Leveraging advanced CMOS processing, silicon photonics is emerging as a pivotal platform to meet the growing demand for lower power consumption and higher bandwidth within the communication band [1–5]. The field is now expanding into the mid-infrared spectrum [6–9], particularly targeting the 2–2.5-μm wavelength range. This range is critical for industrial and medical sensing applications due to the pronounced absorption lines of various gases (e.g., H2S, C2H4, CH4, CO, CO2, N2O) and biomolecules [10,11], making it highly suitable for gas sensing applications. For example, heterogeneous distributed feedback (DFB) laser arrays have been successfully employed for precise gas detection [12,13]. Spectroscopic sensors have demonstrated efficient spectral responses with low insertion loss (2–3 dB) and low crosstalk (20–30 dB) [14]. Moreover, the 2-μm waveband has potential applications in long-haul communications, efficient comb generation, and frequency doubling [15,16].

To date, key components such as waveguides [17], resonators [18], arrayed waveguide gratings [19], and modulators [20] operating at the 2-μm wavelength have been successfully demonstrated. However, significant progress is still required in the development of 2-μm on-chip lasers for applications such as multispecies trace gas spectroscopy and biomolecular detection, which enable high-precision, non-invasive measurements [21,22]. Currently, a mid-infrared laser predominantly utilizes GaSb and InP-based systems [23–26], each offering distinct advantages depending on the integration approach. InP-based lasers, for instance, offer a tunable range of up to 150 nm [27], but are typically limited to wavelengths around 2.3–2.4 μm due to constraints of type-II QW interband transitions [7]. While monolithic integration of InP lasers allows for compact designs, it suffers from high optical losses that significantly degrade performance [28]. On the other hand, GaSb-based lasers can extend into the 2.5–2.6-μm wavelength range. However, they typically require larger external cavity configurations, which complicates on-chip integration [29]. Hybrid integration, which optimizes chip performance and thermal management separately, remains a well-established approach for achieving high-performance lasers. A recent innovation in this field is the hybrid GaSb/Si3N4 lasers, which achieve a tuning range of 170 nm and a maximum power of 6.4 mW at room temperature. However, this approach still faces considerable thermal challenges due to the Si3N4’s low thermo-optic coefficient [30,31]. These limitations underscore the critical need for highly efficient and thermally optimized hybrid 2-μm laser solutions for integration on silicon photonics platforms.

In this paper, we present tunable lasers hybrid-integrated with silicon ring resonators, specifically designed for biomedical applications within the 2-μm range, and holding potentials for future heterogeneous integration using the same platform. Effective coupling between the QW reflective semiconductor optical amplifier (RSOA) chips and the silicon-on-insulator (SOI) waveguides has been achieved via wafer-level assembly and compact packaging. We demonstrate controlled wavelength tuning below 0.1 nm and an efficiency exceeding 1.2 nm/mW within a 25-nm tuning range, while maintaining a side-mode suppression ratio (SMSR) exceeding 40 dB. These features support multi-target detection and high-sensitivity measurements in biosensing applications.

Figure 1(a) shows a schematic of the hybrid tunable laser operating at 2 μm, integrating an RSOA within the passive silicon platform. The RSOA, shown in the top left of Fig. 1(a), has a total thickness of 4.2 μm and provides the necessary cavity feedback and amplification for laser operation. The device employs an edge-coupled ridge waveguide, chosen for its broader bandwidth and minimized reflections. These attributes are achieved through strategic waveguide angling and the application of anti-reflection coatings. In the bottom left of Fig. 1(a), the RSOA is wire-bonded to a thermally conductive substrate to form a chip-on-carrier assembly. This setup unifies the ground connection across all laser units for enhanced electrical stability. The gold-tin eutectic solder layer pre-deposited on the substrate surface facilitates the die-bonding process, improving both thermal and electrical conductivity. Adjacent to the laser module on the substrate, a fast-response glass bead thermistor is secured with thermally conductive aluminum nitride-filled epoxy, critical for temperature monitoring. Furthermore, a copper-tungsten heat spreader mounted to the base of an aluminum enclosure acts as a crucial element of the thermal management system. This spreader works in coordination with the thermoelectric cooler (TEC) to efficiently modulate the device’s operational temperature, ensuring optimal optical performance.

