Open Access
Add to BibSonomyAbstract
We demonstrated the potential application of III-V/polymer nanowires for photonic integrated circuits in a previous paper. Hereby, we report the use of a spot size converter based on 2D reverse nanotaper structure in order to improve the coupling efficiency between the nanowire and optical fiber. A total coupling enhancement of up to a factor 60 has been measured from an 80 nm × 300 nm cross-section tip which feeds an 300 nm-side square nanowire at its both ends. Simultaneously, micro-radius bends have been fabricated to increase the circuit density; for a radius of 5 μm, the 90° bend losses were measured as low as 0.60 dB and 0.80 dB for TE and TM polarizations respectively.
©2007 Optical Society of America
1. Introduction
The key issue for the realization of more sophisticated photonic integrated functions is to improve the compactness of the related photonic integrated circuits (PIC's). This can not be reached using conventional integrated optics components due to the low light confinement in the waveguide core. One solution emerges from the use of a high refractive index contrast between core and cladding of the waveguides; light propagation getting then a very high confinement. In parallel, single-mode operation imposes the waveguide dimensions to be down to the sub-micron scale, leading so to nanowires based components.
The high density of elementary components in PIC's does not only result from the strong decrease of waveguide dimensions and inter-waveguide spacing but relies also on the possibility to achieve very low radius bends.
The main drawback of such nanowires is their poor coupling efficiency with the external world, i.e. optical fibers. The nanowire mode size is indeed at least one hundred times smaller than the one of a fiber. Added to the input reflection, this leads to increase the coupling loss which can be finally up to 30 dB. The obvious solution appears as a taper whose dimensions vary from the fiber ones to the nanowire ones. Compared to the two-dimensional (2D) configuration (only laterally tapered) whose yield is limited by the low thickness of the nanowire, the 3D mode size converter is theoretically more efficient but its length is too high (1 mm) [1] and the backreflections remain high at the input due to the high mode effective index [2]. Other coupling techniques based on grating have been developed but are extremely sensitive to the polarization state and need to be combined with a 2D spot size converter [3,4]. A last solution is a reverse taper structure which consists in gradually reducing the nanowire dimension to define a tip at the launching facet. This device takes benefit from the guided mode delocalization at this tip. Recently, Vivien et al [5] proposed a theoretical study of a high efficiency 3D structure. Since this structure requires extremely high processing control accuracy, the simplest technological coupling method appears as the laterally tapered mode size converter which can be defined in the same processing level as the waveguide. To date, since most of nanowires and nanooptics devices are fabricated in the Silicon On Insulator SOI-based materials [6,7] due to its high technological maturity, all such coupling devices are made of silicon [8–10]. In addition, most of these later structures requires a polymer waveguide overlaying the taper to avoid large substrate leakage.
Since the intrinsic optical properties of silicon make the integration of active functions at telecom wavelengths difficult, III-V semiconductor materials are a substantial alternative to realize both passive and active functions. If the waveguide design is almost the same as the one proposed on SOI line, here cladding and core are constituted of polymer material and III-V semiconductor respectively [11].
We will focus hereby on a III-V/polymer nanophotonic waveguide corresponding to an 300 nm-side square GaAs nanowire embedded in a benzocyclobutene (BCB) matrix. Particularly, we report the results obtained from 2D reverse nanotaper for butt-coupling as well as on low radius bends. Using the developed technology particularly based on polymer wafer bonding, PIC's can be realized whatever the type and the topology of host substrate are.
2. Spot size converter design
The III-V/polymer reverse linear nanotaper structure is presented in Fig. 1.
We chose here GaAs (n ~ 3.37 at 1.55 μm) for the waveguide core but InP could also be used with the same propagation features. Silicon (n ~ 3.47) was chosen as host substrate and BCB (n ~ 1.53) was selected for its properties as thermosetting polymer. The BCB layer (bottom cladding layer) between the core and the Si substrate has to be thick enough to avoid any light leakage (since the mode size will be expanded at taper end in the surrounding BCB compared to waveguide one) into the substrate but not too thick to allow a correct cleaving of the facets; its thickness is about 4 μm. By the way, the top cladding layer is 2μm thick. Since the nanowire height is fixed to 300 nm by the epistructure, the mode expansion area and thus the mode overlap with the fiber only vary with the tip width w.
