Open Access
Add to BibSonomyAbstract
We proposed a silicon-based optical switch with a carrier-plasma-induced phase shifter which employs a silicon-germanium (SiGe) / silicon (Si) hetero-structure in the waveguide core. A type-I hetero-interface formed by SiGe and Si is expected to confine carriers effectively in the SiGe waveguide core. The fabricated Mach-Zehnder optical switch shows a low switching power of only 1.53 mW with a compact phase shifter length of 250 μm. The switching time of the optical switch is less than 4.6 ns for the case of a square waveform driving condition, and 1 ns for the case of a pre-emphasis electric driving condition. These results show that our proposed SiGe/Si waveguide structure holds promise for active devices with compact size and low operation power.
© 2012 Optical Society of America
1. Introduction
Today’s growing broadband network services, i.e. cloud computing, streaming video delivery, etc. are increasing network traffic rapidly. Such a trend will continue for the time being. Dynamic optical path-switching (DOPS) networks through which optical paths are dynamically provisioned and released by individual end users have been proposed to take in today’s rapidly growing network traffic [1]. To build a large scale DOPS network, a lot of large-scale optical switches must be constructed. To produce large scale optical switches at a low cost, the size of each switch element must be small. The larger the scale of the switch is, the faster the switching speed is desired because a multitude of switch elements have to complete switching processes within the interval of signals in the large scale optical switch. Additionally their power consumption must be low.
Opto-electronics on silicon-on-insulator (SOI) substrate can exploit ultra small waveguides owing to a large refractive index difference between silicon (Si) and silicon dioxide (SiO2), and are expected to be integrated with electronic circuits. Therefore, they have recently attracted attention as a new platform for optical integrated devices. Various devices on the Si-platform have been demonstrated to date, i.e. optical modulators [2], photo-detectors [3], and lasers [4].
These capabilities of the optical devices on the Si-platform for high-density opto-electric integration are preferable for realizing large-scale optical switches in DOPS networks. To date, Si nano-wire waveguide 2 × 2 Mach-Zehnder optical switches have been developed as elements of these large-scale switches [5–8]. In particular, a carrier-plasma-induced Si rib waveguide optical switch [7,8] is expected to be a key device for achieving future energy-saving networks, since it has both low switching power and fast switching speed.
In a conventional carrier-plasma-induced Si waveguide optical switch, a forward-biased lateral p-i-n diode structure is used as a phase shifter in the Mach-Zehnder interferometer. In this device, injected and accumulated carriers at the intrinsic waveguide core region vary the refractive index of the media due to the carrier plasma effect. However, since a conventional phase shifter consists of a Si homo junction, carriers injected into the waveguide core cannot accumulate efficiently, especially at high injection condition. According to this fact, the switching power of the carrier-plasma-induced Si optical switch increases when the phase shifter length of the optical switch becomes shorter [8]. Thus, this prevents realization of the small-sized, carrier-plasma-induced Si optical switch with low consumption power.
To improve carrier accumulation efficiency, introduction of a hetero interface, which is commonly used in laser diode structure, is an effective solution. In this report, we propose a carrier-plasma-induced phase shifter, which contains silicon germanium (SiGe) in the core region of the rib waveguide, for low-power operation of a compact optical switch. The SiGe on Si composes a hetero structure and enables efficient carrier confinement. We report on fabrication of a compact optical switch with the proposed structure and demonstration of its low-power and fast-switching characteristics.
2. SiGe nano-wire waveguide structure
The phase shifter of the optical switch controls the phase of optical signal by varying the refractive index of the waveguide. In the Si-based carrier-plasma-induced optical switches, injected carrier densities in the phase shifter waveguides cause plasma dispersion effect and vary the refractive indices. Therefore, it is important for the carrier-plasma-induced optical switch to accumulate carriers in the waveguide core region effectively. The phase shifter of the conventional Si-based optical switches consists of Si homojunction, and there is no bandgap discontinuity. In this case injected carriers in the core region spread easily to the other side, resulting in low carrier density in the core region. On the other hand, by employing a hetero-structure, which is commonly used in laser diodes, injected carriers stay in the intrinsic region which consists of narrower bandgap material because of the potential barrier at the interface. Consequently, higher carrier density can be obtained with lower injection current. Thus, the heterostructure is considered to improve the efficiency of the plasma-effect induced optical switches.
