40 Gbit/s low-loss silicon optical modulators are demonstrated. The devices are based on the carrier depletion effect in a pipin diode to generate a good compromise between high efficiency, speed and low optical loss. The diode is embedded in a Mach-Zehnder interferometer, and a self-aligned fabrication process was used to obtain precise localization of the active p-doped region in the middle of the waveguide. Using a 4.7 mm (resp. 0.95 mm) long phase shifter, the modulator exhibits an extinction ratio of 6.6 dB (resp. 3.2 dB), simultaneously with an optical loss of 6 dB (resp. 4.5 dB) at the same operating point.
©2012 Optical Society of America
On-chip integration of complete optical systems in silicon is the ultimate goal of silicon photonics spanning a range of applications from short link and long-haul communications, to biosciences, medicine and sensing. In the past few years, silicon photonics has continued to make extensive and relevant advances to transform a questionable goal into reality. One of the most critical components of high-speed integrated photonic system is the silicon optical modulator, which is the driving force of optical interconnects to achieve high performance data links . In the last couple of years, it has been demonstrated relevant improvements to speed and performance of all-silicon optical modulators which is taking us closer to a complete integration of optical links. Among the possibilities to achieve high-speed optical modulation in silicon, the plasma-dispersion effect, in which the silicon refractive index is changed by free-carrier concentration variation, is the most common method used. The most published design of silicon optical modulators is the pn. The main difference is the use of various waveguide geometries, and junction orientation [2–12].
The work presented in this paper reports on the latest experimental results of Mach Zehnder modulators with phase shifters of lengths 0.95 mm and 4.7 mm based on carrier depletion in a pipin diode. The optical modulators operate at 40 Gbit/s, with extinction ratio (ER) of 3.2 dB and 6.6 dB respectively. Optical losses at the 40 Gbit/s operating points are as low as 4.5 dB and 6 dB for 0.95 mm and 4.7 mm long active regions, respectively. In particular, in the case of the longer phase shifter, such results demonstrate a very good trade-off between loss and ER, which can be successfully compared with previous reported 40 Gbit/s modulators [9,10]. This reduction of the optical loss can be attributed to the use of a pipin diode as a phase shifter. Indeed, it has been demonstrated that this structure delivers a good compromise between high efficiency, speed, and low optical loss . The results present in this paper are an improvement to the silicon optical modulator demonstrated in , by using a self-aligned fabrication technique in order to achieve an accurate localization of the active p-doped region in the middle of the waveguide and thus, fabrication reliability.
2. Device design and fabrication
The silicon modulator is based on carrier depletion of a pipin diode (shown in Fig. 1(a) ) embedded in an asymmetric Mach-Zehnder interferometer (MZI). The diode is composed of a pin junction that allows a reduction of the optical loss due to the presence of a non-intentionally doped region in the waveguide. A p-doped slit embedded in the intrinsic region of the junction is depleted when a reverse bias is applied to the device, leading to an increase of the effective index variation .
The pipin diode is designed in a single-mode silicon rib waveguide of width 420 nm, height 390 nm, etching depth of 290 nm, generating a slab of 100 nm. Waveguide dimensions are chosen to maximize the optical TE-mode confinement in the depleted region. To reduce the access resistance between the active region and the electrodes, only 800 nm on each side of the rib are etched and p++- and n++-doped regions are used, as illustrated in Fig. 1(a) . Phase shifters of lengths 0.95 mm and 4.7 mm were fabricated. The modulator is biased using coplanar traveling-wave electrodes in ground-signal-ground (GSG) configuration as shown in Fig. 1(b). The GSG configuration was chosen to be compatible with the available RF probes in the experimental set-up. The width of the signal electrode, as well as the gap between the signal and the ground electrodes, were designed to provide an impedance of around 50 ohms. The pipin diode is embedded in both arms of an asymmetric MZI interferometer, but only one arm is connected to the coplanar waveguide electrodes (Fig. 1(c)). Both arms of the MZI are split and combined using star couplers with a reduced area (10 × 2 µm2) . A picture of the device can be seen in Fig. 1(d).
Fabrication of the active region in the center of the waveguide is particularly critical because the best modulation efficiency occurs if the variation of the refractive index region is centered in the rib waveguide. For this reason, a self-aligned fabrication process is used to fabricate the optical modulators. The process illustrated in Fig. 2 , consists in defining the modulator’s waveguide and implantation area of the active region (p-doped slit) in the same lithography step to achieve a perfect alignment between both levels.
