platform. Si₃N₄ is highlighted as a promising material for integrated photonics due to its compatibility with CMOS technology, low optical losses (0.04 dB/cm), wide transparency window (0.25–3.5 µm), achieved through precise control of material deposition percentages, low thermo-optic coefficient (~10⁻⁵/°C), relatively low nonlinear coefficient (2.1 × 10⁻¹⁸ m² W⁻¹), high power handling capability (>100 W), and potential for 3D integration [31,32,33]. Also, comparing the bend and propagation losses among various materials, such as silicon (Si) and indium phosphide (InP), Si₃N₄ offers significant advantages [34].

The structures are designed using a 300 nm thick Si₃N₄ layer on an 8 µm buried oxide (BOX) platform with a 1 μm thick (SiO₂) upper cladding layer. Figure 1a shows a schematic diagram of the triplexer design, outlining the arrangement of the positioning of the diode lasers, S-bends, the coupling region, and the output waveguide with a width of W_o. The triplexer configurations are introduced by considering CE and bending loss. As the wavelength becomes longer, the loss due to bending diminishes [35]. The bend radius in the design is set to R_arc = 5000 µm, with an angle α = 14°, ensuring sufficient space for integration of the micro-on-chip lasers and lower bend loss light propagation. The input and output waveguides have a width of W = W_o = 3 μm, maintained across the S-bend, coupling section, and output waveguide. The coupling lengths L_c1, L_c3l, and L_1-3 are defined at the coupling region, where the gap between all the waveguides is constant and denoted as g.

The triplexer design was performed using the 3D Finite Differences Beam Propagation Method (FD-BPM) [36,37]. The computational mesh grid resolution of the simulation domain was optimized along the three spatial axes of X = 100 nm, Y = 100 nm, and Z = 50 nm. The dispersion formula was used to determine the refractive indices of Si₃N₄ and SiO₂ [38,39]. The tapering of the end of the waveguide is introduced for two reasons: firstly, to avoid potential excitation of higher-order modes, and secondly, to prevent the reflection of light back to the source due to the Si₃N₄-SiO₂ interface at the end of each waveguide. This reflection can lead to undesirable losses and impacts on the overall performance of the triplexer system [40]. The tapering length was optimized to t_p = 10 µm, which reduces the reflection to 1%. The design for Port 1 is modeled to guide the wavelength λ_NH3, with Port 2 serving as the pathway for the wavelength λ_CH4. The electric field profile of each wavelength at the respective ports is depicted in Figure 1b. Figure 1c shows the transmittance spectra for the optimized triplexer, with distinct peaks Port 1 and Port 2, while also giving a broad spectral range via Port 3. Port 1 exhibits a primary peak at 1530 nm, followed by a transmittance peak at 1750 nm. Similarly, Port 2 has a primary transmittance peak at 1670 nm and another 82% transmittance peak at 1395 nm, which aligns with the water vapor (H₂O) absorption line. The bandwidth and free spectral range are both measured at 100 nm for Port 1 and Port 2. The device achieves a maximum transmittance of 90%, 91%, and 88% for wavelengths λ_NH3, λ_CH4, and λ_CO2, respectively.

3. Fabrication and Experimental Setup

A 300 nm thick layer of Si₃N₄ on an 8 µm BOX on a Bulk silicon wafer was purchased from LIONIX International BV. The device was patterned using the processes of electron-beam lithography and dry etching, involving several steps (the detailed fabrication process is discussed in ref. [41]). The fabrication process exhibits a tolerance of ±25 nm of the gap (g) from the designed mask file to the final sample. In this study, we used the actual sample dimensions after fabrication as our experimental parameters. Following inspection with a scanning electron microscope (SEM), a 1 µm thick SiO₂ layer was deposited using plasma-enhanced chemical vapor deposition (PECVD). Various configurations of the device were investigated and implemented to enhance both the fabrication process and transmission performance. The fabrication specifics of the triplexer samples are illustrated in Figure 2, which includes details of the fabricated DC triplexer, featuring a tapered tip of the waveguide corresponding to L_c1, L_c3l, and L_c4, designed to minimize reflections that could potentially destabilize the laser. A straight waveguide with dimensions equal to that of the input/output waveguides of the triplexer (i.e., 3 µm wide and 0.3 µm thick) was fabricated on the same wafer.

