Advances in Optical Technologies

Volume 2008 (2008), Article ID 568936, 10 pages

http://dx.doi.org/10.1155/2008/568936

## On-Chip All-Optical Switching and Memory by Silicon Photonic Crystal Nanocavities

NTT Basic Research Laboratories, NTT Corporation, 3-1 Morinosato-Wakamiya, Atsugi 2430198, Japan

Received 23 December 2007; Accepted 13 April 2008

Academic Editor: D. Lockwood

Copyright © 2008 Masaya Notomi et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

#### Abstract

We review our recent studies on all-optical switching and memory operations based on thermo-optic and carrier-plasma nonlinearities both induced by two-photon absorption in silicon photonic crystal nanocavities. Owing to high-*Q* and small volume of these photonic crystal cavities, we have demonstrated that the switching power can be largely reduced. In addition, we demonstrate that the switching time is also reduced in nanocavity devices because of their short diffusion time. These features are important for all-optical nonlinear processing in silicon photonics technologies, since silicon is not an efficient optical nonlinear material. We discuss the effect of the carrier diffusion process in our devices, and demonstrate improvement in terms of the response speed by employing ion-implantation process. Finally, we show that coupled bistable devices lead to all-optical logic, such as flip-flop operation. These results indicate that a nanocavity-based photonic crystal platform on a silicon chip may be a promising candidate for future on-chip all-optical information processing in a largely integrated fashion.

#### 1. Introduction

It
used to be a great challenge to tightly confine light in a wavelength-scale volume,
which had limited the capability of photonics technologies in various aspects.
Recently, however, an optical resonator with ultrahigh-*Q* (~1 million) and small volume (~)
becomes possible by employing the state-of-the-art photonic crystal
technologies [1–6]. Figure 1 shows an example of high-*Q* nanocavities, which we have recently demonstrated in silicon photonic crystals.
This particular cavity shows a theoretical *Q* of over 10^{8} and an experimental *Q* of 1.3 million with a mode volume of [4, 7].
These values are hardly available in optical systems other than photonic
crystals. Various forms of light-matter interactions are expected to be
enhanced in such high-*Q* nanocavities
having large *Q/V* ratio [8]. Among them, in this article, we
focus on application to all-optical switching and memory operations based on
optical nonlinear interaction. Especially for this, we investigate all-optical operations
based on carrier-induced nonlinearity and examine the features of photonic
crystal nanocavities for such applications.

As has been studied in various forms,
all-optical switches can be realized using optical resonators, where a control
optical pulse induces a resonance shift via optical nonlinear effects. For such
a resonator-based switch, there is a two-fold enhancement in terms of the
switching power if a small cavity with a high-*Q* is employed. First, the light intensity inside the cavity should
be proportional to . Second, the required wavelength shift is
proportional to 1/*Q*. In total, the switching power should be reduced by
(), which can be significantly large for photonic
crystal nanocavities [9].
Although the switching mechanism itself is basically similar to that of
previous resonator-based switches, such as nonlinear etalons [10], this large enhancement has had an
important impact on optical integration since most optical switching components
require too much power for realistic integration. Furthermore, in the case of
carrier-induced nonlinearity or thermo-optic nonlinearity, the smallness of the
device naturally leads to fast operation speed, since the relaxation time of
such processes normally depends on the size. As we will show below, this
advantage is significant for wavelength-sized cavities.

In addition, resonator-based optical switches are well known to exhibit optical bistability [10] and thus they can be used for optical memory and all-optical logic [11]. Such functionality is one of the most important functions missing from existing photonic devices. Thus, we believe that all-optical bistable switches based on photonic crystal cavities are important candidates for future optical integration.

In this paper, we investigate all-optical switching and memory action in silicon photonic crystal nanocavity devices. First, we study the switching action using thermo-optic nonlinearity. Next, we investigate similar switching action using much faster carrier-plasma nonlinearity. Thirdly, we analyze the effect of carrier relaxation process in the switching action and made an attempt to further decrease the relaxation time by ion implantation. Furthermore, we demonstrate bistable memory action employing basically the same nanocavity devices and present an example of design for on-chip all-optical logic circuits consisting of two bistable nanocavities.

