We report on the fabrication and experimental demonstration of optical mode size transformation between standard single-mode fiber and 0.26μm-thick Si-waveguide by 12μm-thick Si/SiO2 multilayer on-chip GRIN lens of lengths 16μm or 24μm butt-joint to 10μm-wide terminated Si-waveguide. The overall coupling loss of the coupler was measured to be 3.45dB in which the Fresnel reflection loss is estimated to be 2dB at the GRIN-lens/air interface. The on-chip integrated GRIN lens opens up the feasibility of a low cost passive aligned fiber-pigtailed electronic-photonics integrated circuits platform.
©2010 Optical Society of America
The extremely high refractive index contrast of Si/SiO2 (Δn = 2) allows the core size of Si-wire waveguide to be shrunk down to sub-micron in dimensions while single-mode propagation can be maintained at 1.3~1.5μm telecommunication wavelengths . By employing mainstream Complementary-Metal-Oxide-Semiconductor (CMOS) compatible processing, Si-wire waveguide fabricated on silicon-on-insulator (SOI) substrate opens the possibility of low cost and high density optoelectronics/photonics integrated circuits (PIC) on such platform . However, the substantial optical mode size mismatch between the Si-wire waveguide on PIC and single-mode optical fiber (core-diameter: 9~10μm) presents a challenge to the acceptance of SOI-based PIC in its applications to communications, optical information processing, and even optical sensing. The key enabler in low cost manufacturing of single-mode-fiber (SMF) pigtailed packaged PIC modules lies in the method and processes to efficiently couple SMF to the PIC chip. In addition, achieving the capability of optical mode size transformation from sub-0.5μm to 8~10μm compatible with the mode-size of standard SMF relaxes the tight fiber placement precision requirement for alignment to PIC, and opens the way to using passive alignment and, hence, low cost manufacturing of fiber-pigtailed modules.
Until now, various approaches have been proposed and demonstrated to couple SMF to Si-waveguide on PIC. These include vertical Si up-taper , Si down-taper , Si down-taper capped with higher index polymer waveguide , amorphous-Si coupler , amorphous hydrated-SiOx graded-index (GRIN) lens [7,8], SiON based GRIN lens , and cascaded linear/non-linear Si up-taper [10,11]. The Si up-taper has disadvantages of being too long (500~1000μm), fabrication method requires complicated lithographical technique, and the coupling efficiency was 45%~60% (coupling loss: 3.46dB to 2.2dB) . Si down-taper requires high resolution lithography to form nano-taper tip and thick bottom SiO2-cladding (≥3μm) to reduce substrate optical leakage loss [4,5]. This can be in conflict with requirement for thinner buried-oxide cladding SOI substrate when electronics circuits are to be built on the PIC. Thus far, only measurements of Si down-taper coupling to small-core SMF (core-diameter: ~5μm) were reported [4,5]. For Si down-taper, the optimal coupling losses for TE and TM polarizations were from 0.4 to 0.5dB , and coupling loss was measured to be 0.8dB . Although the amorphous-Si coupler deposited on Si-wire waveguide can be easily fabricated , the spot size conversion was from 0.5~0.8μm to 3.5μm (and vice versa) in the vertical direction, which was insufficient to couple to standard SMF. The calculated coupling efficiency for amorphous-Si coupler was about 40~45%. The utilization of on-chip GRIN lens is a promising approach for spot-size conversion. One method was to use on-chip amorphous hydrated-SiOx GRIN lens with increasing oxygen content to form GRIN profile in the vertical direction coupled with hemispherical output end-face for lateral beam focusing [7,8]. The amorphous hydrated-SiOx GRIN lens has exhibited mode size conversion from standard SMF (core-diameter:9~10μm) to spot size of 1.8~2.5μm, and the measured coupling loss was 1.5dB . Another approach was to use seven-layer SiON GRIN  lens in which the relative composition of oxygen and nitrogen was varied to change refractive indices in a parabolic GRIN profile in the vertical direction. In this case, the spot size conversion was from standard SMF to 0.9x0.9μm2 SiON waveguide. The best coupling loss for SiON GRIN lens coupler was 0.45dB. However, such method has yet to be shown for focused beam spot size below 0.9μm. Recently, a spot-size converter formed by cascading lateral Si up-taper with vertical non-linear Si up-taper [10,11], was reported. Although the lateral mode size transformation up to 8~9μm to match the mode size of the SMF was demonstrated, the vertical mode size transformations were from 0.5μm to 5μm , and 0.22μm to 2.8μm . While lateral optical mode size transformation can simply be achieved by employing lateral adiabatic Si up-taper, the vertical focused beam spot size from standard SMF via an integrated coupler, hitherto, has been no less than 0.5μm (typically, 0.6~1μm) [3–10]. Recently, super-high numerical aperture (N.A.) GRIN lens with vertical refractive index profile implemented by TiO2/SiO2 dual-materials multilayer structure reported a theoretical vertical focused spot size of 0.53~0.7μm at wavelength of 1.55μm and the calculated coupling efficiency of 95.3% for anti-reflection coated GRIN lens . This is encouraging because vertical GRIN lens profile can be engineered by merely multilayer deposition.