The fabrication of the silicon modulator is carried out in silicon-on-insulator (SOI) wafers with a 2 µm-thick buried oxide (BOX), and a 400 nm-thick silicon film. A SiO2 layer deposited in the silicon substrate is used as a hard mask. Photolithography is used to define simultaneously the rib waveguide and the opening of the central (ultimately p-doped) slit (Fig. 2(a)). The hard mask is completely etched down to the silicon layer (Fig. 2(b)), and a new photolithography step is performed to protect the rib waveguide slit, and to etch the silicon partially to create the rib waveguide (Fig. 2(c)). A last lithography step is then made to define the photoresist window in the middle of the waveguide to allow the implantation of the p-doped slit (Fig. 2(d)). Supplementary classical lithography and implantation steps are used to create all the other doped regions. Device dopant concentration is targeted at 3 × 1017 cm−3 in the slit (boron-doped), 8 × 1017 cm−3 in the p+ region, 1018 cm−3 in the n+ region (phosphorus doped), and 1019 cm−3 in the p++ and n++ regions. All dopant concentrations are analyzed to establish low optical propagation, and high modulation efficiency of the structure, and to reduce access resistances a heavy dopant concentration in the farthermost regions is taken into account . Vias are then patterned, and etched down to the SiO2 cap layer, and silicide is formed with the purpose of obtaining low contact resistance. Finally, a Ti/TiN/AlCu/Ti/TiN metal stack is used to make the electrodes. Overall, CMOS compatible process steps such as deep-UV lithography, ion implantation, and reactive ion etching are performed throughout the entire fabrication process in order to guarantee transferability in high-volume microelectronic manufacturing.
3. Experimental results
The optical transmission spectra of the silicon modulators were measured using a butt-coupling experimental set-up. A linearly TE-polarized light beam was coupled into the waveguide using a polarization-maintaining lensed-fiber, and the output light was collected by an objective which focuses the light on an IR detector. Electrical probes were used to bias the diode.
The transmission of the modulators as a function of the wavelength was recorded for reverse bias of 0, 5 and 10 V. The measured spectra was normalized to transmission waveguides of the same length as the MZI, but without phase shifters, and the resulted maximum transmission (i.e. insertion loss) was −2 dB for the 0.95 mm phase shifter, and −4 dB for the 4.7 mm phase shifter. This normalization method is very useful to separate on-chip loss due to light coupling from the fiber into the rib waveguide; however, in order to determine the actual modulator loss, the passive waveguide loss that was previously removed (as they also occur in the normalization waveguides) was added. By measuring the transmission of passive waveguides of different lengths, this loss was calculated to be less than 0.5 dB/mm. Figure 3 shows the transmission spectra measurements in which passive waveguide loss is included (−0.5 dB, and −2 dB for the 0.95 mm, and 4.7 mm phase shifters, respectively). Therefore, a maximum optical transmission of −2.5 dB and −6 dB was evaluated for the 0.95 mm and 4.7 mm modulator respectively. These values represent the entire on-chip insertion loss (IL) from the optical modulators, including losses in both the star couplers (splitter and combiner), and the phase shifters. From these values, the loss of each coupler was extracted as 0.75 dB, and phase shifter loss of 1 dB/mm was deduced.
The product VπLπ can be deduced from the experimental transmission shifts seen in Fig. 3. A value of 3.5 V × cm is obtained, which is moderately larger than the theoretical value of 1.7 V × cm, showing the possibility to improve the modulator performance by optimizing the dopant concentrations inside the waveguide.