Fig. 2. Mode mismatch loss added to reflection loss versus the tip width w considering a coupling with (a) a cleaved and (b) a lensed fiber.
Simulations were performed to determine the geometrical features of the nanotaper. In a first approximation, simple modal analysis and overlap integral formulation were used. A 3D semi-vectorial imaginary-distance beam propagation method (BPM) [12] was used to get the mode profile. The estimation of the coupling loss Pmr was then calculated taking into account the loss due to the mode mismatch (evaluated by a mode overlap integral calculation) between the fiber and the tip and the backreflections which were estimated assuming a Fresnel reflectivity at the nanotaper input (Fig. 2). This study was performed at 1.55 μm wavelength considering a cleaved fiber with a 1/e2 mode field diameter (MFD) of 9 μm and a lensed fiber having a MFD of 3 μm. Due to the rectangular shape of taper tip, mode profile evolution versus tip width behaves differently in TE and TM modes and so the corresponding mode mismatch loss.
Whatever the polarization is, the lowest Pmr loss is attained for smaller tip widths using the lensed fiber which is obviously linked to its smallest mode dimensions. For each type of fiber, the TE-like mode loss can become less than 1 dB for tip widths lower than 120 nm. Considering the TM mode, the coupling loss is at least 2 dB higher compared to TE mode for w = 120 nm, but this difference rapidly voids as the width is decreased. For such low tip widths, the optical mode extends strongly outside the core into the BCB improving the mode overlap between the tip and the fiber modes. In addition, due to this delocalization, the effective index of the propagating mode is close to the BCB refractive index (1.53) which leads to very low backreflections (around 0.25 dB for each polarization if w is smaller than 100 nm).
Fig. 3. Total nanotaper coupling loss versus the tip width w for 2 nanotaper lengths (90 and 150 μm). A lensed fiber is considered as launching conditions.
Considering the results from the Fig. 2 that show higher coupling efficiency for tip widths smaller than 160 nm, we focused on this dimension range to study, by means of a 3D FD-BPM modeling (along the propagation axis) tool [13,14] and Fresnel reflectivity, the total coupling loss within a nanotaper considering a lensed fiber launch. It was shown that with high refractive index contrast coupling structures, the taper length can be shorted down to 200 μm [8] and even to 40 μm [9]. Considering our study, two coupler lengths have been arbitrarily used: 90 and 150 μm. We defined as the total coupling loss of the nanotaper the sum of the different loss sources which are backreflections, mode mismatch and mode conversion between the taper mode and the waveguide one. The calculated total coupling loss versus tip width and taper length is shown in Fig. 3. The 3D FD-BPM tool gives a more complete assessment of the light coupling efficiency since it takes into account the mode conversion loss arising within the nanotaper propagation. This explains the difference between Fig. 2 and Fig. 3, in particular for the narrow tip widths: the TE-mode conversion loss is larger than the TM-mode one.
Moreover, we can see that there is almost no difference versus nanotaper length. Nevertheless, a small difference (around 1dB) is observed for TE polarization when tip width is below 80 nm that also tends to show that mode conversion loss within the nanotaper for TE-mode is more sensitive (either to tip width or taper length) than for TM-mode.
If the lowest total coupling loss (calculated to be lower than 1 dB) is obtained for different tip width values depending on polarization (w = 120 nm in TE mode and w = 40 nm in TM mode), the 80 nm wide tip is a well trade-off solution in order to limit the nanotaper sensitivity to the polarization without having a high loss (around 2 dB for each polarization state). So we chose this tip width for the nanotaper fabrication.