In this work, we bring the dual-materials multilayer approach further by using high refractive index contrast Si/SiO2 multilayer to form an on-chip GRIN lens structure for optical mode size transformation in the vertical direction . Si/SiO2 not only has the advantage of being CMOS compatible materials, but its high refractive index contrast in the multilayer GRIN lens can potentially bring vertical focused spot size to sub-0.5μm from SMF. While the super-high N.A. GRIN lens in  is symmetrically placed with respect to the semiconductor wire-waveguide, our proposed GRIN lens is asymmetrically placed on Si wire-waveguide which has further advantage of ease of fabrication. In this paper, the physical realization of such Si/SiO2 multilayer super-high N.A. GRIN (MLS-GRIN) lens is reported and the measurements to verify the physics of the MLS-GRIN lens are described. Vertical optical mode size transformation from standard SMF to sub-0.5μm-thick Si waveguide via an on-chip multilayer GRIN lens coupler with length less than 30μm has been achieved.
2. Device structure and simulation
The device structure for mode size conversion from Si wire-waveguide to standard SMF and vice versa is schematically illustrated in Fig. 1(a) . Lightwave on the PIC are guided by Si wire-waveguide with dimension of 0.26μm(thickness)x3μm(width). 3μm is the minimum lateral width achievable for our fabrication using contact-mode photolithography for pattern transfer. The refractive indices of Si (in the SOI) and SiO2 are 3.46 and 1.46, respectively. For this test chip, lateral Si up-taper from width of 3μm to output width W, over a length of 500μm has been utilized for lateral mode size transformation. In this test chip, MLS-GRIN lens stack has been butt-terminated to Si up-tapered waveguides with W’s of 3μm, 6μm and 10μm. In practice, lateral up-taper length can be shorter to minimize chip area. LGRIN refers to the physical length of the MLS-GRIN lens (refer to Fig. 1(a)). The MLS-GRIN lens, which is placed at the end of the lateral up-taper, functions as mode size transformer in the vertical direction. Figure 1(b) shows the schematic vertical cross-section of the MLS-GRIN lens placed on the buried-oxide at the head end tip of the Si wire-waveguide. The MLS-GRIN lens consists of multilayers of amorphous-Si (a-Si) and SiO2 in alternation. The refractive indices of a-Si and SiO2 are 3.2 and 1.46, respectively. These refractive indices have been measured by ellipsometry on single layer of the respective materials. In Fig. 1(b), the different refractive indices of the SOI and a-Si of the multilayer stack are indicated by dark and brighter blue colors, respectively. The total thickness of the multilayer stack is 12μm. This thickness has been confirmed by surface profiler scan of our actual device. The MLS-GRIN lens consists of 40 pairs of Si/SiO2, each pair with varying thicknesses of Si and SiO2. The first pair at the bottom is 2.84μm-thick a-Si and 7nm-thick SiO2. The next pair is 247nm-thick a-Si and 11nm-thick SiO2. The thickness of a-Si gets thinner and of SiO2 gets thicker as one proceeds to the top. The method to generate the layer thicknesses is described in subsequent paragraph and also in . MLS-GRIN lens is not deposited on Si wire-waveguide but butt-terminated to it because a-Si has a lower refractive index in comparison to that of Si in the waveguide.
Anti-reflection coating (ARC), which is yellow in color, is shown to be coated on the output facet of the MLS-GRIN lens (refer to Fig. 1(b)). For maximum transmission of optical power, the refractive index of ARC is 1.73. This is calculated by the square-root of the product of effective refractive index of MLS-GRIN lens at 5μm from the bottom of the GRIN lens stack and index of air (i.e. ), where 3.0 is the effective index of MLS-GRIN medium at 5μm-point (refer to Fig. 1(b)). 5μm-point was chosen because it is the location of the peak optical intensity of the expanded optical beam after passing through the MLS-GRIN lens. An example of dielectric material that can fulfill the refractive index of ARC is SiON. The thickness of the ARC was calculated to be 0.224μm.
