We report on the efficient nonlinear optical interactions in AlGaAs strip-loaded waveguides with a wafer composition specifically designed to increase the nonlinear coefficient. We demonstrate a broad-band self-phase modulation with a nonlinear phase shift up to 6π, and four-wave mixing with a 20-nm tuning range and signal-to-idler conversion efficiency up to 10 dB. Our samples are several times shorter than similar devices used for wavelength conversion by XPM and FWM in previous reports, but the efficiency of the observed effects is similar. Our experimental studies demonstrate the high potential of AlGaAs for all-optical networks.
©2011 Optical Society of America
The Internet bandwidth consumption grows very rapidly with the increasing popularity of different web-based resources such as Facebook and Youtube. The conventional way of processing the optically-transmitted signals by electronic devices cannot keep up with the growing demand and is close to hitting its fundamental limits. A fundamentally different approach to signal processing is needed, and all-optical processing is a promising solution to this problem. However, all-optical networks are still at the early stage of their development. The slow progress in the field is due to the fact that the variety of functions necessary to perform all-optical signal processing cannot be handled by a single material platform at present, as there is no optimal material choice made yet. Narrowing down the choice of technologies in integrated optics by identifying more suitable materials to perform a large variety of all-optical functions will help us speed up the progress in this field.
Silicon has a number of advantages compared to other candidates for all-optical networks [1–7]. It has a very mature low-cost fabrication technology and can be used for combining the optical and electronic devices on the same chip. However, silicon is an indirect-bandgap semiconductor, which means that it is difficult to produce an electrically pumped laser in silicon. A hybrid integration with other materials such as III–V semiconductors is needed, which makes the integrated optical circuits complex and expensive, in addition to creating compatibility problems. A new and very promising candidate for all-optical devices is chalcogenide glasses [8–14]. Out of the varieties of this material system, As2S3 stands out due to its very low linear and nonlinear propagation losses. However, its refractive index cannot be tailored over a large range, which does not allow much flexibility in tailoring all-optical networks.
AlGaAs semiconductor has been termed “the silicon of nonlinear optics”  due to its excellent nonlinear performance [15–19]. In addition to the highest Kerr nonlinearity among the candidates for all-optical signal processing, it also has a high refractive index allowing for a tighter mode confinement for even more efficient nonlinear interactions. Also, the composition of AlGaAs can be adjusted by changing the Al concentration during epitaxial growth. It allows one to vary the refractive index within a broad range of values and provides a lot of flexibility in tailoring the integrated optical devices. AlGaAs is a direct bandgap material and allows one to combine an integrated laser source, low-loss waveguides, and a detector on the same chip without resorting to hybrid integration.
In this paper, we demonstrate the nonlinear performance of AlGaAs strip-loaded waveguides with a specially designed composition suitable for signal regeneration by self-phase modulation (SPM), and efficient wavelength conversion by cross-phase modulation (XPM) and four-wave mixing (FWM). There have been many recent reports of efficient SPM [8, 9, 16, 18, 19], XPM [10,18,20], and FWM [4–7,11–13,21] in different material platforms. SPM with the nonlinear phase shift of 3π/2 , wavelength conversion via XPM , and FWM with the tuning range of 10 nm  have been demonstrated in 5-cm-long non-dispersion-engineered As2S3 chalcogenide glass waveguides. Supercontinuum generation  and broad-band FWM  in dispersion-engineered chalcogenide waveguides with a small effective mode area has also been reported. Broadband FWM [4–6] and demultiplexing of 160 Gb/s signals  has been reported in sub-micron silicon waveguides operated close to zero-dispersion point. Efficient broadband FWM has been demonstrated in high-index doped silica dispersion-engineered waveguides of 45-cm length . Wavelength conversion of 10 Gb/s signal by XPM has been reported in 4.5-mm-long GaAs-AlGaAs waveguides . The above reports reflect a significant improvement in search for an optimal material platform for all-optical networks. However, no optimal choice has been made yet. We believe that our proof-of-principle study of the nonlinear optical performance of AlGaAs reported herein will support our claim that AlGaAs provides a more flexible and reliable solution for all-optical signal processing. We have already demonstrated very promising preliminary results obtained in a 9-mm-long AlGaAs waveguide . This paper is a detailed follow-up study reporting on optimized AlGaAs devices with enhanced nonlinear interactions. The experimental results reported herein display larger spectral broadening in SPM and broader tuning range in FWM compared to the data reported in . In addition, here we demonstrate generation of idler component by FWM both in the longer and shorter wavelength ranges with respect to the pump and signal.
In Section 2, we present a comparison between different material platforms used in integrated optics, including our AlGaAs material with the specially designed composition. In Section 3, we overview design considerations for our AlGaAs composition that we used in the current study. In Section 4, we describe the fabrication procedure for our strip-loaded waveguides. We present the results of the experimental studies in Section 5. Section 6 contains the summary of our studies and conclusions.