Besides, after several tens of microns of propagation in a 300 nm × 300 nm nanowire including or not a nanotaper at its input, the optical power in the waveguide has been normalized to the injected one. The theoretical coupling efficiency enhancement, Ceff has so been defined as the ratio of the normalized power in a nanowire with nanotaper to the one in a nanowire without nanotaper. Light injection is made by means of a lensed fiber. Focusing on an 80 nm tip width, it is evaluated to about 6.4 for a 90 μm-long nanotaper for both polarization states and 5.8 and 6.6 for respectively TE and TM modes when the nanotaper length is 150 μm. The sensitivity to the lensed fiber misalignment with respect to its optimum coupling position was evaluated using 3D FD-BPM. For a 1 dB of additional coupling loss, the tolerance is calculated to be 1.5 and 1 μm in TE and TM polarization respectively.
3. Devices fabrication method
The fabrication process is based on three main steps which are successively (i) the waveguide core patterning using electron beam lithography (EBL) and (ii) reactive ion etching (RIE), and finally (iii) the BCB wafer bonding that will form the bottom cladding layer as well as the binding layer to the host substrate. These processing steps have been optimized to minimize the GaAs core sidewall roughness which is known as the main source of propagation loss through a strong scattering phenomenon [6,7,11].
The required epitaxial structure for the core fabrication is realized by molecular beam epitaxy on a GaAs substrate: it consists of the GaAs core layer (300 nm-thick) grown on an etch stop layer made of GaInP (500 nm-thick). The core patterns (300 nm-wide nanowires and their tapered ends down to 80 nm wide tips) are defined by EBL on a HydrogenSilsesQuioxane (HSQ) resist film [15] which will be used as etching mask (Fig. 4(a)). The GaAs core layer is then defined using a SiCl4/Ar RIE (Fig. 4(b)) [16]. Afterwards, a 2 μm thick BCB (CycloteneTM 3022-35) film was deposited onto host Si and processed GaAs substrates. The host Si substrate was then flipped onto the GaAs epiwafer, taking care to align their cleavage planes. The bonding was realized by firstly applying a pressure of 1 Bar on the assembly at 130°C followed by a hardbake at 250°C; both steps during one hour [17]. By this way, BCB was fully polymerized and both wafers were tightly linked. Next, GaAs growth substrate and GaInP etch-stop layer were removed using chemical etchings. A last BCB layer was spun and hardbaked to fully cover inserted nanopatterns and realize the top cladding layer. Finally, the nanowires and nanotapers exhibited a GaAs core surrounded with about 4 μm and 2 μm of BCB respectively as bottom and top cladding layers. The nanowires with tapered ends were cleaved to form the spot size converter tip; a view of a cleaved waveguide facet is shown in Fig. 4(c).
Fig. 4. SEM views of (a) HSQ pattern of a 90° bend (1μm radius), (b) GaAs 80nm-wide tip after RIE, and (c) cleaved waveguide (tip) facet.
4. Devices characterization
4.1 Measurement setup
The optical connection between the nanowire structures and single-mode fibers were done by butt coupling. The launching fiber was a lensed one (MFD = 3 μm) linked to a laser source (EXFO FLS2600) emitting at 1.55 μm whereas a cleaved fiber (MFD = 9 μm) was used for light collection at the output. A Lefevre loop polarization controller was inserted between the light source and the launching fiber allowing the selection of either TE or TM polarization state. Before aligning the output fiber, a microscope objective (40× magnification) combined with an infrared vidicon tube camera (Hamamatsu C-2741) were aligned at the nanowire output in order to obtain the near field pattern. The polarization state was checked by inserting a free space optics polarizer in front of the camera and visualizing the near field spot intensity. All measurements have been made maximizing the output power for a given polarization state; the rigorous analysis of the misalignment effect has not been performed since our test bench does not allow sufficiently accurate positioning recording.
4.2 Reverse nanotapers
On Fig. 5, the near field pattern of output waveguides are shown (TE mode). For comparison, “classical” (4 μm wide) integrated optics waveguide output is exhibited. Obviously, the same infrared camera sensitivity is kept for all structures.
We can clearly notice the mode expansion into the BCB matrix due to the reverse nanotaper comparing the near field pattern of these both nanostructures: the smaller one corresponds to the 300 nm × 300 nm nanowire and the larger one to the tip of a nanotaper (80 nm × 300 nm). By comparison to the classical integrated optics waveguide reference, this delocalization all around the tip gives rise to a circular mode profile of several micrometers.