In the design of the MLS-GRIN lens structure, the asymmetric GRIN profile is first engineered by multilayer thin films in which the refractive index of each film, ni, is a constant in each layer but varies from film to film in a near-parabolic profile as we move up the stack. For the asymmetric GRIN lens, the GRIN profile has decreasing refractive index from bottom to the top of the stack such that ni > ni + 1.. As lightwave propagates in the GRIN lens medium, in order for the lightwave to be practically unperturbed by the thin film interfaces such that it sees a continuous GRIN profile, the thickness of each film, h (h = D/N) is such that it is about one tenth of the optical wavelength in the GRIN lens material. D is the overall stack thickness and N is the total number of layers. LGRIN, the physical length of the MLS-GRIN lens is also the focal length of the GRIN lens. To generate the near-parabolic refractive index profile for aberration free propagation in the lens, refractive index n1 of the first bottom layer is chosen to be nearly the same as the waveguide material. The tip of Si-waveguide at the point of entry into the MLS-GRIN lens is treated like a point source of light. The design rule was established by optical ray tracing (refer to Fig. 1(c)). The refractive index of the second layer n2 can be determined by employing the condition that light ray is refracted in horizontal direction after traversing a distance, LGRIN, of the MLS-GRIN lens from the point source of light. The incident angle of light from layer n1 to n2, is the critical angle, θcrit, as governed by Snell’s law n1sin(θcrit) = n2sin(90°) (Fig. 1(c)). From Snell’s law, n1 and n2 can be related to the physical dimension parameters LGRIN and h by,Fig. 1(c), for ray that traverses across two thin films, refracts and exits in the horizontal direction at the interface to third thin film, it traces out two horizontal segments in the MLS-GRIN lens Lf1 and Lf2, in which the sum of them is LGRIN. In general, a ray of light that traverses (i-1) number of thin films and exits parallel at the interface to the i-th thin film, traces out (i-1) number horizontal segments: Lf1, Lf2,…, Lfi-1, in which the sum of them is LGRIN,
The refractive index ni for the i-th layer can be calculated from Eq. (2). The refractive index profile of the multilayer thin films can, thus, be generated when D, LGRIN, and n1, refractive index of the first layer, are given.
The refractive index of j-th layer nj can, then, be approximated by a pair of high and low refractive index thin films. The average refractive index of the pair can be evaluated by weighted average by their thicknesses of their respective dielectric permittivities. The permittivity can be written as square of the refractive index (εj = nj 2). The permittivity of j-th film can be written as,13]. The refractive index for each j-th layer against vertical distance in X-direction can be found in Fig. 2(a) of , and the near-parabolic refractive index profile of aberration-free MLS-GRIN lens against vertical distance is shown in Fig. 6 of . By finite-difference time domain (FDTD) simulation of lightwave propagation in this 12μm-thick MLS-GRIN lens, the focal length LGRIN, has been optimized to be 19.7μm to transform vertical optical mode size from Si-waveguide of thickness 0.26μm to 8~10μm at the output of the MLS-GRIN lens, and vice versa.
The LUMERICAL 2D-FDTD method  has been used to verify the design for the propagation of polarized light pulse through the MLS-GRIN lens for both cases of light expansion from Si wire-waveguide through the MLS-GRIN lens into air, and light compression from SMF through the MLS-GRIN lens and into the Si wire-waveguide in the vertical XZ-plane of the MLS-GRIN lens. The light from standard SMF is represented by a Gaussian light source at the output facet of the MLS-GRIN lens (Fig. 2(a)). The length of the MLS-GRIN lens used in the simulation was 19.7μm. The ARC with refractive index of 1.73 and thickness of 0.224μm, was included at the output facet of MLS-GRIN lens. Figure 2(a) shows the structure of the MLS-GRIN lens, which is butt-joint to the 0.26μm-thick Si-waveguide. Refractive indices of Si in the waveguide and in the MLS-GRIN lens are 3.46 and 3.2, respectively, in the simulation. Simulation was done for polarized light with E-field aligned in the Y-direction (perpendicular to the XZ-plane or point out of the page) with reference to axes in Fig. 1(a). This is the TE-mode in the Si-waveguide.
