We have experimentally demonstrated broadband tuneable four-wave mixing in AlGaAs nanowires with the widths ranging between 400 and 650 nm and lengths from 0 to 2 mm. We performed a detailed experimental study of the parameters influencing the FWM performance in these devices (experimental conditions and nanowire dimensions). The maximum signal-to-idler conversion range was 100 nm, limited by the tuning range of the pump source. The maximum conversion efficiency, defined as the ratio of the output idler power to the output signal power, was −38 dB. In support of our explanation of the experimentally observed trends, we present modal analysis and group velocity dispersion numerical analysis. This study is what we believe to be a step forward towards realization of all-optical signal processing devices.
© 2015 Optical Society of America
Since the inception of the Internet, the growth of bandwidth consumption shows no signs of abating. Optical networks are moving from relatively simple point-to-point arrangements to reconfigurable wavelength-routed architectures. The present role of the large-scale photonic integration in optical communications is to replace multiple sets of bulk components in optical-to-electrical-and-back-to-optical (OEO) converters at the network nodes with a single optical chip accommodating all their functionalities. Transmission of data at rates above 1 Tbit/s in a single channel is challenging as the pulse width in such data streams is only a few hundred femtoseconds. At such ultrahigh data rate, all-optical signal processing becomes necessary to overcome the bandwidth limitations of electronics.
At present, signal processing at the network nodes is mostly performed electronically. The future role of photonic integration in optical communications is to enable all-optical signal processing to minimize the need for OEO at the network nodes. Remarkable progress in this direction has evolved in recent years [1–6]. Numerous all-optical functions have been demonstrated [1, 5]; among them are all-optical logic gates, label switching, analog-to-digital conversion, wavelength conversion, tuneable optical delay lines, and 3R regeneration technique, capable of retiming, reshaping and reamplifying the signal. Most of these functions are still at the level of their first laboratory demonstrations and need to be developed further for practical applications, while some others, such as various temporal waveform shaping operations and on-chip pulse metrology, are yet to be demonstrated.
The all-optical operations rely on all-optical wavelength conversion which can be implemented using nonlinear optical effects that modify the spectrum of the optical signal. Among such effects are either second-order nonlinear interactions resulting in a sum-frequency generation followed by a difference-frequency generation (SFG/DFG) [2,5], or the third-order effects, such as cross-phase modulation (XPM) or four-wave mixing (FWM) that rely on the intensity-dependent refractive index n2 (Kerr coefficient) . Among the material platforms and devices used for demonstrating wavelength conversion are lithium niobate , semiconductor optical amplifiers (SOAs) , silicon , chalcogenide glass , and AlGaAs semiconductor passive waveguides [6,9]. Lithium niobate exhibits strong second-order nonlinearity and is suitable for wavelength conversion by SFG/DFG, while the rest of the materials have large values of the Kerr coefficient. Silicon and SOA, however, suffer from the carrier dynamics that causes inter-channel cross-talk and limits the operation speed of the devices (these adverse effects are worse in silicon due to a strong two-photon absorption leading to a free-carrier absorption). None of the listed materials, except for SOAs and AlGaAs, can emit light and permit flexibility in terms of adjusting the operational wavelength and refractive index contrast of the waveguides. In addition, fabrication of chalcogenide glass waveguides requires a sophisticated and expensive purification procedure. AlGaAs has been termed “the silicon of nonlinear optics”  because of its excellent nonlinear performance [6, 9–22]. Together with the highest value of intensity-dependent refractive index n2 among the materials for wavelength conversion , it exhibits low linear and nonlinear propagation losses in the telecommunication (telecom) spectral range (1400–1600 nm). Efficient FWM , self-phase modulation , and wavelength conversion based on XPM [15, 16] and FWM  have been demonstrated in AlGaAs waveguides with relatively large dimensions (∼ 1 µm). These structures are, however, far from optimal, as they do not allow for dense photonic integration and dispersion management. Ultracompact AlGaAs waveguides with submicron dimensions (“nanowires”) have a potential of larger density of integration on an optical chip. They also offer dispersion management for phase matched nonlinear optical interactions such as FWM . Second-harmonic generation , cross-phase