#### 2. All-Optical Bistable Switching by Thermo-Optic Nonlinearity

As
described above, photonic crystal nanocavities have a promising potential for
all-optical switching applications. To experimentally confirm this, we have investigated
all-optical bistable switching operations employing the thermo-optic
nonlinearity induced by two-photon absorption (TPA) in silicon [12]. For this study, we
employed an end-hole shifted four-point Si photonic crystal cavity integrated
with input/output photonic crystal waveguides (shown in Figure 2(a)) [13, 14]. Figure 2(b) shows transmission spectra with various input power at 1.5 *μ*m wavelength region. As clearly seen in the
graph, the transmission spectrum is shifted to longer wavelength (red-shifted).
This shift is due to thermo-optic nonlinearity induced by two-photon absorption
in the cavity. At input power higher than 10 *μ*W, there is an abrupt kink in the spectrum. This
kink corresponds to bistable switching. To see this more directly, we measured
the output power as a function of the input power at various detuning
conditions, as shown in Figure 2(c). Now, it is clear that this device exhibits
bistable switching. The most noteworthy point regarding this switching is its
switching power, which is as small as 40 *μ*W. This value is remarkably smaller that of
bulk-type thermo-optic nonlinear etalons (a few to several tens mW) [15]
and also smaller than that of recent miniature-sized thermo-optic silicon
microring resonator devices (~0.8 mW) [16]. In addition, it is important to note that TPA occurs only in the cavity (the
linear absorption of silicon at this wavelength is negligible), and, therefore,
we can easily integrate this device with transparent waveguides in the same
chip.

Our cavity shown in Figure 2(a) was intentionally designed to have two resonant modes, and, therefore, we can perform all-optical switching using independent two inputs with different wavelengths (signal and control), which is normally required for signal processing. Since bistable switching for one resonant mode should influence the transmission intensity of the other resonant mode in the same cavity, we can easily realize such operation. As we reported in [12], we observe basically similar bistable switching behavior in the relation between the input power for the control mode and the output power for the signal mode. This operation is basically what we expect for the so-called all-optical transistors, and will be basis for various logic functions. The detail of this operation is described in [12]. For example, we demonstrated that we can amplify an AC signal using this device.

Although the bistable operation itself is
similar to that of nonlinear etalon switches, these photonic crystal switches
can be clearly distinguished in terms of the operating power and capability for
integration. The mode volume of this cavity is only approximately 0.1 *μ*m^{3}. This small footprint itself is of
course advantageous for integration, but it is also beneficial for reducing the
switching speed because our device is limited by the thermal diffusion process.
To demonstrate these characteristics, we measured temporal response of the operation
for switch-off and switch-on processes, as shown in Figures 3(a) and 3(b). The
relaxation time of our switch is approximately 100 nanoseconds, which is much
shorter than that of conventional thermo-optic switches (~milliseconds). This
amazingly fast thermo-optic switching is primarily due to the smallness of our
cavity. In addition, we can estimate the required switching energy from the
switch-on measurement. The deduced smallest value is as small as 11 pJ as shown
in Figure 3(c).

#### 3. All-Optical Switching by Carrier-Plasma Nonlinearity

These thermo-optic nonlinear bistable switches clearly demonstrate that large photonic crystal cavities are very effective in improving the operation power and speed. However, the speed itself is still not very fast, which is limited by the intrinsically slow thermo-optic effect. To realize much faster all-optical switches, here we employ another nonlinear effect, namely, the carrier-plasma effect [17]. This process is also based on the same TPA process in silicon. Thus most of the arguments concerning their advantages are similar to that for thermo-optic nonlinearity. For this experiment, we used similar photonic crystal cavity devices with a control pulse input. If the duration of the control pulse is sufficiently short, we can avoid thermal heating and may be able to observe only carrier-plasma nonlinearity. In fact, we observed a clear blue shift in the resonance when we injected a 6-picosecond pulse into this device, which is consistent with the expected shift induced by the carrier-plasma nonlinearity. Figure 4 shows the time-resolved output intensity for the signal mode when a 6-picosecond control pulse is input [17]. We observed clear all-optical AND-type switching from OFF to ON for the detuning of 0.45 nm (Figure 4(a)) and NOT-type switching from ON to OFF for the detuning of 0.01 nm (Figure 4(b)). The required switching energy is only a few hundred fJ, which is much smaller than that of ring-cavity-based silicon all-optical switches [18]. If we take the coupling efficiency between the cavity and waveguide into account, the actual pulse energy used for the switching is less than 10 fJ. This extremely small switching energy is attributed to large ratio in our cavity.

