An all-optical packet switching using bistable photonic crystal nanocavity memories was demonstrated for the first time. Nanocavity-waveguide coupling systems were configured for 1 × 1, 1 × 2, and 1 × 3 switches for 10-Gb/s optical packet, and they were all operated with an optical bias power of only a few μW. The power is several magnitudes lower than that of previously reported all-optical packet switches incorporating all-optical memories. A theoretical investigation indicated the optimum design for reducing the power consumption even further, and for realizing a higher data-rate capability and higher extinction. A small footprint and integrability are also features of our switches, which make them attractive for constructing an all-optical packet switching subsystem with a view to realizing optical routing on a chip.
© 2015 Optical Society of America
Photonic technologies have been developed for long-haul telecom networks and datacom networks such as in datacenters, and we are now entering the era of ultrashort-range networks in microprocessors. Several aspects of this development, including the miniaturization, power reduction, and bandwidth enhancement of optical processing components, are crucial as regards introducing them into a chip-com network (chip-to-chip and intra-chip) and constructing a photonic network-on-chip (PhNoC) architecture [1, 2]. Specifically, an ultrafast all-optical switch without electrical processing will provide the most fundamental function in photonics. There have already been some reports of all-optical switches that enable the bit-by-bit control of optical signals with an order of Gb/s repetition rate [3, 4], which might be applicable for photonic logic functions or computing. In addition to such bit-by-bit control, optical switching for a granula of optical data (i.e. an optical packet) is also required for efficient data routing. In telecom and datacom networks, optical packet switch (OPS) is now replacing optical path switching and optical burst switching [5–7], and in a similar way, it is worth developing a chip-com system and PhNoC for optical routing in a many-core CMOS architecture. Considering the demands in such a small world, electrical switch should be eliminated because of its bandwidth limitation and high power consumption at a high bit rate.
Various forms of OPS systems that can handle optical packets with bit rates of several tens of Gb/s have been developed as switch subsystems in photonic routers [5–8]. One conventional type of OPS is an 1 × N switch constructed by combining a 1 × N optical coupler and gate arrays based on a semiconductor optical amplifier (SOA) or electro-absorption (EA) switch, and they have been developed as N × N integrated switches . Phased-array-based optical switches are another type that have a similar number of input/output (I/O) ports and are expected to be applicable to wavelength-division multiplexing (WDM) . These devices can operate with bit-rate and format transparency, a high extinction ratio, and in particular they can compensate for optical loss or even obtain a net gain when they include an SOA. However, these devices require electrical control of optical gain and/or phase for a large number of waveguides, and therefore their power consumption exceeds 10 mW and they occupy a large footprint of a few mm2. Electro-optic (EO) switches consisting of a 1 × 2 microring resonator switch or a 2 × 2 Mach-Zehnder-based switch can be used to configure a 4 × 4 optical routing switch [9–11] with a smaller footprint of 0.07 mm2 for PhNoC applications, although they still consume power at the mW level. The switching speed and power of these OPS systems are limited by the photoreceiver and electrical header processor that are used as an electrical switch to route the payload data.
An attractive form of OPS is an all-optical packet switch (AOPS), in which an optical header signal can quickly and directly switch the optical packet without optical-to-electrical (O/E) conversion. In particular, all-optical memories are important building blocks as regards the implementation of AOPSs because they can store the optical header information while the memory output is used to selectively route the optical payload to the output port. Some all-optical memories have been demonstrated using an injection-locking semiconductor laser, in which an optical pulse can latch the bistable memory state [7, 12, 13]. Fast switching at a few tens of picoseconds is possible with such a device, and the footprint will be several hundred μm2 if a small microring resonator is employed. However, the power consumption is still of mW order because of the need for lasing. In addition, these devices need a subsequent separate optical switch such as an MZI-based all-optical switch, which is fed by the output of the all-optical memory to route the optical payload data. Hence, the previous AOPSs consist of a combination of a header memory part and payload switching part. It would be better in terms of high integrability if both parts could be achieved in a single device. The most significant bottleneck for OPS/AOPS is their mW-level power consumption, which makes them difficult to employ for chip-com applications. Using the above background as a basis, we aimed to realize an AOPS that had (i) an all-optical latching ability for both memory holding and optical packet switching, (ii) a routing ability with output-port selectivity, (iii) a low operation power, (iv) a small footprint, and (v) a fast switching speed.