Figure 1(b) shows single-mode lasing achieved via the Vernier effect within a double-loop micro-ring resonator (MRR) system. The system is designed to offer a free spectral range (FSR) of 80 nm that exceeds the amplified spontaneous emission (ASE) of the gain chip. The ring resonators have radii of 130.69 μm and 128.43 μm [32]. A 2×2 Mach–Zehnder interferometer (MZI) broadband switch is fabricated by cascading two directional 3-dB couplers in a back-to-back configuration. The lengths of the MZI’s arms are precisely matched to maintain equilibrium. A two-level taper transitioning to an inverse taper effectively addresses the mode field size mismatch between the silicon waveguide and the RSOA. Furthermore, a calculated 15.4° angling of the waveguide relative to the polished facet reduces back reflections and boosts edge-coupling efficiency with the RSOA channel. Heater pads located on top of the rings, the MZI arm, and waveguide facilitate the tuning of the wavelength, mirror, and phase through the thermal-optic effect.

Figures 1(c) and 1(d) present effective mode confinement within both the RSOA and the silicon waveguide, achieving an edge-coupling efficiency exceeding 89%. Top-view scanning electron microscopy (SEM) images in Fig. 1(e) and Fig. 1(f) present the detailed design of the inverse taper and the ring coupler, respectively. The fabrication process of the passive silicon platform is shown in Fig. 1(g). The process begins with the etching of vertical channels to enable plasma-assisted wafer bonding with a 4-inch SOI wafer, which incorporates a 3-μm-thick SiO2 layer designed to minimize losses at the 2-μm wavelength. Following this, rib and strip waveguides are formed using CF4/SF6/Ar inductively coupled plasma (ICP) etching, reaching depths of 231 nm and 500 nm, respectively. The process concludes with the deposition of a 1-μm-thick SiO2 layer over the waveguides for protection, followed by the deposition of probe pad metals (Au/Ti) and heater metals (Pt/Ti), which are essential for electrical connectivity and thermal tuning within the device.

Figure 2(a) shows the setup for characterizing passive devices. In Fig. 2(b), the left part shows the modulation depth of the ring resonator, measured by an arbitrary waveform generator (AWG). The right part of Fig. 2(b) presents response waveforms at various AWG frequencies, showing a 3-dB modulation bandwidth of 66 kHz. Figure 2(c) illustrates a peak ring resonance with a cavity Q of approximately 1×105 and an extinction ratio (ER) of 11.6 at a wavelength near 2.020 μm. Figure 2(d) validates the effectiveness of side-mode resonance suppression, assuming a theoretical coupling coefficient (κ) of 5%. The crosstalk is presented in Figs. 2(e) and 2(f) between two resonators, Ring1 and Ring2, where crosstalk shifts have been efficiently mitigated for precise wavelength tuning to sub-0.1-μm accuracy. Figure 2(e) shows the spectral shifts of all rings following a 20-mW heating of Ring1, with shifts of 0.38 nm and 0.04 nm for Ring1 and Ring2, respectively. Figure 2(f) examines the resonance shift across all rings due to the heating of Ring1, showing a linear relationship between crosstalk and heating. The crosstalk impact is roughly an order of magnitude smaller for the non-heated rings compared to the primary ring. This characteristic is crucial for systems where multiple resonators are closely packed but must operate independently without interference.

Figure 3(a) presents the transmission characteristics during coarse tuning, achieved through the red shift by Ring1 and blue shift via Ring2. A consistent coarse tuning efficiency of 1.18 nm/mW for both shifts is observed upon altering power to one ring, as detailed in the inset of Fig. 3(a). Figure 3(b) illustrates the transmission spectrum of the fine-tuning mode. By simultaneously heating both rings, fine adjustment to the peak shift is achieved with precision as small as 0.02 nm. The inset of Fig. 3(b) demonstrates a linear model to estimate power effect on the wavelength shifts during fine tuning, with an estimated sensitivity of approximately 19.4 pm/mW.

Figure 4(a) shows the measurement system of the hybrid laser and Fig. 4(b) depicts the schematic of the edge-coupling process used to integrate the RSOA with the silicon waveguides. Thermal sweeping is conducted using a customized probe card positioned on the heater pads. Figure 4(c) presents a single-mode lasing spectrum with an SMSR exceeding 40 dB. This is achieved by fine-tuning the heaters to align the Vernier scale of Ring1 and Ring2 to 0.17 mW and 1.43 mW, respectively. The light-current-voltage (L-I-V) characteristics, depicted in Fig. 4(d), are measured by incrementally increasing the injection current while maintaining the heater powers at the initial levels. Although the emission wavelength remains around 2.02 μm, fluctuations in the injection current induce temperature variations that affect the effective cavity length due to thermo-optic effect. This leads to phase mismatches that require periodic realignment for optimal phase congruence. Each injection current setting requires precise calibration of the laser’s phase shifter to optimize the output power and align with the maximum reflectivity of the Vernier configuration. This optimization produces a stable L-I-V curve [Fig. 4(d)], characterized by a threshold injection current of 92 mA and a slope efficiency of 5.83 mW/A. The roll-off in power beyond 400 mA, attributed to thermal effects, highlights the importance of thermal management in maintaining laser performance. Good metal contacts are evidenced by a differential series resistance of approximately 4 Ω and a turn-on voltage of 0.6 V, confirming the effectiveness of the electrical interfaces in the device.

