We present a broadband 2x2 electro-optic silicon switch with an ultralow switching power and fast switching time based on a Mach-Zehnder interferometer (MZI). Forward-biased p-i-n junctions are employed to tune the phase of silicon waveguides in the MZI, to achieve a π-phase switching power of 0.6 mW with a drive voltage 0.83 V with a MZI arm length of 4 mm. The 10%-90% switching time is demonstrated to be 6 ns. Optical crosstalk levels lower than −17 dB are obtained for an optical bandwidth of 60 nm. The free carrier induced silicon refractive index change is extracted from the experimental results for the concentration range from 1016 to 1017 cm−3. We find that at the concentration of 1016 cm−3, the index change is about twice that calculated by the commonly used index change equation.
©2010 Optical Society of America
Silicon-based optical interconnects are expected to provide high bandwidth and low power consumption for chip-level communication, due to their electronics integration capability, proven manufacturing record and price volume curve [1–5]. On-Chip optical network architects have been proposed by several groups [5–8]. A key optical component to realize reconfigurable communication is a silicon optical switch, with a broad bandwidth to facilitate wavelength division multiplexing (WDM) optical links, a low switching power and a sub-microsecond switching time. A comb switching technique using a large-radius silicon ring resonator to realize multi-wavelength operation was proposed in  wherein the free spectral range of the ring matches the wavelength spacing in a WDM optical link. Subsequent demonstrations have verified the switching functionality of multiple wavelengths for such resonator-based switches [10,11]. However, it is known that such rings must be tuned to compensate fabrication errors and temperature variations. In order to provide ‘true’ broadband operation free from these effects, Mach-Zehnder interferometers (MZI) can be employed to achieve either thermo-optic [12,13] or free-carrier-injection electro-optic switches [14,15]. The switching time of thermo-optic silicon switches is usually longer than a few microseconds. Free-carrier driven MZI based switches have been demonstrated to realize a large bandwidth and a fast switching time simultaneously. The switching power for such switches was reported to be as low as 3.1 mW . With this amount of power, self-heating effects, which act in the opposite direction of free carrier induced index change, would still degrade the optical performance .
In this paper, we report carrier-injection MZI switches with a 0.6 mW of π-phase switching power at a drive voltage 0.83 V. The arm length of MZIs needs to be about 4 mm to achieve this low power. The 10%-90% switching time is demonstrated to be 6 ns. Optical crosstalk levels lower than −17 dB are obtained for an optical bandwidth of 60 nm. The possible origin of such low power switching is explored. The silicon refractive index change is extracted from the experimental results for the free carrier concentration range from 1016 to 1017 cm−3 (as far as we know, this is the first report on silicon index change in this concentration range). It is to be noted that at the concentration of 1016 cm−3, the index change is about 2 times of that calculated by the commonly used index change equation.
2. Device structure and fabrication
A symmetric MZI, with two 2x2 3 dB couplers at input and output ends, functions as a 2x2 switch if the phase of either of the two arms can be tuned (Fig. 1(a) ). The 2x2 3 dB couplers are realized by multimode interference couplers (MMIs) which are expected to have higher fabrication tolerance than directional couplers. The silicon waveguide in MZIs has a cross section of 0.50 µm x 0.25 µm with a 0.05 µm slab. The buried oxide thickness is 3 µm and the top cladding oxide thickness is 1.2 µm. In order to achieve phase modulation, a p-i-n junction is created across the silicon waveguides in the two arms of the MZIs (Fig. 1(b)). Both p and n have a doping level of 1020 cm−3 to minimize the series resistance of the junction, which is critical to minimize unwanted thermal effects due to the current flow through the device. The device fabrication is similar to our silicon modulators presented in Ref . We have fabricated MZIs with a range of arm lengths, from a few hundred microns to 6 mm. Figure 1(a) shows an example with an arm length of 0.25 mm. When the MZI arm length is a few mms, it can be curled together to reduce device footprint, forming a spiral shape by taking advantage of the fact that such waveguides can have a tight bend with a radius down to a few microns. For example, In Fig. 1(c) and (d), a 2 mm long silicon waveguide with p-i-n junction is packaged to an area about 150 µm by 150 µm.
