A distributed feedback Bragg grating waveguide all-optical switch design relying on the ultrafast cross modulation effect of the intersubband transitions in InGaAs/AlAs/AlAsSb coupled double quantum wells is demonstrated. The pump-induced phase modulation to the signal light is converted to intensity modulation efficiently on chip with the help of the grating structures. To our best knowledge, the switching dynamic characteristics of such design are reported for the first time. With a 400 μm long grating waveguide, 3 dB modulation depth with switching energy of 5.5 pJ and recovery time of 4.5 ps is obtained for the switch-off operation.
© 2011 OSA
During the past decade, all-optical switch (AOS) designs based on near-infrared saturable intersubband transitions (ISBTs) between the conduction band levels of semiconductor quantum wells (QWs) have attracted a lot of attention due to their ultrafast and large optical nonlinearities . Recently, a novel and more efficient modulation phenomenon was reported in the InGaAs/AlAs/AlAsSb coupled double QWs (CDQWs) system. A strong intersubband (ISB) absorption of the transverse-magnetic (TM)-polarized control light can initiate ultrafast cross phase modulation (XPM) to the co-propagating transverse-electric (TE)-polarized signal light. The real carrier-induced XPM effect originates from two different mechanisms, the plasma dispersion effect [2–4] and the interband dispersion effect [5,6], both of which have been thoroughly studied lately. The XPM efficiency of ~0.03 to 0.5 rad/pJ has been achieved experimentally in the InGaAs/AlAsSb CDQWs of an optical ridge waveguide (WG) geometry [2,5,6]. With an off-chip Mach-Zehnder interferometer (MZI) configuration, high-speed all-optical devices for wavelength conversion , demultiplexing  and modulation  have been demonstrated.
In this article, a compact on-chip AOS design of a deeply-etched distributed feedback (DFB) grating WG is presented. The core of the slab WG is a InGaAs/AlAs/AlAsSb CDQWs stack exhibiting ISBTs. The principle of this AOS design is to convert the weak XPM effect to a intensity change with the help of the frequency-dependent transmission properties of the DFB grating structure. As illustrated in Fig. 1 , the pumping of carriers by a TM control light induces refractive index changes to the materials, which are experienced by the succeeding TE probe (or signal) light. Therefore, the Bragg wavelength of the DFB grating WG gets shifted for the TE signal light. The AOS can work under the transmission mode or the reflection mode depending on detecting the transmitted or reflected signal. With the help of the transmittance contrast between the passbands and stopband, the signal light can be either switched on or switched off according to the initial choices of the signal wavelength and the function mode. In this fashion, the TM pump light modulates the transmission properties of the TE signal light. Although it is possible to operate the AOS under the reflection mode, where the reflected signal is extracted by an optical circulator or polarizing beam splitter, in this study, only operation of the AOS under the transmission mode is discussed. The presented AOS may work with single color, and the control and signal pulses can be simply discriminated by the orthogonal polarization. Additional advantages can be the low insertion loss of the TE probe light and subsequently the superior signal-to-noise ratio. The use of the DFB grating structures have been suggested for AOS before [10–13], however, the detailed analysis of the switching characteristics of an ISBT-based DFB grating WG AOS are demonstrated for the first time to our best knowledge. The design presented in this paper differs from the devices reported earlier in  and  mainly in two aspects: (i) its ultrafast real carrier transition-based material refractive nonlinearities and (ii) stronger XPM efficiency. Both aspects are crucial for ultrafast switching applications. Switching energy for an on-off modulation about 3 dB of 5.5 pJ is obtained for our device. The 1/e on-off response time is 9 ps. The present monolithic DFB grating AOS device will be useful to various high-speed photonic data processing applications, including the all-optical clock extraction, all-optical signal format conversion, optical data storage, time-division-demultiplexing, wavelength conversion, and so on.
