In the last few years Mid InfraRed (MIR) photonics has received renewed interest for a variety of commercial, scientific and military applications. This paper reports the design, the fabrication and the characterization of SiGe/Si based graded index waveguides and photonics integrated devices. The thickness and the Ge concentration of the core layer were optimized to cover the full [3 - 8 µm] band. The developed SiGe/Si stack has been used to fabricate straight waveguides and basic optical functions such as Y-junction, crossings and couplers. Straight waveguides showed losses as low as 1 dB/cm at λ = 4.5 µm and 2 dB/cm at 7.4 µm. Likewise straight waveguides, basic functions exhibit nearly theoretical behavior with losses compatible with the implementation of more complex functions in integrated photonics circuits. To the best of our knowledge, the performances of those Mid-IR waveguides significantly exceed the state of the art, confirming the feasibility of using graded SiGe/Si devices in a wide range of wavelengths. These results represent a capital breakthrough to develop a photonic platform working in the Mid-IR range.
© 2014 Optical Society of America
In the last few years Mid InfraRed (MIR) photonics has received renewed interest for a variety of commercial, scientific and military applications [1–4]. In particular, the impressive progression in the development of compact widely tunable quantum cascade lasers (QCL) opened the way to the commercial exploitation of a number of MIR based techniques typically relegated to a lab environment because of size, cost or operation constrains. Furthermore, with the availability of QCL sources that can continuously cover the full MIR spectrum up to 12 µm or beyond  many other applications domains can be today foreseen. As suggested by Soref , the replication of the development scheme undertaken for near-IR data communications, where almost any photonic devices has been integrated in Si, would be tremendously beneficial for future applications where size portability and versatility are an issue.
However, while MIR sources are mature enough for commercialization, achievements are much scarcer concerning fast efficient detectors and low losses passive elements for beam handling and guiding. Some results on low-losses integrated passive devices fabricated on SOI/SOS were recently published [6, 7]. Despite the attractive performances of these devices issued from the natural extension of the platforms used for telecom applications, the operational range is substantially limited by the presence of SiO2, which is known to have strong adsorption above 3.6 µm and OH bonds absorbing at lower wavelengths limiting any practical application in this wavelength region .
In this paper we report the design, the fabrication and the characterization of SiGe/Si based graded index waveguides and Photonics Integrated Circuit – PIC – devices. The thickness and the Ge concentration of the core layer were optimized to cover the full [3 - 8 µm] spectrum. Using our 200 mm Si manufacturing facilities, we have developed a SiGe/Si mid-IR photonics platform with a set of design rules suitable to different waveguides structures and to investigate the potential of this material. In the following, details about the design, the fabrication and test of basic and complex functions realized according to this platform rules will be reported and discussed.
Si1−xGex alloys have been investigated as possible material for microelectronic applications since the late 1950s. Novel depositions techniques capable to grow epitaxially this material on Si substrates at temperatures as low as 700 °C, thus preventing dopant redistribution in existing structures, are available since the early nineties . The interest in SiGe for integrated photonic applications in the Near-IR and MIR range is threefold: first, according to , when SiGe used together with Si as a cladding, the range of usability can be extend up to 8 µm, making this material a perfect candidate for applications involving QCL sources over a broad range. Second, many optical properties, such as the n index or bandgap, can be continuously varied and finely tuned with the Ge content in the alloy . Finally, as reported in [12,13], SiGe exhibits superior non-linear properties which might be advantageously used in many applications where high detection performances are required.
2. Waveguide design and modeling
Achieving low losses operation in the whole 3 to 8 µm bandwidth with a single stack of layers is an ambitious challenge. In a step index waveguide of a given dimension, the electromagnetic field progressively expands out of the core into the cladding as the wavelength increases, with the detrimental increase of losses. Losses are also due to the core/cladding interface roughness, scaling with the square of the electric field difference and Δ(n2) at the core/cladding interface . Therefore, full operability over a wide range of wavelengths can be achieved only by adapting simultaneously the size and the thickness of the core layer. If the fabrication process can be adapted for a specific set of wavelengths, the overall fabrication process becomes complex and unreliable when the bandwidth to be addressed simultaneously is several µm wide.