The input coupling efficiency enhancement CeffL is obtained by the comparing the output power from nanowires with input nanotaper to nanowires without the nanotaper (lensed fiber launching is used for both cases). The ratio of these two output powers gives the value of CeffL. Measurement results are given in Table 1.
Fig. 5. Near field views, in TE polarization, of different waveguide structures showed in inserts; (a) classical reference InP-based waveguide, (b) 300 nm × 300 nm nanowire, (c) 80 nm × 300 nm reverse nanotaper tip.
First, it is noted that the input coupling enhancement decreases with increasing the nanotaper length. Besides calculated and measured values are in close agreement. However the coupling efficiency enhancement is globally higher for TE polarization state. One cause could be the leakage of the optical mode into the substrate as demonstrated by modeling results in Ref. [9] because of the vertical rectangular shape of the GaAs core along the nanotaper which induces mode delocalization. Nevertheless, our modeling results do not point out a significant excess loss due to this leakage for our nanotaper structures showing that a distance of 4.2 μm between the host substrate and the nanowire is large enough. The second cause can be related to the sidewall roughness which leads to higher scattering propagation loss [6,7,11] as the nanotaper length is raised. Thirdly, we observe from the Fig. 3 that a little change of the tip width (several nanometers) around 80 nm during the technological process can result in a relatively high change of the coupling for both polarizations.
Table 1. Measured and calculated input coupling efficiency enhancements (input nanotaper coupled to a lensed fiber)
Concerning the output, light is collected by a cleaved fiber and similarly to CeffL, the coupling efficiency enhancement CeffC is evaluated by comparing the output power issued from nanowires incorporating or not an output nanotaper (input taper is always present). Table 2 reports both CeffC and Ctot (CeffL are obtained from corresponding Table I values). We can notice that CeffC (nanotaper to a cleaved fiber coupling) is greater than CeffL (lensed fiber to nanotaper coupling) mainly because the coupling efficiency of the nanowire without nanotaper is very small when using a cleaved fiber.
Table 2. Measured coupling efficiency enhancements using only output nanotaper coupled to a cleaved fiber (CeffC) and both input and output nanotapers (Ctot)
Furthermore, CeffC value is greater for the largest length (300 μm). The scattering loss is here counterbalanced by a lower mode conversion loss between the modes of the tip and the waveguide owing to a low transition slope of the nanotaper which allows a more efficient mode matching.
Finally, taking into account both nanotapers (coupled to a lensed fiber at the input and to a cleaved fiber at the output), the total coupling enhancement that can be obtained is about 60 (17.8 dB) for TE mode and around 40 (16 dB) for TM mode.
In a previous paper [11], we evaluated the propagation loss of a straight 300 nm × 300 nm cross-section nanowire using the cut-back method. It has been found to be around 5 dB/mm for both polarizations. From this cut-back method, the total coupling loss has been estimated to 6.5 dB ± 2 dB and 12 dB ± 2 dB for TE and TM polarization states respectively. These values confirm the ones of the coupling efficiency enhancement showing a better efficiency for the TE mode. As our measurement set-up uses two different kinds of fibers, lensed and cleaved for respectively launching and collecting light, input and output coupling losses can not be extracted separately.
4.3 Bends
To obtain the excess loss of a bend, we have compared the measured insertion loss between a straight waveguide and a waveguide of same length (1) containing two 90° bends separated by a distance of 50 μm. The measured loss is presented in tables 3 and 4. The excess (Lexcess) and total (Ltot) losses are linked by:
where Lstraight is the linear propagation loss (~ 5 dB/mm [11]) in a straight nanowire.