2.1 ISBT QW design
The symmetric AlAs0.56Sb0.44/In0.78Ga0.22As CDQW system consist two QWs with an effective width of 1.93 nm separated by 1.45 nm wide AlAs barrier. The Indium composition in the InGaAs QWs is designed to be large in order to increase the XPM efficiency . The CDQWs were Si δ-doped to an effective volume carrier concentration of 1 × 1019 cm−3 . The CDQWs are simulated by solving the one-dimensional Schrödinger equation self-consistently . Four dipole-allowed subbands in the conduction band exist as illustrated in Fig. 2 (a) . The symmetric nature of the QW does not allow direct ISBTs between energy levels from the same parity under the selection rule, so the transitions |1>→|3> and |2>→|4> are dipole forbidden. State |1>↔|2> and |3>↔|4> are strongly coupled and almost resonant with longitudinal phonons (LO phonon), respectively. The calculated ISBTs agree well with the intersubband absorption peaks measured in the polished sample with 45° edge multi-pass scheme by Fourier transform infrared (FTIR) spectroscopy. As shown in Fig. 2 (b), there is one absorption peak centered at 1765 nm corresponding to the transition |2>→|3>. Another weaker absorption peak (owing to a smaller dipole moment) corresponding to the transition |1>→|4> locates at 1550 nm. The overlap of the two wide absorption spectra is mainly caused by QW-barrier interface roughness. The measured 3-dB ISB absorption saturation intensity for TM modes is deducted to be 115 MW/cm2 characterized with a 150 fs ultrashort pulse from an optical parametric oscillator (OPO).
2.2 XPM effect
One of the two mechanisms contributing to the XPM effect is the carrier plasma dispersion effect, which causes the change of refractive index owing to the electron redistribution among the highly nonparabolic subbands of the CDQW conduction band. Figure 3 (a) shows the theoretical analysis of the calculated index change after pumping based on the dynamic modeling with four-level optical Bloch equations . The approach is similar to the rate equation model described in Ref . The longitudinal decay times including the phonon-assisted scattering effects used in the dynamic modeling are computed as τ12 = 69 fs, τ13 = 3.77 ps, τ14 = 3.96 ps, τ23 = 4.57 ps, τ24 = 3.87 ps, τ34 = 1.19 ps [15,16]. The equilibrium quasi-Fermi level lies above the band energy of state |2>. The excitation is a Gaussian-shape pulse of 150 fs full-width-at-half-maximum (FWHM) for illustrating the fast material response. By using the above set of parameters and a fitted dephasing time of ~24 fs, the calculated saturation intensity for the transition |1>→|4> is 113.2 MW/cm2, which is in agreement with the experimental value. According to the Drude model, the overall change in refractive index is calculated with the equation [1,2]Figure 3 (a) plots the time response of the index change Δn under the 3-dB saturation pumping light intensity, a net index change of 5.1 × 10−3 is excited, which is in the same order of magnitude as reported in Ref [1–4]. The 1/e exponential decay rate is 1.87 ps. Figure 3 (b) shows the dependence of index change on the increasing input pumping intensities. The plasma dispersion effect trends to saturate with pumping light intensity similar to the ISBT nonlinear absorption phenomenon, which is used for the more conventional cross absorption modulation (XAM)-based AOS . Moreover, as one important feature of the XPM effect under study, it is not fully dependent on the ISB absorption in the unsaturated regime. A stronger ISB absorption is not necessary to bring in a higher XPM efficiency [3,4].
The other mechanism is the interband transition (IBT)-induced refractive index change. Because the CDQW is compressively strained, the top-most valance band is a heavy-hole-like band so that the IBT for TE-polarized light is allowed after pumping . The interband dispersion can be obtained via Kramers-Krönig relationship [1,5]. The peak linear absorption coefficient at the IBT bandgap energy is assumed to be αIBT = 1 × 104 cm−1 , and the corresponding linear index change can be calculated simply by using the formula below assuming a Lorentzian-shape absorption spectrum :Fig. 4 (a) . Under 3-dB saturation pumping intensity, the index change of 2.4 × 10−3 is estimated at 0.8 eV probe photon energy, which is half of the value computed by Eq. (2). The TE interband absorption is controlled by the strongly-coupled ISB absorption of the pumping light ; hence, as a fast and simple approximation approach we adopt the TM pumping absorption recovery response as the time evolution of the instantaneous IBT-induced index change. As shown in Fig. 4 (b), the peak of the change is fixed, and the exponential recovery time is 1.12 ps which exhibit faster response than that of the carrier plasma effect. Taking both two XPM mechanisms into account, the peak overall index change is predicted to be positive 7.6 × 10−3. This XPM efficiency is one order of magnitude higher than the reported Kerr effect in III-V group semiconductors, such as InGaAsP bulk material with bandgap of 1.41 μm .