2. Materials for wavelength conversion and all-optical networks
Wavelength conversion is an efficient technique for optical signal regeneration and demultiplexing. It can be implemented by cross-phase modulation (XPM) or four-wave mixing (FWM), which are the nonlinear optical processes relying on the intensity-dependent refractive index n 2. That is why highly nonlinear materials with the large and fast Kerr nonlinearity are of special interest. Together with a high value of n 2, a nonlinear material should also exhibit low linear and nonlinear propagation losses in the telecom spectral range (1400 - 1600 nm). This includes the material absorption and nonlinear absorption, which are critical at the high intensities required. The performance of the nonlinear material can be evaluated in terms of the figure of merit (FOM) T , defined as
In Table 1, we summarize the values of the linear refractive index n 0, nonlinear refractive index n 2, TPA coefficient α 2, and the parameter T from Eq. (1) for different materials used in integrated optics, including Al0.18Ga0.82As used as the guiding layer in our devices. Silicon has a large refractive index, large Kerr nonlinearity and negligible linear losses in the telecom range. However, its TPA coefficient is larger than those of As2S3 chalcogenide glass and Al0.18Ga0.82As resulting in T > 1. As2S3 chalcogenide glass displays a very good nonlinear performance, confirmed by numerous experimental data [8–13]. GaAs semiconductor used as a guiding layer in  has a very large Kerr nonlinearity and very low linear absorption which allowed the authors to obtain a significant XPM in a 4.5-mm-long sample. However, the TPA coefficient of this material is very high, which results in an unacceptably high value of T. It is possible to make use of the high Kerr nonlinearity of GaAs by shifting its band gap below 750 nm to eliminate the nonlinear absorption in the telecom range. It is achievable, for example, by introducing some aluminum during the epitaxial growth . Thus, Al0.18Ga0.82As with the shifted band gap has the potential for being used in all-optical networks due to its large Kerr coefficient, small TPA coefficient, flexibility in tailoring waveguides of different geometries and higher mode confinement that can be achieved in this material due to its large refractive index. The value of the refractive index of AlGaAs can be adjusted in the range between 2.90 and 3.38 by changing the Al concentration from 100% to 0, which offers one a lot of freedom in designing various AlGaAs integrated optical components. Neither silicon nor chalcogenide glass have this advantage. In fact, an efficient SPM [16,18,19] and XPM  have already been demonstrated in AlGaAs ridge waveguides  and nanowires  despite the high propagation loss in the latter case. This motivated us to take a step further in optimizing the performance of AlGaAs, making use of its flexibility and tailorability.
3. Design of AlGaAs wafer composition
The goal of the present study is to demonstrate the potential of AlGaAs for wavelength conversion. Due to limitations of the fabrication facilities available to us, we narrowed down our present scope to testing non-dispersion-engineered micron-size AlGaAs waveguides which are much easier to fabricate. At the same time, we were aiming at producing the maximum light confinement in the waveguides to minimize the effective mode area, A eff, given by17]. By minimizing the effective mode area one can maximize the nonlinear coefficient γ, given by
Our AlGaAs wafer composition is outlined in Fig. 1, together with the schematic of the strip-loaded waveguide geometry used in our experiments. We designed the wafer composition according to the following considerations.
- The guiding layer was chosen to have 18% of aluminum, because the previous studies [15, 17] have shown that, at this Al concentration, both the two-photon absorption (2PA) and three-photon absorption (3PA) are minimized at 1550 nm, as the spectral window between the 2PA and 3PA peaks overlaps with the telecom wavelength range. The refractive index associated with the Al0.18Ga0.82As composition is 3.2813 at 1550 nm. The thickness of the guiding layer was set to 1 μm to minimize the effective mode area.
- We choose the composition of the lower cladding to be Al0.5Ga0.5As to maximize the refractive index contrast with the core layer. The refractive index of the lower cladding at 1550 nm is 3.1218, resulting in an index contrast of Δn ≈ 0.16 with the guiding layer. Such a high index contrast allowed us to push the mode up to ensure that there is no loss due to the leakage to the higher-index GaAs substrate. At the same time, it allowed us to achieve a more circular mode shape to minimize the mode shape mismatch when coupling the focused Gaussian beam to the waveguide. The thickness of the lower cladding was designed to be 4 μm to isolate the guiding layer from the GaAs substrate to minimize the leakage loss.
- We chose the upper cladding to be Al0.24Ga0.76As, which yields a moderate index contrast Δn ≈ 0.03 with the guiding layer at 1550 nm. A higher index contrast could have resulted in a multimode operation in our waveguides. The thickness of the upper cladding was designed to be 1.5 μm. This provided us with some flexibility in choosing the etching depth during the waveguide fabrication.