All-optical memories can also be constructed by using a nonlinear cavity that induces a resonant wavelength shift by optical injection. Of all the cavity structures, photonic crystal (PhC) nanocavities have proved attractive because they can realize high quality optical cavities (the maximum Q is over a million) with a small modal volume (V) comparable to (λ/n)3 (λ is the wavelength of light, n is the refractive index). Since optical nonlinearity can be greatly enhanced in high-Q and small-V cavities, a very low energy/power consumption can be expected if we apply them to all-optical nonlinear devices such as switches and bistable memories [3, 14–16]. Moreover, they can be effectively coupled with single-mode PhC waveguides, which means that it is easy to integrate these elements monolithically. As a result, we can expect to realize an AOPS on a chip operating with an ultralow power consumption and a small footprint. We achieved an ultra-compact buried heterostructure (BH) consisting of InGaAsP in a thin InP-PhC waveguide by using the etching and regrowth technique . This novel BH technique provided strong carrier confinement in the InGaAsP region and efficient heat dissipation to the surrounding InP region. A PhC cavity including InGaAsP-bulk nonlinear material has an all-optical memory operating at a power as low as 30 nW thanks to the strong carrier-based nonlinearity .
In this study, we employed an ultrasmall BH-PhC nanocavity as an AOPS based on the all-optical bistable memory operation. Our PhC nanocavity acts both as an optical memory and an optical packet switch, and thereby several types of cavity-waveguide coupling system were investigated to demonstrate 1 × 1, 1 × 2, and 1 × 3 switches. In other words, we demonstrated a monolithically-integrated output-port-selective optical switch having a latch function, for the first time. This confirms the scalability for realizing an integrated 1 × N switch. The operating power needed for switching a 10-Gb/s optical packet ranged from several hundred nW to a few μW, and so the energy-per-bit required for data routing was only about 100 aJ/bit. This power/energy consumption is more than two orders of magnitude lower than that of previously reported AOPSs. We also investigated theoretically the performance obtainable with an optimally designed cavity-waveguide coupling system. Our investigation indicated the possibility of achieving a high bit-rate capability and a high extinction ratio while maintaining an ultralow power consumption.
2. All-optical packet switching based on a bistable cavity
Figure 1 is a simple schematic of an AOPS based on the optical bistable cavity. The fundamental configuration is a single cavity coupled with an I/O waveguide as shown in Fig. 1(a), in which light at the resonant wavelength can be transmitted through the cavity by resonant tunneling. The cavity exhibits a hysteresis response in the input-to-output light characteristics (Fig. 2(c)) thanks to carrier-induced optical nonlinearities. This allows bistable states to be switched by injecting a bias light, a writing pulse, and a reset pulse (temporarily turning off the bias light) [14, 15], as shown in the middle inset in Fig. 1(a). The transmissionlevels can be changed between upper and lower states because of the shift in the resonant wavelength. Therefore, when a data light with appropriate wavelength detuning is additionally injected into the cavity, the transmission level can be changed corresponding to the bistable states. This switching scheme is shown on the right in Fig. 1(a), indicating that the data light can be transmitted simply by injecting a write pulse to turn the memory on to the upper state and conversely by injecting a reset pulse to return the memory to the lower state and block the data light transmission.
Although Fig. 1(a) shows gate switching that turns the optical packet transmission level on and off (1 × 1 switch), it can be developed into an output-port-selective switch (1 × 2 and 1 × 3 switch) by changing the cavity-waveguide coupling configuration, as shown in Fig. 1(b) and 1(c). The cavity is coupled with a bus waveguide and a drop waveguide, and the packet data can be switched to a through port or drop port by switching the bistable states. When the number of cavities with an identical resonant wavelength is increased, the packet data can be selectively switched to one of the drop ports by injecting a write pulse into the corresponding cavities to turn it onto the bistable upper state. Figure 1(c) shows the case of a 1 × 3 switch. Even for a large number of cavities, the power consumption should not be greatly increased, because just a single cavity need be turned on for data switching.