2.3 DFB WG design
The slab WG is composed of the following layer stacks, a 105 nm thick InGaAsP (n = 3.35) lower capping layer above the InP substrate (n = 3.14) bottom cladding layer, 70 periods of CDQWs (averaged n = 3.18) as core layer, 150 nm thick InGaAsP (n = 3.35) upper capping layer and 500 nm InP (n = 3.17) top cladding layer. The thickness of each capping or cladding layer is optimized to obtain a high mode confinement factor . The 2D TE fundamental mode profile simulated with a beam propagation method mode solver for a WG width of 1.2 μm is shown in Fig. 5 (a) . The TE and TM mode confinement factors are 52.4% and 51.6% in the core region, respectively. The calculated effective indices are 3.118 and 3.128 for TE and TM modes, respectively. The structural birefringence indicates that the locations of the Bragg wavelengths for both modes are not degenerated. Hence, it is possible to adopt TM pump and TE probe light with identical wavelength, and the pulses are thereafter discriminated by polarization.
As illustrated in Fig. 6 , the band edge of DFB grating WG is shifted by pumping, which leaves an operating window of good switching performance for the signal pulse. The magnitude of the shift is determined by the pumping conditions and the XPM efficiency as the intrinsic material properties. A proper probe signal pulse has to fit into this window as shown Fig. 6 (a), and then it can be switched with a high extinction ratio by the control pulse. For a probe pulse of shorter duration, the spectral bandwidth of the probe pulse might be wider than the operating window, which leads to two situations. The probe signal can be switched completely as shown in Fig. 6 (b), however, the DFB grating structure may act as a narrow-passband filter and induce distortion to the un-switched signal pulse by the corresponding straighter band edge. The opposite case is as shown in Fig. 6 (c). The signal pulse is pushed away from the band edge and can pass through the device distortionless; nevertheless, part of the pulse stay outside the switching window and consequently suffers from inefficient switching and poor extinction ratio. Hence, the extinction ratio and the bandwidth of the operating pulse which determines the AOS device switching speed can be a trade-off. In this respect, the Sech2-shape pulse is superior to Gaussian-shape pulse for ultrafast operation on a bandwidth-limited device due to its smaller minimal time-bandwidth product (TBWP). It is worth mentioning that as long as the slope of the band edge is steeper than that of the corresponding probe pulse, the required band edge shift for a complete switching remains the same as the pulse bandwidth. If the band edge of the grating WG fits exactly to the probe pulse spectrum as shown in Fig. 6 (a), the required band edge shift for switching is minimal for any extinction ratio. Or in other word, the extinction ratio is maximized with a fixed band edge shift, because the switching window overlaps the pulse spectrum in the most efficient way. Meanwhile, the steeper the band edge, the less dispersion is introduced to the pulse by the grating WG. If the continuous-wave (CW) signal light was adopted for applications as wavelength-conversion  or modulation , the narrow-band operation situation would be substantially relieved. Figure 5 (b) illustrates the schematic drawing of the lateral 1st order Bragg grating WG. The position, the transmission drop, the bandwidth of the stopband and the steepness of the band edges can be adjusted by the design parameters, including the grating periods Λ, grating length L and grating depth ΔW.
As shown in Fig. 5 (c), simple 2D finite-difference time-domain (FDTD) simulations are performed to predict the TE mode band edge shift with respect to the relative effective index change Δneff assuming a constant pumping condition. The DFB WGs are 40 μm long, 2 μm wide and with grating depth 0.8 μm. The band edge position is defined as the wavelength corresponding to the 3 dB drop to the transmittance of the passband. The dependency agrees well with the relationship of Δλ~Δneff /n eff × λ. A stronger index change indicates a larger shift of the band edges, and the DFB grating device is easier to support an ultrafast transform-limited light pulse. For instance, an index change of Δneff = 0.002 corresponds to a band edge (upper one in wavelength of the DFB grating WG) shift of Δλ = 0.94 nm, which is roughly the FWHM bandwidth of 1 ps Sech2-shape or 3.8 ps Gaussian-shape pulses at 1550 nm. As a stricter criterion, the term of pulse technique bandwidth may be applied, which is defined as the bandwidth within which 95% of pulse energy remains. The corresponding Gaussian-shape pulse durations with these technical bandwidths are plotted in Fig. 5 (c) as well. They demand a larger index change. These theoretical results imply the potentials for high-speed operation (a few hundred GHz bandwidth) of the DFB grating WG AOS if working with appropriate ultrafast pulses.