Figure 2(b) shows the optical power of forward propagating lowest order guided mode of the Si-waveguide with E-field aligned in the Y-direction. Optical power propagates through the MLS-GRIN lens and the ARC into the air with relatively little reflected power. Figure 2(c) shows the optical power of backward propagating Gaussian light source (beam diameter: ~9μm) with E-field aligned in Y-direction. Optical beam propagates through the MLS-GRIN lens and focuses into the 0.26μm-thick Si-waveguide. The red arrow shows the direction of travel of the lightwaves. Figure 2(d), (e) and (f) show the Ey, Hz, and Hx field profiles for backward propagating Gaussian light beam. Therefore, it has been verified by simulation that the proposed MLS-GRIN lens can perform bi-directional vertical optical mode size transformation from standard SMF to 0.26μm-thick Si-waveguide for TE-polarization. This is about 35x mode size transformation in vertical direction. The key focus of this work is the demonstration of a working prototype device for vertical optical mode size transformation over a short distance of ~20μm for TE-polarization because Si-waveguide on our PIC platform is based on TE-polarization. Polarization dependence characterization data for our device will be provided in future publications.
3. Device fabrication
The MLS-GRIN lens on Si waveguide was fabricated using 4inch SOITEC-SOI with thickness of the SOI thinned from 340nm to 260nm by thermal oxidation. The buried oxide (BOX) of the substrate has a thickness of 1μm. 15~20nm of thermal oxide was left un-removed on the SOI in the last thermal oxidation step. 200nm of silicon nitride was deposited on the wafers by low-pressure chemical vapor deposition (LPCVD) to function as hard-mask for Si-waveguide formation. The intermediate oxide was to relieve the stress of the nitride on the SOI layer. The Si waveguides were patterned on the wafer using the first mask by contact-mode photolithography employing AZ5214E positive photo-resist. The minimum reproducible feature width for such photolithography is 2~3μm. The width of the Si-waveguide for our sample has been designed to be 3μm which is multimode waveguide. It is not critical that the Si waveguide is multimode as observation of vertical mode size transformation is our main interest. The silicon nitride hard-mask and bottom 15~20nm of thermal oxide was first etched by reactive ion etching (RIE). Without removal of photo-resist, the SOI was etched by C4F8/SF6/O2 inductive coupled-plasma (ICP) RIE to form vertical side wall Si-waveguide. After removal of photo-resist by O2 plasma and silicon nitride by phosphoric acid, 400nm of passivation silicon dioxide by plasma-enhanced chemical vapor deposition (PECVD) was deposited on the Si-waveguide. Then, 50~60nm of Ni was deposited on the passivation oxide. Subsequently, rectangular openings of 30μm(longitudinal)x60μm(lateral) were made by photolithography at the Si-waveguide tips such that the overlap of rectangular opening and Si-waveguide tip is equal to the LGRIN. Areas outside the rectangular openings were covered by photo-resist. Ni and passivation oxide in the openings were etched by RIE to expose the Si-waveguide tips. The Si-waveguide tips were, then, selectively removed by RIE while underlying oxide remained intact. After removal of photo-resist and appropriate cleaning, the Si/SiO2 multilayer was blanket deposited on the wafer. The Si/SiO2 multilayer was deposited by ion-assisted deposition (IAD) on the wafer. The multilayer Si/SiO2 deposited in the openings adheres strongly to the substrate since the underlying material is SiO2. Ni does not provide good adhesion to the deposited Si/SiO2 multilayer. Si/SiO2 selectively lifted off from the Ni surface. To further assist the lift-off of Si/SiO2 multilayer, SU-8 negative photo-resist was spin-coated on the substrate. Photolithography was performed to expose those areas of the SU-8 resist where there were Ni underneath. After resist development, samples were dipped into heated Remover-PG solvent to lift-off both the SU-8 resist together with the Si/SiO2 multilayer that did not stick well to the Ni surface. Only Si/SiO2 multilayer on the exposed openings at the Si waveguide tips remained attached to the substrate and the waveguide tips. The height of the MLS-GRIN lens stack was measured to be 12μm by surface profiler. Finally, 1.5μm of passivation oxide was deposited on the sample to protect the MLS-GRIN lens. Individual chips were diced and edge-polished by diamond lapping films to achieve the desired LGRIN for the MLS-GRIN lens.
Figure 3(a) shows the scanning electron micrograph (SEM) of the MLS-GRIN lens butt-joint to the tip of the Si-waveguide. Figure 3(b) shows the front-side view of the MLS-GRIN lens. Take note that the Si up-taper is at the background. Figure 3(c) shows a higher magnification of the front-side view of the MLS-GRIN lens. The Si/SiO2 multilayer is situated on the BOX. Two types of Si up-tapers are present on this lot. These are up-tapers that terminate with widths of 6μm and 10μm, respectively. For Si up-tapers that are terminated with 6μm-width, the length of the MLS-GRIN lens was measured to be 16μm. Si up-taper that are terminated with 10μm, the length of MLS-GRIN lens was measured to be 24μm. Due to the imprecise nature of edge-polishing method, focal length LGRIN of 19.5μm was not achieved for these samples. A CMOS compatible and higher throughput fabrication process to integrate the MLS-GRIN lens to Si-waveguide on PIC including the method for more precise control of LGRIN using ICP/RIE etching of the Si/SiO2 multilayer will be reported in future publications. For these future works, edge-polishing method will not be used.