3. Cavity configuration and static bistable characteristics
These AOPSs were realized with a nanocavity-waveguide coupling system on a two-dimensional PhC platform. We chose InGaAsP as the core material for the PhC, because we previously found this to be one of the most efficient carrier-induced nonlinear materials around 1.55 μm when the composition was optimized [3, 18]. Figure 2(a) shows a scanning electron micrograph image of top and cross-sectional views of a fabricated sample. We employed a buried heterostructure (BH) design for the cavity structure. Namely, we embedded a tiny bulk InGaAsP active region (4 × 0.3 × 0.15 μm3) with a photoluminescence peak at 1.45 μm in an InP PhC line defect waveguide. This ultracompact InGaAsP region was successfully buried in the waveguide by using a butt-joint regrowth process. We found that the carrier relaxation time τc in a BH-PhC nanocavity was ~7 ns , which proves that the carriers were strongly confined and that no undesirable non-radiative recombination centers were generated at the regrowth interface. The air hole diameter and the lattice constant of the PhC were 225 and 420 nm, respectively. The widths of the InP and InGaAsP-embedded regions were changed to 1.1W0 and 0.98W0, respectively, where W0 is the basic line defect width defined as the removal of one row of air holes. A high-Q nanocavity mode was then formed as a result of local index modulation by embedding the InGaAsP in the InP. The required footprint for our nanocavity including the surrounding PhC area was less than 100 μm2.
Figure 2(b) shows the transmission spectra of CW light at different input powers. The loaded Q factor in the linear regime is 16,000. At shorter wavelengths the spectra become distorted by the photo-generated carriers as the input power is increased. A bistable window clearly appears if an additional pump light is injected. Figure 1(c) shows the hysteresis response when the laser wavelength was detuned by δ from the resonance, and Pout versus Pin was measured for upward/downward sweeps of the power level. The smallest bistable threshold power was approximately −46 dBm (25 nW).
4. Measurement setup and all-optical gate (1 × 1) switching
Figure 3(a) shows the experimental setup for acquiring the AOPS operation, and Fig. 3(b) and 3(c) show the input waveform of the bias light, write pulse, reset pulse, and packet data. All the light inputs were obtained by modulating the continuous-wave (CW) light with a lithium niobate (LN) modulator, where the pulse width for the write pulse was 100 ps, the reset pulse was 50 ns, and the packet data were 27-1 PRBS NRZ data with a 10 Gb/s repetition. To avoid data loss, the guard time of the packet should exceed the switching time, which is determined by the carrier relaxation time (~7 ns in this device). In this experiment, we set a sufficiently large guard time of 50 ns (equal to the reset pulse width). The optical power of the write pulse was appropriately adjusted with an erbium-doped fiber amplifier (EDFA), a band-pass filter (BPF) with a 0.3-nm spectral width, and a variable optical attenuator (VOA), while the optical power of the bias light and packet data was adjusted by changing the power of the CW laser source. The actual optical power and the wavelength of each light were adjusted in each switching scheme and are described later. These light sources were combined with a coupler and set with transverse electric polarization, which is defined as the electric field in the plane of the PhC slab. Then, these lights were coupled via a lensed fiber into a 3-μm-wide PhC waveguide and an intermediate spot size converter so that the light could be input efficiently into the PhC W1 waveguide. A reset pulse was applied with a period of 1 μs, and packet data were injected between each reset pulse, as shown in Fig. 3(b). The write pulse was injected immediately after the reset pulse as shown in Fig. 3(c), when applying the memory switching. The output lights from the device were observed through the EDFA, BPF with a 0.3-nm spectral width, and sampling oscilloscope. The output waveforms of the bias light and packet data were selectively monitored with a BPF and a sampling oscilloscope. When observing the waveform of a multiple-output-port device with a drop port and a through port, each port can be selected by changing the position of a lensed fiber. In the measurement, an additional loss of −22 dB occurred primarily due to the coupling loss between the optical fiber and the PhC waveguide. Note that our devices are directed to use in on-chip communication system rather than to use for an external fiber system, and therefore the optical power injected into the cavity was estimated by measuring the power in the fiber and the insertion loss of −11 dB. Unfortunately, on the other hand, such a high coupling loss prevents us from measuring thedevices in a rigorous way such as a bit error rate and an eye diagram. We evaluated a switch operation with an averaged optical waveform in the following experiment, which should be enough for the proof of principle in switch operation.