modulation , and continuous-wave (cw) FWM at fixed wavelengths  have been demonstrated in these structures. Our team was the first to report on the experimental demonstration of broadly tuneable FWM in AlGaAs nanowires . In a more recent report, Porkolab, et al. have demonstrated AlGaAs nanowires with the widths 700 nm and higher with record-low propagation loss of 0.56 dB/cm . This study confirms that AlGaAs nanowires have nearly as much potential in terms of low-loss propagation as silicon sub-micron devices, provided that the fabrication procedure is well established. The same team of authors has also demonstrated superior performance of FWM with cw beams in their AlGaAs waveguides with the conversion range spanning 80 nm and the conversion efficiency of −7.8 dB in . The present study is unique in the sense that the widths of our AlGaAs nanowires that were tested for their FWM performance fell in the range 400–650 nm, which is smaller compared to those reported in . Besides, we have conducted a detailed investigation of the influence of different experimental conditions and waveguide dimensions on the performance of FWM. Our work is thus an original contribution that demonstrates the potential of AlGaAs nanowires as wavelength converters in future optical communication networks, and outlines trends in FWM efficiency depending on various parameters.
2. Fabrication of the devices
We used as a starting material an AlGaAs wafer with a specifically designed composition, grown by metalorganic chemical vapour deposition (MOCVD) at the Canadian Microelectronics Corporation (CMC). The wafer composition and device structures are outlined in Fig. 1(a). The composition of the guiding layer in our devices contained 18% of aluminum. This concentration ensured low linear and two-photon absorption . Our sample had a series of nanowires with different widths and lengths. The devices were defined on the AlGaAs wafer using direct writing by electron-beam lithography in HSQ e-beam resist. The layer of HSQ was deposited on top of the substrate by spin-coating; the height of the layer was 190 nm. After writing and developing the patterns, we proceeded with inductively coupled plasma (ICP) etching to define the profile in AlGaAs. The etching was performed at Sherbrooke University, using a well developed etching procedure . The simulated fundamental mode pattern and SEM picture of the cross-section of a typical nanowire are presented in Figs. 1(b) and 1(c), respectively.
To simplify coupling in and out from the devices, we have made 2-µm-wide coupling waveguides at the input to the devices, followed by tapers reducing the waveguide dimension down to the width of the nanowire, and then expanding it again at the output. The top view of the devices is shown in Fig. 2. The widths of our nanowires were in the range between 300 and 800 nm with a 50-nm step. The nanowires with the lengths of 0 (taper-to-taper), 1, 2, and 4 mm were defined for each width. The entire sample containing all the devices was 7-mm-long, and the lengths of the individual nanowires were controlled through the variation in the lengths of the coupling regions. The role of the “taper-to-taper” devices and the straight 2-µm-wide waveguides was two-fold. First of all, these devices were used for the loss measurement so that we could estimate the linear propagation loss of the nanowires. In addition, they also were tested for the FWM signal so that we could find out how significant was the contribution to the overall signal from these parts of the devices, and how much of it could be attributed to the nanowires.
3. Numerical analysis
3.1. Modal analysis
We have performed a modal analysis for the nanowires with the widths ranging between 300 nm and 800 nm for both the fundamental TE and TM modes. We found that, for a 300-nm-wide nanowire, the fundamental TE mode cannot be supported at the wavelengths longer than 2.2 µm. In fact, the experimentally observed cut off wavelength was around 1600 nm due to the drastic increase of the propagation loss for the longer wavelengths. In Fig. 3, we plot the effective mode area for the fundamental TE and TM modes as a function of the waveguide width at the fixed wavelength 1550 nm [Fig. 3(a)], and as a function of wavelength for the TE mode for different values of the nanowire width [Fig. 3(b)]. These data were obtained from the modal analysis performed with Lumerical Mode Solutions software. It can be seen from Fig. 3(a) that the modal area expands for the waveguide widths smaller than 300 nm (TE mode) and 200 nm (TM mode) as the fundamental mode in these devices approaches the cut-off range. The mode area for 300-nm-wide nanowires exhibits a more rapid growth with the increase of the wavelength compared to the trends for the wider nanowires, as shown in Fig. 3(b). This observation agrees with the drastic increase in the measured propagation loss in the range of wavelengths beyond 1570 nm due to the approaching cut-off.
