Fiber-to-chip fusion splicing for low-loss photonic packaging

Silicon photonic devices are poised to enter high-volume markets such as data communications, telecommunications, biological sensing, and optical phased arrays. Permanently attaching a fiber to a photonic chip with high optical efficiency still, however, remains a challenge [1–4]. Silicon photonics leverages the mature electronics fabrication infrastructure because of its compatibility with complementary metal oxide semiconductor (CMOS) fabrication. During the past few decades, the basic building blocks of silicon photonic devices have been demonstrated: modulators, detectors, switches, filters, and lasers. One of the main challenges remaining is a packaging method to permanently attach optical fibers to photonic chips with high optical efficiency and high speed, while maintaining compatibility with CMOS processing without introducing changes to the fabrication process or consuming a significant area on the chip [1,5–9].

Techniques to package the fiber to the chip rely on bulky fixtures, metallic or glass ferrules, or specialized fibers. Numerous methods for high-efficiency coupling of light from an optical fiber to a photonic chip use gratings for coupling light from the top of the chip or waveguide nanotaper-based couplers to couple light from the edges [1,5,7–46]. While grating couplers have larger alignment tolerance and give optical access from the top of the chip, they have a narrow bandwidth and typically require bulky fixtures to attach them to the chip [24,26,28,29,37,38,40,47]. Edge couplers are broadband, but have a tighter alignment tolerance. They typically need bulky fixtures, high NA fibers, or lensed fibers to attach them to the chip. In these techniques, the fibers are usually permanently attached to the chips using optical adhesives [1,11,12,20,21,25,28,30,41]. Fixtures and lensed fibers, however, significantly increase the packaging cost.

We present a low-cost, robust, and low-loss packaging technique of permanent optical edge coupling between a fiber and a chip using fusion splicing that is scalable for high-volume manufacturing and has a larger alignment tolerance. Our approach consists of a cantilever-type silicon dioxide waveguide [7,42,48], which is mode-matched to a single-mode fiber on one side and to a waveguide nanotaper on the other.

The oxide waveguide is permanently fused to the optical fiber (Fig. 1) using a ${\mathrm{CO}}_2$ laser via radiative energy. The fusion splice between the fiber and the chip forms a permanent bond and decreases coupling losses by eliminating the Fresnel reflections at both oxide–air interfaces and the gap between the fiber and the oxide taper. This method is compatible with different types of inverse nanotapers (e.g., linear taper, metamaterial taper) [11,49,50] since the oxide waveguide geometry can be engineered to match the nanotaper mode profile. The oxide waveguide is carved out from the upper and under claddings of the device. It is compatible with standard foundry processes and does not require adding or removing steps from the typical fabrication process. The proposed method does not require special blocks or fiber holders to hold the fibers and uses standard, cleaved optical fibers. The silicon dioxide mode converter is designed to efficiently couple to a cleaved optical fiber on one of its ends (Fig. 2, plane A) and to a waveguide nanotaper on the other end (Fig. 2, plane B). The mode conversion process occurs in two stages: from the waveguide (mode size $<1\mathrm{\mu m}$) to the waveguide nanotaper plane B and then from the oxide mode converter to the optical fiber (mode size of 10.4 μm, plane B). The geometry of the oxide mode converter is engineered to optimize coupling by matching the modes of the optical fiber with the oxide mode converter. We vary the oxide mode converter width (from fiber to chip coupling, plane A, to mode converter to waveguide coupling, plane B) adiabatically, to maximize the mode overlap area between the fiber on one end and the waveguide taper on the other. The optimum dimensions of the oxide mode converter widths depend mainly on the refractive index of the guiding medium and the thickness of oxide cladding. We set its length to 10 μm for mechanical stability and negligible transmission loss. This coupling method can be used in fusing any device that has a cladding of silicon dioxide, including, for example, devices based on silicon, silicon nitride, and lithium niobate on insulator. We isolate the oxide mode converter from the silicon substrate to prevent a loss of light to the substrate. Due to the physical size mismatch between the fiber and the oxide mode converter, we reinforce the splice with a UV curable optical adhesive. Controlling the refractive index of the optical adhesive will further tailor the properties of the oxide taper mode and improve the coupling to the fiber [7].