For the gate switching using a single cavity shown in Fig. 2, the wavelength detuning amounts of the bias light, write pulse, and packet data were set at −0.4, −0.1, and −0.5 nm, respectively, from the original resonant wavelength. The optical powers for the bias light and the packet data were set at 800 nW and 1.0 μW, respectively, and the energy for the write pulse was 13 fJ. Figure 4 shows the gate switching results, and reveals the output waveform of the bias light and the packet data. When the write pulse was injected, the transmission level of the bias light became higher, which indicated that the bistable memory was turned on. This also increased the transmission level of the packet data, because the packet data wavelength was set close to the bias light and so the change in the transmission level was also similar. The extinction ratio of the packet data between the on and off states was more than 10 dB. This clearly shows that both memory holding and packet data switching could be all-optically achieved with a single nanocavity. We observed the head/end of each packet, and confirmed there was no data loss because of the sufficiently long guard time (50 ns). The total power consumption of a few μW for the operation is three orders of magnitude lower than that of laser-based all-optical memories.
5. 1 × 2 port-selective switching with single nanocavity
When configuring an AOPS system for routing an optical packet to different output ports by using gate switches, we generally require the combination of a 1 × N splitter and a corresponding number of gate switches. However, such a broadcast-and-select configuration suffers from significant optical loss if there is no optical amplifier. A low-loss and output-port selective AOPS is desirable for practical use. To this end, our nanocavity was coupled with a bus waveguide and a drop waveguide, as shown in Fig. 5(a), to switch the optical packet to the through and drop ports. The nanocavity consists of an InGaAsP-embedded region with a length of 6a (~2.5 μm) in an InP PhC waveguide with a length of 25a. The cavity is coupled with the bus and drop waveguides with a separation of five rows of air holes for each waveguide. Figure 5(b) shows the transmission spectrum of the through and drop ports. There are two cavity modes, namely the 0th and 1st cavity modes, which are the longer- and shorter-wavelength modes, respectively. As shown in the magnified spectrum, the loaded Q factors for the 0th and 1st modes were 11300 and 19300, respectively, and both modes exhibited bistable behavior when the input optical power was increased. For a packet switching experiment, the bias light was assigned to the 0th mode, and the write pulse and the packet data were assigned to the 1st mode. This difference in assignment is effective for easily realizing data extraction from output light mixed with bias light by using a BPF or an add-drop multiplexer. Figure 5(c) shows the hysteresis response when the optical output for the drop port was observed and the laser wavelength was detuned by δ from the resonance of the 0th mode, which clearly indicated a bistable window at an input power larger than −26 dBm (2.5 μW).
The AOPS operation was demonstrated as shown in Fig. 6. The wavelength detuning of the bias light was −0.3 nm from the original resonance of the 0th mode, and those of the write pulse and packet data were 0 nm and −0.2 nm, respectively, from the original resonance of the 1st mode. The optical powers for the bias light and packet data were set at 2.3 and 3.0 μW, respectively, and the energy for the write pulse was 34 fJ. When the write pulse is injected, the transmission into the drop port becomes higher for both the bias light and the packet data, as similar to the situation in the gate switch. When the write pulse is not injected, the memory state is kept at a low level and hence both lights go to the through port. The extinction ratios for the packet data during switching were approximately 8 and 5 dB for the drop ports and the through port, respectively. The drop efficiency was as low as −12 dB even for the bistable on state. This low extinction ratio and drop efficiency are caused by the unbalanced coupling Q, although they would be greatly improved if the coupling Q were appropriately designed, as discussed in section 7.