3.2. Dispersion analysis
Group-velocity dispersion (GVD) is the key reason placing the limit on the tuneability of the FWM process. If the frequency components interacting in the process of FWM are far enough apart so that their group velocities differ significantly, a walk-off effect occurs, in the process of which these components get separated in time and don’t interact efficiently as they travel through the nonlinear waveguide. GVD can be mathematically described in terms of the GVD parameter β2, which is the second derivative of the propagation constant with respect to frequency: β2 = d2β/dω2. One of the benefits offered by AlGaAs nanowires with the dimensions we consider in this study is the fact that they exhibit zero-dispersion points where the GVD parameter β2 changes sign within the telecom frequency range . The range of wavelengths in the vicinity of a zero-dispersion point experiences the smallest walk-off effect and can be used for efficient FWM. It is thus important to be aware of the dispersion behaviour of the nonlinear optical waveguides in the wavelength range of interest. In order to explain some trends observed in the process of measuring the FWM signal, we performed numerical analysis of GVD for our AlGaAs nanowires. In this section, we report the results of the analysis.
In Fig. 4, we display GVD curves for the nanowires of different diameters as a function of wavelength for the TE [Fig. 4(a)] and TM [Fig. 4(b)] fundamental modes. The wavelength range of interest spans between 1450 and 1580 nm, which corresponds to the range of conversion in our FWM experiments. The strongest FWM efficiency was observed with a 550-nm-wide nanowire. According to Fig. 4, this device, indeed, exhibits the smallest GVD which is relatively flat in the wavelength range of interest. This also holds for the nanowires with the widths 600 and 650 nm. The GVD for the fundamental TM mode is very large and positive in the wavelength range of interest. This explains why the FWM efficiency for this polarization was measured to be relatively low. On the other hand, zero-GVD for the TM polarization is achievable at the longer wavelengths. There could be a potential for dispersion management at TM polarization in that wavelength range.
We also present the GVD as a function of the nanowire width at a fixed wavelength 1550 nm (see Fig. 5). The graph shows that zero-GVD point at this wavelength occurs for a nanowire with the width close to 600 nm, which is in agreement with the earlier GVD analysis repotted for similar structures .
The walk-off length is the characteristic interaction length for the wavelengths participating in the four-wave mixing process. It answers the question on what length scale the signal and pump wavelengths, separated by the difference ∆λ = |λp −λs|, can co-propagate and interact efficiently before they encounter a significant walk-off due to the dispersion present in the medium. We used the simulated values of the dispersion coefficient D = (−2πc/λ2)β2 in order to estimate the walk-off length
Some other characteristic lengths useful for the analysis of our experimental observations include the effective length
In Table 1, we combine all the characteristic lengths and the nonlinear coefficient γ for the nanowire dimensions that were used in our comparative experimental studies. Among those are 1-mm-long 400, 550, and 650-nm-wide nanowires, as well as a 2-mm-long 550-nm-wide nanowire (two bottom lines in the table). The parameters are calculated for the two pump wavelengths settings, corresponding to 1505 nm and 1525 nm. It is clear from the table that the walk-off length is not the factor limiting the performance of our devices: all the values of Lw are minimum one order of magnitude larger than the lengths of the nanowires used in our experiments. On the other hand, the characteristic lengths Leff and LNL are on the order of the lengths of the nanowires, and thus represent more significant figures for the assessment of the nonlinear optical performance of our devices.