4. Experimental verifications of physics of the MLS-GRIN lens
In this section, the experimental data to verify the physics of the MLS-GRIN lens are presented and discussed. In addition, base on the experimental propagation loss data of the Si waveguide on the chip, the coupling loss of the MLS-GRIN lens to standard SMF was estimated.
4.1 Mode expansion from Si-waveguide through MLS-GRIN lens to air
The devices-under-test (DUT) consist of three groups of Si-waveguides. First group consists of 3μm-wide Si-waveguides that terminate with widths of 3μm (i.e. no lateral up-taper). Second group consists of 3μm-wide Si-waveguides up-taper to widths of 6μm termination at both ends. The lengths of these Si-waveguides are 4189μm. The third group consists of 3μm-wide Si-waveguides up-taper to 10μm-wide termination at both ends. The lengths of these Si-waveguides are 2152μm. The lateral Si up-taper length is 500μm for all waveguides. At output ends of these Si-waveguides, MLS-GRIN lens are deposited after exposed Si-waveguide tips are removed by RIE. Figure 4 shows the schematic diagram of a typical DUT which consists of SOI-based Si-wire waveguide with MLS-GRIN lens termination.
As shown in Fig. 4, discrete objective lens (60X, N.A. = 0.65) was employed to couple TE-polarized 1550nm laser light into the Si-waveguide at input end without MLS-GRIN lens. From Fig. 5 (a) to (c) , discrete objective lens (30X, N.A. = 0.4) was utilized to image the optical near-field pattern of the optical mode at the termination facet of the MLS-GRIN lens onto the IR-camera. Figure 5(a) shows the image of the optical mode at the output of 3μm-wide waveguide without MLS-GRIN lens, and Fig. 5(b) shows the near-field pattern of the optical mode at the output of MLS-GRIN lens deposited on 3μm-wide waveguide termination. It can be observed that the centre of brightness of the beam in the original 3μm-wide waveguide is shifted upward due to the effect of the MLS-GRIN lens. This upward shift of the centre of intensity peak has been shown previously in device simulation (refer to Fig. 2(b)). Figure 5(c) shows the near-field pattern of the optical mode at the output facet of MLS-GRIN lens deposited on 6μm-wide waveguide termination. The centre of brightness of the beam not only shifted upward, but also broadened laterally. For Fig. 5(d) and (e), 60X magnification (N.A. = 0.65) discrete objective lens was utilized for imaging on IR-camera. Figure 5 (d) and (e) show the near-field pattern of optical modes of 10μm-wide Si-waveguide termination without and with MLS-GRIN lens, respectively. Optical mode is expanded vertically above the red line which indicates the baseline level of the Si-waveguide on the PIC. In Fig. 5(e), the extend of vertical expansion is larger than the lateral extend. This is because the MLS-GRIN is 12μm in thickness deposited on 10μm-wide Si-waveguide termination. In summary, the results in this section show that MLS-GRIN lens expanded the optical mode vertically ready to be coupled to SMF.
4.2 Beam collimation of the optical output from MLS-GRIN lens
To further investigate the property of the output beam from the MLS-GRIN lens, SMF-core was placed facing the output end of the MLS-GRIN lens, as shown in Fig. 6 (a) to (c). IR laser light propagated through the Si-waveguide, MLS-GRIN lens (LGRIN = 24µm), the air gap, into the SMF-core. The optical power coupled into the SMF was recorded by an IR-power meter connected to the SMF. Figure 6 (d) shows the measured optical power in the SMF against wavelength of the IR-light for position-(a), (b) and (c) of the SMF. The air-gap distances between MLS-GRIN output facet and the SMF were 5μm, 30μm, and 52μm for positions-(a), (b), and (c), respectively. It can be observed that average received optical power by the SMF did not reduce significantly with distance of the air-gap even up to 52μm. This shows that the optical beam from MLS-GRIN lens was well-collimated prior to entering the SMF. Fabry-Perot resonance peaks observed in these optical power curves further support the evidence of collimated beam since the optical energy resonates between the output facets of MLS-GRIN lens and SMF. These experimental results also show that the MLS-GRIN lens has large longitudinal tolerance.