We performed the design of the wafer composition using the commercial mode solver Lumerical MODE Solutions. While trying to minimize the effective mode area, we were aiming at obtaining a single-mode operation at the fundamental TE and TM modes. In our strip-loaded AlGaAs waveguides the fundamental TE and TM modes experience a slightly different confinement. Generally, the fundamental TM mode is more confined than the TE mode. In Fig. 2, we show the intensity distributions of the two fundamental modes for a 2-μm-wide waveguide with the ridge height h = 1.2μm. One can see from the simulation results that the TM mode has a more circular shape and is more pushed upwards compared to the TE mode which is more expanded horizontally. As a result, the coupling into the TM mode is expected to be more efficient due to a smaller mode shape mismatch with respect to a focused laser beam. The calculated values of the effective mode area are 4.91μm2 and and 4.35μm2 for the TE and TM modes, respectively. These values of A eff are smaller than those in previously designed AlGaAs strip-loaded waveguides having different wafer compositions. The AlGaAs composition containing 25% of aluminum in the claddings and 18% of aluminum in the 1.5-μm-high guiding layer typically results in A eff ≈ 12μm2 for the fundamental mode . An optimized wafer composition having 30% and 40% of aluminum in the upper and lower claddings, respectively, and a 1-μm-high core with 18% of aluminum has been considered in , resulting in A eff ≈ 6μm2. Since our A eff is even smaller than that reported in , we expect a more efficient nonlinear interaction in our AlGaAs waveguides with the improved wafer composition.
The AlGaAs wafer with the designed composition has been grown by a metalorganic chemical vapor deposition (MOCVD) technique at the Canadian Photonics Fabrication Centre (CPFC). A 200-nm-thick layer of SiO2 was deposited on the surface of the wafer after the growth to serve as a hard mask for the AlGaAs etching. The grown wafer had a slightly different composition compared to the targeted one: the guiding layer contained 20% of aluminum. We have revisited our effective mode area calculations based on the modified wafer composition. In Table 2, we show a summary of the values of the effective mode area for the fundamental TM modes for the strip-loaded waveguides with the as-grown wafer composition for different values of the ridge height h and width w. The shaded area on the graph denotes the waveguide dimension space for which a single-mode behavior is expected, according to the simulations. The range of the waveguide widths is limited by the mask pattern, containing the waveguides with the widths between 1.5 and 5 μm with the half-a-micron increment step. The heights of the ridge correspond to different AlGaAs etching depths that we can adjust in the process of fabrication. It can be seen from the table that higher ridges (or deeper etching) result in stronger mode confinement, which is good for the efficiency of the nonlinear interactions. At the same time, the probability of ending up with a multimode waveguide also increases, as in practice there are always deviations from the design. In this sense, there is a trade off between the small effective mode area and the reliable single-mode operation. Based on the information given in Table 2, we decided to work with the etching depths not larger than 1.2 μm to ensure a single-mode operation in our devices.
4. AlGaAs waveguide fabrication
We have prepared four AlGaAs waveguide samples using slightly different fabrication procedures. For two of the four samples, that we name Sample 1 and Sample 2, we patterned the waveguides in AlGaAs by a standard photolithography procedure using S1818 commercial positive photoresist. We spin-coated the photoresist on top of the 200-nm-thick SiO2 layer covering the AlGaAs wafer. After exposing the surface of the photoresist with a UV light through a photomask with the waveguide pattern, we developed the samples and etched the SiO2 layer using a reactive ion etching (RIE) with inductively coupled plasma (ICP). After that, the samples were treated in buffered oxide for 30 sec to smoothen the edges of the SiO2 pattern after the dry etching. The photoresist was stripped off, and the resulting SiO2 pattern served as the etching mask for patterning the waveguides in AlGaAs. The AlGaAs etching was performed using a Trion Minilock etcher via RIE with BCl3 and chlorine and ICP. Sample 1 and Sample 2 were etched for 45 and 50 sec to the depths 1.15 and 1.22 μm, respectively. Different etch depths resulted in different effective mode areas in the waveguides of the samples (see Table 2).
Another sample, that we name Sample 3, was prepared by photolithography, but without the SiO2 hard mask. We stripped off the SiO2 layer by treating a piece of the wafer in buffered oxide for 5 min prior to the photolithography. Then the photolithography was performed in a similar way with the S1818 photoresist. We used the resulting photoresist pattern as a mask for the AlGaAs etching. We etched Sample 3 for 40 sec, which resulted in a 1-μm etch depth.
For the three above samples we used a photomask consisting of nine series of waveguides with different widths ranging from 1.5 to 5 μm with the 0.5-μm increment step between the adjacent series. Each series contained 5 identical waveguides. Since the limit of the photolithography resolution for the equipment and photoresist we used is 1 μm, many 1.5-μm and some 2-μm waveguides in our samples turned out to be damaged and could not be optically characterized. That is why, we prepared one more sample that we name Sample 4, using the electron beam lithography to produce the finer waveguides without any damage. We used the negative HSQ electron beam resist. The SiO2 layer was preliminary stripped off with the buffered oxide. The patterned HSQ served as the etching mask for the AlGaAs. The same etching procedure was used with the resulting etch depth 1.15 μm. Sample 4 contains the waveguides with the widths ranging between 1 and 3.5 μm.