We also calculated the normalized phase matching parameter for the FWM for different values of the nanowire width for both TE and TM polarization (shown in Fig. 6). The phase matching parameter is defined as 
4. Linear characterization
The linear characterization of the coupling and propagation loss in the devices has been performed using Fabry-Perot method of fringes. We used 2-µm-wide waveguides to evaluate the propagation loss per unit length associated with the 2-µm-wide coupling regions. The measured value was around 9 dB/cm for both TE and TM fundamental modes. We also measured the coupling loss (6.2 dB per facet). Measuring the overall loss of the devices containing tapered regions only (“taper-to-taper” or “zero-length” nanowires) allowed us, after subtracting the propagation loss of the 2-µm-waveguide sections and the coupling loss, to estimate the loss associated with the taper regions. Measured in such a manner value of the taper loss was 1 dB per device. This value was measured to be the same for the whole range of widths of the nanowires present in our sample.
The above measurements allowed us to estimate the propagation loss of the nanowires themselves, which we present in Fig. 7. The part (a) of Fig. 7 displays the propagation loss of the nanowires measured at 1550 nm as a function of the nanowire width, while the part (b) represents the propagation loss as a function of wavelength for a 400-nm-wide nanowire. The data are shown for both TE and TM modes. It is intuitive that the smaller nanowires exhibit higher propagation loss. The propagation loss for the TM mode is lower because of the lower refractive index contrast this mode “sees” . Also, longer wavelengths exhibit higher propagation loss because they are closer to the cut-off range of the nanowires. In such a way, this experimental part allowed us to fully quantify different sources of losses associated with the signal propagation through the devices containing the nanowires.
The primary reason for the high propagation losses in our devices is the strong light scattering off the sidewall defects. Such defects appear in the process of waveguide fabrication. Their influence is especially strong due to the high refractive index contrast at the interface between the semiconductor and air. Reducing the loss is possible via coating the devices with a thin layer of a dielectric material so that the refractive index contrast in the immediate vicinity of the semiconductor boundaries is lower .
5. Four-wave mixing experiments
5.1. Experimental setup
The nonlinear optical characterization of the devices involved conducting four-wave mixing experiments. The schematic of the experimental setup is shown in Fig. 8. We used the output of a Coherent optical parametric oscillator (OPO) as a pump source. The device was pumped by a Ti:Sapphire laser, operating at 800 nm, and produced up to 300 mW of the output power in the wavelength range between 1500 and 1600 nm. The optical pulses originating from the OPO had 2-ps FWHM temporal duration and followed at the repetition rate 76.6 MHz. The signal was obtained from a JDSU tuneable cw laser, amplified by an Amonics erbium-doped fibre amplifier (EDFA) to the power level up to 2 W. The tuning range of the signal beam was constrained by the operation range of the EDFA, capable of amplifying the radiation with the wavelengths between 1535 and 1565 nm. Both the cw laser and EDFA were fibre-coupled; the radiation coming from EDFA was decoupled from the fibre to free space with a collimator and combined with the pump radiation using a non-polarizing 50 % beam splitter (marked “50 %” on the figure). The polarization states of both beams were set independently by a pair of a polarizing beam splitter (PBS) and a half-wave plate (HWP), placed in the corresponding arm. As the polarization of light originating from the cw laser and EDFA was scrambled, we placed a fibre-based polarization controller (PC) at the output of the EDFA to maximize the signal power after the free-space polarization control devices. After being combined with a non-polarizing beam splitter, the pump and signal beams were simultaneously coupled into the integrated optical devices placed into a micro-positioning coupling stage with piezo-controller. The two 40× microscopic objectives were used for coupling the light into and out from a particular waveguide on the chip. The power and spectrum of the radiation at the output of a waveguide were measured using Ando optical spectrum analyzer (OSA) and an IR photodetector (“signal”), respectively. The power levels of the pump and signal were monitored with a beam sampler (BS) reflecting around 10 % of the optical power to the reference IR photodetector (marked as “reference” on the sketch) before the beams entered the integrated optical devices.