We propose a phase shifter structure for optical switches with a heterostructure. Figure 1(a) shows a cross-sectional view of the phase shifter we propose. The structure is basically a conventional nano-wired rib waveguide structure with lateral p-i-n junction formed on a silicon-on-insulator (SOI) wafer. A SiGe layer is employed on the Si slab and forms a waveguide core. Both slab regions next to the SiGe core are doped p-type and n-type for carrier injection. The SiGe layer is pseudomorphically grown on the Si layer and compressively strained. The structure can be fabricated in production facilities of silicon photonic devices including Ge photo detectors or SiGe BiCMOS devices.
Fig. 1 (a) Cross-sectional view of optical switch device with SiGe waveguide core, (b) Band diagram of Si/SiGe.
The interface of the strained SiGe layer and unstrained Si is a type-I heterointerface [9], and for the case of Ge content of 0.1, the offset energy of the conduction band (ΔEc) and valence band (ΔEv) are roughly 30 meV and 100 meV, respectively, as shown in Fig. 1(b). Owing to this bandgap discontinuity, carriers are expected to be confined in a SiGe region with narrower bandgap. For the waveguide structure, we employed the Si slab thickness of 50 nm, the SiGe layer thickness of 200 nm and the waveguide rib width of 480 nm, respectively.
To confirm carrier confinement effect of the proposed structure, carrier densities in the waveguide core region were calculated for the case of the proposed structure and a conventional Si-based structure under the same injection condition. In the calculation, both n+ and p+ doping concentration of 1×1020 cm−3 are assumed. Figure 2 shows the calculated carrier density along the z-axis in Fig. 1 for the proposed SiGe/Si structure and conventional Si structure at an injection current density of 1 mA/mm. As can be seen in Fig. 2, the carrier density in the SiGe layer is much higher than that in the conventional Si structure. The proposed structure is not a standard double heterostructure which is commonly utilized in laser diodes; however, injected carriers are efficiently confined at the SiGe layer as the band edge energies of the SiGe layer are lower than those of the Si layer for both conduction and valence bands. Such large carrier density causes large refractive index change by plasma effect, and it is expected to reduce operation power of the optical switch. Figure 3 shows calculated switching power of the proposed SiGe/Si optical switch and conventional Si switch with various phase shifter length. In the calculation, we assumed both n+ and p+ doping concentrations are 1×1019 cm−3, and the coefficients of the refractive index change for SiGe and Si conform to well known free carrier or Drude formula for Si. From the Fig. 3, for the case of conventional Si structure, the switching power increases when the phase shifter length becomes shorter. This trend shows that the carrier density in the core region saturates for injection current because of absence of the carrier confinement. By employing the SiGe/Si structure, carrier confinement by the heterostructure causes high carrier density and lower switching power is expected especially at short phase shifter length.
Fig. 2 Accumulated carrier density in the intrinsic region of the proposed SiGe/Si structure and conventional Si structure.
Fig. 3 Calculated switching power of the optical switches with proposed SiGe/Si structure and conventional Si structure with various phase shifter length.
3. Device structure
We fabricated a 2 × 2 Mach-Zehnder optical switch with the proposed SiGe/Si phase shifter. A plan view of the fabricated optical switch is shown in Fig. 4. The SiGe layer was grown on the 50-nm-thick SOI layer using ultra-high vacuum chemical vapor deposition (UHV-CVD). The waveguides were patterned by electron-beam (EB) lithography. Two phase shifters are combined using 50% multi-mode interference (MMI) couplers and form a Mach-Zehnder interferometer (MZI). The input and output ports form tapered waveguides to ensure fiber coupling. The length of the phase shifters is 250 μm.
4. DC measurements
At first we measured the DC switching characteristics of the fabricated MZ optical switch. The 1550 nm TE polarized light entered the input port a1. Transmitted light power at both output ports was measured while changing the forward bias voltage applied to the Arm 2 phase shifter. The voltage to the Arm 1 phase shifter was not applied (kept at 0 V). Figure 5 shows the DC extinction curves of the optical switch in terms of the driving electric power of the phase shifter, which is the product of the injection current and the applied voltage. When the phase shifter is not driven (the driving electric power of 0 mW), the light injected into input port a1 outputs from output port b1 (“Bar” state), and the transmittance is normalized by setting this power as 0 dB. As shown in the figure, the extinction curves vary in a complementary manner. At the driving electric power of 1.53 mW, a corresponding operation current of 1.55 mA and voltage of 0.99 V, the output port switched to port b2, and the port b1 output power was minimized (“Cross” state), namely, a π-phase shift has been achieved with switching power of only 1.53 mW.