In Table 3, we summarize the parameters of the four samples that we prepared for the nonlinear optical characterization. The table gives the sample processing description, together with the etching depth (or the waveguide ridge height h) and the length of the sample L. In Fig. 3, we display the scanning electron microscope (SEM) images of the waveguides in the samples. It can be seen from Figs. 3(a) and 3(b) that the quality of Sample 1 and Sample 2 is satisfactory, except for some sidewall roughness and a slight deviation of the waveguide shapes from the rectangular shape. Sample 4 [Fig. 3(d)] also displayed some sidewall roughness and a significant distortion of the waveguide shape. However, Sample 3 turned out to have a perfect rectangular waveguide shape and very smooth sidewalls [see Fig. 3(c)]. According to our multiple etching tests with the use of different etching masks, the difference in the etching quality between the waveguide samples can be fully attributed to the conditions of the etching chamber that can vary significantly day-to-day.
5. Experimental results
5.1. Loss measurement
We measured the propagation loss in our waveguides using the Fabry-Perot method . For this measurement we used a JDS Uniphase cw laser, tunable in the range between 1520 and 1570 nm with the maximum output power 5 mW. The reflectivity at the interfaces between AlGaAs and air was estimated to be 30%. In all our experiments we used end-fire coupling of light into the waveguides with microscope objectives. Using a high-quality 40× diode objective lens, we achieved the coupling efficiency η > 0.5 for some of our waveguides, estimated from the coupling loss (in dB)Table 4, we summarize the losses, coupling efficiency, effective mode area and corresponding values of the nonlinear coefficient γ, obtained from Eq. (3), for the fundamental TM mode for the waveguides used in the nonlinear optical characterization. We found that the waveguides with the widths around 2 μm exhibit the strongest nonlinear interactions. The smaller waveguides displayed much higher propagation losses, and the larger waveguides had larger effective mode area resulting in less efficient nonlinear interactions. Most of the waveguides that we used for the nonlinear optical characterization had the propagation losses higher that 3 dB/cm, except for Sample 3 with exceptionally good waveguide profile and smooth sidewalls, exhibiting the propagation loss only 2 dB/cm. With Sample 4, we were limited to testing a 3-μm-wide waveguide, as the propagation loss for the smaller-width waveguides was too high. All the waveguides in our samples were single-mode, except for the ones with the widths larger than 4 μm in Sample 1 and Sample 2 that have larger etch depths and, consequently, higher mode confinement. Even though the propagation losses for the TE mode were lower in all waveguides, the nonlinear interactions were observed to be not as strong as those for the TM mode. We attribute it to the fact that the effective mode area for the fundamental TE mode is larger than that for the TM mode, and the coupling efficiency to the TE mode is lower due to the more elliptical mode shape and, consecutively, larger mode shape mismatch with the focused laser beam. We thus present the results of the nonlinear optical characterization for the fundamental TM mode only.
5.2. Experimental setup
The schematic of the experimental setup is shown in Fig. 4. The nonlinear characterization was performed using a Coherent Mira laser system consisting of a tunable Ti:sapphire laser, generating 2-ps FWHM Gaussian-like pulses, and an optical parametric oscillator (OPO) used for the frequency conversion of 800-nm Ti:sapphire pulses to the telecom range. We tuned the OPO output in the range between 1500 and 1600 nm by changing the wavelength of the Ti:sapphire laser. The average power of the OPO was 300 mW. The repetition rate of the laser was 35.6 MHz. For XPM and FWM experiment we also used a tunable cw laser as a probe. As the duty cycle of our pump laser was low, we used an erbium-doped fiber amplifier (EDFA) to amplify the probe. The cw and pulsed laser beams were combined in a non-polarizing 50% beam splitter cube (50% NPBS in Fig. 4). The polarizations of the pump and probe were adjusted independently by the half-wave plates (HWP) and polarizing beam splitter cubes (PBS) installed in each beam’s arm. A telescope formed by the lenses L1 and L2 with the focal lengths 10 cm and 5 cm, respectively, was used to reduce the OPO beam diameter for a better matching with the beam diameter of the cw probe and a better coupling into the sample. We used a piezo-driven Elliot coupling unit consisting of 3D position stages for the sample and the microscope objectives. We coupled the pump and probe into the sample using 40× diode objective lens, and collected the output from the waveguide with a 20× objective, sending it to an IR camera, a spectrum analyzer, or a power meter. Using flip mirrors (FM), we switched between the three devices.
5.3. Self-phase modulation
In Fig. 5, we present the best self-phase modulation data obtained in three out of four samples. Sample 2 had the lowest coupling efficiency and the shortest length. As a result, the maximum power coupled into the waveguide was much lower, and the SPM was not as significant as that for the rest of the samples. We thus do not present the SPM data for this sample.