5.2. Four-wave mixing spectra
Prior to performing a detailed study of the FWM in different nanowires, we first maximized four-wave mixing signal by establishing the right experimental conditions. We first looked at the polarization of the pump and signal, setting it to either TE or TM state and measuring the four-wave mixing spectra for several nanowires. In Fig. 9, we show a typical example of the measurement outcome, obtained for a 400-nm-wide 1-mm-long nanowire when the pump wavelength was set to 1525 nm. In this and any other FWM spectrum plot, presented in this section, we show simultaneously the spectra of the pump, various peaks corresponding to the cw signal (placed all together on the same graph), and the generated idler peaks corresponding to the signal peaks. The central peak on the graph is the pump spectrum (shown with an arrow). The narrow peaks on the long-wavelength side correspond to the signal. The shorter-wavelength peaks represent the generated idler. The pedestal around the signal peaks is the unfiltered amplified spontaneous emission of the EDFA. The grey shaded areas on the graph represent a FWM peak appearing as a consequence of the secondary peak in the pump spectrum (pump artifact). At some wavelength settings, the OPO produced a double-peak output with the extra peak appearing at the longer-wavelength side. The two peaks interacted in the samples to produce a FWM-generated idler peak that we shade on the graphs in order not to confuse it with the idler obtained from the FWM produced by the cw signal and OPO beams. In this, as well as all other cases, the TE-polarized pump and signal exhibited a significantly higher FWM conversion efficiency and FWM tuning range compared to that demonstrated by the TM-polarized pump and signal, despite the higher propagation loss for the TE polarization. This result can be explained by the fact that the TE mode is more confined that the TM mode (the effective mode area for TE-polarized light is smaller), and by the fact that dispersion management in these structures in the wavelength range of the FWM conversion is achievable for the TE polarization only. It is especially clear from Fig. 6 where we show the normalized phase matching term as a function of the wavelength detuning: the TM mode has a very narrow wavelength range within which it is possible to observe FWM. Besides, there is no perfect phase matching for the TM polarization. For the rest of the measurements, we set both the pump and signal polarizations to the TE state.
We also looked into optimizing the power level of the signal and pump. Maximizing the signal power has resulted in maximizing the idler. We thus set the signal to the maximum value corresponding to the in-waveguide power level around 18 mW (after accounting for the coupling efficiency). The pump power, on the other hand, had an optimum value, increasing the pump power beyond which did not result in any increase in the idler power. In Fig. 10, we show FWM spectra collected from a 550-nm-wide 2-mm-long nanowire at three values of the in-waveguide pump peak power, as labeled on the plots. In this and most other measurements, the FWM idler kept increasing with the pump peak power until its value reached 1.8 W. When we increased the value of the pump power beyond 1.8 W, the level of the generated idler remained the same, while the pump, signal, and idler spectra exhibited broadening due to SPM and XPM nonlinear effects. This spectral broadening could degrade the performance of a potential wavelength converter due to the overlap of the spectra of the adjacent wavelength channels. Besides, it is unnecessary to pump the samples beyond the optimum power level. These measurements helped us to make a judgement regarding the optimal value of the in-waveguide pump peak power for the rest of the FWM studies.
Next question that we attempted to answer was related to the significance of the contribution to the overall FWM signal coming from the 2-µm-wide portions corresponding to the coupling in and out waveguide regions. In Fig. 11, we present a comparison between the performance of a 2-µm-wide straight waveguide spanning the entire length of the sample [Figs. 11(a) and 11(d)], a similar waveguide tapered down to the width of 550 nm, and immediately tapered back up to 2 µm [“taper-to-taper,” Figs. 11(b) and 11(e)], and a waveguide containing a 1-mm-long 550-nm-wide nanowire [Figs. 11(c) and 11(f)]. The data are presented for the two settings of the pump wavelength: 1505 nm [Figs. 11(a)–11(c), 100-nm conversion range], and 1525 nm [Figs. 11(d)–11(f), 60-nm conversion range]. By “conversion range” we mean the difference in the wavelength between the signal and generated idler beams. Clearly, most of the FWM signal can be attributed to the 2-µm-wide regions when the pump wavelength is set to 1525 nm, as there is only a small difference in the signals collected from the straight 2-µm-wide waveguides and nanowires. On the other hand, the presence of the nanowires becomes crucial when the conversion range is wide. The reason to it is the smaller value of the GVD and higher modal confinement associated with the nanowires. The conclusion one can draw based on the data presented in Fig. 11 is that, despite the much higher propagation loss of the nanowires and the fact that the 2-µm-wide regions are approximately 6 times longer compared to the nanowire length, the FWM signal associated with the nanowire is significant and becomes dominant for the broader conversion range.