To evaluate the high-speed performances of the modulators, the electro-optic bandwidths as well as the eye diagrams were measured. For the electro-optic bandwidth measurements, an Agilent N4373 Lightwave Component Analyzer (LCA) was used to measure the optical response as a function of frequency. For the eye diagram measurements, a Centellax TG1P4A 40 Gbit/s pseudo random binary sequence (PRBS) source with a 215-1 pattern length was used. The output signal was connected to an amplifier which provided a 7 V peak-to-peak amplified signal. For both experiments, a bias tee was used to add a reverse dc bias to guarantee reverse bias operation of the pipin diode, and the electrical bias was applied to the silicon modulator using high-speed RF probes. A 50 ohm load was connected to the output electrode of the phase shifter via a DC block. The light output signal from the modulator was fed through an Erbium Doped Fiber Amplifier (EDFA) and then through a tunable wavelength filter to eliminate a large part of the amplified spontaneous emission noise of the EDFA. The amplified light output was measured using an Agilent Infinium DCA-J 86100C oscilloscope with an 86106B optical module. Despite of the presence of the optical filter, a significant residual amplified spontaneous emission noise level is seen on both eye diagrams mainly due to a large optical loss in the coupling from the fiber into the rib waveguide, which was not optimized for this demonstration. The measured optical responses as a function of the electrical signal frequency are shown in Fig. 4(a) . The measured curves were normalized by the response at 100 MHz. A flat response is shown for the 0.95 mm long device, with a 3 dB optical bandwidth of 40 GHz. For the 4.7 mm long device, a 3 dB bandwidth of 20 GHz is obtained. The bandwidth reduction from 0.95 mm to 4.7 mm long device is explained by the propagation of the RF electrical signal along the electrodes. Indeed electrical S parameters of both device lengths (Fig. 4(b) and Fig. 4(c)) show similar S21 behavior in comparison with optical response. For the shorter device, the optical response drop around 15 GHz is probably due to impedance mismatching at this frequency, which is confirmed by a decrease of S21 and an increase of S11. However despite this moderate ripple, S11 parameter below 10 dB up to 50 GHz is achieved for both devices lengths. Finally, the high values of the measured optical bandwidth (40 GHz for 0.95 mm long device and 20 GHz for 4.7 mm long device) indicate a proper design of the RF electrodes as well as a minimized contribution of electrical loss coming from SOI substrate, by the use of 2 µm thick buried oxide silica and high resistivity silicon substrates.
40 Gbit/s optical eye diagrams were then measured in order to evaluate the modulator performances in a data transmission configuration (Fig. 5 ). An ER of 3.2 dB was obtained with the 0.95 mm-long modulator, and it reaches 6.6 dB with the 4.7 mm-long modulator. It is important to notice that the maximum level of the eye diagram in the longer device corresponds to the maximum transmission level of the modulator. In other words, the wavelength was chosen to provide maximum optical transmission, so there was no additional optical loss contribution associated to the optical modulator working close to its minimum optical transmission level. Additionally, saturation of the signal can be seen at the bottom of the eye diagram (Fig. 5(b)), which corresponds to the noise limitation due to the residual spontaneous emission of the EDFA, indicating that the ER value of 6.6 dB is not limited by the device itself, but by the experimental conditions. This result is compatible with the dc transmission (Fig. 3(b)), in which an ER larger than 15 dB can be expected, using a 7 Vpp RF signal without any additional optical loss, and with a 3dB optical bandwidth of 20 GHz measured for this device (Fig. 4(a)). For the 0.95 mm-long device, the eye diagram was measured at a wavelength that does not correspond exactly to the maximum transmission level of the modulator. In that case, 2 dB additional loss was tolerated in order to achieve the 3.2 dB ER. In summary, the 0.95 mm long optical modulator simultaneously delivered a 3.2 dB ER, and a 4.5 dB optical loss, whereas the 4.7 mm long modulator simultaneously delivered a 6.6 dB ER with a 6 dB optical loss. Finally, it can be noticed that the rise /fall time of approximately 25 ps that is seen in the eye diagrams comes exclusively from the rise/fall time of the electrical signal delivered by the PRBS source. These results show significant performance of the pipin silicon optical modulators. Table 1 reports on the performance comparison of 40 Gbit/s optical modulators based on MZI reported up to now. For a comparison purposes, we report measured VπLπ, 40 Gbit/s ER, on-chip IL (at the maximum transmission wavelength), and optical loss at the 40 Gbit/s operating point, when reported. In all cases, an RF signal of approximately 6-7 Vpp was used. The table shows that the pipin diode delivers among the largest ER, and one of the lowest on-chip IL. Moreover, it is the only structure in which losses increased moderately with increasing active region length.