In the legend in Fig. 5, we show the values of the in-waveguide peak powers and the corresponding nonlinear phase shifts, ϕNL, estimated from the output spectra . The maximum ϕNL, obtained in Sample 1, Sample 3, and Sample 4 for the incident OPO signal at 1550 nm were 4.5π, 5π, and 5.5π, respectively [see Fig. 5(a)–5(c)]. The factors influencing the nonlinear performance of a waveguide are the sample length, the propagation loss, the maximum peak power, and the effective mode area that affects the value of the nonlinear coefficient. It is convenient to compare the nonlinear performances of the samples in terms of the nonlinear length,Eqs. (5) and (6), we find that L eff/L NL ≈ 24, 20, and 32 for Sample 1, Sample 3, and Sample 4, respectively. The larger the ratio L eff/L NL, the larger the nonlinear phase shift that could be achieved in the sample. This is valid for Sample 4 having the largest ratio L eff/L NL ≈ 32 and, consequently, the largest nonlinear phase shift. However, there is a slight disagreement between the values of the ratios and the phase shifts for Sample 1 and Sample 3. We attribute it to an uncertainty related to quantifying different losses in our experiment, resulting in errors of the coupled-in peak power values and the lengths L NL and L eff.
We also show the nonlinear phase shift obtained in Sample 4 at the OPO wavelength 1565 nm. The ϕNL up to 6π has been observed for that wavelength at the same in-waveguide maximum peak power as for the 1550 nm wavelength. The larger phase shift could be attributed to the slightly lower propagation loss at the longer wavelengths. To the best of our knowledge, this is the largest value of the SPM-induced nonlinear phase shift reported in a non-dispersion-engineered waveguide.
For the wafer composition, we expected the nonlinear absorption to be negligible. However, measuring the output power as a function of the incident power, we noticed some nonlinear behavior due to the nonlinear absorption. By plotting the inverse transmission 1/T as the function of the incident peak intensity we obtained a nonlinear dependence, indicating that the intensity-dependent absorption is governed by the nonlinear effect higher than the third-order. Plotting 1/T 2 as the function of the peak intensity squared, we obtained a straight line (see Fig. 6). It suggests that the primary nonlinear absorption mechanism is due to the 3PA, which is a χ (5) effect. We evaluated the 3PA coefficient to be α (3) ≈ 0.08 ± 0.03 cm3/GW2. This value is in excellent agreement with the one reported earlier for AlGaAs with 20% of Al (the composition of our guiding layer) . The total power loss due to the nonlinear absorption did not exceed 15% in all our experiments. The samples did not exhibit any damage sign at the operated power level.
It is informative to compare the impact of the nonlinear absorption in AlGaAs to that in silicon and GaAs. However, the precise values of α (3) for silicon and GaAs are unavailable at telecom C-band wavelengths as the dominant nonlinear absorption mechanism for these materials at that wavelength range is TPA. There was an attempt to estimate the value of 3PA coefficient of silicon at 1500 nm, resulting in α (3) ≈ 0.07 cm3/GW2 . The measured 3PA coefficient of AlGaAs at three different Al doping levels shows a tendency to increase with the decrease in Al concentration . This calls for a conclusion that α (3) of GaAs is at least as large as that of Al0.2Ga0.8As. Even if we choose not to rely on these estimates of α (3) in silicon and GaAs and compare the values of α (2) in these materials with α (3) I 0, taking some realistic value of peak intensity that is practical for the power levels in communication systems (e.g., 1 GW/cm2), we will find that the expected power loss due to the nonlinear absorption in Al0.2Ga0.8As (α (3) I 0 ≈ 0.08 cm/GW) is still smaller than that expected in GaAs (α (2) ≈ 15 cm/GW) and silicon (α (2) ≈ 0.5 cm/GW). This comparison thus confirms that the adverse effect of the non-linear absorption on the performance of our devices based on Al0.2Ga0.8As is much smaller than that in the case of silicon or pure GaAs.
5.4. Cross-phase modulation and four-wave mixing
FWM is a parametric nonlinear process which requires an appropriate phase matching between the interacting beams. It is governed by the energy conservation law in the form
In Fig. 7 we present XPM and FWM obtained in 2-μm-wide waveguides in Sample 2 and Sample 3. We limit our report to presenting the FWM and XPM data for these two samples only, as they displayed the most efficient FWM. In Fig. 7(a), we show an example of the changes in the spectrum of the cw signal due to the nonlinear interaction with the high-peak-power OPO pulses. The XPM manifests itself by the appearance of the side wings around the cw peak, while the FWM causes the generation of an idler peak on the opposite side of the OPO peak. In the example shown in Fig. 7(a) the wavelength of the OPO is 1565 nm, and the signal and idler are centered at 1550 nm and 1580 nm, respectively, which agrees with Eq. (8).