The switching time of our device is approximately from 50 picoseconds to 300 picoseconds depending on the detuning and control pulse energy. Considering the fact that the carrier lifetime in silicon is normally very long (~microseconds), this switching time is also surprisingly fast. This apparently indicates that the switching time is not limited by the carrier recombination time of bulk silicon. We analyzed the switching behavior employing simplified rate equations for the photon and carrier density in a photonic-crystal nanocavity incorporating the effective carrier relaxation time. We found that this simple model can reproduce the experimental result fairly well. Figure 5 shows simulated and experimental results for the same parameter condition. In this calculation, we assumed the carrier relaxation time of 80 picoseconds. This suggests that the effective carrier relaxation time for this all-optical switch is in the range between 50 to 100 picoseconds. This fast carrier relaxation may be attributed to the short diffusion time for generated carriers. Note that the photon lifetime of our cavity is approximately 10 picoseconds, and thus the operation speed of our device is limited by the carrier relaxation time. This carrier relaxation time is much shorter than that in other silicon photonic microdevices [18]. That is, the small footprint of the device is again effective in improving the operating speed.

#### 4. Effect of Carrier Relaxation Process on Switching Operation

In Section 3, we demonstrate that silicon all-optical switches based on
carrier-plasma nonlinearity can operate at a significantly fast speed, which is
attributed to the fast carrier relaxation process. We regard that this fast
carrier relaxation is possibly due to the fact that the diffusion process is an
efficient relaxation channel for nanocavities. To verify this explanation, we
performed numerical simulations for the carrier diffusion process in silicon
photonic crystal cavities [19]. In this simulation, we assumed an initial carrier distribution determined
from the optical intensity distribution of the cavity mode, and numerically
solved two-dimensional diffusion equations with assuming a realistic photonic
crystal cavity structure which is the same as that used in the experiment. The
side-wall nonradiative recombination process at the air-hole surface is
incorporated in the equation as effective surface recombination rate (*S*), and the effect of the top-surface
nonradiative recombination process is incorporated as an effective carrier
lifetime parameter which is determined from another calculation for a simple
slab. We calculated the time-dependent resonance wavelength shift using
instantaneous carrier concentration in the cavity. The calculated snapshots of
the carrier distribution at *t* = 0, 8,
and 24 picoseconds are shown in Figures 6(a), 6(b), and 6(c). These snapshots
clearly show that the initial distribution is rapidly spread as a result of
diffusion. This rapid diffusion results rapid switching recovery. Figure 6(d)
shows the calculated shift of the resonance wavelength for the cavity. The
initial wavelength shift caused by the carrier plasma shift is recovered
quickly. As shown in the figure, the nonradiative recombination does not
seriously affect the initial recovery. Therefore, we believe that the fast
operation of our optical switches is explained by the rapid carrier diffusion
process.

If we wish to increase the operation speed
further, we have to somehow decrease the carrier relaxation time. Although
there are several ways to do so, we have recently employed an Ar-ion
implantation process in order to introduce extremely fast nonradiative
recombination centers into silicon. If the carrier recombination time becomes
faster than the diffusion time, we can expect an improvement in the operation
speed. When we implanted silicon photonic crystal nanocavity switches with
Ar^{+} dose of cm^{−2} and an acceleration voltage of 100 keV, we observed a significant improvement in switching speed as shown in Figures
7(a), 7(b). The implantation condition was carefully determined so as to keep
almost the same cavity *Q* but to
significantly decrease the carrier lifetime in the cavity. In the case of
detuning for an NOT gate (Figure 7(a), it was reduced from 110 picoseconds to
50 picoseconds. In the case of detuning for an AND gate, the switching time was
reduced from 220 picoseconds to 70 picoseconds (Figure 7(b)) [20].
As shown in Figure 7(c), the ON/OFF ratio is mostly the same between two
conditions. In addition, the required switching energy was almost the same as
that without the implantation. Although the reduction of the carrier lifetime
may lead to increase in the switching energy, this change is not significant as
far as the carrier lifetime is longer than the photon lifetime and the control
pulse length (which was the case in our experiment). These results clearly
demonstrate the effect of ion-implantation on the improvement in the switching
speed.