6. 1 × 3 port-selective switching with two nanocavities
To demonstrate the scalability of our AOPS, we coupled two nanocavities to a bus waveguide and a drop port for each cavity, as shown in Fig. 7(a). The cavity structure is the same as that of the 1 × 2 switch, and the second cavity is placed opposite the first one across the bus waveguide with a separation of 35a in the x direction. Figure 7(b) shows the transmission spectra for the through port and two drop ports. There are two cavity modes, in which the resonant wavelength of the 1st cavity mode is identical for two cavities. Note that no cavity-cavity coupling or subsequent resonance split was observed, which suggested that the cavity loss due to the out-coupling loss rate into the drop port and the internal cavity loss rate are greater than the cavity-cavity coupling rate. For a packet switching experiment as described later, the bias light and the packet data light were assigned to the 1st mode, and the write pulse was assigned to the 0th mode. Figure 7(c) shows the hysteresis responses for the output light from two drop ports, for which the laser wavelength was detuned from the resonance of the 1st mode. There are power ranges in which both cavities exhibited a bistable window at the same wavelength. These conditions can be useful for selective and exclusive bistable switching by writing one of the cavities. The resonant wavelength of the 0th cavity mode is different for the two cavities, which means that selective switching is possible even with the injection of a write pulse into the bus waveguide. (Even if the resonant wavelength of the 0th cavity mode is also identical for the two cavities, selective switching is possible by injecting a write pulse into each drop port.)
AOPS operation was demonstrated as shown in Fig. 8. The wavelength detuning of the bias light and packet data were −0.3 and −0.2 nm from the resonance of the 1st mode, and those of the write pulse for the two cavities was 0 nm from the resonance of the 0th mode for each. The bias light power was set at 6.9 μW, which can maintain the bistable state for a single cavity but cannot maintain it for both so that only one of the cavities can be selectively switched on. The optical power for the packet data was set at 10.0 μW, and the energy for the write pulse was 130 fJ. The packet data were switched into the drop ports for the selective injection of a write pulse into the corresponding cavity or into the through port for no write pulse injection. This indicated a 1 × 3 switch operation as shown in Fig. 1(c). The extinction ratio for the packet data during the switching is as low as 4 and 1.4 – 3 dB for the drop ports and the through port, respectively, and the drop efficiency was as low as −20 dB. As with the 1 × 2 switch, these low extinction and drop efficiency values are caused by an imbalance in the optical coupling between the bus waveguide and the drop waveguide. The coupling Q with a bus waveguide Qbus_c and with a drop waveguide Qdrop_c were estimated to be 1 × 105 and 1 × 106, respectively, with an intrinsic cavity loss Qint of 3 × 104 and an absorption loss Qabs of 1 × 105. This coupling imbalance resulted in a drop efficiency of −20 dB, and subsequently the extinction ratio for the through port was less than 3 dB even at the resonant wavelength, as suggested by the transmission spectrum of Fig. 7(b). To optimize the drop efficiency and extinction ratio for the switching, Qbus_c and Qdrop_c should be balanced, and both Q values should be sufficiently higher than Qint, as discussed in the next section. However, our experimental results for 1 × 2 and 1 × 3 switches provided the first clear evidence that an optical packet can be switched to a different direction by changing and holding the bistable memory states. It also suggests the feasibility of an extended configuration for a 1 × N optical packet switch using integrated multiple nanocavities.