Although using SiGe/Si waveguides helps in minimizing the material losses in the 3-8 µm range, the specific index profile in the core layer has been designed to minimize diffusion losses over a large wavelength bandwidth. Taking into account the process capabilities available in our IC pilot line, we have designed a specific linearly graded stack suitable to be used in conjunction with QCL sources, where the Ge concentration has been ramped up to 50%. The proposed profile is sketched in Fig. 1, showing the graded index variation in the growth direction (perpendicular to the (001) substrate surface) and a step index profile in the in-plane direction. The idea is to (i) properly tailor the waveguide width along to the in-plane direction with UV-lithography and etching and (ii) limit diffusion losses in the perpendicular direction via the use of an interface-free graded index stack suitable to cover the whole range of operability required.
The optical properties of this stack for large bandwidth singlemode and low loss operation have been modeled using RSoft mode solver . The refractive index of the Si-Ge alloy was calculated by the following linear formula nSi(1-x)Gex = x nGe + (1-x) nSi, where nSi and nGe were taken from . An optimal match between the laser sources and SiGe/Si waveguides was found for stacks where the maximum Ge concentration was 43%. The optical behavior of this stack is summarized in Fig. 2, where the operating wavelength range is reported as a function of the waveguide size. The blue line shows the limit below which multi-mode operation starts to appear. The orange line sets the limit above which the electromagnetic optical field extends well beyond the core region and light confinement is not enough to provide low losses guiding. For a waveguide of a given size, the operation range with acceptable losses may extend between these two limits. Furthermore, the same vertical stack can be dimensioned for singlemode operation by changing the waveguide width, with good guiding properties for wavelengths in the [3 to 8 µm range], enabling an easy co-integration of optical devices at different wavelengths on the same chip. It should be noted that these simulations were done in semi-vectorial mode for quasi TM polarization as it is intended to be butt-coupled with TM single polarization QCL arrays.
3. Stack growth and device fabrication
To grow the thick SiGe core layers with the triangular Ge concentration profile, a 200 mm Epi Centura Reduced Pressure – Chemical Vapor Deposition (RP-CVD) tool from Applied Materials has been used at the pressure of 20 Torr. To promote the glide of misfit dislocations in the relaxing layers and benefit from high growth rates, the growth temperature was as high as 850 °C to 900 °C for SiGe [16–18]. To change the relative concentration of Ge, the ratio between germane and dichlorosilane was gradually varied during growth. Secondary Ion Mass Spectroscopy (SIMS) depth profiling of the Si and the Ge atoms was used to determine the Ge concentration profile in those 3 µm thick SiGe layers. The Ge content was controlled, with a rather linear increase up to 40% at in the first half of the profiled layer and a quasi-symmetric decrease back to 0% in the second half. A typical Ge profile concentration as function of the depth is shows Fig. 3.
As shown in Fig. 4, the core of the waveguide has been patterned by conventional photolithography and deep reactive ion etching process (D-RIE) down into the Si substrate. Then, a 10 µm thick Si cap layer was deposited at 850°C, 20 Torr by RP-CVD technique. The use of SiH4 precursor allowed the conformal epitaxial growth of the Si capping layer all over the patterned structure including the edges of the SiGe waveguides. Chemical mechanical Polishing (CMP) was finally performed to recover a flat surface. A typical cross sectional image of fully processed waveguide is displayed in the inset of Fig. 4.
The thickness of the Si capping layer after CMP was designed to allow complete optical isolation of our photonic integrated device from the surrounding environment and, depending on the wavelength operational range, it has been varied between 5 and 10 µm. It is worth to note that, with the process integration scheme used, Si wafers are flat after the whole PIC fabrication process and the topmost Si layer is compatible with standard IC and Micro-Electro-Mechanical Systems/Micro-Opto-Electro-Mechanical Systems fabrication processing. It thus allows the further integration of other functional devices on the same substrate.
4. Device characterization
Using the process flow described above we have fabricated two sets of devices suitable for operation in the MIR band (i.e. at 4.5 µm and 7.4 µm). According to the optical properties measured on the SiGe/Si waveguide stack, the waveguide width was 3.3 µm and 7.0 µm for single mode operation at λ = 4.5 µm and λ = 7.4 µm, respectively.
Losses in straight waveguides were evaluated by measuring the finesse of the Fabry-Perot resonators formed by cleaving two opposite ends of the waveguides. Conventional cut-back methods are indeed not suitable for propagation loss measurements of low-loss semiconductor waveguides. This method, established in the late 80s [19,20], has been recently upgraded for use with polychromatic light with either a thermal source or a DFB-QCL in pulsed mode operation . As sketched in Fig. 5, when a pulsed DFB-QCL source is shined through the waveguide, the change in wavelength during the pulse results into Fabry Perot fringes out of the optical cavity defined the facets. Since the modulation intensity is not related to the amount of light coupled into the waveguide, this technique fits very well with our measurement requirements where the lack of fibers for this wavelength range and the size of the waveguides close to the diffraction limit makes beam injection problematic and poorly reproducible.