As described in Section 2, our sample contained nanowires of various widths and lengths. We investigated how the nanowire dimensions affect the FWM performance by measuring the FWM spectra for the nanowires of different widths and equal length (Fig. 12), and for the nanowires of equal widths and different lengths (Fig. 13). In Fig. 12, we show the FWM spectra for the 1-mm-long nanowires with the widths 400 nm [Fig. 12(a)], 500 nm [Fig. 12(b)], 550 nm [Fig. 12(c)], and 650 nm [Fig. 12(d)] under similar experimental conditions (pump power and wavelength). Comparing the performance of the nanowires, we conclude that, despite the higher value of the propagation loss, the 400-nm-wide nanowire performs nearly as well as the 550-nm-wide nanowire, while the performance of the 650-nm-wide nanowire is significantly worse as it does not demonstrate as wide FWM conversion range and as high idler peaks. It is rather surprising that the 400-nm-wide nanowire with relatively high propagation loss and dispersion demonstrated such a strong idler. We turn for the interpretation of these results to the values of the characteristic lengths listed in Table 1. As was already mentioned, the walk-off length, which is the shortest for the 400-nm-wide nanowire, is still one order of magnitude longer compared to the lengths of the nanowires in this study. We thus can conclude that this parameter is not the bottleneck in the FWM performance. On the other hand, the effective and nonlinear lengths are comparable to the lengths of the nanowires, and their influence on the FWM is by far more significant. Based on the ratio Leff/LNL, reported in the table, we can conclude that the 400-nm-wide nanowire is, indeed, expected to demonstrate a relatively high FWM conversion efficiency, despite the high dispersion associated with this device. The ratio Leff/LNL for a 650-nm-wide nanowire is, on the other hand, significantly lower, which explains the relatively poor performance demonstrated by this device.
The length of the nanowires could become a limiting factor affecting the efficiency of the FWM as the propagation loss in our devices was measured to be extremely high. In order to experimentally test whether the optimal length falls in the range of the nanowires’ lengths present in our sample, we measured the FWM spectra of the “taper-to-taper” device (zero-length nanowire) [Figs. 13(a) and 13(d)], 1-mm-long [Figs. 13(b) and 13(e)], and 2-mm-long nanowires [Figs. 13(c) and 13(f)] with the widths 550 nm. The 4-mm-long nanowires exhibited such high overall propagation loss that we were unable to observe any clear mode pattern at the output of such devices with the widths smaller than 800 nm. The measurements were performed for two values of the pump wavelengths corresponding to 1505 nm [Figs. 13(a)–13(c)] and 1525 nm [Figs. 13(d)–13(f)]. Based on the spectra shown in Figs. 13(d)–13(e), all the devices performed equally well when the conversion range was shorter. The data obtained for the pump wavelength 1525 nm confirm the earlier observation that when the conversion range is relatively narrow, most of the FWM signal can be attributed to the 2-µm-wide waveguide portions of the devices. The benefit of the longer devices becomes evident at the 1505-nm pump wavelength setting, as they yield longer interaction lengths with a smaller value of the overall dispersion, which is crucial for the broader conversion range [Figs. 13(a)–13(c)]. These observations agree with the fact that the ratio Leff/LNL, reported in Table 1, is highest for the 2-mm-long nanowires.
We next experimented with different settings for the pump wavelength to achieve tuneable FWM in different ranges. The spectra for a 400-nm-wide 1-mm-long nanowire are presented in Fig. 14. In Figs. 14(a)–14(c), we demonstrate the signal conversion to the shorter wavelengths exhibiting smaller propagation loss. As a result, the FWM in these cases was more efficient. In Figs. 14(d) and 14(e), we show the FWM spectra for the cases when the pump wavelength was longer compared to the tuning range of the signal, resulting in the conversion to the longer-wavelength idler components. These components experienced higher values of the propagation loss and, accordingly, the FWM efficiency was observed to be lower. Tuning the pump to the wavelength above 1575 nm did not result in any measurable FWM signal on the longer-side of the spectrum.