Finally, we theoretically investigated the potential performance of our AOPS with achievable cavity parameters. Figure 9(a), 9(b), and 9(c) show the simulated drop efficiency ηdrop, the cavity linewidth Δλcav, and the bistable threshold power Pth, respectively, as a function of the coupling ratio Q between a cavity-to-drop waveguide (Qdrop_c) and a cavity-to-bus waveguide (Qbus_c), where the ratio between the intrinsic cavity loss (Qint) and Qbus_c is taken as a parameter. We assumed both Qbus_c and the absorption (Qabs) to be 1 × 105, which we roughly estimated from the experimental drop spectrum in Fig. 7. The bistable threshold power Pth is given by14]. The equation suggests that a large Q/V ratio is advantageous in terms of realizing a low power consumption, and a PhC nanocavity is therefore suitable for this term. In this calculation, we assumed ħω = 0.8 eV, τc = 7 ns, V = 0.15 μm3, Γ = 0.4, and σc = 8 × 10−26 m3. As shown in Fig. 9(a)-9(c), the experimental condition was fitted with Qdrop_c/Qbus_c = 10 and Qint/Qbus_c = 0.3, which suggested a significant degradation in both ηdrop and Pth. The large Qdrop_c/Qbus_c might result from an asymmetric spatial cavity-field overlap with the bus waveguide and the drop waveguide in the present device. This is the reason for low transmission level for the drop ports, as shown in Fig. 5(b) and 7(b). As seen in the Fig. 9(a), however, balanced optimization would be obtained by making Qdrop_c/Qbus_c around 1 and Qint/Qbus_c as high as possible. For our device, by designing these ratios (Qdrop_c/Qbus_c = 0.5 and Qint/Qbus_c = 10) we can improve ηdrop and Pth to more than −7 dB and less than 100 nW, respectively, while maintaining Δλcav at around 10 GHz. The maximum possible ηdrop would be −3 dB even if we assume that there is no absorption in the present configuration, and in this case, the optical crosstalk into the through port would be −6 dB and the optical reflection into the input port would also be −6 dB. Although these crosstalks are an intrinsic problem for a three-port coupling system, this problem can be overcome by modifying the cavity-waveguide configuration. One way is to block the through port to make it just as a mirror, as shown in the inset of Fig. 9. Each cavity works like a gate switch, as described in section 4, and has a balanced coupling Q. Although the distance from the cavity to the mirror should be designed to obtain an appropriate Qbus_c as discussed in , this configuration can potentially enhance ηdrop up to unity and correspondingly reduce the optical crosstalk into the other drop ports. One of the output ports (through port) is sacrificed to realize this configuration. However, as mentioned later, the number of cavities can be increased without a large increase in power consumption. The dashed red curves in Fig. 9 show the simulated results for this configuration, where we also assumed possible values of Qbus_c = 1 × 104 and Qabs = 1 × 105. This indicates that an ηdrop of almost unity, a Δλcav of more than 40 GHz, and a Pth of less than 1 μW are simultaneously expected around Qdrop_c/Qbus_c = 1.
AOPSs generally must meet certain requirements for optical packet routing applications such as a low power consumption, small size, fast switching, scalability, and data format transparency. The AOPS configuration we demonstrated using bistable PhC nanocavities has certain significant features even when introduced at the chip level. A possible low power consumption of less than 1 μW is several orders lower than those of the previously reported AOPS devices, which are of mW order [7, 12]. Although a bit error rate and an eye diagram measurements are more desirable in the power comparison, the switched waveform without large distortion in our experiment suggests the availability as an AOPS device. Even if the number of the cavities is increased to configure a 1 × N switch, the bias power should not increase linearly and should be almost constant. This is because the bias light wavelength is detuned from the resonant wavelength and light is coupled into only one of the cavities during memory holding. (If the wavelength detuning is sufficiently large, the coupling of bias light into the other cavities should be negligible.) On the other hand, we may need to consider the integration of heaters beside each cavity in getting a coincidence of resonant wavelength. For example, a 4.8-μm-diameter silicon microring consumes 50 μW for a thermal resonance tuning of 1 nm . However, the cavity field of our PhC nanocavity is concentrated in the vicinity of a small InGaAsP region with a footprint of less than 2 μm2, and therefore the tuning power should also be significantly smaller than those of other structures in the same way as the operating power.
The fast switching speed can reduce the shortest guard time of the packet to enhance the data throughput. However, the switching speed and the power consumption are generally in trade-off relation each other. In our experiment, the switching speed is dominantly limited by the carrier lifetime τc, although the long τc would reduce the bistable threshold power, as indicated by Eq. (1). One effective approach is the integration of an electrical junction [21, 22] to allow us to apply an electric field just at the reset timing to instantaneously sweep the carriers. This would result in both keeping the low power consumption and enabling fast switching.
In addition, a wavelength-division multiplexing (WDM) system will be configured by monolithically integrating cavities with a different resonant wavelength and sharing a single bus waveguide as we previously demonstrated [23, 24]. A large free-spectral range (FSR) of more than 10 nm thanks to the ultrasmall nanocavity is desirable in terms of increasing the number of operative channels. We have recently reported that when we embed a three-hole-missing (L3) nanocavity in the same InGaAsP region and adjust the position of the surrounding air holes we obtain an InGaAsP volume of 0.09 μm3, a footprint including the surrounding PhC area of 40 μm2, a Qint of more than 2 × 105, and an FSR of more than 40 nm . Such a nanocavity can make it feasible to obtain the optimum condition discussed in Fig. 9, a lower aggregate power consumption, and large-scale WDM integrability. Our system will also allow data-format and data-rate transparency if the rate is within the cavity linewidth. Our system is very effective even with the emergence of new data formats in various forms of optical communication systems.