Table 3. Measured 90° bend losses in TE polarization
Table 4. Measured 90° Bend Losses in TM Polarization
When light is propagating into a curved waveguide, mode shifts towards its outer edge and is consequently perturbed; this effect is accentuated when the bend radius is decreased. So the excess bending loss includes a mode radiation loss, an additional scattering loss caused by the increased interaction of the mode with sidewall surface roughness and coupling loss at the connection between bend and straight waveguide [18,19]. For both TE and TM polarizations, the lowest and highest excess bending loss were measured for a radius of 5 μm and 40 μm respectively. Focusing on the linear excess loss, it increases as long as the radius goes down. This confirms the effect of mode shifting towards the nanowire outer sidewall. The low linear excess loss of the 40 μm-radius bend is not enough low to compensate the longer interaction between the optical mode and the outer roughened sidewall due to the very large bend length compared to the one of the other curves. This is in agreement with already reported behaviors showing that an optimum value of bend radius should exist [18,19]; this value is a trade-off between loss due to the mode shifting (that increases when radius decreases) and scattering loss of the waveguide (that increases when radius increases since waveguide is longer and which is favored by the mode shifting). These values of loss and so the optimum bend radius are strongly dependent upon the fabrication procedure and in particular the sidewall roughness.
The 5 μm radius exhibits the lowest total loss with 0.60 dB and 0.80 dB for TE and TM polarizations respectively; this points out the polarization dependence and the highest loss for TM mode as shown by Sakai et al. [20] for such small radii.
5. Conclusion
We have presented two passive optical devices which are basic components of PIC's but which required to be adapted to be used with III-V/polymer nanowires in order to develop III-V nanophotonic circuits. The first one is a spot size converter (nanotaper) with a short length based on a laterally reverse tapered structure. The optical mode is tightened from the tip to the 300 nm × 300 nm square nanowire. Modeling showed that the coupling efficiency is the same for both polarizations for an 80 nm tip width and a nanotaper length as low as 90 μm. In this case, the coupling loss is calculated to be down to 2 dB with a high misalignment tolerance. Characterization pointed out that even if the polarization insensitivity was not preserved, the global coupling efficiency including both input and output nanotapers can be improved up to 17.8 dB for TE mode and 16 dB for TM mode. Moreover, in spite of the same theoretical results obtained for both polarizations for an 80 nm wide tip, experimental coupling efficiencies are polarization dependent. To enhance the nanotaper coupling, a further improvement of the III-V technological patterning steps could reduce the sidewalls roughness and result in a better definition of the tip. Besides an alternative way could consist in favoring one polarization state against the other by choosing the appropriate tip width (Fig 3).
The second device is a micro-radius bend which allows shortening the propagation distances in the photonic integrated circuits. For a radius of 5 μm, the 90°-bend loss was measured as low as 0.60 dB and 0.80 dB for TE and TM modes respectively.
Acknowledgments
The authors thank Francis Mollot and the Epitaxy group from IEMN for the supply of epitaxial wafers and IRCICA for its support.
References and links
1. M. Fritze, J. Knecht, C. Bozler, C. Keast, J. Fijol, S. Jacobson, P. Keating, J. Leblanc, E. Fike, B. Kessler, M. Frish, and C. Manolatou, “Fabrication of three-dimensional mode converters for silicon-based integrated optics,” J. Vac. Sci. Technol. B 21, 2897–2902, (2003). [CrossRef]