In summary, silicon optical modulators based on a reverse biased pipin diode embedded in a Mach-Zehnder interferometer working at 40 Gbit/s were demonstrated. Two devices were presented, one with an ER of 3.2 dB and an optical loss of 4.5 dB, and another one with an ER of 6.6 dB and an optical loss of 6 dB. Further improvements can be considered, including push-pull operation to increase the ER. While there is still room for improvement, the pipin structure presented offers the performance required for future silicon photonic applications.
The research leading to these results has received funding from the European Community's under grant agreement n° 224312 HELIOS and from the French National Research Agency (ANR) under project SILVER. The authors thank Agilent Technologies and Catherine Maurin for the loan of the Lightwave Component Analyzer.
References and links
1. M. Paniccia and R. Won, “Integrating silicon photonics,” Nat. Photonics 4(8), 498–499 (2010). [CrossRef]
2. A. Liu, L. Liao, D. Rubin, H. Nguyen, B. Ciftcioglu, Y. Chetrit, N. Izhaky, and M. Paniccia, “High-speed optical modulation based on carrier depletion in a silicon waveguide,” Opt. Express 15(2), 660–668 (2007). [CrossRef] [PubMed]
3. D. Marris-Morini, L. Vivien, J.-M. Fédéli, E. Cassan, P. Lyan, and S. Laval, “Low loss and high speed silicon optical modulator based on a lateral carrier depletion structure,” Opt. Express 16(1), 334–339 (2008). [CrossRef] [PubMed]
4. D. Marris-Morini, L. Vivien, G. Rasigade, E. Cassan, J.-M. Fédéli, E. Cassan, X. Le Roux, P. Crozat, S. Maine, A. Lupu, P. Lyan, P. Rivallin, M. Halbwax, and S. Laval, “Recent progress in high speed silicon-based optical modulators,” Proc. IEEE 97(7), 1199–1215 (2009). [CrossRef]
5. J. W. Park, J.-B. You, I. G. Kim, and G. Kim, “High-modulation efficiency silicon Mach-Zehnder optical modulator based on carrier depletion in a PN Diode,” Opt. Express 17(18), 15520–15524 (2009). [CrossRef] [PubMed]
6. N.-N. Feng, S. Liao, D. Feng, P. Dong, D. Zheng, H. Liang, R. Shafiiha, G. Li, J. E. Cunningham, A. V. Krishnamoorthy, and M. Asghari, “High speed carrier-depletion modulators with 1.4V-cm VπL integrated on 0.25µm silicon-on-insulator waveguides,” Opt. Express 18(8), 7994–7999 (2010). [CrossRef] [PubMed]
7. M. R. Watts, W. A. Zortman, D. C. Trotter, R. W. Young, and A. L. Lentine, “Low-voltage, compact, depletion-mode, silicon Mach-Zehnder modulator,” IEEE J. Sel. Top. Quantum Electron. 16(1), 159–164 (2010). [CrossRef]
8. L. Liao, A. Liu, D. Rubin, J. Basak, Y. Chetrit, H. Nguyen, R. Cohen, N. Izhaky, and M. Paniccia, “40 Gbit/s silicon optical modulator for high speed applications,” Electron. Lett. 43(22), 1196–1197 (2007). [CrossRef]
9. D. J. Thomson, F. Y. Gardes, Y. Hu, G. Mashanovich, M. Fournier, P. Grosse, J.-M. Fédéli, and G. T. Reed, “High contrast 40Gbit/s optical modulation in silicon,” Opt. Express 19(12), 11507–11516 (2011). [CrossRef] [PubMed]
11. G. Rasigade, D. Marris-Morini, L. Vivien, and E. Cassan, “Performance evolutions of carrier depletion silicon optical modulators: from PN to PIPIN diodes,” IEEE J. Sel. Top. Quantum Electron. 16(1), 179–184 (2010). [CrossRef]
12. G. Rasigade, M. Ziebell, D. Marris-Morini, J.-M. Fédéli, F. Milesi, P. Grosse, D. Bouville, E. Cassan, and L. Vivien, “High extinction ratio 10 Gbit/s silicon optical modulator,” Opt. Express 19(7), 5827–5832 (2011). [CrossRef] [PubMed]
13. G. Rasigade, X. Le Roux, D. Marris-Morini, E. Cassan, and L. Vivien, “Compact wavelength-insensitive fabrication-tolerant silicon-on-insulator beam splitter,” Opt. Lett. 35(21), 3700–3702 (2010). [CrossRef] [PubMed]