The past decade has witnessed tremendous development of silicon nitride (${{\rm Si}_3}{{\rm N}_4}$) in photonic areas, with innovations in nonlinear photonics [1], optical sensing [2], etc. However, the lack of a fully integrated laser with high performance prohibits the large-scale integration of ${{\rm Si}_3}{{\rm N}_4}$ waveguides into complex photonic integrated circuits (PICs). All existing active applications still require off-chip laser sources or butt-coupled gain medium with low-volume production, large coupling loss and high cost of packaging [3–5].

A heterogeneously fully integrated ${{\rm Si}_3}{{\rm N}_4}$ laser uses compound waveguide coupling through hybrid evanescent modes within the heterogeneous layers, or via efficient mode transitions between the compound waveguiding layers including ${{\rm Si}_3}{{\rm N}_4}$. A ${{\rm Si}_3}{{\rm N}_4}$ cavity-based laser would benefit from the low propagation loss and low thermal sensitivity of ${{\rm Si}_3}{{\rm N}_4}$ waveguides. Low wavelength variation versus temperature distinguishes a ${{\rm Si}_3}{{\rm N}_4}$ Bragg grating-based laser from a laser coupled to an external ${{\rm Si}_3}{{\rm N}_4}$ waveguide.

However, in current mature heterogeneous integrated lasers working around 1550 nm, the III-V epitaxial layer has $\sim 2\;{\unicode{x00B5}{\rm m}}$ thickness and a slab mode refractive index around 3.2, while the refractive index of ${{\rm Si}_3}{{\rm N}_4}$ is around 2. Even extreme tapering of the thick III-V epitaxial layer is unable to facilitate efficient mode coupling between them within a ${\rm III} {\text -} {\rm V}/{{\rm Si}_3}{{\rm N}_4}$ structure. To overcome this, we demonstrate a ${\rm III} {\text -} {\rm V}/{\rm Si}/{{\rm Si}_3}{{\rm N}_4}$ laser structure using multilayer heterogeneous integration that employs multiple mode transitions; from a gain section III–V/Si hybrid waveguide, transitioning to a Si waveguide, through a ${\rm Si}/{{\rm Si}_3}{{\rm N}_4}$ transition to the ${{\rm Si}_3}{{\rm N}_4}$ waveguide. As illustrated in Fig. 1(a), the ${{\rm Si}_3}{{\rm N}_4}$ passive layer is deposited and processed first while the Si layer and subsequent III–V epitaxial layer are transferred on top of the ${{\rm Si}_3}{{\rm N}_4}$ waveguides via wafer bonding and processed at the backend. The required silicon dioxide (${{\rm SiO}_2}$) cladding for a low loss ${{\rm Si}_3}{{\rm N}_4}$ waveguide consists of a lower thermal oxide cladding and an upper cladding of deposited spacer oxide, together with a VIA oxide layer which is also used for the laser passivation. Figure 1(b) shows a three-dimensional (3D) schematic of the laser. A 1.5 mm long hybrid section with indium phosphide based multiple quantum well (InP MQW) on Si provides the optical gain while the laser mirrors are formed by a narrow-band ${{\rm Si}_3}{{\rm N}_4}$ spiral shaped distributed Bragg grating reflector (DBR) and a broadband tunable Si loop mirror on the two ends. Details of the InP/Si hybrid section design can be found from previous works [6]. Formed by circular ${{\rm Si}_3}{{\rm N}_4}$ posts placed along the curved waveguide [Fig. 1(c)], the spiral DBR length is 20 mm within a footprint of only ${3.5\;{\rm mm}} \times {3.6\;{\rm mm}}$ using a low loss ${{\rm Si}_3}{{\rm N}_4}$ waveguide (2.8 µm wide and 90 nm thick). The grating period is 526 nm, providing a reflection peak near 1550 nm. The gap between the grating post and waveguide is increased from 920 nm to 944 nm along the spiral to keep the grating unchirped during the decrease of spiral waveguide radius. Another waveguide spiral with its radius decreasing from 100 µm to 40 µm is used as a low-reflection waveguide radiation terminator. With a larger spiral design this terminator could be replaced by a spiral waveguide with opposite curvature to form a laser output for light passing through the grating. The inset picture taken by an infrared (IR) camera shows the ${{\rm Si}_3}{{\rm N}_4}$ spiral grating radiation during lasing. A dual-level Si taper with low reflection is used to adiabatically couple the Si waveguide fundamental transverse electric (TE) mode to the ${{\rm Si}_3}{{\rm N}_4}$ waveguide fundamental TE mode through the hybridized ${\rm Si}/{{\rm Si}_3}{{\rm N}_4}$ fundamental TE mode in the taper area, over a length of 200 µm [Fig. 1(d)]. A thermo-optic tuner is placed between the ${\rm Si} {\text -} {{\rm Si}_3}{{\rm N}_4}$ taper and the InP/Si gain section to enable in-cavity phase tuning. The laser output is taken after the Si loop mirror for characterization. SEM images in Fig. 1(e) show a cross-sectional view of the hybrid InP/Si region, hybrid ${\rm Si}/{{\rm Si}_3}{{\rm N}_4}$ region and tilted top view of the InP-Si taper.