 

Fig. 1. 3D model of fiber-to-chip packaging using fusion splicing. The red spot shows the fusing location (not to scale).

Download Full Size | PPT Slide | PDF

 

Fig. 2. Schematic representation of a packaged device using a silicon dioxide mode converter fused to SMF-28 fiber. The side view shows an undercut silicon substrate, which isolates the oxide mode converter from the chip. The top view of the method shows a microscope image of the fabricated device, with the silicon dioxide mode converter, nitride taper, and the undercut silicon substrate.

Download Full Size | PPT Slide | PDF

We simulate the coupling loss for TE polarization between a silicon nitride waveguide nanotaper and the optical fiber using the eigenmode expansion method (FIMMPROP by Photon Design) with mode overlaps. The waveguide nanotaper is 0.18 μm wide at the tip, 100 μm long, and has a linear profile. We calculate the coupling loss between the optical fiber and the input of the oxide mode converter plane A (Fig. 2) by launching a Gaussian mode with a 10.4 μm diameter into the mode converter. Our calculations (Fig. 3) show that coupling improves as the oxide cladding thickness increases for an oxide mode converter width of between 13 and 15 μm. The optical properties of the adhesive used to stabilize the splice can further tailor the mode matching and enhance the coupling by expanding the mode even in cases with thinner cladding thicknesses [7]. The losses at the output of the mode converter, plane B, (Fig. 2) are calculated by launching the fundamental mode of the input oxide mode converter waveguide (Fig. 4). The coupling loss at plane B (Fig. 2) is dominated by the nanotaper design and is weakly dependent on the oxide cladding thickness. The calculated 1 dB penalty misalignment tolerance between the fiber and the oxide mode converter is $+/-2.5\mathrm{\mu m}$ and $+/-2.4\mathrm{\mu m}$ in the horizontal and vertical directions, respectively (Fig. 5). This tolerance compares favorably with other edge coupling methods [1,12,20,21,25,30,41]. The oxide mode converter is polarization insensitive; hence, the polarization dependence is due to the waveguide nanotaper design. Losses are 0.2 dB higher if designed for simultaneous TE and TM operation.

 

Fig. 3. Simulated coupling loss between the fiber and the mode converter, plane A as a function of oxide mode converter input width. A minimum coupling loss of 0.3 dB can be achieved with a 14 μm oxide mode converter width and 11 μm total (top plus bottom) oxide cladding thickness with an adhesive of refractive index 1.4.

Download Full Size | PPT Slide | PDF

 

Fig. 4. Simulated coupling loss between the oxide mode converter and the waveguide nanotaper, plane B as a function of output mode converter width. Loss is sensitive to mode converter width and only weakly dependent on total cladding thickness.

Download Full Size | PPT Slide | PDF

 

Fig. 5. Simulated coupling loss due to misalignment between the fiber and oxide mode converter. The tolerance for 1 dB loss is $+/-2.5\mathrm{\mu m}$ and $+/-2.4\mathrm{\mu m}$ in the horizontal and vertical direction, respectively.

Download Full Size | PPT Slide | PDF

We fabricate a silicon nitride photonic chip using standard, CMOS-compatible, microfabrication techniques. For optimum results, a cladding thickness of 11 μm was selected with 14 μm of input mode converter width (A) and 12 μm of output mode converter width (B). A 5.5 μm layer of silicon dioxide is deposited via a plasma-enhanced chemical vapor deposition (PECVD) and 325 nm of silicon nitride are deposited via low-pressure chemical vapor deposition (LPCVD). The waveguides are patterned using standard DUV optical lithography at 248 nm using an ASML stepper. After etching in an inductively coupled plasma reactive ion etcher (ICP-RIE) with a ${\mathrm{CHF}}_3/{\mathrm{O}}_2$ chemistry, the devices are clad with 5.8 μm of oxide using a plasma-enhanced chemical vapor deposition (PECVD). We then pattern and etch the oxide mode converter similarly to the waveguide step. After dicing, we remove the silicon substrate using xenon difluoride (${\mathrm{XeF}}_2$) dry etch to optically isolate the oxide mode converter from the silicon substrate.