We show the tunability of the FWM process in Sample 2 [see Fig. 7(b)] and Sample 3 [see Fig. 7(c)] for the case of the OPO wavelength centered at 1565 nm. In this case, the cw signals with the range of wavelengths within the gain spectrum of the EDFA is converted to the idler peaks with the longer wavelength. The limiting factors affecting the efficiency of FWM are the limited EDFA gain spectral range, material losses, and the walk off between the pump and generated idler due to the group velocity dispersion. The material dispersion of AlGaAs is D mat = −1000 ps/nm/km, and the waveguide dispersion in our samples is relatively small, only slightly changing the total dispersion at 1550 nm. That change resulted in the simulated total dispersion of −910 ps/nm/km at 1550 nm. The characteristic walk off length, indicating the length scale of the efficient nonlinear interactions in the sample, is given in terms of the total dispersion D = D mat + D wg, including the contributions of the material and waveguide dispersion, and the exponential pump pulse half-width as
The tunability range of the FWM in Sample 2 and Sample 3 is measured to be 14 nm and 20 nm, respectively. In the latter case, it was limited by the EDFA gain spectrum and might have been broader. The signal-to-idler conversion efficiency as large as 10 dB and 8 bB was observed in Sample 2 and Sample 3, respectively. The maximum separation of 60 nm between the signal and idler wavelengths was observed in Sample 3, limited by the EDFA tunability range.
Tuning the OPO to 1547 nm, we also observed the wavelength conversion by the FWM from the EDFA range to the shorter-wavelength range in Sample 2 [see Fig. 7(d)]. However, the material dispersion at the shorter wavelengths is larger, and the level of the EDFA noise at that wavelength range is higher, which resulted in the maximum signal-to-idler conversion efficiency only 5 dB.
In all our XPM and FWM experiments the in-waveguide peak power for the pump constituted 40 W. The average in-waveguide power for the signal beam was 130 mW. Both signal and pump beams were TM-polarized.
6. Discussions and conclusions
We have demonstrated broadband self-phase modulation, efficient cross-phase modulation and four-wave mixing in AlGaAs strip-laded waveguides. Our devices were fabricated by a combination of photolithography and dry etching in a wafer with the composition specifically designed to minimize the linear and nonlinear propagation losses, and to maximize the nonlinear coefficient. Our experimental results confirm the potential of AlGaAs for nonlinear integrated optical devices and the flexibility of this material in tailoring the optical properties of the devices. None of other candidates for integrated optics have the benefit of combining the excellent nonlinear properties with low linear and nonlinear losses and highly tailorable refractive index.
To compare the nonlinear performance of our AlGaAs devices to that of other materials, we recall the results on SPM in chalcogenide glass, reported in . The authors have observed the maximum nonlinear phase shift ϕNL = 3π/2 at the peak power of 55 W in a 5-cm-long chalcogenide glass waveguide. In our experiment, we measured the SPM-induced ϕNL = 2π at the in-waveguide peak power as low as 46 W [see Fig. 5(b)], and much larger ϕNL at higher peak powers without a sign of an optical damage. Given that the propagation losses in our samples were higher than expected, and the lengths of the samples were only 1.3–2.7 cm, we conclude that the potentials of AlGaAs for the nonlinear optical interactions are very high. The SPM performance of our sample is comparable to that reported in a 2-cm-long AlGaAs waveguide sample earlier . The nonlinear phase shift as large as 2.5π with the in-waveguide peak power 53 W was observed , increasing to 5π for higher power levels. Both the present and earlier results indicate that AlGaAs has much to offer to all-optical signal processing.
Our XPM and FWM data also compare well with those reported earlier in silicon  and chalcogenide glass . In , the authors report on the broad-band parametric gain on a silicon photonics chip via FWM with the conversion efficiency up to 5 dB. The devices studied in  were dispersion-engineered sub-micron waveguides. In , the authors report on FWM in 5-cm-long As2S3 chalcogenide glass waveguides with the 10-nm tunability range of the signal wavelength and the maximum FWM conversion efficiency 14 dB. In our samples, we have achieved maximum conversion efficiency of only 10 dB, but the tunability range of our FWM was as large as 20 nm, limited by the tuning range of the EDFA. In , the authors used pulsed signal, which helped them to increase the conversion efficiency of their FWM. By minimizing the losses and optimizing the coupling, we can achieve a higher conversion efficiency in AlGaAs waveguides in future.