The optical characterization of the fabricated samples was performed using a free-space optics measurement setup described in Figure 2d. Each device port was measured separately. The light source was focused on input ports and the output (at W_o) was collected using lensed fibers. Two different lensed fibers were used during characterization: the first, made from SMF-28, for input wavelengths of 1530 nm and 1653.7 nm, while the second covered the wavelength of 2003 nm. For characterizing Port 1 and Port 2, which correspond to λ_NH3 and λ_CH4, respectively, a single-wavelength (Yenista Optics, Singapore) tunable laser source (TLS) was employed. Once the alignment was completed, the TLS was replaced with a broadband light source (Amonics ALSD, Phoenix, AZ, USA) (1500 nm–1660 nm). The resulting output spectra were analyzed with a power meter and an optical spectrum analyzer (OSA).

In the second configuration, dedicated to the characterization of λ_CO2, the sample was initially aligned with a 1630 nm TLS source, to which the IR camera is relatively sensitive. This IR camera was employed to ensure precise alignment between the triplexer sample’s input/output and the lensed fiber, which is fundamental to these measurements. The TLS was then replaced with a 2 µm diode laser. The wavelength and power of the laser are controlled by the laser driver. The sample was tested using a lensed fiber suitable for the 2 µm range and the output beam was collected at W_o using another 2 µm compatible lensed fiber. The output was measured using a photodiode connected to an oscilloscope.

4. Results

4.1. Optimization of Coupling Length (Lc₁) for λ_NH3 and λ_CH4

The primary focus of the design optimization is to achieve high coupling efficiency for λ_NH3 and λ_CH4, ensuring maximum transmittance at the output for both wavelengths. The critical parameters influencing this process are g and L_c1. As the propagating electric field profile shown in Figure 1b indicates, light introduced at Port 1 propagates through an S-bend waveguide and couples into the Port 2 waveguide at every 330 µm of coupling length. Similarly, light launched at Port 2, corresponding to λ_CH4, transfers its energy completely at every 250 µm of coupling length, (see Equation (1)). After reaching three coupling lengths for λ_NH3 and four coupling lengths for λ_CH4, both wavelengths converge at a common length, forming a devoid-of-light region. This region can be strategically used to efficiently couple λ_CO2 and further prevent the interaction of λ_NH3 and λ_CH4 at the L_1-3.

4.1.1. Simulation Results

A computed fundamental mode corresponding to each wavelength (λ_NH3 and λ_CH4) was launched at the input Ports 1 and 2, respectively. A pathway monitor was placed at the W_o and was normalized to the launch power. The analysis primarily examines the effect of changing g by adjusting the distance between Port 1 and Port 2, evaluated as a function of the wavelength launched at Port 1. The transmittance spectra corresponding to the g variations, with a fixed length of L_c1 = 1030 µm, L_c3l = 200 µm, and L_1-3 = 160 µm, are illustrated in Figure 3a. The solid line indicates the transmittance for the wavelength emitted from Port 1, while the dotted line represents the transmittance for the wavelength emitted from Port 2. The spectator demonstrates a redshift of 35 nm per increment of 25 nm in the g dimensions for the wavelength emitted from Port 1. Likewise, a 25 nm g alteration results in a similar 35 nm shift in the spectra for the wavelength emitted from Port 2. Conversely, the transmittance spectra for changes in length L_c1 for both laser Ports 1 and 2 were analyzed with a fixed g = 300 nm, L_c3l = 200 µm, and L_1-3 = 160 µm, as illustrated in Figure 3b. The spectra show a blueshift as L_c1 increases, shifting by 15 nm for every 30 µm change in length.