Although, in principle Mach-Zehnder interferometers should be in a “Cross” state when no phase shift is given, our device was in a “Bar” state due to phase error caused by insufficient fabrication accuracy. The phase error is expected to be reduced by using a higher technology node [10]. And the phase error can also be reduced by using width-tolerant waveguide structure such as shallow etched waveguide structure [11]. However, our device is expected to has larger tolerance than conventional silicon device, because we can choose shorter phase shifter than that of conventional Si optical switch to obtain the same operation power, and the shorter phase shifter produces the smaller phase error. Even if there will be still remaining phase error, our device requires lower operation power than that of conventional silicon optical switches to adjust such remaining phase error.
The total insertion loss of the device is about 8.0 dB at 1550 nm for the “Bar” state. This value is almost the same as the conventional Si-based devices. The main loss source is the coupling loss to optical fiber and is estimated to about 3 dB per facet. The excess loss of the MMI coupler is estimated to less than 1 dB. The propagation loss of the SiGe waveguide is sufficiently low and is to be less than 1 dB for total 3-mm-long waveguide. It is the same level as our standard silicon nano-wired waveguide. This fact shows the bandgap reduction in SiGe does not affect the optical loss at 1550 nm wavelength range. For the case of the “Cross” state, the insertion loss at 1550 nm is about 9.5 dB. There is the inevitable free carrier plasma loss of about 1.5 dB.
Broadband wavelength applicability is one of the important characteristics of optical switches for network application. Figure 6 shows the transmission spectra of the fabricated optical switch in the “Bar” and “Cross” states described previously. The transmittance is normalized by the maximum transmittance in the “Bar” state. The figure indicates that the device has flat transmission over the full C-band range within +/−1.3 dB deviation. In addition, crosstalk (XT) for both the “Bar” state and the “Cross” state are less than −19.7 dB over the full C-band range. These characteristics are comparable to the other reported Si-based 2 × 2 Mach-Zehnder optical switches [6–8]. The optical bandwidth is determined mainly by the wavelength dependence of the MMI couplers and phase error between the two arms. The central wavelength and 1 dB transmission bandwidth of the MMI coupler used in the device are about 1500 nm, and about 90 nm, respectively. The MMI structure hasnot been adjusted for the SiGe/Si waveguide yet, so the center wavelength can be adjusted by optimization of the structure, because its structural tolerance is much wider than that of the phase error.
We compare switching power of the fabricated SiGe/Si optical switch with the other Si-based and III–V semiconductor based carrier-plasma-induced broadband optical switches. Figure 7 shows the relation between the switching power and the phase shifter length for the optical switches. As can be seen in Fig. 7, the switching power increases with the phase shifter length decreasing for the case of the conventional Si optical switches. On the other hand, the fabricated SiGe/Si optical switch has a low switching power of 1.53 mW with a short phase shifter of 250 μm, combining small-size and power-saving features. The switching power of the SiGe/Si optical switch is about the half of that of the conventional Si optical switch with similar phase shifter length. This trend almost agrees with the theoretical prediction discussed in Section 2. In addition, SiGe/Si optical switch is expected to reduce device size further with almost the same switching power.
Fig. 7 Relationships between switching power and phase shifter length for various carrier-plasma-induced broadband optical switches. ‘Si’ represents the conventional Si optical switch reported by Zhou al. [12], Campenhout et al. [7] and Dong et al. [8]. ‘SiGe’ represents this work. There is a plot for the optical switch made up of the III–V semiconductor reported by Ueda et al. [13], indicated with ‘III–V’ as a reference.