The maximum FWM conversion efficiency that we have achieved in our samples was −38 dB (defined as the ratio of the output idler power to the output signal power). This value was achieved for a 550-nm-wide 2-mm-long nanowire. The broadest conversion range (the wavelength difference between the signal and generated idler components) was demonstrated to be 100 nm, limited by the tuning range of the pump on the shorter-wavelength side. The estimated values for the conversion efficiency as the function of the pump wavelength for that nanowire are shown in Fig. 15. Different curves correspond to different signal wavelengths, as shown in the legend. The corresponding idler wavelengths span between 1455 and 1615 nm, covering a 160-nm band. The conversion efficiencies follow the model trend outlined in Fig. 6 for the normalized phase mismatching parameter, except for the 1525-nm pump wavelength at which the contribution to the idler power originating from the 2-µm-wide coupling waveguide sections is significant. The drop in efficiency at the pump wavelengths 1565 and 1575 nm could be explained by the higher propagation losses for the idler components spanning the range 1580–1615 nm in this case (see Fig. 7).
In conclusion, we have demonstrated a broadband tuneable four-wave mixing in AlGaAs nanowires with different widths and lengths. The fabrication process of such devices is not well established yet: there are many challenges, such as relatively high waveguide sidewall roughness and scattering, that need to be addressed. Successful attempts to reduce propagation loss in deeply-etched AlGaAs waveguides have already been reported by some researchers [20,22]. Despite these limitations, the strong Kerr nonlinearity made it possible to demonstrate FWM with a broad tuning range in these devices, and to study the parameters influencing its efficiency. There is a large potential for AlGaAs sub micron waveguides to become practical passive wavelength converters and building blocks of more sophisticated optical signal processing devices. This motivates on-going endeavour at addressing their fabrication challenges.
The authors would like to acknowledge Canadian Microelectronic Corporation (CMC) for the wafer growth. KD is thankful to Mitacs Elevate postdoctoral fellowship she has been receiving while performing the experiments.
References and links
1. B. J. Eggleton, T. D. Vo, R. Pant, J. Schr, M. D. Pelusi, D. Y. Choi, S. J. Madden, and B. Luther-Davies, “Photonic chip based ultrafast optical processing based on high nonlinearity dispersion engineered chalcogenide waveguides,” Laser Photon. Rev. 6, 97–114 (2012). [CrossRef]
2. C. Langrock, S. Kumar, J. E. McGeehan, A. Wilner, and M. M. Fejer, “All-optical signal processing using χ(2) nonlinearities in guided-wave devices,” J. Lightwave Technol. 24, 2579–2592 (2006). [CrossRef]
3. 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]
4. F. Morichetti, A. Canciamilla, C. Ferrari, A. Samarelli, M. Sorell, and A. Melloni, “Travelling-wave resonant four-wave mixing breaks the limits of cavity-enhanced all-optical wavelength conversion,” Nat. Commun. 2, 296–303 (2011). [CrossRef] [PubMed]
5. L. Yan, A. E. Willner, X. Wu, A. Yi, A. Bogoni, Z.-Y. Chen, and H.-Y. Jiang, “All-optical signal processing for ultrahigh speed optical systems and networks,” J. Lightwave Technol. 30, 3760–3770 (2012). [CrossRef]
6. B. M. Cannon, T. Mahmood, W. Astar, P. Apiratikul, G. Porkolab, P. Boudra, T. Mohsenin, C. J. Richardson, and G. M. Carter, “All-optical amplitude-phase transmultiplexing of RZ-OOK and RZ-BPSK to RZ-QPSK by polarization-insensitive XPM using a nonlinear birefringent AlGaAs waveguide,” J. Lightwave Technol. 31, 952–966 (2013). [CrossRef]