2. A. N. M. M. Masum Choudhury, T. R. Stanczyk, D. Richardson, A. Donval, R. Oron, and M. Oron, “Method of improving light coupling efficiency between optical fibers and silicon waveguides,” IEEE Photon. Technol. Lett. 17, 1881–1883 (2005). [CrossRef]
3. S. Lardenois, D. Pascal, L. Vivien, E. Cassan, S. Laval, R. Orobtchouk, M. Heitzmann, N. Bouzaida, and L. Mollard, “Low-loss submicrometer silicon-on-insulator rib waveguides and corner mirrors,” Opt. Lett. 28, 1150–1152 (2003). [CrossRef] [PubMed]
4. D. Taillaert, W. Bogaerts, P. Bienstman, T.F. Krauss, P. Van Daele, I. Moerman, S. Verstuyft, K. De Mesel, and R. Baets, “An out-of-plane grating coupler for efficient butt-coupling between compact planar waveguides and single-mode Fibers, ”IEEE J. Quantum Electron. 38, 949–955 (2002). [CrossRef]
5. L. Vivien, S. Laval, E. Cassan, X. Le Roux, and D. Pascal, “2-D taper for low-loss coupling between polarization-insensitive microwaveguides and single-mode optical fibers,” J. Lightwave Technol. 21, 2429–2433, (2003). [CrossRef]
6. T. Tsuchizawa, K. Yamada, H. Fukuda, T. Watanabe, J. I. Takahashi, and M. Takahashi et al., “Microphotonics devices based on silicon microfabrication technology,” IEEE J. Sel. Top. Quantum Electron. 11, 232–240 (2005). [CrossRef]
7. P. Dumon, W. Bogaerts, V. Wiaux, J. Wouters, S. Beckx, J. Van Campenhout, D. Taillaert, B. Luyssaert, P. Bienstman, D. Van Thourhout, and R. Baets, “Low-loss SOI photonic wires and ring resonators fabricated with deep UV lithography,” IEEE Photon. Technol. Lett. 16, 1328–1330 (2004). [CrossRef]
8. T. Shoji, T. Tsuchizawa, T. Watanabe, K. Yamada, and H. Morita, “Low loss mode size converter from 0.3 μm square Si wire waveguides to singlemode fiber, ”Electron. Lett. 38, 1669–1670 (2002). [CrossRef]
9. V. R. Almeida, R. R. Panepucci, and M. Lipson, “Nanotaper for compact mode conversion,” Opt. Lett. 28, 1302–1304 (2003). [CrossRef] [PubMed]
10. S. J. McNab, N. Moll, and Y. A. Vlasov, “Ultra-low loss photonic integrated circuit with membrane-type photonic crystal waveguides,” Opt. Express 11, 2927–2939 (2003), http://www.opticsexpress.org/abstract.cfm?id=77823 [CrossRef] [PubMed]
11. D. Lauvernier, S. Garidel, M. Zegaoui, J.P. Vilcot, and D. Decoster, “GaAs/polymer optical nanowires: fabrication and characterization,” Electron. Lett. 42, 217–219 (2006) [CrossRef]
12. J. C. Chen and S. Jungling, “Computational of high-order waveguide modes by imaginary-distance beam propagation method,” Opt. Quantum Electron. 26, 199–205 (1994). [CrossRef]
13. Y. Chung and N. Dagli, “An assessment of finite difference beam propagation method,” IEEE J. Quantum Electron. 26, 1335–1339 (1990). [CrossRef]
14. H. Deng, G. H. Jin, J. Harari, J. P. Vilcot, and D. Decoster, “Investigation of 3-D semivectorial finite-difference beam propagation method for bent waveguides, ”J. Lightwave Technol. 16, 915–922 (1998). [CrossRef]
15. D. Lauvernier, J. P. Vilcot, M. Francois, and D. Decoster, “Optimization of HSQ resist e-beam processing technique on GaAs material,” Microelectron. Eng. 75, 177–182 (2004). [CrossRef]
16. D. Lauvernier, S. Garidel, C. Legrand, and J. P. Vilcot, “Realization of sub-micron patterns on GaAs using a HSQ etching mask,” Microelectron. Eng. 77, 210–216 (2005). [CrossRef]
17. D. Lauvernier, J. P. Vilcot, S. Garidel, S. Mc Murtry, and D. Decoster, “Benzocyclobutene wafer bonding for III-V nanophotonic guiding structures, Electron. Lett. 41, 1170–1171 (2005) [CrossRef]
18. R.J. Deri and E. Kapon, “Low-loss III-V semiconductor optical waveguides,” IEEE J. Quantum Electron. 27, 626–640 (1991) [CrossRef]
19. Y.A. Vlasov and S.J. McNab, “Losses in single-mode silicon-on-insulator strip waveguides and bends,” Opt. Express 12, 1622–1631 (2004), http://www.opticsexpress.org/abstract.cfm?id=79595 [CrossRef] [PubMed]
20. A. Sakai, T. Fukuzawa, and T. Baba, “Estimation of polarization crosstalk at a micro-bend in Si-photonic wire waveguide,” J. Lightwave Technol. 22, 520–525 (2004) [CrossRef]