5. High speed switching characteristics
Since our device utilizes carrier plasma effect as an index modulation principle, the device is expected to achieve ns-order operation. The switching speed was evaluated by applying an electric signal to the Arm 2 phase shifter. The time-resolved relative transmittance waveforms are shown in Fig. 8. The low level voltage of the electric pulses was 0.55 V (for the “Bar” state) and the high level voltage was 0.75 V (for the “Cross” state) under 50 Ω termination. Since actual device’s impedance was not 50 Ω, actual voltages applied to the device are different from these values. It is presumed that almost the same voltage was applied as DC condition at both the “Cross” and “Bar” states. The 10–90% transient time of the electric pulses is within 1 ns. For the “Bar” state–“Cross” state transition, 10–90% fall time of the b1 transmittance was 4.6 ns. For the “Cross” state–“Bar” state transition, 10–90% rise time of the b1 transmittance was 1.5 ns. These fast switching characteristics are applicable for packet switching application of more than 10 Gbps data routing system, in which less than 10 ns switching is required in terms of fitting into the frame gap between the packets. Our device has a hetero-barrier in the structure; however, the barrier does not affect the switching speed significantly, maintaining high-speed switching characteristics similar to the other Si-based carrier-plasma-induced optical switches.
Meanwhile, the transition from the “Bar” state to the “Cross” state corresponds to the carrier injection side and the transition from the “Cross” state to the “Bar” state corresponds to the carrier extraction side, respectively. As can be seen in Fig. 8, switching speed for the carrier injection side is slower than that for the carrier extraction side. This trend coincides with the other Si-based optical switches. [7,8] This is considered to be caused by capacity variation during driving the pin junction. Thus, the capacitance increases with the carrier density increasing during the carrier injection, hence the operation speed expressed by RC time constant is reduced. In contrast, the capacitance decreases during the carrier extraction, hence the switch can operate at relatively high speed.
Figure 9 shows logarithmic plot of time-resolved relative transmittance waveforms of the optical switch. From Fig. 9, crosstalks are kept around −20 dB at the both states as it is for DC condition, though, slight fluctuation can be seen. Such a fluctuation deteriorates the crosstalk of the switch during high speed operation. This may be attributed to the thermal drift by self heating of foward biased p-i-n diode. To confirm this, we have measured switching characteristics of similar switch in thermo-optic mode by utilizing contact region as a heater, and typical switching power of 30 to 40 mW is obtained. The switching power of the SiGe/Si optical switchi of about 1.5 mW is less than 5% of the thermal switching power. At the crostalk level of about −20 dB, the phase shift of 5% is estimated to about 2 to 3 dB transmittance shift. The experimental variation around 3 dB is likely thermal drift. Further reduction of the switching power, and improvement of heat dissipation by a heat sink near by the phase shifter are effective to remove or lower the thermal drift.
To be applied to packet switching for more than 40 Gbps data, switching time of less than 1 ns is required. By small signal analysis, the frequency bandwidth of our device is estimated to be a few hundred MHz. To obtain faster switching, the driving signal is required to be emphasized at high frequency, i.e. pre-compensation technique [14]. We used a circuit like Fig. 10 to obtain a pre-compensation signal. The driving electric signal and the output optical waveforms are shown in Fig. 11. The leading and trailing edges of the driving electric signal are emphasized by the pre-compensation circuit, and a peak amplitude of about 2Vpp was obtained. By using this electric waveform for the driving signal of the optical switch, high speed switching with a leading and trailing time of less than 1 ns is obtained.
Fig. 11 Dynamic switching characteristics of fabricated optical switch driven by using the pre-compensation circuit.
6. Conclusion
To reduce the switching power and the size of Si-based optical switches essentially, we proposed a SiGe Nano-wire waveguide phase shifter. The proposed structure utilizes carrier confinement effect of the heterostructure and can obtain high carrier density in the intrinsic core region. We fabricated a carrier-plasma-induced Si-based Mach-Zehnder switch with the proposed phase shifter. The fabricated optical switch shows a low switching power of only 1.53 mW with short phase shifter length of 250 μm. The presented operation power is the lowest among those of the other carrier-plasma-induced optical switches with the same phase shifter length. The device also shows a switching time of less than 4.6 ns which is at the same level as conventional Si optical switches. These results demonstrate that our proposed SiGe/Si structure is promising for the optical switches and the other silicon photonic active devices, enabling to reduce energy consumption with a compact device size.
Acknowledgments
Part of this work was carried out under the Project of AIST ( National Institute of Advanced Industrial Science and Technology), supported by the Special Coordination Funds for Promoting Science and Technology of MEXT.