The samples were grown by molecular beam epitaxy (MBE) on double side polished (100)-oriented InP Fe-doped substrates using As2, Sb2, and P2 group V species with valved cracker sources at 400°C. Details of the growth procedure of strained InGaAs/AlAsSb layers with strain-compensating AlAs layers can be found elsewhere . The InGaAsP guiding layers and InP cladding layers are grown on both sides of the QW core region to form the slab WG.
The WG design is fabricated with our in-house fabrication technology. The DFB grating WG and tapered deep trench access WGs are patterned in a single step of e-beam lithography, then transferred to a SiNx hardmask by reactive ion etching (RIE) and finally etched into the semiconductor slabs by inductively coupled plasma RIE (ICP-RIE) with a Cl2/Ar/N2 chemistry. The SiNx hardmask is finally removed with a SF6/O2 plasma to avoid damaging of the QW layers by wet etchants. Side and top view scanning electron microscope (SEM) images of the fabricated samples are shown in Fig. 7 . The depth of deeply etched WG is more than 3 μm. The access taper is 150 μm long, and the width is tapered down from 5 μm to that of the DFB WG for an efficient coupling. The DFB WG is placed just after the short access taper in order to reduce the undesired parasitic pump power absorption by the absorbing access structures. The overall sample length is 1.3 mm. It is notably that the active/passive section integration is not really compulsory to perform for such switch design as long as the DFB grating structures are placed to the front input side of the chip. Also as another merit, the access WGs after the DFB grating WGs can be used as the integrated on-chip TE-pass polarizers to filter out the TM pump light. The TE probe signal light may transmit through the rest of the sample transparently once the residue pump light is absorbed completely.
For the time-resolved pump-probe measurement, a mode locked fiber laser (Pritel) with 5-6 ps (depending on the pumping condition) pulses at a repetition rate of 20 MHz is employed. Light from one of the two polarization maintaining (PM) output ports is set to TM polarization, amplified by a PM Erbium doped fiber amplifier (EDFA) and used as the pump pulse. Light of the other laser output port is attenuated and used as much weaker probe pulse with a constant energy. A variable time delay is inserted into the probe pulse to adjust the relative arriving time related to light from the other branch. In addition, the polarization of probe light is rotated by 90 degree with a half-wave plate to TE polarization and purified with a sheet polarizer (PER~30 dB). The orthogonal pump and probe pulses are then recombined together with a compact PM fiber coupler, and butt-coupled to the chip facet through a PM lensed fiber. The output light from the sample is collected by a microscope objective, filtered with a TE-pass Glan-Thomson polarizing prism (PER>30 dB) to eliminate the possible residue TM pump light, and finally detected by a power meter. All fibers in the setup, especially for transmitting amplified high intensity pulse, are kept as short as possible to avoid excessive broadening and nonlinear effects to the optical pulses. The setup is shown schematically in Fig. 8 . For the static characterization using the endfire transmission technique , the signal light is replaced by TE-polarized CW light with tunable wavelength within the range 1470–1630 nm (Agilent 81940A and 81980A). The polarization extinction ratio (PER) of the whole measurement setup is above 20 dB.