#### 5. All-Optical Memory Operation and Logic Circuit

In
the same way as thermo-optic switching, carrier-plasma switching also provides
bistable operation. Figure 8 shows bistable operations realized by employing a
pair of set and reset pulses [21].
When a set pulse is fed into the input waveguide, the output signal is switched
from OFF to ON and remains ON even after the set pulse exits (green curve).
When a pair of set and reset pulses is applied (as shown in Figure 8(a)), the
output is switched from OFF to ON by the set pulse and then ON to OFF by the
reset pulse (blue curve). This is simply a memory operation using optical
bistability. The energy of the set pulse is less than 100 fJ, and the DC bias
input for sustaining the ON/OFF states is only 0.4 mW. These small values are
primarily the results of the large *Q*/*V* ratio of the photonic
crystal cavity. It is worth noting that the largest should
always result in the smallest switching power, but the operation speed can be
limited by *Q*. In the present situation, the switching speed is still
limited by the carrier relaxation time, and thus a large is
preferable. In the case when the photon lifetime limits the operation speed, we
have to choose appropriate loaded *Q* for the required speed. Even in such a case, it is better to have high unloaded *Q* because loaded *Q* can be controlled by changing the cavity-waveguide coupling, and
high unloaded *Q* means low loss of the
device. The best design of out device would be a device with the smallest
volume, the lowest transmission loss, and the designated *loaded Q* (depending on the operation speed). The lowest loss with the designated *loaded**Q* can be obtained only when we employs an
ultrahigh *unloaded**Q* cavity. Compared with other types of
all-optical memories, this device has several advantages, such as small
footprint, low-energy consumption, and the capability for integration. The fact
that all the light signals used for the operation are transparent in waveguides
is important for the application, which is fundamentally different from bistable-laser-based
optical memories.

In the above, we showed that a single photonic crystal cavity coupled to waveguides functions as a bistable switch or a memory. If we couple two or more bistable cavities, we can create much more complex logic functions [11], in the same way as with transistor-based logic in electronics. As an example, here we show our numerical design for an all-optical flip-flop consisting of two bistable cavities integrated in a photonic crystal. It has been proposed that all-optical flip-flops be realized by using two nonlinear etalons with appropriate cross-feedback [22], but this proposal is unsuitable for on-chip integration. Here, we propose a different design using two photonic crystal nanocavities [23].

A typical example of flip-flop operation in the
high-speed information processing is a retiming circuit, which corrects the
timing jitter of an information bitstream and synchronizes it with the clock
pulses. This function is normally accomplished by high-speed electronic
circuits, but if it can be done all optically, it will be advantageous for
future ultrahigh-speed data transmission.
Here, we propose a simple flip-flop design for realizing the retiming
function. Figure 9(a) shows a design for the retiming circuit. The coupled
cavities (C1 and C2) have one common resonant mode ( nm, ) extended to
both cavities and two modes ( nm, , and nm, ) localized in
each cavity. Here, we use two bistable switching operations for C1 and C2. The
cross-feedback is realized as follows. C1 is switched ON only when and are both applied (P_{IN}1 and P_{IN}2
are ON). C2 is ON only when are applied (P_{IN}3 is ON) and
simultaneously is supplied from C1 (which means C1 is ON).
Thus the output signal of (P_{OUT}3) becomes ON only if P_{IN}3
is turned ON when C1 is already ON in advance. These results achieve retiming
process. We set P_{IN}1 and P_{IN}3 as two different clock
signals as shown in Figure 9(b), and assume P_{IN}2 to be bit-stream
NRZ (nonreturn-to-zero) data with finite timing jitter. The resultant P_{OUT}3
is precisely synchronized to the clock signals and is actually an RZ
(return-to-zero) data stream converted from P_{IN}2 with jitter
corrected.