3. Characterization of a 2x2 switch with an arm length of 4 mm
3.1. Switching power
Low switching power is experimentally demonstrated in a MZI switch with an arm length of 4 mm. We test this device using an optical detector and a tunable laser source together with a voltage-current source-meter. In this measurement, the input wavelength is fixed at 1520 nm and the applied voltage is scanned from 0.6 V to 1 V (the p-i-n junction will turn on at about 0.7 V). The electrical current is measured at the same time and the electrical power is calculated from the product of the measured current and the applied voltage. Figure 2(a) and (b) show the experimental results of the normalized optical power at two outputs as a function of voltage and electrical power, respectively. With a power of 0.6 mW at a voltage of 0.83 V, the optical power is observed to switch from output 2 to output 1, namely, a π-phase shift has been realized. With a power of 1.8 mW at the voltage of 0.88 V, a 2π-phase shift can be achieved. In the case that fabrication errors cause random phase variations between the two arms of MZIs, a 2π-phase shift can compensate any initial phase difference. The sub-1 V drive voltage is useful in the design of electrical drivers to minimize power and retain compatibility with CMOS voltage scaling . The total optical power drops as the injected electrical power rises due to free-carrier absorption. The extinction ratios also decrease due to increased power imbalance between the two MZI arms since only one active arm experiences free carrier absorption loss.
3.2. Optical performance of 2x2 switching
Based on these results, we define the switch ‘off’ and ‘on’ states corresponding to applied voltage of 0 V and 0.83 V, respectively. The steady-state transmission spectra Tij (i, j = 1, 2) for both states were collected, and are shown in Fig. 3 for the wavelength range from 1480 nm to 1540 nm. The central wavelength of the MMIs is 1510 nm, off the nominal design target of 1550 nm as a result of fabrication errors and non-idealities. The output power is normalized to the total power at the ‘off’ state at a wavelength of 1510 nm. The optical crosstalk between the two outputs is found to be below −17 dB for both ‘off’ and ‘on’ states in the wavelength range from 1480 nm to 1540 nm. This optical bandwidth is mainly limited by the MMI bandwidths. We note that the use of wavelength-insensitive couplers can extend the working bandwidth to 110 nm . In addition, wavelength-insensitive couplers or y-junction couplers may have better fabrication tolerance to achieve the designed wavelength operation than MMIs.
Insertion loss is another important application parameter. While it is difficult to measure absolute insertion loss, we estimate the insertion loss from the MMIs to be about 1.0 dB (0.5 dB each) at a wavelength of 1510 nm by comparing the output power with a reference straight waveguide. This measurement is approximate and may incur some inaccuracy due to coupling variations between the different devices. The free carrier loss for a π-phase shift is about 0.7 dB, obtained from the difference between T11 ‘on’ state and T12 ‘off’ state in Fig. 3. The propagation loss of 4 mm silicon waveguide is about 1.5 dB. Therefore, the total device insertion loss is estimated about 3.2 dB at a wavelength of 1510 nm. Except for the free carrier loss associated with switching, we note that all other losses can be reduced. The intrinsic loss from MMIs can be less than 0.1 dB based on the simulation, and the propagation loss with a similar waveguide geometry was found to be as low as ~1 dB/cm in the literature. Therefore, for such switches, it may be possible to achieve ~1 dB insertion loss even with an arm length of 4 mm.
3.3. Switching speed
For a free-carrier injection device, it is well known that the speed is mainly limited by free-carrier recombination time. We measured the switching time for the device by driving the device with a 20 MHz square-wave voltage signal, with a peak-to-peak voltage 0.83 V and a d.c. bias of 0.42 V, shown in Fig. 4(a) . The rise and fall times of the electrical signal is about 25 ps. The optical transmission was measured by a high speed detector, capable of detecting 40 Gbps optical signal. The optical waveforms, presented in Figs. 4(b)-(d), demonstrate a 10%-90% switching time of 6 ns and 0.4 ns, corresponding to the free carrier injection process (from 0 V to 0.83 V) and extraction process (from 0.83 V to 0 V), respectively. The free carrier extraction time is much faster than the injection time due to the built-in potential of the p-i-n junction. Similar switching behaviors have been observed in previously reported free-carrier-injected switches .