The operating power in our experiment was relatively high compared to what it could be if the propagation loss were smaller. Comparing the propagation losses in Sample 3 that has the best-quality waveguide profile and the smoothest side walls to those in Sample 4 of the worst quality, we arrive at the conclusion that the sidewall roughness is not the primary contribution to the propagation loss in our devices. Indeed, the mode in a strip-loaded waveguide is buried underneath the ridge, so, it does not “sense” much of the sidewall roughness. This suggests that the primary cause of the high propagation loss in our sample was the epilayer roughness and wafer defects, as well as possible ion implantation during the reactive ion etching of AlGaAs. By improving the epitaxial growth procedure for the future AlGaAs wafer it is possible to eliminate most of the excessive propagation loss. For comparison, we refer the reader to  in which the authors report the propagation loss in AlGaAs waveguide as low as 0.4 dB/cm. This confirms that the limitations in the performance of our samples are not of fundamental nature and can be eliminated in future. Growing a defect-free wafer in future should help us to minimize the linear propagation losses. The nonlinear loss in our experiment, associated with the three-photon absorption, is nearly negligible and could further be minimized by the lower operation powers.
Our experimental results confirm the exceptionally good nonlinear performance of AlGaAs. Despite the aforementioned limitations, our samples demonstrate a nonlinear performance comparable to that of the high-quality chalcogenide glass waveguides with very low linear and nonlinear propagation losses. Further research and attempts to improve the performance of the AlGaAs devices will eventually lead to demonstrating high-quality robust tailorable integrated optical chips for all-optical signal processing. We believe that our results sufficiently demonstrate the significance of AlGaAs for all-optical networks.
We acknowledge CMC for the wafer growth. The photolithography was performed at the ECTI fabrication facility at the University of Toronto. This work was supported by NSERC.
References and links
3. B. G. Lee, B. A. Small, K. Bergman, Q. Xu, and M. Lipson, “Transmission of high-data-rate optical signals through a micrometer-scale silicon ring resonator,” Opt. Lett. 31, 2701–2703 (2006). [CrossRef] [PubMed]
4. R. Salem, M. A. Foster, A. C. Turner, D. F. Geraghty, M. lipson, and A. L. Gaeta, “Signal regeneration using low-power four-wave mixing on silicon chip,” Nat. Photonics 2, 35–38 (2007). [CrossRef]
5. B. J. Lee, A. Biberman, A. C. Turner-Foster, M. A. Foster, M. Lipson, A. L. Gaeta, and K. Bergman, “Demonstration of broadband wavelength conversion at 40 Gb/s in silicon waveguides,” IEEE Photon. Technol. Lett. 21, 182–184 (2009). [CrossRef]
6. W. Mathlouthi, H. Rong, and M. Paniccia, “Characterization of efficient wavelength conversion by four-wave mixing in sub-micron silicon waveguides,” Opt. Express 16, 16735–16745 (2008). [CrossRef] [PubMed]
7. F. Li, M. Pelusi, D.-X. Xu, A. Densmore, R. Ma, S. Janz, and D. J. Moss, “Error-free all-optical demultiplexing at 160 Gb/s via FWM in a silicon nanowire,” Opt. Express 18, 3905–3910 (2010). [CrossRef] [PubMed]
8. V. G. Ta’eed, M. Shokoon-Saremi, L. Fu, D. J. Moss, M. Rochette, I. C. M. Littler, B. J. Eggleton, Y. Ruan, and B. Luther-Davies, “Integrated all-optical pulse regenerator in chalcogenide waveguides,” Opt. Lett. 30, 2900–2902 (2005). [CrossRef] [PubMed]
9. M. R. E. Lamont, B. Luther-Davies, D.-Y. Choi, S. Madden, and B. J. Eggleton, “Supercontinuum generation in dispersion engineered highly nonlinear (γ = 10 W−1m−1 As2S3 chalcogenide planar waveguide,” Opt. Express 16, 14938–14944 (2008). [CrossRef] [PubMed]
10. V. G. Ta’eed, M. R. E. Lamont, D. J. Moss, B. J. Eggleton, D.-Y. Choi, S. Madden, and B. Luther-Davies, “All optical wavelength conversion via cross phase modulation in chalcogenide glass rib waveguides,” Opt. Express 14, 11242–11247 (2006). [CrossRef] [PubMed]