References and links
1. S. Namiki, T. Hasama, M. Mori, M. Watanabe, and H. Ishikawa, “Dynamic optical path switching for ultra-low energy consumption and its enabling device technologies,” in International Symposium on Applications and the Internet (SAINT 2008), 393–396 (2008).
2. S. Akiyama, T. Kurahashi, T. Baba, N. Hatori, T. Usuki, and T. Yamamoto, “A 1 V peak-to-peak driven 10-Gbps slow-light silicon Mach–Zehnder modulator using cascaded ring resonators,” Appl. Phys. Express 3, 072202 (2010). [CrossRef]
3. J. Liu and J. Michel, “High performance Ge devices for electronic-photonic integrated circuits,” ECS Trans. 16, 575–582 (2008). [CrossRef]
4. D. Liang and J. E. Bowers, “Recent progress in lasers on silicon,” Nat. Photonics 4, 511–517 (2010). [CrossRef]
5. T. Chu, H. Yamada, S. Ishida, and Y. Arakawa, “Compact 1 × N thermo-optic switches based on silicon photonic wire waveguides,” Opt. Express 13, 10109–10114 (2005). [CrossRef] [PubMed]
6. R. Kasahara, K. Watanabe, M. Itoh, Y. Inoue, and A. Kaneko, “Extremely low power consumption thermooptic switch (0.6 mW) with suspended ridge and silicon-silica hybrid waveguide structures,” in 34th European Conference on Optical Communication 2008, (ECOC 2008), Vol. 5, pp. 55–56 (2008).
7. J. V. Campenhout, W. M. Green, S. Assefa, and Y. A. Vlasov, “Low-power, 2 × 2 silicon electro-optic switch with 110-nm bandwidth for broadband reconfigurable optical networks,” Opt. Express 17, 24020–24029 (2009). [CrossRef]
8. P. Dong, S. Liao, H. Liang, R. Shafiiha, D. Feng, G. Li, X. Zheng, A. V. Krishnamoorthy, and M. Asghari, “Submilliwatt, ultrafast and broadband electro-optic silicon switches,” Opt. Express 18, 25225–25231 (2010). [CrossRef] [PubMed]
9. L. Yang, J. R. Watling, R. C. W. Wilkins, M. Boriçi, J. R. Barker, A. Asenov, and S. Roy, “Si/SiGe heterostructure parameters for device simulations,” Semicond. Sci. Technol. 19, 1174–1182 (2004). [CrossRef]
10. S. K. Selvaraja, W. Bogaerts, P. Dumon, D. V. Thourhout, and R. Baets, “Subnanometer linewidth uniformity in silicon nanophotonic waveguide devices using CMOS fabrication technology,” IEEE J. Sel. Top. Quantum Electron. 16, 316–324 (2010). [CrossRef]
11. D.-J. Kim, J.-M. Lee, J. H. Song, J. Pyo, and G. Kim, “Crosstalk reduction in a shallow-etched silicon nanowire AWG,” IEEE Photon. Technol. Lett. 20, 1615–1617 (2008). [CrossRef]
12. G.-R. Zhou, M. W. Geis, S. J. Spector, F. Gan, M. E. Grein, R. T. Schulein, J. S. Orcutt, J. U. Yoon, D. M. Lennon, T. M. Lyszczarz, E. P. Ippen, and F. X. Käertner, “Effect of carrier lifetime on forward-biasedsilicon Mach-Zehnder modulators,” Opt. Express 16, 5218–5226 (2008). [CrossRef] [PubMed]
13. Y. Ueda, S. Nakamura, S. Fujimoto, H. Yamada, K. Utaka, T. Shiota, and T. Kitatani, “Polarization-independent low-crosstalk operation of InAlGaAs-InAlAs Mach-Zehnder interferometer-type photonic switch with hybrid waveguide structure,” IEEE Photon. Technol. Lett. 21, 1118–1120 (2009). [CrossRef]
14. F. Gan, S. Spector, M. Geis, M. Grein, R. Schulein, J. Yoon, T. Lyszczarz, and F. Kartner, “Compact, low-power, high-speed silicon electro-optic modulator,” in Conference on Lasers and Electro-Optics, 2007 (CLEO 2007), paper CTuQ6 (2007).