4. Switching power versus device length and silicon index change computation
The switching power drops if the arm length of MZI switches is increased. We measured the π-phase switching power and current for about 50 devices with different arm lengths and presented the average switching power and current in Fig. 5 . For device lengths of 3 mm and 4 mm, both straight arms and spiral-type arms were fabricated and measured. No appreciable power difference was measured between these two device types. This device-length dependence may be mainly attributed to the nonlinear relationship of the index change versus the free carrier concentration. The injected carrier concentration in our devices is at the level of 1016 to 1017 cm−3. To our best knowledge, there are no previous reports on index change at this concentration in the literature. Soref and Bennett’s widely cited equation on the free carrier induced index change,17]. In another paper , the authors reported index change for carrier concentration from 1013 to 1015 cm−3 and found that the index change is about 5 to 10 times larger in this concentration range than that calculated by the above equation. We are lead to believe that at lower concentration levels than 1017 cm−3, Eq. (1) may underestimate the index change.
Using our MZI switches, we are able to compute the index change as a function of carrier concentration. The π-phase switching powers (Pπ) and currents (Iπ) for different device lengths have been measured. The injected carrier concentration is expressed byEquations (2)-(4) are sufficient to extract the silicon index change as a function of free carrier concentrations. However, two parameters, τ and PT, are still missing at this moment. Fortunately, τ can be experimentally measured and is determined to be ~3 ns in our devices. We also indirectly measured PT from other test devices with same waveguide structures and determined that it is about 13 mW. As the Pπ is already less than 1 mW for an arm length of 2 mm, the accuracy of PT is not very critical for long devices (or low carrier concentration). With all this information, we compute the silicon refractive index change as a function of carrier concentration, from the switching power and current shown in Fig. 5. The results are presented in Fig. 6 and the inset shows the ratio between the index changes from our experiment to that simulated by Eq. (1). It is seen that for a carrier concentration larger than 1017 cm−3, the index change ratio is close to 1 (meaning that our experimental results agree with that calculated from Eq. (1)), indicating that our calculation method and the measured values of τ and PT are reasonably accurate. However, if the carrier concentration is lower than 1017 cm−3, the difference between our experiment and that calculated by Eq. (1) grows with decreasing carrier concentration. At a carrier concentration of about 1016 cm−3, the index change is about 2 times that calculated by Soref and Bennett’s equation. The index change includes the contribution from both electrons and holes. Our method cannot separate them. However, from Eq. (1) which is applicable if the carrier concentration is larger than 1017 cm−3, it is seen that the lower the carrier concentration, the larger the index change from holes than from electrons with the same concentration. With a concentration of 1017 cm−3, the contribution due to holes is 4 times that due to electrons with the same concentration (From Eq. (1)). We therefore expect that for carrier concentrations between 1016 cm−3 and 1017 cm−3, the electrons change the index much less than holes with the same carrier concentration. However, more theoretical and experimental work needs to be done to confirm this.
In this paper, we present 2x2 silicon electro-optic switches with a submilliwatt switching power (0.6 mW), a broad bandwidth (60 nm), and ultrafast speed (6 ns). From these switches, we extract the silicon refractive index change as a function of free carrier concentration in the range of 1016 to 1017 cm−3. Such low-power and ultrafast silicon switches are particularly useful to realize reconfigurable communication in on-chip optical networks.
The authors acknowledge partial funding of this work by DARPA MTO office under UNIC program supervised by Dr. Jagdeep Shah (contract agreement with SUN Microsystems HR0011-08-9-0001). The authors greatly acknowledge Dr. W. Qian, Dr. C.-C. Kung, Dr. J. Fong, Dr. Ning-Ning Feng, and Dr. B. J. Luff from Kotura Inc. for their work in fabricating of the device and revising the manuscript, and Dr. J. E. Cunningham and Dr. K. Raj from Sun Labs at Oracle for helpful discussions. The views, opinions, and/or findings contained in this article/presentation are those of the author/presenter and should not be interpreted as representing the official views or policies, either expressed or implied, of the Defense Advanced Research Projects Agency or the Department of Defense. Approved for Public Release, Distribution Unlimited.
References and links
1. R. A. Soref, “The past, present and future of silicon photonics,” IEEE J. Sel. Top. Quant. Electron. 12(6), 1678–1687 (2006). [CrossRef]
2. L. C. Kimerling et al.., “Electronic–photonic integrated circuits on the CMOS platform,” Proc. SPIE 6125, 6–15 (2006).