Figure 4(e) shows the tuning characteristics of the laser, beginning with the fine-tuning of the phase shifter and the MZI heater to maximize the output power at a stable injection current of 200 mA. Once optimal power levels are achieved, power adjustments to the ring resonator are made to facilitate precise wavelength tuning. For fine tuning, increasing the heater power simultaneously in both rings achieves an efficiency of 20 pm/mW. During tuning, it is crucial to control the phase shift of the laser cavity and maintain TEC temperature stability to prevent mode-hopping, as the longitudinal modes are highly sensitive to temperature fluctuations. Smooth and stable tuning without mode-hopping can be achieved by gradually adjusting the phase shift while ensuring constant TEC regulation, thereby maintaining mode stability throughout the tuning process. For a coarse red shift, precise power allocation is crucial, with Ring1 requiring 4 mW and Ring2 needing 2 mW for each tuning step. Conversely, coarse blue shift tuning requires an inverse adjustment in the heating mechanism. By combining these two tuning methods, a wide range of precise wavelength control can be achieved. Figure 4(f) shows the laser’s robust tuning capability, maintaining single-mode operations with an SMSR exceeding 40 dB across a 25 nm span. An essential factor that impacts laser performance is the coupling loss between the chips. The presence of an air gap between the chips, which integrates into the laser cavity, may increase cavity losses through butt-coupling. Proper chip alignment is crucial to minimize this coupling loss, which significantly impacts the laser’s threshold and output power. Fiber-chip coupling losses, influenced by the efficiency of edge-coupling calibration, typically range from 3 to 7 dB. In addition, thermal expansion from injection current can disturb the well-aligned coupling channel, and stage vibrations may further exacerbate these effects, posing challenges in maintaining consistent and precise coupling alignment. These factors underscore the importance of careful thermal and mechanical design in the optimization of hybrid silicon photonics devices.

The 25-nm tuning range is primarily constrained by the RSOA’s gain bandwidth of ∼40nm, restricting spectral coverage. Additionally, the low coupling ratio of the ring resonator impacts output power and tuning efficiency, while thermal crosstalk between the Vernier rings may introduce minor instabilities. To address these limitations, employing RSOAs with broader gain bandwidth and optimizing coupling ratios could extend the tuning range, albeit potentially at the expense of SMSR. Additionally, advanced packaging techniques could improve coupling stability and minimize mechanical drift, pushing the tuning range closer to the theoretical limits of the silicon waveguide platform.

Table 1 summarizes the performance of state-of-the-art tunable lasers operating in the 2–3-μm region. Our hybrid tunable laser achieves the highest tuning efficiency compared to other technologies, offering precise control over the tuning interval—essential for high-accuracy biosensing applications. Additionally, the laser design ensures a stable output power with a low threshold current of 92 mA, enhancing both energy efficiency and operational stability. Unlike the commonly used 220-nm SOI waveguides, our platform utilizes a thicker 500-nm silicon layer. This thickness improves the index matching with the III-V layers, facilitating efficient and low-loss optical mode transfer between the active (III-V) and passive (silicon) layers. Consequently, our platform offers a robust design space for the development of future Si/III-V heterogeneous photonic devices.Table 1.

Literature Overview of the State-of-the-Art Mid-infrared Lasers^a

In this work, we demonstrated a 2.0-μm hybrid tunable laser integrated on a silicon platform, extending the operational wavelength for applications spanning sensing, environmental surveillance, medical diagnosis, etc. Through the strategic optimization of multi-ring silicon resonators, the laser achieves a tuning range over 25 nm and maintains an SMSR of over 40 dB. Its capacity for ultra-precise tuning down to tens of picometers and efficient power utilization broadens the applicability of silicon photonics within the 2.0-μm band, facilitating scalable, cost-effective solutions in these critical areas. Moving forward, our efforts will concentrate on refining the design of Vernier rings to mitigate thermal crosstalk, improving packaging techniques to optimize coupling efficiency, enhancing the gain spectrum of the RSOA design to expand the tuning range, integrating on-chip wavelength division multiplexing networks [34] to enhance scalability, and adopting a heterogeneous integration approach to create more compact, multifunctional PICs [1,35].

Acknowledgment. The authors are grateful to Xiangpeng Ou and William He for helpful discussions and the support from the nanofabrication facilities at the University of California Santa Barbara.