4. Results and discussion
4.1 Static state characterization
The measured transmission spectra of CW TE signal light without pumping are shown in Fig. 9 (a) and (b) . The oscillations shown in the raw transmission spectra with periodicity of 0.8 nm are attributed to the Fabry-Pérot interference between DFB grating WGs and access WGs. The transmission contrast between the stopband and passband is more than 20 dB for the DFB structure of width W = 2 μm, length L = 400 μm and grating depth ΔW = 0.8 μm. Shorter length and small grating depth may reduce the contrast. In particular, the band edges locating in longer wavelength (dielectric band edges) of the DFB structures exhibit sharper changes in transmission than those of the band edges in shorter wavelength (air band edges). The reason for the degradation of air band edge steepness may lie in that these band edges are more sensitive to the slightly sloped grating sidewalls (>85°) and other fabrication inaccuracies. The stopband positions for TM-polarized light are difficult to measure due to the ISB absorption. However, according to the theoretical calculation, the spacing between stopbands of different polarized light is roughly 5 nm, which is wide enough to ensure that TM-polarized light pulse is not guided on its stopband. As shown in Fig. 9 (c), with the presence of pulsed pump light, the transmission curves experience red shift as predicted. It is known that the two photon absorption (TPA) induces negative index change ; on the contrary, both the linear interband dispersion and plasmas dispersion effects generate positive index change in this CDQW material system. Hence, the air band edge can be used for switch-on operation of the AOS while dielectric band edge is for switch-off operation. The interband absorption of the signal light is enhanced near the DFB grating WG band edge due to the mild slow-light effect, and increasing with pumping power. It is worth mentioning that the purpose of the experiment as shown in Fig. 9 (c) is to demonstrate the band edge shift can be induced by pumping rather than evaluating the device performance because the pumping by the low repetition rate mode-locked laser light is pretty inefficient although its peak power is sufficiently high. As a fair estimation, the signal transmittances extracted by dynamic pump-probe studies (described in section 4.2) of pulsed probe light with two central wavelengths λ = 1552.1 nm and 1552.2 nm are plotted together in Fig. 9 (c) (solid symbols; Different colors correspond to the power intensities). The shift becomes much more distinct. However, it is still very difficult to extract the precise relationship of the effective index change induced band edge shift and the associated input pumping light intensity because of the nonlinear absorption of the TM pump light along the complex periodic DFB grating structures. A comprehensive study of this relationship is beyond the scope of the present paper and left for future work. The incoupling losses at the chip facet are estimated to be 6 dB measured from the reference WGs. The pure linear insertion losses for TE modes through the 400 μm long DFB WG is 6.8 dB (with wavelength above the stopband), which is extracted from cut-back analysis. In addition to some intrinsic losses of the DFB grating WGs (e.g. in-plane and out-of-plane scattering losses), excess losses can be attributed to the fabrication imperfections, e.g. the WG sidewall surface roughness. These non-idealities can be improved by technology optimization. The coupling losses from the access to the DFB WG are 6 dB in total, which may be also further reduced substantially by an optimized tapering design at the interfaces . The switching energy is defined as the injected TM pump light energy deducting the linear access losses to the DFB WG.
4.2 Dynamic switching characterization
As discussed previously, although the used CDQW system generally supports XPM relaxation around a few picoseconds, the periodic DFB grating design may still impose inherent limitation on the switching speed. Therefore, firstly, the DFB grating WG has to be carefully designed for a suitable band edge slope; secondly, both the central wavelength and the bandwidth of the signal pulse should be cautiously chosen for an ultrahigh-speed switching operation. In our measurement of the structures shown in Fig. 9 (c), the central wavelength of the signal pulse is 1552.1 nm with a bandwidth around 0.7 nm. The short signal pulse fits the transmission resonance, which ensures a transmission with minimal distortion and losses . As shown in Fig. 9 (d), the FTIR characterization proves that the TE signal light may transmit through the DFB grating WG almost intact. Figure 10 (a) and (b) show the experimental switching dynamics of the DFB AOS. Under two-pulse operation, the walk-off between the pump and probe pulses can be tuned to maximize the modulation effect . For switch-off operation, the switching energy is about 5.5 pJ for about a 3 dB on-off ratio and 11.6 pJ for 5 dB. The monoexponential 1/e decay and on-off response times for 3 dB ratio are 4.55 ps and 9 ps, respectively. The ultrafast response indicates a switching bandwidth beyond 100 GHz. This energy consumption is comparable to the ~4-6 pJ for 3 dB modulation of the XAM-type AOS with ISBTs in InGaAs/AlAsSb CDQWs . Although the present switch design incorporating XPM effect does not yield reduced switching energy at the moment, its potentials and especially its specific advantages, like single color operation, simple control/signal pulse discrimination by