3. B. Jalali, M. Paniccia, and G. Reed, “Silicon photonics,” IEEE Microw. Mag. 7(3), 58–68 (2006). [CrossRef]
4. D. A. B. Miller, “Device requirements for optical interconnects to silicon chips,” Proc. IEEE 97, 1166–1185 (2009). [CrossRef]
5. A. V. Krishnamoorthy, R. Ho, X. Zheng, H. Schwetman, J. Lexau, P. Koka, G. Li, I. Shubin, and J. E. Cunningham, “Computer systems based on silicon photonic interconnects,” Proc. IEEE 97, 1337–1361 (2009). [CrossRef]
6. A. Shacham, K. Bergman, and L. P. Carloni, “Photonic networks-on-chip for future generations of chip multiprocessors,” IEEE Trans. Comput. 57(9), 1246–1260 (2008). [CrossRef]
7. J. Ahn, M. Fiorentino, R. G. Beausoleil, N. Binkert, A. Davis, D. Fattal, N. P. Jouppi, M. McLaren, C. M. Santori, R. S. Schreiber, S. M. Spillane, D. Vantrease, and Q. Xu, “Devices and architectures for photonic chip-scale integration,” Appl. Phys., A Mater. Sci. Process. 95(4), 989–997 (2009). [CrossRef]
8. C. Batten, A. Joshi, J. Orcutt, A. Khilo, B. Moss, C. W. Holzwarth, M. A. Popovic, H. Li, H. I. Smith, J. L. Hoyt, F. X. Kartner, R. J. Ram, V. Stojanovic, and K. Asanovic, “Building Many-Core Processor-to-DRAM Networks with Monolithic CMOS Silicon Photonics,” IEEE Micro 29(4), 8–21 (2009). [CrossRef]
10. B. G. Lee, A. Biberman, P. Dong, M. Lipson, and K. Bergman, “All-optical comb switch for multiwavelength message routing in silicon photonic networks,” IEEE Photon. Technol. Lett. 20(10), 767–769 (2008). [CrossRef]
11. Y. Vlasov, W. M. J. Green, and F. Xia, “High-throughput silicon nanophotonic wavelength-insensitive switch for on-chip optical networks,” Nat. Photonics 2(4), 242–246 (2008). [CrossRef]
13. Y. Shoji, K. Kintaka, S. Suda, H. Kawashima, T. Hasama, and H. Ishikawa, “Low-crosstalk 2 x 2 thermo-optic switch with silicon wire waveguides,” Opt. Express 18(9), 9071–9075 (2010). [CrossRef] [PubMed]
14. J. Van Campenhout, W. M. Green, S. Assefa, and Y. A. Vlasov, “Low-power, 2 x 2 silicon electro-optic switch with 110-nm bandwidth for broadband reconfigurable optical networks,” Opt. Express 17(26), 24020–24029 (2009). [CrossRef]
15. B. G. Lee, J. Van Campenhout, A. V. Rylyakov, C. L. Schow, W. M. J. Green, S. Assefa, M. Yang, F. E. Doany, C. V. Jahnes, R. A. John, J. A. Kash, and Y. A. Vlasov, “Broadband silicon photonic switch integrated with CMOS drive electronics,” in Proceedings of Conference on Quantum electronics and Laser Science Conference (CLEO/QELS 2010), paper CThJ1.
16. P. Dong, S. Liao, D. Feng, H. Liang, D. Zheng, R. Shafiiha, C.-C. Kung, W. Qian, G. Li, X. Zheng, A. V. Krishnamoorthy, and M. Asghari, “Low Vpp, ultralow-energy, compact, high-speed silicon electro-optic modulator,” Opt. Express 17(25), 22484–22490 (2009). [CrossRef]
17. R. A. Soref and B. R. Bennett, “Electrooptical effects in silicon,” IEEE J. Quantum Electron. 23(1), 123–129 (1987). [CrossRef]
18. L. S. Yu, Q. Z. Liu, Z. F. Guan, and S. S. Lau, “Direct measurement of the refractive index change of silicon with optically injected carriers,” Appl. Phys. Lett. 68(11), 1546 (1996). [CrossRef]