The design is highly sensitive to the value of the gap, with changes in the order of tens of nanometers having a significant effect on transmission, potentially resulting in losses exceeding 25%. The optimal g-value is determined to be 300 nm by considering the high transmittance for both wavelengths, the region is highlighted with vertical blue and red boxes in Figure 3a. On the other hand, changes in L_c1 for the waveguide exhibit a relatively flat response to variations in transmittance within the range of tens of micrometers. In simulations, L_c1 is optimized to 1030 µm at a fixed g-value of 300 nm. After optimization, numerical simulations achieve a transmittance of 90% for a wavelength of 1530 nm and 91% for a wavelength of 1653.7 nm.

4.1.2. Experimental Results

The fabricated sample collection includes samples with varying values of g, while keeping constant L_c1 = 1030 µm and L_c3l = 240 µm. Additionally, the collection includes samples with varying L_c1 values, with a fixed g = 275 nm and L_c3l = 180 µm. The measured values were normalized relative to straight waveguides (mentioned in Section 3). The normalized data were further refined using a smoothing spline with a smoothing parameter of 0.82. This process was employed to extract the underlying trend in the spectra while minimizing the impact of noise introduced during normalization, alignment, and source. The spectral shift for g = 275 nm and g = 300 nm at a fixed L_c1 = 1030 µm was analyzed by considering a broadband excitation at the input ports of the triplexer, as depicted in Figure 4a, showing normalized transmittance. Modifications to the g-value result in a redshift of 60 nm (65 nm) when exciting Port 1 (Port 2). For a 300 nm gap at Port 2, the output wavelengths exceed the operational range of the Amonics ALSD. The target wavelengths are marked with blue and red vertical boxes. The normalized transmittance obtained for these gap values was 45% and 31%, respectively. As shown in Figure 4b, Port 1’s spectrum exhibited a blueshift with increasing length L_c1, with peak shifts of 25 nm for every 30 µm increment, consistently with simulation results. For Port 2, the selected spectra were obtained for L_c1 values of 1000 µm and 1030 µm. Further decreasing L_c1 would result in output wavelengths exceeding the operational range of the Amonics ALSD, which is typically around 1660 nm. The distance between the transmittance peaks was measured at 100 nm. Similarly, the wavelength emitted from Port 2 also demonstrated a blueshift of 25 nm for every 30 µm change in length, maintaining a peak-to-peak separation of 100 nm.

The change in L_c3l is linearly affected with the transmittance of λ_NH3 and λ_CH4. As illustrated in Figure 4b, an optimal configuration was identified with L_c1 set at 1000 µm and g of 275 nm and reduced L_c3l to 180 µm. The influence of cascaded coupling length L_c3l is an important parameter, as it also serves as the path for λ_CO2. The transmittances obtained for λ_NH3 and λ_CH4 were 82% and 73%, respectively.

4.2. Optimization of Coupling Length (L_c3l) for λ_CO2

The triplexer design is optimized for the coupling length L_c3l to ensure high coupling efficiency across all three wavelengths. The wavelength λ_CO2 is injected at Port 3. To prevent interactions with the S-bend in the pathway following Port 3, a cascaded waveguide section with length L_c3l is introduced. This length is carefully selected not only to maintain efficient coupling for λ_CO2 but also to avoid crosstalk with the other wavelengths, λ_NH3 and λ_CH4, which effectively minimizes the coupling to the L_c3l and L_c4 waveguides. The value of L_c3l corresponds to the coupling length for λ_CO2. The coupling length of the waveguide for L_c3l is determined by calculating the coupling length within a four-waveguide coupled system. The pathway for this wavelength is structured to extend from L_c4 through Port 3, proceed to L_c3l, traverse the L_c1 waveguide, and finally return to W_o, as shown in Figure 1b. Simulation and characterization results for varying coupling lengths L_c3l with λ_CO2 = 2003 nm launched at Port 3 are shown in Figure 5. The plot exhibits a bell-shaped trend as L_c3l varies from 30 µm to 400 µm. The dashed lines show the simulation results for changes in the g-value with respect to the coupling length L_c3l. As the g-value increases in increments of 25 nm, the maximum transmittance shifts by a length of 20 µm in L_c3l. This effect is attributed to the changes in the effective refractive index (n_eff) of the coupled device. Transmittance improves with increasing g-values, since the fixed (L_c1 = 1000 µm) is not optimal at lower g-values. The device achieves its highest transmittance from simulation of 88% at the optimized coupling parameters of L_c3l = 200 µm, L_1-3 = 180 µm, and g = 300 nm.

