1. Introduction

Silicon is a powerful platform for high performance photonic devices, and high quality factor silicon ring resonators are of particular interest for nonlinear optical applications due to silicon’s high intrinsic nonlinearity [1, 2]. However, in order for silicon resonators to provide an ideal platform for nonlinear phenomena, high quality factor is in itself insufficient: silicon resonators must also have high modal confinement and small ring size.

Ridge etched waveguides have low losses, but present low lateral confinement of the optical mode [3], which requires resonators to have large bending radii to avoid significant radiation losses [4]. In contrast, etched strip waveguides present high lateral confinement of the mode, which enables resonators to have small bending radii down to 5 microns [5]. However strip waveguides suffer from high propagation losses in bends (typically 2–4 dB/cm in the highest quality factor etched ring resonators [6]). Therefore neither the etched ridge waveguides nor the etched strip waveguides can serve as an ideal platform for high quality factor and at the same time high confinement resonators.

Ring resonators based on etched silicon high confinement waveguides are inherently limited in quality factor due to the imperfections introduced by the etching process, with the highest demonstrated Q of 420,000 [5]. Plasma etching introduces roughness into the waveguide side-walls, and can also induce absorption sites into the silicon [7] that will increase losses [8, 9].

Optical losses in high confinement resonators can in principle be mitigated by using a non-etching process to create the waveguide sidewalls. Of particular potential is the use of thermal oxidation to form silicon waveguides and resonators. The highest quality factors achieved in silicon ring resonators have been accomplished using etchless processes [10, 11]. To date, there have been two types of etchless ring resonators demonstrated: ridge waveguides made with the etchless process [10], and very thin etchless waveguides [11, 12]. Both designs have shown intrinsic quality factors of over a million; however, this has been at the cost of ring size and optical confinement. The etchless ridge resonator was only demonstrated at a bending radius as large as 400 μm in order to avoid radiation losses. The thin etchless resonator, due to the small waveguide size, has optical modes that are large compared with the waveguide [12].

The delocalization of the optical mode and large ring size in previously demonstrated etchless resonators means that the resonator is unsuitable for nonlinear optical devices, however this disadvantage is not fundamental to the etchless process, as high confinement etchless waveguides have been recently shown in [13]. We demonstrate here a silicon ring resonator fabricated with an etchless process that has high quality factor (760,000) and high modal confinement and localization, in a compact platform (50 μm radius ring). This process enables high aspect ratio waveguides with a high level of control of the structure dimensions.

2. Fabrication of high confinement etchless resonators

The device was fabricated on a silicon-on-insulator (SOI) wafer, with a top silicon layer of 500 nm, and a buried oxide of 3 μm. We used low-pressure chemical vapor deposition (LPCVD) to deposit 200 nm of silicon nitride (Si₃N₄) as a mask (see Fig. 1(a)). The mask layer was patterned using electron beam lithography with ma-N 2405 resist, and etched in a reactive ion etching (RIE) tool using CF₄ gas (see Fig. 1(b)). The resist was then stripped, followed by wet and dry thermal oxidation of the silicon (see Fig. 1(c)). We used a process of 20 minutes of dry oxidation, followed by 85 minutes of wet oxidation, and then 20 minutes of dry oxidation. The oxidation was done at 1200 degrees C. The silicon nitride was left in place. The device was clad with 2 μm of plasma enhanced chemical vapor deposition (PECVD) silicon dioxide (see Fig. 1(d)).

 

Fig. 1 Process flow for high confinement etchless waveguide. (a) LPCVD deposition of 200 nm silicon nitride on top of 500 nm silicon-on-insulator (SOI) platform. (b) Patterning and etching of the silicon nitride using electron beam lithography and ma-N 2405 photoresist. (c) Waveguide after selective thermal oxidation of the silicon. (d) Deposition of 2 μm PECVD oxide as cladding.

Download Full Size | PDF

3. Device design

The physical waveguide structure is determined by the silicon nitride width and by the wet and dry oxidation times. The SEM image of the resulting waveguide structure is shown in Fig. 2(a). We used commercial software (Silvaco Athena) to determine the dimensions of the waveguide structure; the simulated TM-polarized mode at a wavelength of 1.55 μm is shown in Fig. 2(b). The effective indices of the optical modes of the waveguide are 3.0 for the TE mode and 2.9 for the TM mode, indicating a high degree of confinement for both polarizations. The waveguide is multimode for both polarizations, with six modes for each polarization. The ring is designed to operate for the fundamental TM mode.

 

Fig. 2 (a) False colored SEM image of the cross-section of the silicon waveguide structure. In this image, the silicon nitride is not visible. (b) Simulated Poynting vector of the fundamental TM optical mode.

Download Full Size | PDF

We patterned the nitride mask to a width of 1 μm for the waveguides and ring resonators, with a gap of 300 nm in the mask for the waveguide-resonator coupling regions (see Fig. 3(a)). For critical coupling of the TM mode, we designed the device with 5 μm long coupling regions.

 

Fig. 3 (a) Optical microscope image of the racetrack resonator. (b) SEM image of the cross-section of the silicon structure. The different thickness in the coupling region achieved by the slower oxidation rate in the gap region.

Download Full Size | PDF

From simulation, coupling of the fundamental TE mode into the ring is significantly stronger than that of the fundamental TM mode. In the coupling region, about 0.5% of TM light is coupled into the ring, whereas 5% of TE light is coupled. The difference in coupling arises from the 140 nm silicon slab between the waveguides, as seen in Fig. 3.(b). Horizontally polarized light (TE) strongly interacts with the slab, whereas vertically polarized light (TM) does not. For this coupling length, the fundamental TE mode is very over coupled.

Coupling into and out of the waveguide is accomplished with a silicon nanotaper [14], with a patterned nitride mask width of 200 nm.

4. Results and discussion

We demonstrate a high quality factor of 760,000 in a high confinement etchless silicon ring resonator. To test the device, we couple light from a tunable laser light source through a polarization maintaining fiber into a lensed fiber. The lensed fiber couples light into a silicon nanotaper. The light is collected from the output of the chip by another lensed fiber, and its power measured by a photodetector. We observe a transmission spectrum with clear resonances, and varying resonance extinction, as seen in Fig. 4(a). We believe this is caused by wavelength dependent optical mode mixing in the coupling region shown in Fig. 3(b).

 

Fig. 4 (a) Normalized transmission spectrum for TM polarized light. The spectrum is normalized to the average off-resonance transmission. The resonance peaks with high extinction correspond to the fundamental TM mode. The resonant peaks with the different FSR (with extinctions less than 3 dB) correspond to the higher order TM modes. (b) Normalized transmission spectrum of the resonance at λ_o = 1533.063 nm

Download Full Size | PDF

We measured an intrinsic quality factor of Q_i = 760, 000 for the fundamental TM mode. The propagation loss in the ring is given by [15]:

5. Conclusion

We have designed and demonstrated a high-Q factor etchless silicon ring resonator with high confinement of the fundamental TM mode. We achieved a high intrinsic quality factor of 760,000 without sacrificing modal confinement and have demonstrated significant lower losses over traditional high confinement etched waveguide resonators. We believe this technology will provide an invaluable platform for nonlinear optics, due to the high quality factor and confinement, but also to any application where minimizing crosstalk without compromising quality factor is important.