 

Fig. 1. Schematics of (a) multilayer heterogeneous integration, (b) ${\rm III} {\text -} {\rm V}/{\rm Si}/{{\rm Si}_3}{{\rm N}_4}$ laser (c) ${{\rm Si}_3}{{\rm N}_4}$ spiral DBR followed by a ${{\rm Si}_3}{{\rm N}_4}$ waveguide terminator, (d) ${\rm Si} {\text -} {{\rm Si}_3}{{\rm N}_4}$ taper. (e): SEM image of fabricated laser.

Download Full Size | PDF

Figure 2(a) shows the continuous-wave (CW) light–current–voltage (LIV) characteristics of the III-${\rm V}/{\rm Si}/{{\rm Si}_3}{{\rm N}_4}$ laser operated at 20°C stage temperature. The lasing threshold is 75 mA with a peak on-chip output power over 0.5 mW for 320 mA gain current. The differential resistance is about 2.5 Ω. The lasing peak wavelength during the LIV sweep is also recorded using a wavemeter. As the gain current increases, the output power and peak wavelength see several discontinuities following the same trend evidencing a ‘cycled’ mode hop, which is typical for a DBR laser. The mode hop behavior as well as power fluctuation mainly depends on the in-cavity optical phase. At 190 mA gain current, the power and peak wavelength show strong hysteresis with phase tuning [Fig. 2(b)].

 

Fig. 2. (a) LIV characteristics and corresponding peak lasing wavelength. (b) Laser power and peak lasing wavelength hysteresis dependence on the phase current. (c) Single-mode optical spectrum with gain current of 160 mA, inset shows measured normalized reflection spectra of the ${{\rm Si}_3}{{\rm N}_4}$ spiral grating. (d) Lasing wavelength dependence on stage temperature. (e) Laser frequency noise.

Download Full Size | PDF

The ${{\rm Si}_3}{{\rm N}_4}$ spiral grating provides a narrowband filter together with high extinction ratio. This results in a large lasing side mode suppression ratio (SMSR) of over 58 dB [Fig. 2(c)].

InP lasers and InP/Si lasers are normally quite temperature sensitive as they both have large thermo-optic coefficients (dn/dT). By comparison, the thermo-optic coefficients of ${{\rm Si}_3}{{\rm N}_4}$ and ${{\rm SiO}_2}$ are an order of magnitude smaller. Thus, a laser cavity based on a ${{\rm Si}_3}{{\rm N}_4}$ waveguide DBR will be far less sensitive to temperature variations than one based on Si or InP. We compared our laser with an extended-DBR InP/Si laser with a 15 mm long Si Bragg grating and the results are shown in Fig. 2(d). With a temperature change from 10°C to 55°C, the wavelength shift of the ${\rm InP}/{\rm Si}/{{\rm Si}_3}{{\rm N}_4}$ laser is only 0.47 nm (10.46 pm/°C), while the InP/Si laser wavelength shift is 3.3 nm (73.18 pm/°C), over ${7\times}$ difference.

We measured the frequency noise of our laser as shown in Fig. 2(e). With no tuning of the Si loop mirror, it provides $\sim{0.3}$ power reflectivity. The lowest obtained white-noise-limited frequency noise level is about ${2000}\;{{\rm Hz}^2}/{\rm Hz}$, giving a Lorentzian linewidth of 6 kHz. This is further reduced to $\sim{1300}\;{{\rm Hz}^2}/{\rm Hz}$ and 4 kHz respectively by tuning the Si loop mirror reflectivity to maximum to increase the photon density in the laser cavity. In this fabrication process run, the spacer oxide thickness is thinner than designed after planarization, resulting in a large mode overlap with the VIA oxide, leading to a propagation loss of 0.43 dB/cm, which can be reduced to around 0.001–0.01 dB/cm in the future [7]. The further reduced loss will enable lasers with a longer ${{\rm Si}_3}{{\rm N}_4}$ spiral grating and weaker coupling constant $\kappa$ to significantly reduce the laser linewidth and increase the output power [5,6,8]. We anticipate this new ${\rm III} {\rm -} {\rm V}/{\rm Si}/{{\rm Si}_3}{{\rm N}_4}$ heterogeneous platform can take narrow-linewidth semiconductor lasers to a new level.

The successful demonstration of multilayer heterogenous integration is a key step towards fully exploiting the capabilities of wafer bonding technology. Our approach could enable a new class of devices that require low loss waveguides and are presently not fully integrated, including optical clocks, optical gyroscopes and ultra-narrow-linewidth Brillouin lasers.