11. M. D. Pelusi, V. G. Ta’eed, M. R. E. Lamont, S. Madden, D.-Y. Choi, B. Luther-Davies, and B. J. Eggleton, “Ultra-high nonlinear As2S3 planar waveguide for 160-Gb/s optical time-division demultiplexing by four-wave mixing,” IEEE Photon. Technol. Lett. 19, 1496–1498 (2007). [CrossRef]
12. M. Galili, J. Xu, H. C. H. Mulvad, L. K. Oxenlowe, A. T. Clausen, P. Jeppesen, B. Luther-Davies, S. Madden, A. Rode, D.-Y. Choi, M. Pelusi, F. Luan, and B. J. Eggleton, “Breakthrough switching speed with an all-optical chalcogenide glass chip: 640 Gbit/s demultiplexing,” Opt. Express 17, 2182–2187 (2009). [CrossRef] [PubMed]
13. F. Luan, M. D. Pelusi, M. R. E. Lamont, D.-Y. Choi, S. Madden, B. Luther-Davies, and B. J. Eggleton, “Dispersion engineered As2S3 planar waveguides for broadband four-wave mixing based wavelength conversion of 40 Gb/s signals,” Opt. Express 17, 3514–3520 (2009). [CrossRef] [PubMed]
14. V. G. Ta’eed, M. Shokooh-Saremi, L. Fu, I. C. M. Littler, D. J. Moss, M. Rochette, B. J. Eggleton, and Y. Ruan, “Self-phase modulation-based integrated optical regeneration in chalcogenide waveguides,” IEEE J. Sel. Top. Quantum Electron. 12, 360–370 (2006). [CrossRef]
15. G. I. Stegeman, A. Villeneuve, J. Kang, J. S. Aitchison, C. N. Ironside, K. Al-Hemyari, C. C. Yang, C.-H. Lin, H.-H. Lin, G. T. Kennedy, R. S. Grant, and W. Sibett, “AlGaAs below half bandgap: the silicon of nonlinear optical materials,” Int. J. Nonlinear Opt. Phys. 3, 347–371 (1994). [CrossRef]
16. A. Villeneuve, J. S. Aitchison, B. Vogele, R. Tapella, J. U. Kang, C. Trevino, and G. I. Stegeman, “Waveguide design for minimum nonlinear effective area and switching energy in AlGaAs at half the bandgap,” Electron. Lett. 31, 549–551 (1995). [CrossRef]
17. J. S. Aitchison, D. C. Hutchings, J. U. Kang, G. I. Stegeman, and A. Villeneuve, “The nonlinear optical properties of AlGaAs at the half band gap,” IEEE J. Quantum Electron. 33, 341–348 (1997). [CrossRef]
18. D. Duchesne, R. Morandotti, G. A. Siviloglou, R. El-Ganainy, G. I. Stegeman, D. N. Christodoulides, D. Modotto, A. Locatelli, C. De Angelis, F. Pozzi, and M. Sorel, “Nonlinear photonics in AlGaAs photonics nanowires: self phase and cross phase modulation,” International Symposium: Signals, Systems and Electronics, 2007, 475–478. [CrossRef]
19. G. A. Siviloglou, S. Suntsov, R. El-Ganainy, R. Iwanow, G. I. Stegeman, D. N. Christodoulides, R. Morandotti, D. Modotto, A. Locatelli, C. De Angelis, F. Pozzi, C. R. Stanley, and M. Sorel, “Enhanced third-order nonlinear effects in optical AlGaAs nanowires,” Opt. Express 14, 9377–9384 (2006). [CrossRef] [PubMed]
20. W. Astar, P. Apiratikul, T. E. Murphy, and G. M. Carter, “Wavelength conversion of 10-Gb/s RZ-OOK using filtered XPM in a passive GaAs-AlGaAs waveguide,” IEEE Photon. Technol. Lett. 22, 637–639 (2010). [CrossRef]
21. A. Pasquazi, Y. Park, J. Azana, F. Legare, R. Morandotti, B. E. Little, S. T. Chu, and D. J. Moss, “Efficient wavelength conversion and net parametric gain via four wave mixing in a high index doped silica waveguide,” Opt. Express 18, 7634–7641 (2010). [CrossRef] [PubMed]
22. K. Dolgaleva, W. C. Ng, L. Qian, J. Aitchison, M. Camasta, and M. Sorel, “Broadband self-phase modulation, cross-phase modulation, and four-wave mixing in 9-mm-long AlGaAs waveguides,” Opt. Lett. 35, 4093–4095 (2010). [CrossRef] [PubMed]
23. M. R. E. Lamont, B. Luther-Davies, D.-Y. Choi, S. Madden, X. Gai, and B. J. Eggleton, “Net-gain from a parametric amplifier on a chalcogenide optical chip,” Opt. Express 16, 20374–20381 (2008). [CrossRef] [PubMed]
24. S. Gehrsitz, F. K. Reinhart, C. Gourgon, N. Herres, A. Vonlanthen, and H. Sigg, “The refractive index of AlxGa1–xAs below the band gap: accurate determination and empirical modeling,” J. Appl. Phys. 87, 7825–7837 (2000). [CrossRef]
25. R. J. Deri and E. Kapon, “Low-loss III–V semiconductor optical waveguides,” IEEE J. Quantum Electron. 27, 626–640 (1991). [CrossRef]
26. G. P. Agrawal, Nonlinear Fiber Optics, 3rd ed. (Academic Press, 2001).
27. N. Vermeulen, C. Debaes, and H. Thienpont, “Modeling mid-infrared continuous-wave silicon-based Raman lasers,” Proc. SPIE 6455, 64550U (2007). [CrossRef]
28. M. A. Foster, A. C. Turner, J. E. Sharping, B. S. Schmidt, M. Lipson, and A. Gaeta, “Broad-band optical parametric gain on a silicon photonic chip,” Nature 441, 960–963 (2006). [CrossRef] [PubMed]