4.3 Coupling from SMF to Si-waveguide through MLS-GRIN Lens (Mode compression)
To show that the MLS-GRIN lens can indeed contract mode size vertically, laser light source of random polarization at 1550nm was coupled into the MLS-GRIN lens from SMF, which was placed with its core (diameter: 9~10μm) in proximity to the end-face of the MLS-GRIN lens, as shown in Fig. 7(a) and (b) . Figure 7(b) shows the magnified optical micrograph of the SMF placed at the input end of the waveguide with MLS-GRIN lens. In this case, the MLS-GRIN lens has length LGRIN of 16μm, and total Si-waveguide length is 4189μm. IR-lens objective (30X, N.A. = 0.4) was used to image the near-field pattern of the optical mode at the output end of the waveguide onto the IR-camera. No MLS-GRIN lens was placed at the waveguide output end, which terminates at width of 3μm. Figure 7(d) shows the near-field pattern of the optical mode at the Si-waveguide output for light that passed from SMF, contracted through the MLS-GRIN lens and passed through the Si-waveguide. Figure 7(c) shows the image of the SMF output facet placed at the same image plane as the waveguide output facet in Fig. 7(d), for 2μW of optical power in the SMF. Comparison of optical spot sizes, as shown in both images, shows that the size contraction in the vertical direction is slightly more than 10 times. Since the optical spot diameter of the SMF is almost equal to the diameter of the SMF-core (9~10μm), the optical spot diameter at the output of Si-waveguide as imaged by the IR-camera should be about one micron or less. This size we believe is not the true size of the optical mode at the output facet of the Si-waveguide since, as reported in , the optical spot diameter as imaged by the optical imaging system is always larger than the true spot diameter due to limitation of objective lens N.A. of 0.4. Figure 7(e) shows the SEM image of the Si-waveguide termination at the output end. The Si-waveguide has a vertical thickness of 0.26μm, which is the thickness of the SOI. Since optical power has been observed in the Si-waveguide output in Fig. 7(d), it can be inferred that the true vertical spot size should be in the order of the thickness of the Si-waveguide at the output. It is also observed that the Si-waveguide has a good vertical side-wall with 400nm of PECVD oxide covered above it. The SEM image also shows 50~60 nm of Ni on the PECVD oxide. In this fabrication process, Ni served as anti-adhesion layer when the MLS-GRIN lens was deposited. After all the Si/SiO2 multilayers above the Ni were selectively lifted-off, 1.2μm of PECVD passivation oxide was blanket deposited on the whole chip to protect the MLS-GRIN lenses on the chip without removal of Ni. The Ni will be removed in the future to reduce propagation loss in the Si-waveguide. In comparison with Fig. 7(c), the MLS-GRIN lens contracted the optical mode size vertically from that of the SMF to sub-micron in size equivalent to the thickness of the SOI.
4.4 Estimation of MLS-GRIN lens to single-mode fiber coupling losses
To estimate the coupling loss from MLS-GRIN lens to SMF, the DUT employed and the experimental set-up is shown schematically in Fig. 8(b) . The DUT consists of 3μm-wide Si-waveguide up-tapered to 10μm-wide termination at both ends. The output (right) end has MLS-GRIN lens fabricated on it. The total length of the Si-waveguide inclusive of the MLS-GRIN lens is 2152μm, and the MLS-GRIN lens has length of LGRIN = 24μm. Conical-tip lens fiber-probe (N.A. better than 0.5) was used to couple wavelength tunable laser into the Si-waveguide, and standard SMF was used to couple light out from the MLS-GRIN lens. For the DUT in Fig. 8(b), a simple Eq. (4) to show the relationship of coupling losses at both ends of the DUT, propagation loss and the overall insertion loss is provided (FP: fiber-probe),
The MLS-GRIN lens to SMF coupling loss was evaluated by subtracting the fiber-probe input coupling loss (left-side) and propagation loss through the waveguide from the overall insertion loss of the device in the set-up shown in Fig. 8(b).