We designed this function in a photonic crystal
slab system, and numerically simulated its operation using the FDTD method. The
structural parameters are shown in the figure caption. We assumed realistic
material parameters (with a Kerr coefficient (m^{2}/V^{2}), a typical value for AlGaAs) and the
instantaneous driving power is assumed to be 60 mW for all three inputs. Figure 9(b) shows three input signals (a data
stream with jitter, and two clock pulses), and the output from P_{D}(P_{OUT}3).
As seen in this plot, P_{OUT}3 is the RZ signal of the input with the
jitter corrected. We confirmed that the operation speed corresponds to 50 GHz
operation. Note that this work was intended to demonstrate the operation
principle and the structure has not yet been optimized. We expect ~mW operation
may be possible after the optimization since our rough estimation shows that a
single cavity switch can operate at sub-mW input power for the similar
condition. In addition, if we manage to employ the carrier-plasma effect for
this operation (as was done in Figure 8), we can expect further decrease in
terms of the input power.

In addition to this work, we have reported another design of all-optical logic circuit which is mostly equivalent to an SR flip-flop employing two bistable cavities [8].

#### 6. Summary

Recent rapid progress in photonic crystal nanocavities is enabling low-power, all-optical switching and memory actions on a silicon chip. We have shown that our experimental demonstration of all-optical switching operation using thermo-optic nonlinearity and carrier-plasma nonlinearity, both based on two-photon absorption in silicon. For both cases, we observed significant decrease in the switching power (energy) and also significant increase in the switching speed. In this article, we numerically investigated the diffusion process concerning photonic-crystal nanocavities, which is fundamentally different from that in conventional optical devices with much larger size. Moreover, we have demonstrated their potential for optical logic by a combination of bistable elements. For the application in all-optical processing, a photonic crystal has several important advantages, such as small footprint, small energy consumption, high speed, and significant integratability. Thus we believe that it may be a promising candidate for all-optical information processing on a silicon chip.