As discussed in , the fringe contrast is directly related to losses by the formula K = (Tmax-Tmin)/(Tmax + Tmin) = 2R’/(1 + R’2) where K is the fringe contrast, R’ = Re-αL with R the waveguide facet reflectivity, L the length of the waveguide and α the propagation loss coefficient. In this configuration, the source of losses is twofold: the propagation losses in the waveguide structure itself and the reflection from the cleaved facet. If the latter contributor is known to be close to 0.3 and can be estimated by measuring samples of different lengths, the former contributor is predominant in long enough waveguides. Small facet loss variations have then a negligible impact on the overall losses values, which mainly come from the waveguide itself.
According to the procedure described above, straight waveguides were measured using QCL sources emitting at 4.5 and 5.65 µm for 3.3 µm wide devices and emitting at 7.4 µm for 7.0 µm wide devices. Losses < 1 dB/cm at 4.5 µm, <1.5 dB/cm at 5.65 µm and < 2 dB/cm at 7.4 µm were measured in SiGe waveguides fabricated using this process flow. To the best of our knowledge, the results presented here above are state of the art performances for Mid-IR waveguides, confirming the suitability of single stack, linearly graded SiGe/Si devices for operation in a wide range of Mid-IR wavelengths.
Based on these waveguides, we have designed several basic optical guided functions such as S-Bend waveguides, Y-junctions, crossing waveguides, and symmetric and asymmetric evanescent couplers. Micrographs of these basic functions are reported in Fig. 6.
To measure the losses introduced by the different optical functions, we used a near field imaging setup based a cryogenic cooled quantum-well infrared photodetector (QWIP) camera. Despite this camera has a central working wavelength of 8.5 µm, the sensitivity at 4.5 µm is high enough, to image the output of the waveguides. It is otherwise linear with the impinging flow and extremely robust to optical damage. The output of the different optical functions could therefore be directly imaged and the relative transmitted intensity calculated. The device layout used for the optical functions measurement is schematized in Fig. 7: the light entering the waveguide was split by a Y-junction and a reference output path was added next to each set of optical functions. Transmission losses could therefore directly be estimated by comparing the measured beam intensity values. For very low loss optical functions, such as S-Bend or crossing waveguides the measurement accuracy has been increased by designing several identical functions in series. The results obtained on the different functions are summarized in Table 1 and the behavior of each type of function is discussed in the following.
This function was used as splitter in each test layout. The behavior of a Y junction is explained in detail in , complete description of the Y junction geometry is shown in Fig. 8(a) and the parameter values are detailed in Table 2. However, it is worth to note that one of the critical point during the fabrication of this device is the control of the ridge shape at the separation between the two arms of the junction. In any practical application the limited resolution of the photolithographic and etch processes over thick layers may round the junction and the subsequent cladding deposition may results in voids. To circumvent these technological constrains a 2 µm wide rectangular shaped notch had been added at the arms separation. The results of the characterization were compared with the behavior modeled by semi-vectorial BPM tools: losses are above the expected theoretical losses (0.4 dB) but the excellent balancing allows the easy measurements of other functions.
As waveguide manufacturing is addressed by planar technology, crossing of waveguides in the X junction is mandatory to reduce the footprint of a photonic circuit. The influence of the crossing angle on the losses has been described extensively in [23,24]. Close to 4.5 µm, the radius of curvature is still small and it is possible to maintain 90° angles to minimize losses without increasing too much the footprint of the circuit. However, to minimize further diffraction induced losses we have locally enlarge the width of the waveguide at the waveguide intersection. To find the best configuration, three different waveguide widths have been tested, namely 3.3, 5 and 7 µm for the X-junction 1, 2 and 3, respectively (see details in Table 3). Note however that the width of the waveguide linking the x functions has been kept to 3.3 µm in order to ensure single mode propagation. Results show that losses are as low as 0.1 dB per crossing independently of the waveguide width at the intersection between guides.
At 7.4 µm, the waveguide cross section is more rectangular and the mode larger. Thus the minimum allowed radius of curvature is higher and reducing the angle of the crossing waveguide becomes important. Again we evaluated 3 different angles between crossing waveguides: 70, 80 and 90° respectively for the X-junctions 1, 2 and 3. Results show that, as one would expect, losses decreases as the angle increases, but losses are still acceptable even for the 70° X junction.