Firstly, propagation loss of the 3μm-wide Si-waveguide up-tapered with 10μm-wide terminations at both ends but without any MLS-GRIN lens on the waveguide was evaluated by using Fabry-Perot spectrum [16,17]. A pair of lensed fiber-probes was used to couple randomly polarized wavelength tunable laser into and out of the Si-waveguide, as shown in Fig. 8(a). The Fabry-Perot spectrum for wavelengths from 1550nm to 1551nm was recorded and is shown in Fig. 9 . From the peak-to-valley intensity ratio of 1.47~1.52, the propagation loss was evaluated to be 18.5 ± 1dB/cm [16,17]. In a separate experiment, discrete objective lenses (60X, N.A. = 0.65) were used to couple wavelength tunable laser light of specific polarization into the same waveguide to measure propagation losses. The propagation losses for TE and TM were measured to be 19 ± 1dB/cm and 14 ± 1dB/cm, respectively. For waveguide length of 2152μm, the optical power propagation loss at random polarization was calculated to be 4dB. In comparison to Si-waveguide propagation loss values in , our large propagation loss values are accounted for by the 60nm of Ni above the 400nm of PECVD oxide overlaid on the Si-waveguide. Such Ni will be removed in future processing.
Next, using the same set-up shown in Fig. 8(a), randomly polarized laser was coupled into and out of the Si-waveguide using lensed fiber-probes. Transmitted optical power for wavelengths from 1510nm to 1600nm was recorded using an IR-sensitive photo-detector synchronized to the wavelength tunable laser. The transmitted optical power spectra were collected for cases of with and without the Si-waveguide device. Insertion loss of the Si-waveguide was evaluated by dividing the transmitted optical power with the waveguide device by that without the waveguide device. Figure 10 shows the insertion loss spectrum of the lensed fiber-probe coupled Si-waveguide shown at Fig. 8(a), and the insertion loss was measured to be 21.5dB. Since the set-up shown at Fig. 8(a) is symmetrical, the fiber-probe to waveguide coupling loss is evaluated to be 8.75dB after subtracting propagation power loss from the overall insertion loss and divided by two, i.e. (21.5dB – 4dB)/2 = 8.75dB.
Figure 11 shows the insertion loss spectrum measured for lensed fiber-probe coupled Si-waveguide at the input and SMF coupled to MLS-GRIN lens at the output. The experimental set-up is shown at Fig. 8(b). Insertion loss was evaluated by dividing the transmitted optical power by the optical power without the waveguide device with fiber-probe in close proximity to and in maximum transmission into the SMF. The insertion loss degrades as wavelength moves away from 1550nm. This is because the optical output in the experimental set-up was optimally adjusted at 1550nm, and degradation of insertion loss was due to defocusing of the fiber-probe at the input. MLS-GRIN lens to SMF facet Fabry-Perot resonance peaks are also observed on the insertion-loss spectrum. The best insertion loss from Fig. 11 is 16.2 dB. After subtracting 8.75dB and 4dB, the fiber-probe coupling loss and power loss due to propagation over 2152μm of the waveguide respectively, the best estimated MLS-GRIN lens to SMF experimental coupling loss is 3.45dB.
By FDTD method, the theoretical coupling efficiency for lightwave from Si-waveguide expanding through the MLS-GRIN lens, exiting the MLS-GRIN lens output facet into the air, and entering into the SMF, has been calculated to be 60% (coupling loss of 2.2 dB) at 1550nm (refer to ). The 2.2dB loss can be accounted for by Fresnel reflection losses at the MLS-GRIN-lens/air and air/SMF interfaces for the collimated beam from the MLS-GRIN lens. Since the air/SMF interface typically has a Fresnel loss of 0.15~0.2dB, the MLS-GRIN-lens/air interface accounts for 2dB of coupling loss. If Fresnel reflection loss at MLS-GRIN lens/air interface is minimized by deposition of ARC at the output facet, the best estimated MLS-GRIN lens to SMF coupling loss that can be attained is 1.45dB.
Since the optical mode size from MLS-GRIN lens has been expanded to 9~10μm, in the order of the mode size of the SMF-core, it is expected that the MLS-GRIN lens to SMF has reasonably good lateral and vertical misalignment tolerance. By simulation, it has also been calculated that the additional coupling loss due to ± 1μm lateral and vertical misalignment of MLS-GRIN to SMF is 0.5dB for both cases .
We fabricated a compact Si/SiO2 multilayer on-chip GRIN lens in conjunction with the Si-waveguide on PIC using multilayer lift-off method, and experimentally demonstrated optical mode size transformation from SMF-core (diameter: 9~10μm) to 0.26μm-thick Si-waveguide, and vice versa. The on-chip GRIN lens has thickness of 12μm and lengths of 16μm or 24μm, and it is butt-joint to the Si-waveguides with termination widths of 3μm, 6μm or 10μm. The output beam from the MLS-GRIN lens showed fairly collimated character as the average optical power coupled into the SMF facing the MLS-GRIN lens does not deteriorate even up to distance of 52μm. The best measured coupling loss from MLS-GRIN lens to SMF without ARC is 3.45dB. Assuming that the Fresnel reflection loss at the MLS-GRIN/air interface can be minimized by ARC, the best attainable estimated coupling loss between MLS-GRIN lens and SMF is 1.45dB.