#### References

- B.-S. Song, S. Noda, T. Asano, and Y. Akahane, “Ultra-high-
*Q*photonic double-heterostructure nanocavity,”*Nature Materials*, vol. 4, no. 3, pp. 207–210, 2005. View at Publisher · View at Google Scholar - S. Noda, M. Fujita, and T. Asano, “Spontaneous-emission control by photonic crystals and nanocavities,”
*Nature Photonics*, vol. 1, no. 8, pp. 449–458, 2007. View at Publisher · View at Google Scholar - E. Kuramochi, M. Notomi, S. Mitsugi, A. Shinya, T. Tanabe, and T. Watanabe, “Ultrahigh-
*Q*photonic crystal nanocavities realized by the local width modulation of a line defect,”*Applied Physics Letters*, vol. 88, no. 4, Article ID 041112, 3 pages, 2006. View at Publisher · View at Google Scholar - T. Tanabe, M. Notomi, E. Kuramochi, A. Shinya, and H. Taniyama, “Trapping and delaying photons for one nanosecond in an ultrasmall high-
*Q*photonic-crystal nanocavity,”*Nature Photonics*, vol. 1, no. 1, pp. 49–52, 2007. View at Publisher · View at Google Scholar - R. Herrmann, T. Sünner, T. Hein, A. Löffler, M. Kamp, and A. Forchel, “Ultrahigh-quality photonic crystal cavity in GaAs,”
*Optics Letters*, vol. 31, no. 9, pp. 1229–1231, 2006. View at Publisher · View at Google Scholar - T. Tanabe, A. Shinya, E. Kuramochi, S. Kondo, H. Taniyama, and M. Notomi, “Single point defect photonic crystal nanocavity with ultrahigh quality factor achieved by using hexapole mode,”
*Applied Physics Letters*, vol. 91, no. 2, Article ID 021110, 3 pages, 2007. View at Publisher · View at Google Scholar - T. Tanabe, M. Notomi, E. Kuramochi, and H. Taniyama, “Large pulse delay and small group velocity achieved using ultrahigh-
*Q*photonic crystal nanocavities,”*Optics Express*, vol. 15, no. 12, pp. 7826–7839, 2007. View at Publisher · View at Google Scholar - M. Notomi, T. Tanabe, A. Shinya et al., “Nonlinear and adiabatic control of high-
*Q*photonic crystal nanocavities,”*Optics Express*, vol. 15, no. 26, pp. 17458–17481, 2007. View at Publisher · View at Google Scholar - M. Soljačić and J. D. Joannopoulos, “Enhancement of nonlinear effects using photonic crystals,”
*Nature Materials*, vol. 3, no. 4, pp. 211–219, 2004. View at Publisher · View at Google Scholar · View at PubMed - H. M. Gibbs,
*Optical Bistability: Controlling Light with Light*, Academic Press, Orlando, Fla, USA, 1985. - S. D. Smith, “Optical bistability, photonic logic, and optical computation,”
*Applied Optics*, vol. 25, no. 10, pp. 1550–1564, 1986. View at Google Scholar - M. Notomi, A. Shinya, S. Mitsugi, G. Kira, E. Kuramochi, and T. Tanabe, “Optical bistable switching action of Si high-
*Q*photonic-crystal nanocavities,”*Optics Express*, vol. 13, no. 7, pp. 2678–2687, 2005. View at Publisher · View at Google Scholar - S. Mitsugi, A. Shinya, E. Kuramochi, M. Notomi, T. Tshchizawa, and T. Watanabe, “Resonant tunneling wavelength filters with high
*Q*and high transmittance based on photonic crystal slabs,” in*Proceedings of the 16th Annual Meeting of IEEE Lasers and Electro-Optics Society (LEOS '03)*, vol. 1, pp. 214–215, Tucson, Ariz, USA, October 2003. View at Publisher · View at Google Scholar - M. Notomi, A. Shinya, S. Mitsugi, E. Kuramochi, and H.-Y. Ryu, “Waveguides, resonators and their coupled elements in photonic crystal slabs,”
*Optics Express*, vol. 12, no. 8, pp. 1551–1561, 2004. View at Publisher · View at Google Scholar - G. R. Olbright, N. Peyghambarian, H. M. Gibbs, H. A. MacLeod, and F. Van Milligen, “Microsecond room-temperature optical bistability and crosstalk studies in ZnS and ZnSe interference filters with visible light and milliwatt powers,”
*Applied Physics Letters*, vol. 45, no. 10, pp. 1031–1033, 1984. View at Publisher · View at Google Scholar - V. R. Almeida and M. Lipson, “Optical bistability on a silicon chip,”
*Optics Letters*, vol. 29, no. 20, pp. 2387–2389, 2004. View at Publisher · View at Google Scholar - T. Tanabe, M. Notomi, S. Mitsugi, A. Shinya, and E. Kuramochi, “All-optical switches on a silicon chip realized using photonic crystal nanocavities,”
*Applied Physics Letters*, vol. 87, no. 15, Article ID 151112, 3 pages, 2005. View at Publisher · View at Google Scholar - V. R. Almeida, C. A. Barrios, R. R. Panepucci, and M. Lipson, “All-optical control of light on a silicon chip,”
*Nature*, vol. 431, no. 7012, pp. 1081–1084, 2004. View at Publisher · View at Google Scholar · View at PubMed - T. Tanabe, H. Taniyama, and M. Notomi, “Carrier diffusion and recombination in photonic crystal nanocavity optical switches,”
*Journal of Lightwave Technology*, vol. 26, no. 11, pp. 1396–1403, 2008. View at Google Scholar - T. Tanabe, K. Nishiguchi, A. Shinya et al., “Fast all-optical switching using ion-implanted silicon photonic crystal nanocavities,”
*Applied Physics Letters*, vol. 90, no. 3, Article ID 031115, 3 pages, 2007. View at Publisher · View at Google Scholar - T. Tanabe, M. Notomi, S. Mitsugi, A. Shinya, and E. Kuramochi, “Fast bistable all-optical switch and memory on a silicon photonic crystal on-chip,”
*Optics Letters*, vol. 30, no. 19, pp. 2575–2577, 2005. View at Publisher · View at Google Scholar - H. Tsuda and T. Kurokawa, “Construction of an all-optical flip-flop by combination of two optical triodes,”
*Applied Physics Letters*, vol. 57, no. 17, pp. 1724–1726, 1990. View at Publisher · View at Google Scholar - A. Shinya, S. Mitsugi, T. Tanabe et al., “All-optical flip-flop circuit composed of coupled two-port resonant tunneling filter in two-dimensional photonic crystal slab,”
*Optics Express*, vol. 14, no. 3, pp. 1230–1235, 2006. View at Publisher · View at Google Scholar