We demonstrated an AOPS by using integrated PhC nanocavities for the first time. The nanocavities provide both all-optical memory holding and switching operation for 10-Gb/s optical packet, and 1 × 1, 1 × 2, and 1 × 3 switches were configured with an ultralow power consumption of a few μW or less. This corresponds to energy-per-bit data switching of the order of 100 aJ/bit. These power and energy are more than two orders of magnitude lower than those of previous AOPSs. The footprint of less than 100 μm2 is also one order of magnitude smaller than previous switches. This confirms the feasibility of realizing a 1 × N switch by integrating a larger number of cavities without any large increases in power and area. Our theoretical investigation indicated that the optimum design for a nanocavity-waveguide coupling system can improve performance in terms of drop efficiency close to unity, and realize a data bandwidth of more than 40 GHz, and an operating power of less than 1 μW. We believe that this AOPS can provide an all-optical switching subsystem in a photonic router, with which to replace an OPS system combined with optical-to-electrical converters, towards the realization of a chip-scale photonic network.
We thank T. Tamamura, H. Onji for help in fabricating the device.
References and links
1. M. Notomi, K. Nozaki, A. Shinya, S. Matsuo, and E. Kuramochi, “Toward fJ/bit optical communication in a chip,” Opt. Commun. 314, 3–17 (2014). [CrossRef]
2. 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]
3. K. Nozaki, T. Tanabe, A. Shinya, S. Matsuo, T. Sato, H. Taniyama, and M. Notomi, “Sub-femtojoule all-optical switching using a photonic-crystal nanocavity,” Nat. Photonics 4(7), 477–483 (2010). [CrossRef]
4. T. Hirooka, M. Okazaki, T. Hirano, P. Y. Guan, M. Nakazawa, and S. Nakamura, “All-Optical Demultiplexing of 640-Gb/s OTDM-DPSK Signal Using a Semiconductor SMZ Switch,” IEEE Photonics Technol. Lett. 21(20), 1574–1576 (2009). [CrossRef]
5. S. C. Nicholes, M. L. Masanovic, B. Jevremovic, E. Lively, L. A. Coldren, and D. J. Blumenthal, “An 8 x 8 InP Monolithic Tunable Optical Router (MOTOR) Packet Forwarding Chip,” J. Lightwave Technol. 28(4), 641–650 (2010). [CrossRef]
7. K. Takeda, M. Takenaka, M. Raburn, Y. Kanema, J. S. Barton, X. L. Song, and Y. Nakano, “Dynamic operation of all-optical flip-flops with distributed Bragg reflectors for self-routing of 10-Gbit/s optical packets,” Jpn. J. Appl. Phys. 46(3A), 1028–1032 (2007). [CrossRef]
8. S. Tanaka, S. H. Jeong, S. Yamazaki, A. Uetake, S. Tomabechi, M. Ekawa, and K. Morito, “Monolithically Integrated 8:1 SOA Gate Switch With Large Extinction Ratio and Wide Input Power Dynamic Range,” IEEE J. Quantum Electron. 45(9), 1155–1162 (2009). [CrossRef]
9. N. Sherwood-Droz, H. Wang, L. Chen, B. G. Lee, A. Biberman, K. Bergman, and M. Lipson, “Optical 4x4 hitless slicon router for optical Networks-on-Chip (NoC),” Opt. Express 16(20), 15915–15922 (2008). [CrossRef] [PubMed]
10. A. Biberman, B. G. Lee, N. Sherwood-Droz, M. Lipson, and K. Bergman, “Broadband Operation of Nanophotonic Router for Silicon Photonic Networks-on-Chip,” IEEE Photonics Technol. Lett. 22(12), 926–928 (2010). [CrossRef]