The research is substantially supported by Agency for Science and Technology for Research (A*STAR) of Singapore. We thank the Institute of Microelectronics of A*STAR for usage of their facilities in dicing and lapping.
References and links
1. B. Jalali and S. Fathpour, “Silicon photonics,” J. Lightwave Technol. 24(12), 4600–4615 (2006). [CrossRef]
2. C. Gunn, “CMOS photonics for high-speed interconnects,” IEEE Micro 26(2), 58–66 (2006). [CrossRef]
3. A. Sure, T. Dillon, J. Murakowski, C. Lin, D. Pustai, and D. Prather, “Fabrication and characterization of three-dimensional silicon tapers,” Opt. Express 11(26), 3555–3561 (2003). [CrossRef] [PubMed]
5. T. Shoji, T. Tsuchizawa, T. Watanabe, K. Yamada, and H. Morita, “Low loss mode size converter from 0.3μm square Si wire waveguides to singlemode fibres,” Electron. Lett. 38(25), 1669–1670 (2002). [CrossRef]
6. A. Delâge, S. Janz, B. Lamontagne, A. Bogdanov, D. Dalacu, D.-X. Xu, and K. P. Yap, “Monolithically integrated asymmetric graded and step-index couplers for microphotonic waveguides,” Opt. Express 14(1), 148–161 (2006). [CrossRef] [PubMed]
7. K. Shiraishi and C. S. Tsai, “A micro light-beam spot-size converter using a hemicylindrical GRIN-slab tip with high-index contrast,” J. Lightwave Technol. 23(11), 3821–3826 (2005). [CrossRef]
8. H. Yoda, H. Ikedo, T. Ketsuka, A. Irie, K. Shiraishi, and C. S. Tsai, “A high performance micro-GRIN-chip spot-size converter formed with focused ion-beam,” IEEE Photon. Technol. Lett. 18(14), 1554–1556 (2006). [CrossRef]
9. R. Sun, V. Nguyen, A. Agarwal, C. Y. Hong, J. Yasaitis, L. Kimerling, and J. Michel, “High performance asymmetric graded index coupler with integrated lens for high index waveguides,” Appl. Phys. Lett. 90(20), 201116 (2007). [CrossRef]
10. K. Shiraishi, H. Yoda, A. Ohshima, H. Ikedo, and C. S. Tsai, “A silicon-based spot-size converter between single-mode fibers and Si-wire waveguides using cascaded tapers,” Appl. Phys. Lett. 91(14), 141120 (2007). [CrossRef]
11. H. Yoda, K. Shiraishi, A. Ohshima, T. Ishimura, H. Furuhashi, H. Tsuchiya, and C. S. Tsai, “A two-port single-mode fiber-silicon wire waveguide coupler module using spot-size converters,” J. Lightwave Technol. 27(10), 1315–1319 (2009). [CrossRef]
13. Q. Wang, Y. Huang, T.-H. Loh, D. K. T. Ng, and S.-T. Ho, “Thin-film stack based integrated GRIN coupler with aberration-free focusing and super-high NA for efficient fiber-to-nanophotonic-chip coupling,” Opt. Express 18(5), 4574–4589 (2010). [CrossRef] [PubMed]
14. T. H. Loh, Q. Wang, K. T. Ng, and S. T. Ho, “Design and fabrication of multilayer Si/SiO2 super-high N.A. GRIN lens for nano-waveguide to optical fiber coupling,” in Frontiers in Optics, Technical Digest (CD) (Optical Society of America, 2009), paper CThK2. http://www.opticsinfobase.org/abstract.cfm?URI=CLEO-2005-CTuT2
15. LUMERICAL FDTD SOLUTIONS, http://www.lumerical.com
16. R. G. Walker, “Simple and accurate loss measurement technique for semiconductor optical waveguide,” Electron. Lett. 21(13), 581–583 (1985). [CrossRef]
17. G. Tittelbach, B. Richter, and W. Karthe, “Comparison of three transmission methods for integrated waveguide propagation loss measurement,” Pure Appl. Opt. 2(6), 683–700 (1993). [CrossRef]
19. Q. Wang, T. H. Loh, K. T. Ng, and S. T. Ho, “Design and analysis of optical coupling between silicon nanophotonic waveguide and standard singlemode fiber using an integrated asymmetric Super-GRIN lens,” IEEE J. Sel. Top. Quantum Electron. (to be published). [PubMed]