We fuse a SMF 28 cleaved fiber with a mode field diameter of 10.4 μm to the photonic chip by using a standard research ${\mathrm{CO}}_2$ laser of 40 W from Synrad and reinforce the splice with an optical adhesive. Silicon dioxide strongly absorbs light at a 10.6 μm wavelength from the ${\mathrm{CO}}_2$ laser. The laser radiatively heats the silicon dioxide fiber as well as the cladding of the chip. Radiative heating for splicing leaves no residue behind [51–53]. The laser beam from the ${\mathrm{CO}}_2$ laser is focused to a spot of 45 μm using a ZnSe aspheric lens ($f=20\mathrm{mm}$, $\mathrm{NA}=0.5$) and aimed at the fiber–chip interface with a collinear red diode laser for alignment. The laser beam is incident at an angle of 30°, which gives clear line of sight to the full chip facet. To fuse the fiber to the oxide mode converter, we irradiate the spot with 9 W of laser power for 0.5 s. We did not see any effect on the waveguide due to the heating, as the melting point of silicon nitride is around 1900°C and that of silicon dioxide is 1700°C.

We demonstrate a minimum loss of 1.0 dB per facet with a 0.6 dB penalty over 160 nm bandwidth near the C-band on a standard, cleaved SMF-28 fiber fused to the silicon nitride photonic chip. This result represents a significant improvement from our initial demonstration of this technique [48] due to the optimization of the mode properties.

To measure the coupling loss, we first measure the optical power exiting the input fiber. We align cleaved fibers to the oxide mode converters at the input and output of the chip and measure a loss of 2.1 dB per facet. Note that the total loss is 4.6 dB, and both coupling regions are the same within fabrication variation and 0.4 dB are waveguide propagation losses. We measure the loss of 0.4 dB through the nitride waveguide by collecting the output with a microscope objective in place of the output cleaved fiber and comparing the two measurements. After fusing the fiber to the mode converter on the input side, we measure a total loss of 3.8 dB. We subtract the 2.1 dB loss from the output fiber and the 0.4 dB loss from the waveguide to find the coupling loss of the fused fiber of 1.3 dB (Fig. 6). The loss is reduced after fusing because Fresnel reflections are eliminated (approximately 0.3 dB of loss) and the small gap between fiber and mode converter disappears. We apply an optical adhesive with a refractive index of 1.3825, as specified by the manufacturer, to stabilize the splice. The optical adhesive expands the mode and reduces the coupling loss to 1.0 dB.

 

Fig. 6. Coupling loss per facet as a function of wavelength. The measured loss is 1.0 dB per facet with a 0.6 dB penalty over 160 nm bandwidth. The error bars of $+/-0.1\mathrm{dB}$ are due to fluctuations in the power measurements.

Download Full Size | PPT Slide | PDF

The total coupling loss is the sum of the losses from the fiber to oxide mode converter coupling and the oxide mode converter to waveguide nanotaper (loss in Fig. 3 plus loss in Fig. 4). The loss at the fiber-to-oxide mode converter can be minimized by increasing the oxide cladding thickness or by optimizing the refractive index of the optical adhesive. In general, both methods are used together to minimize coupling loss; however, in cases when the cladding thickness can’t be changed, such as in silicon on insulator-based devices where the buried oxide is typically 3 μm or less, choosing the optimum optical adhesive is critical to achieve low loss. The loss between the oxide mode converter and the waveguide nanotaper can be minimized through a suitable choice of nanotaper tip width, taper geometry, and taper type (e.g., continuous versus metamaterial). The packaging method described here is flexible to accommodate different design choices depending on the photonic platform being considered. It is compatible with any platform that uses silicon dioxide as a cladding material.

Fiber-to-chip fusion splicing has the potential to enable high throughput optical packaging with a robust, high efficiency, and low-cost solution. The method is compatible with multiple photonic platforms, such as silicon, silicon nitride, and lithium niobate, which use silicon dioxide as the cladding. We envision that this method can be fully automated to enable highly efficient fiber-to-chip coupling in high-volume applications and can be extended to passive alignment techniques.