To obtain theoretical bending losses lower than 0.1 dB/cm, the calculated radius of curvature for S-bends was 550 µm for 4.5 µm waveguides and 1350 µm for 7.4 µm waveguides. Different geometries for S-bends junction were design, fabricated and tested: According to the schematic drawing of Fig. 9, constant radius of curvature, cosine bend, sine bend, and continuous curvature bends with different widening coefficients were implemented as to interconnect the waveguide. Constant curvature S-bend design is self – explanatory. Sine and Cosine S-bend were designed as given by RSoft suite. P-curve bends were designed according to [22, 23] where the L parameter has been obtained by minimization of losses along the P-curve path, the widening coefficient K is defined by δρ = K C where δρ is the waveguide widening in µm and C is the curvature along the path in µm−1. As discussed in , these parameters identify an optimized set of polynomial curves (P-curves) that minimize the losses at the junction between waveguides of different curvatures. The parameterization of the P-curve design is summarized in Table 4.
The obtained results show that losses in bent waveguides can be reduced by a factor of 6 between simple concatenation of constant radius paths and optimized paths using P-curves, confirming the effectiveness of the design methodology.
4.4. Evanescent couplers
3dB evanescent couplers were designed at both 4.5µm and 7.4µm wavelengths using standard coupled mode theory and refined by BPM modeling. Both symmetric and asymmetric couplers were designed. Details on the coupler geometry are given in Fig. 8(b) and the corresponding dimensions are shown in Table 5. Figure 10 shows the transmission of the different couplers for wavelengths close to 4.5 µm, as calculated from BPM modeling. As showed in this picture, the asymmetric coupler design is less sensitive to the working wavelength and better fit to the requirements of wide operational range PICs. The coupler design to work at 7.4 µm showed the same kind of behavior, exhibiting splitting ratios close to the values predicted by modeling. However, these structures are extremely sensitive to the fabrication tolerances and discrepancies between experimental results and theory might be due to an over-etching during processing.
The design, fabrication and optical characterization of SiGe/Si based graded index waveguides and photonics devices working in the [3 - 8 µm] wavelength range was discussed. Straight waveguides based on optimized SiGe/Si stacks showed losses as low as 1 dB/cm at λ = 4.5 µm and 2 dB/cm at 7.4 µm. Likewise straight waveguides, basic functions like Y-junctions, crossings and couplers, have exhibited nearly theoretical behavior, with losses compatible with the implementation of more complex optical functions. As far as we know, we have obtained state the art performances for Mid-IR waveguides, confirming the feasibility of using a graded SiGe/Si stack to fabricate devices able to operate in a wide range of wavelengths.
Furthermore, in spite of the limited number of functions designed so far, the layer stack, the process flow and the test protocol might represent a core of key know-how to pave the route for a photonic platform and a set of design rules suitable to develop photonic integrated circuits working in the Mid-IR range.
This work was partially funded by the Clarity (Grant agreement No 288304) and Doggies (Grant agreement No 285446) European Projects. The Micro and Nanofabrication Platform of CEA-Leti MINATEC is also gratefully acknowledged for the support provided during the fabrication of the devices.
References and links
1. R. A. Soref, “Mid-infrared photonics in silicon and germanium,” Nat. Photonics 4(8), 495–497 (2010). [CrossRef]
2. B. G. Lee, M. A. Belkin, R. Audet, J. MacArthur, L. Diehl, C. Pflügl, F. Capasso, D. C. Oakley, D. Chapman, A. Napoleone, D. Bour, S. Corzine, G. Höfler, and J. Faist, “Widely tunable single-mode quantum cascade laser source for mid-infrared spectroscopy,” Appl. Phys. Lett. 91(23), 231101 (2007). [CrossRef]
3. S. D. Gunapala, S. V. Bandara, J. K. Liu, J. M. Mumolo, and C. J. Hill, “Toward dualband megapixel QWIP focal plane arrays,” Proc. SPIE 6542, 65420W (2007).