polarization, and low insertion loss, still make it very attractive as competing technology for practical application. Comparing to the off-chip MZI-based AOS configuration, the implementation to convert XPM to XAM of the DFB grating design is much easier . It is not required to set additional initial phase bias (optical delay between two branches of the MZI) to generate the necessary phase shift . As shown in Fig. 10 (a), the DFB grating WG AOS design is possible to achieve even higher on-off ratio, however, it is observed that the signal transmission cannot return to the ground level owing to the ISBT-induced nonlinear interband absorption in addition to the linear interband absorption . This nonlinear interband absorption appears obviously with pump light intensity above a certain threshold value, and scales proportionally to the square of the TM-polarized pump light peak intensity. Furthermore, the relaxation time for the nonlinear IBT is in the order of a few hundred picoseconds . This slow process also gives rise to biexponential relaxation, and degrades the operating speed of the AOS device due to the interference among signal channels. To demonstrate the switch-on operation at the air band edge, the corresponding central wavelength of the light is adjusted to 1549.2 nm. As shown in Fig. 10 (b), the signal power first drops due to the interband absorption, and then increases because the index change effect in the DFB grating structure plays a dominant role. The off-on ratio is bounded to 1.5 dB. This smaller ratio is caused by two effects. One is the smaller transmission change on the flatter air band edges. Higher contrast of signal transmittance can be obtained by optimizing the DFB grating structure design as well as improving the fabrication quality. The other reason is the instantaneous interband absorption counteracts the power enhancement due to the index change effect. This impact could be weakened by proper QW design to enhance solely the carrier plasma effect  and by reducing the required pumping power because the nonlinear interband absorption is found to be increased with pumping intensity . It is worthwhile reminding that it is also possible to carry out the DFB grating WG AOS for switch-on operation under the reflection mode, which theoretically provides the same performance as the switch-off operation under the transmission mode. In the aspect of switching mechanism, it is foreseeable that there is no clearly unsolvable fundamental issue to restrain the switch performance by the device design.
4.3 Numerical analysis based on one-dimensional coupled-mode theory
According to the coupled mode theory (CMT), the transmission at the band edge follows the equations :Figure 11 (a) shows both the simulated and experimental transmission spectra with a reasonable good agreement except some slight mismatch on the passband. The transmission fringes, which appear usually in the transmission spectra of the DFB grating WGs, have been eliminated substantially by the passive propagation losses. The light intensity transmission coefficient T is given as T = t × t*. Finally, the pulse energy is obtained by integrating the intensities in frequency domain according to the Parserval principle. The extinction ratio of the switch η is computed as the radio for the output pulse energies with (Iout_w/) and without pumping (Iout_w/o):
Figure 11 (b) plots the dependence of switch extinction ratios on the effective index changes with signal/probe light at λ = 1552.1 nm. Effective index changes of 6.5 × 10−4 and 1 × 10−3 are necessary to generate 3 dB and 5 dB ratios, respectively. With the computed mode confinement in the active layers of 17% and 6 ps pulse excitation, the CDQWs material modeling predicts theoretically the switching energies of 2.54 and 4.96 pJ to generate the required effective index changes. The higher switching energies obtained from experiments can be explained by two reasons: 1) the pumping-assisted nonlinear absorption of the signal light. This effect reduces the index changes by the carrier plasma dispersion effect . Moreover, it also introduces simultaneous negative index change contrary to the other two XPM mechanisms. Hence, higher pump energies are required to generate more index changes so as to compensate this effect; 2) as the other reason, because the pumping light is attenatuated nonlinearly propagating through the device, the light power may not produce index change as high as possible over the whole length. A much more complex distributed model is required in the future to precisely analyze the device performance. Figure 11 (c) presents the switch extinction ratios as a function of the DBR grating WG length with a fixed index change of 1.0 × 103. The results show that longer WGs provide better extinction ratios, however, such improvement is getting weakened with WG length. The reason lies in the higher passive propagation losses. Therefore, an optimum length should be chosen based on systematic consideration of both the insertion losses and switching ratios. Also, further developed fabrication technology may reduce the undesired excess losses of both pump and probe light, and thereby improve the switch performance in terms of the switching energy and insertion loss. Figure 11 (b) and (d) shows the switch performance for different signal light wavelengths. The extinction ratio and the insertion losses are the trade-off for probing signal light with different wavelengths. The results agree with the prediction from the operation principle of DFB grating WG switch discussed previously.