11. M. Yang, W. M. J. Green, S. Assefa, J. Van Campenhout, B. G. Lee, C. V. Jahnes, F. E. Doany, C. L. Schow, J. A. Kash, and Y. A. Vlasov, “Non-Blocking 4x4 Electro-Optic Silicon Switch for On-Chip Photonic Networks,” Opt. Express 19(1), 47–54 (2011). [CrossRef] [PubMed]
12. K. Huybrechts, T. Tanemura, Y. Nakano, R. Baets, and G. Morthier, “40-Gb/s All-Optical Packet Switching With a Distributed-Feedback Laser as All-Optical Flip-Flop,” IEEE Photonics Technol. Lett. 21(11), 703–705 (2009). [CrossRef]
13. L. Liu, R. Kumar, K. Huybrechts, T. Spuesens, G. Roelkens, E. J. Geluk, T. de Vries, P. Regreny, D. Van Thourhout, R. Baets, and G. Morthier, “An ultra-small, low-power, all-optical flip-flop memory on a silicon chip,” Nat. Photonics 4(3), 182–187 (2010). [CrossRef]
14. K. Nozaki, A. Shinya, S. Matsuo, Y. Suzaki, T. Segawa, T. Sato, Y. Kawaguchi, R. Takahashi, and M. Notomi, “Ultralow-power all-optical RAM based on nanocavities,” Nat. Photonics 6(4), 248–252 (2012). [CrossRef]
15. A. Shinya, S. Matsuo, T. Yosia, T. Tanabe, E. Kuramochi, T. Sato, T. Kakitsuka, and M. Notomi, “All-optical on-chip bit memory based on ultra high Q InGaAsP photonic crystal,” Opt. Express 16(23), 19382–19387 (2008). [CrossRef] [PubMed]
16. C. Husko, A. De Rossi, S. Combrie, Q. V. Tran, F. Raineri, and C. W. Wong, “Ultrafast all-optical modulation in GaAs photonic crystal cavities,” Appl. Phys. Lett. 94(2), 021111 (2009). [CrossRef]
17. S. Matsuo, A. Shinya, T. Kakitsuka, K. Nozaki, T. Segawa, T. Sato, Y. Kawaguchi, and M. Notomi, “High-speed ultracompact buried heterostructure photonic-crystal laser with 13 fJ of energy consumed per bit transmitted,” Nat. Photonics 4(9), 648–654 (2010). [CrossRef]
18. B. R. Bennett, R. A. Soref, and J. A. Delalamo, “Carrier-induced change in refractive-index of InP, GaAs, and InGaAsP,” IEEE J. Quantum Electron. 26(1), 113–122 (1990). [CrossRef]
21. L. D. Haret, X. Checoury, Z. Han, P. Boucaud, S. Combrié, and A. De Rossi, “All-silicon photonic crystal photoconductor on silicon-on-insulator at telecom wavelength,” Opt. Express 18(23), 23965–23972 (2010). [CrossRef] [PubMed]
22. K. Takeda, T. Sato, A. Shinya, K. Nozaki, W. Kobayashi, H. Taniyama, M. Notomi, K. Hasebe, T. Kakitsuka, and S. Matsuo, “Few-fJ/bit data transmissions using directly modulated lambda-scale embedded active region photonic-crystal lasers,” Nat. Photonics 7(7), 569–575 (2013). [CrossRef]
23. K. Nozaki, E. Kuramochi, A. Shinya, and M. Notomi, “25-channel all-optical gate switches realized by integrating silicon photonic crystal nanocavities,” Opt. Express 22(12), 14263–14274 (2014). [CrossRef] [PubMed]
24. E. Kuramochi, K. Nozaki, A. Shinya, K. Takeda, T. Sato, S. Matsuo, H. Taniyama, H. Sumikura, and M. Notomi, “Large-scale integration of wavelength-addressable all-optical memories in a photonic crystal chip,” Nat. Photonics 8(6), 474–481 (2014). [CrossRef]
25. E. Kuramochi, E. Grossman, K. Nozaki, K. Takeda, A. Shinya, H. Taniyama, and M. Notomi, “Large Q factor enhancement of Ln nanocavity by a unified hole-shifting rule ” in Conference on Lasers and Electro-Optics Pacific Rim (CLEO-PR), Kyoto, WI2–2 (2013).