5. B. G. Lee, M. A. Belkin, C. Pflügl, L. Diehl, H. A. Zhang, R. M. Audet, J. MacArthur, D. P. Bour, S. W. Corzine, G. E. Hufler, and F. Capasso, “DFB quantum cascade laser arrays,” IEEE J. Quantum Electron. 45(5), 554–565 (2009). [CrossRef]
6. M. M. Milosević, M. Nedeljkovic, T. M. Ben Masaud, E. Jaberansary, H. M. H. Chong, N. G. Emerson, G. T. Reed, and G. Z. Mashanovich, “Silicon waveguides and devices for the mid-infrared,” Appl. Phys. Lett. 101(12), 121105 (2012). [CrossRef]
7. F. Li, S. D. Jackson, C. Grillet, E. Magi, D. Hudson, S. J. Madden, Y. Moghe, C. O’Brien, A. Read, S. G. Duvall, P. Atanackovic, B. J. Eggleton, and D. J. Moss, “Low propagation loss silicon-on-sapphire waveguides for the mid-infrared,” Opt. Express 19(16), 15212–15220 (2011). [CrossRef] [PubMed]
8. E. D. Palik, Handbook of Optical Constants of Solid (Academic, 1985).
9. J.-M. Hartmann, “Epitaxy of strained Si/SiGe heterostructures,” in Silicon Technologies Ion Implantation and Thermal Treatment, A. Baudrant, ed. (John Wiley, 2013)
10. R. Soref, S. J. Emelett, and W. Buchwald, “Silicon waveguided components for the long-wave infrared region,” J. Opt. A Pure Appl. Opt. 8(10), 840–848 (2006). [CrossRef]
11. P. Barritault, M. Brun, P. Labeye, O. Lartigue, J. M. Hartmann, and S. Nicoletti, “Mlines characterization of the refractive index profile of SiGe gradient waveguides at 2.15 µm,” Opt. Express 21(9), 11506–11515 (2013). [CrossRef] [PubMed]
12. N. K. Hon, R. Soref, and B. Jalali, “The third-order nonlinear optical coefficients of Si, Ge, and Si1-xGex in the midwave and longwave infrared,” J. Appl. Phys. 110(1), 011301 (2011). [CrossRef]
13. M. A. Ettabib, K. Hammani, F. Parmigiani, L. Jones, A. Kapsalis, A. Bogris, D. Syvridis, M. Brun, P. Labeye, S. Nicoletti, and P. Petropoulos, “FWM-based wavelength conversion of 40 Gbaud PSK signals in a silicon germanium waveguide,” Opt. Express 21(14), 16683–16689 (2013). [CrossRef] [PubMed]
14. See for example A. W. Snyder and J. D. Love, Optical Waveguide Theory (Chapman and Hall, 1983).
15. RSoft is a trademark of Synopsys OSG, http://optics.synopsys.com/about/
16. Y. Bogumilowicz, J. M. Hartmann, G. Rolland, and T. Billon, “SiGe high temperature growth kinetics in Reduced Pressure – chemical vapor deposition,” J. Cryst. Growth 274(1–2), 28–37 (2005). [CrossRef]
17. F. Gonzatti, J. M. Hartmann, and K. Yckache, “Low and high temperature boron and phosphorus doping of Si for junctions and MEMS purposes,” ECS Trans. 16(10), 485–493 (2008). [CrossRef]
18. Y. Bogumilowicz, J. M. Hartmann, F. Laugier, G. Rolland, T. Billon, N. Cherkashin, and A. Claverie, “High germanium content SiGe virtual substrates grown at high temperatures,” J. Cryst. Growth 283(3–4), 346–355 (2005). [CrossRef]
19. R. Regener and W. Sohler, “Loss in low-finesse Ti:LiNbO3 optical waveguide resonators,” Appl. Phys. B 36(3), 143–147 (1985). [CrossRef]
20. E. Kapon and R. Bhat, “Low-losses single-mode GaAs/AlGaAs optical waveguides grown by organometallic vapor phase epitaxy,” Appl. Phys. Lett. 50(23), 1628–1630 (1987). [CrossRef]
21. E. Peter, S. Laurent, C. Sirtori, M. Carras, J. A. Robbo, M. Garcia, and X. Marcadet, “Measurement of semiconductor waveguide optical properties in the mid-infrared wavelength range,” Appl. Phys. Lett. 92(2), 021103 (2008). [CrossRef]
22. F. Ladouceur and P. Labeye, “A new general approach to optical waveguide path design,” J. Lightwave Technol. 13(3), 481–492 (1995). [CrossRef]
23. P. Labeye, “Composants optiques intégrés pour l’interférométrie astronomique,” Ph.D. thesis dissertation, http://arxiv.org/abs/0904.3030.
24. F. Ladouceur and J. D. Love, “X-junctions in buried channel waveguides,” Opt. Quantum Electron. 24(12), 1373–1379 (1992). [CrossRef]