In this article, the DFB AOS design based on cross modulation in InGaAs/AlAs/AlAsSb CDQWs exhibiting ISBTs has been discussed. The pump-induced phase modulation to the signal light is converted to intensity modulation efficiently on chip with the help of the DFB grating structures. The presented AOS can be operated in both on-off and off-on regimes depending on the band edges. Switching energy for an on-off 3 dB modulation of 5.5 pJ is achieved. The effective 1/e recovery time is 4.55 ps. Nevertheless, limited by the accessible test equipments we could not test our devices with pulses of shorter duration (~3-4 ps) and smaller minimal TBWP (e.g., Sech2-shape) for higher speed demonstration. Further improvements on the AOS performance could be achieved by modified CDQW design with InAlAs or AlGaAs coupling barrier. The XPM efficiency is enhanced substantially as reported in . In addition, through adopting membrane-type DFB grating-based design, the cross phase-intensity modulation conversion can be enhanced by the strong-index-contrast membrane system and the high mode confinement in the active layer. Subsequently, the ISBT-induced interband absorption of the signal, which constrains the performance of the DFB AOS design, will be also highly depressed due to the reduced switching energy .
This work is supported by the ETH research grant TH-19/04-3.
References and links
1. H. Ishikawa, “Ultrafast all-optical signal processing devices”, chapter 5, ISBN 978–0-470–51820–5, Wiley (2008).
2. H. Ishikawa, H. Tsuchida, K. S. Abedin, T. Simoyama, T. Mozume, M. Nagase, R. Akimoto, T. Miyazaki, and T. Hasama, “Ultrafast all-optical refractive index modulation in intersubband transition switch using InGaAs/AlAs/ AlAsSb quantum wells,” Jpn. J. Appl. Phys. 46(8), L157–L160 (2007). [CrossRef]
3. C. G. Lim, “Plasma dispersion effect in heavily doped antimony-based passive optical waveguides,” Appl. Phys. Lett. 92(20), 203508 (2008). [CrossRef]
4. C. G. Lim, “Characteristics of the ultrafast all-optical cross-phase modulation in InGaAs/AlAs/AlAsSb coupled double-quantum-well optical waveguides,” J. Appl. Phys. 107(10), 103109 (2010). [CrossRef]
5. G. W. Cong, R. Akimoto, M. Nagase, T. Mozume, T. Hasama, and H. Ishikawa, “Mechanism of ultrafast modulation of the refraction index in photoexcited InxGa1−xAs/AlAsySb1−y quantum well waveguides,” Phys. Rev. B 78(7), 075308 (2008). [CrossRef]
6. G. W. Cong, R. Akimoto, K. Akita, S. Gozu, T. Mozume, T. Hasama, and H. Ishikawa, “Experimental and theoretical study of cross-phase modulation in InGaAs/AlAsSb coupled double quantum wells with a AlGaAs coupling barrier,” Phys. Rev. B 80(3), 035306 (2009). [CrossRef]
7. H. Tsuchida, T. Simoyama, H. Ishikawa, T. Mozume, M. Nagase, and J.- Kasai, “Cross-phase-modulation-based wavelength conversion using intersubband transition in InGaAs/AlAs/AlAsSb coupled quantum wells,” Opt. Lett. 32(7), 751–753 (2007). [CrossRef]
8. R. Akimoto, T. Simoyama, H. Tsuchida, S. Namiki, C. G. Lim, M. Nagase, T. Mozume, T. Hasama, and H. Ishikawa, “All-optical demultiplexing of 160 - 10 Gbit/s signals with Mach-Zehnder interferometric switch utilizing intersubband transition in InGaAs/AlAs/AlAsSb quantum well,” Appl. Phys. Lett. 91(22), 221115 (2007). [CrossRef]
9. K. S. Abedin, G. W. Lu, T. Miyazaki, R. Akimoto, and H. Ishikawa, “High-speed all-optical modulation using an InGaAs/AlAsSb quantum well waveguide,” Opt. Express 16(13), 9684–9690 (2008). [CrossRef]
10. S.-H. Jeong, H.-C. Kim, T. Mizumoto, J. Wiedmann, S. Arai, M. Takenaka, and Y. Nakano, “Polarization-independent all-optical switching in a nonlinear GaInAsP-InP high mesa waveguide with vertically etched Bragg reflector,” IEEE J. Quantum Electron. 38(7), 706–715 (2002). [CrossRef]
11. T. Mizumoto, Y. Akano, K. Tamura, and M. Yoshimura, “DFB waveguide all-optical switching devices employing pump-induced refractive index change in GaInAsP,” Adv. Mater. Res. 31, 206–208 (2008). [CrossRef]
12. T. Mizumoto, J.-K. Seo, and N. Tanaka, “All-optical inverting response in a GaInAsP/InP DFB waveguide,” in Proceeding of the 18th Annual Meeting of the IEEE Lasers and Electro-Optics Society. (Institute of Electrical and Electronics Engineers, New York, 2005), pp. 148–149.