The normalized characterization results, relative to the straight waveguides (mentioned in Section 3), are presented for a set of triplexer devices in Figure 5, with a red curved solid line “o” marker. These results correspond to a range of L_c3l values, from L_c3l = 30 μm to L_c3l = 400 μm, with fixed values of g = 275 nm and L_c1 = 1000 μm, optimized for Port 3 to target λ_CO2. The configuration of the 2003 nm diode laser characterization approach from Figure 2b is used here. The device exhibits a greater sensitivity to L_c3l in experimental conditions than in simulations, with a tolerance of 25 µm in the full width at half maximum of L_c3l. The optimal transmittance is achieved at L_c3l equal to 180 µm, yielding a measured transmittance of 91% at a gap of 275 nm, in good agreement with the simulated maximum transmittance.

5. Discussion

The core of the design focuses on maintaining the critical parameter—the gap between the waveguides—and optimizing the coupling length. To achieve this, the waveguides are fixed at a width of 3 µm, ensuring the phase-matching condition necessary for efficient evanescent field coupling. The transmittance obtained from all characterizations determines the overall coupling efficiency of the triplexer. Furthermore, the simulation results align well with the characterization trends, validating the design approach. The development of high-resolution EBL has significantly enhanced the device’s repeatability and reliability by ensuring precise nanoscale patterning and minimizing fabrication-induced variations. The estimated propagation loss is 4 dB/cm at 1550 nm and further decreases with an increase in n_eff. Optical losses observed during measurements can be attributed to several practical reasons. These include the difficulties associated with effectively coupling light into the chip’s input port, using a free-space end-fire setup with lenses to align the optical fiber mode with the waveguide mode. Additionally, fabrication defects, unwanted reflections within the waveguide structure resulting from cleaved facets, and the attenuation of higher-order modes in the tapering region further exacerbate these losses. Collectively, these elements account for the losses documented in the measurements. The emerging challenge is the precise integration of the chip-scale diode laser into the specified ports on the chip.

6. Conclusions

In this work, a triplexer based on a compact directional coupler structure has been extensively studied, demonstrating enhanced power coupling ratios specifically designed for the absorption wavelengths of ammonia (λ_NH3 = 1530 nm), methane (λ_CH4 = 1653.7 nm), and carbon dioxide (λ_CO2 = 2003 nm). The different coupler configurations were investigated taking into account both bend losses and coupling losses. Numerical analysis revealed high coupling efficiencies for the fundamental modes: 90% at λ_NH3, 91% at λ_CH4, and 88% at λ_CO2. Experimental measurements further validate these findings, demonstrating coupling efficiencies of 82% at λ_NH3, 73% at λ_CH4, and 91% at λ_CO2. This improvement in coupling efficiency is critical for enhancing the sensitivity and selectivity of trace gas detection. The use of the triplexer will allow a single spectroscopy system to detect multiple gases, eliminating the need for multiple single-wavelength systems. The hybrid integration of lasers with the triplexer will provide improved compactness. This novel approach streamlines the detection process and improves the overall performance and adaptability of gas-sensing applications.