13. M. Nagase, R. Akimoto, K. Akita, H. Kawashima, T. Mozume, T. Hasama, and H. Ishikawa, “Fabrication of All-Optical Switch Based on Intersubband Transition in InGaAs/AlAsSb Quantum Wells with DFB Structure,” in Proceedings of 20th IEEE International Conference on Indium Phosphide and Related Material. (Institute of Electrical and Electronics Engineers, New York, 2008), pp. 1092–8669.
14. Y. Fedoryshyn, M. Beck, P. Kaspar, and H. Jaeckel, “Characterization of Si volume- and delta-doped InGaAs grown by molecular beam epitaxy,” J. Appl. Phys. 107(9), 093710 (2010). [CrossRef]
15. Quantum well simulations were performed with the use of the software provided by Professor J. Faist from ETH Zürich, Switzerland.
16. P. Ma, Y. Fedoryshyn, and H. Jäckel, “Detailed analysis of all-optical switches based on intersubband cross modulation in InGaAs/AlAs/AlAsSb double coupled quantum wells,” to be published.
17. T. Simoyama, S. Sekiguchi, H. Yoshida, J.- Kasai, T. Mozume, and H. Ishikawa, “H. Ishikawa, “Absorption dynamics in all-optical switch based on intersubband transition in InGaAs–AlAs–AlAsSb coupled quantum wells,” IEEE Photon. Technol. Lett. 19(8), 604–606 (2007). [CrossRef]
18. S. L. Chuang and D. Ahn, “Optical transitions in a parabolic quantum well with an applied electric field-analytical solutions,” J. Appl. Phys. 65(7), 2822 (1989). [CrossRef]
19. J. Kasai, T. Mozume, H. Yoshida, T. Simoyama, A. V. Gopal, and H. Ishikawa, “Optical quality improvement of InGaAs/AlAs/AlAsSb coupled double quantum wells grown by molecular beam epitaxy,” Phys. Status Solidi C 1(2), 368–371 (2004). [CrossRef]
20. Y. Fedoryshyn, P. Strasser, P. Ma, F. Robin, and H. Jäckel, “Optical waveguide structure for an all-optical switch based on intersubband transitions in InGaAs/AlAsSb quantum wells,” Opt. Lett. 32(18), 2680–2682 (2007). [CrossRef]
21. P. Cristea, Y. Fedoryshyn, and H. Jäckel, “Growth of AlAsSb/InGaAs MBE-layers for all-optical switches,” J. Cryst. Growth 278(1-4), 544–547 (2005). [CrossRef]
22. P. Ma, P. Kaspar, Y. Fedoryshyn, P. Strasser, and H. Jäckel, “InP-based planar photonic crystal waveguide in honeycomb lattice geometry for TM-polarized light,” Opt. Lett. 34(10), 1558–1560 (2009). [CrossRef]
24. M. Scalora, R. J. Flynn, S. B. Reinhardt, R. L. Fork, M. J. Bloemer, M. D. Tocci, C. M. Bowden, H. S. Ledbetter, J. M. Bendickson, J. P. Dowling, and R. P. Leavitt, “Ultrashort pulse propagation at the photonic band edge: Large tunable group delay with minimal distortion and loss,” Phys. Rev. E Stat. Phys. Plasmas Fluids Relat. Interdiscip. Topics 54(2), R1078–R1081 (1996). [CrossRef]
25. C. G. Lim, “Effects of two-photon absorption on pump-induced refractive-index change in AlAsSb–InGaAs–AlAs optical waveguides,” IEEE J. Quantum Electron. 45(5), 523–530 (2009). [CrossRef]
26. M. Davanço, A. M. Xing, J. Raring, E. Hu, and D. Blumenthal, “Detailed characterization of slow and dispersive propagation near a mini-stop-band of an InP photonic crystal waveguide,” Opt. Express 13(13), 4931–4938 (2005). [CrossRef]