7. R. W. Boyd, Nonlinear Optics, 3rd ed. (Academic Press, 2008).
8. C. Meuer, C. Schmidt-Langhorst, H. Schmeckebier, G. Fiol, D. Arsenijević, C. Schubert, and D. Bimberg, “40 Gb/s wavelength conversion via four-wave mixing in a quantum-dot semiconductor optical amplifier,” Opt. Express 19, 3788–3798 (2011). [CrossRef] [PubMed]
10. 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. Sibbett, “AlGaAs below half bandgap: the silicon of nonlinear optical materials,” Int. J. Nonlinear Opt. Phys. 3, 347–371 (1994). [CrossRef]
12. 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]
13. D. Duchesne, K. A. Rutkowska, M. Volatier, F. Légaré, S. Delprat, M. Chacker, D. Modotto, A. Locatelli, C. De Angelis, M. Sorel, D. N. Christodoulides, G. Salamo, R. Arès, V. Aimez, and R. Morandotti, “Second harmonic generation in AlGaAs photonic wires using low power continuous wave light,” Opt. Express 19, 12408–12417 (2011). [CrossRef] [PubMed]
14. 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]
15. W.-C. Ng, Q. Xu, K. Dolgaleva, S. Doucet, D. Lemus, P. Chretién, W. Zhu, L. A. Rusch, S. LaRochelle, L. Qian, and J. S. Aitchison, “Error-free 0.16π-XPM-based all-optical wavelength conversion in a 1-cm-long AlGaAs waveguide,” in 23rd Annual Meeting of IEEE Photonics Society, 2010.
16. 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]
17. P. Apiratikul, W. Astar, T. E. Murphy, and G. M. Carter, “10-Gb/s wavelength and pulse format conversion using four-wave mixing in a GaAs waveguide,” IEEE Photon. Technol. Lett. 22, 872–874 (2010). [CrossRef]
18. K. Dolgaleva, W.-C. Ng, L. Qian, J. S. Aitchison, M. C. 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]
19. C. Lacava, V. Pusino, P. Minzioni, M. Sorel, and I. Cristiani, “Nonlinear properties of AlGaAs waveguides in continuous wave operation regime,” Opt. Express 22, 5291–5297 (2014). [CrossRef] [PubMed]
20. G. A. Porkolab, P. Apiratikul, B. Wang, S. H. Guo, and C. J. Richardson, “Low propagation loss AlGaAs waveguides fabricated with plasma-assisted photoresist reflow,” Opt. Express 22, 7733–7743 (2014). [CrossRef] [PubMed]
21. K. Dolgaleva, P. Sarrafi, P. Kultavewuti, J. S. Aitchison, L. Qian, M. Volatier, R. Arès, and V. Aimez, “Highly efficient broadly tunable four-wave mixing in AlGaAs nanowires,” in CLEO Conference, San Jose, CA (USA), 2013.
22. P. Apiratikul, J. J. Wathen, G. A. Porkolab, B. Wang, L. He, T. E. Murphy, and C. J. Richardson, “Enhanced continuous-wave four-wave mixing efficiency in nonlinear AlGaAs waveguides,” Opt. Express 22, 26814–26824 (2014). [CrossRef] [PubMed]
23. M. Volatier, D. Duchesne, R. Morandotti, R. Arès, and V. Aimez, “Extremely high aspect ratio GaAs and GaAs/AlGaAs nanowaveguides fabricated using chlorine ICP etching with N2-promoted passivation,” Nanotechnology 21, 134014 (2010). [CrossRef]
24. H. Hu, H. Ji, M. Galili, M. Pu, C. Peucheret, H. Christian, H. Mulvard, K. Yvind, J. M. Hvam, P. Jeppersen, and L. K. Oxenløwe, “Ultra-high-speed wavelength conversion in a silicon photonic chip,” Opt. Express 21, 19886–19894 (2011). [CrossRef]
25. C. Poulton, C. Koos, M. Fujii, A. Pfrang, T. Schimmel, J. Leuthold, and W. Feude, “Radiation modes and roughness loss in high-index-contrast waveguides,” IEEE J. Sel. Topics Quantum Electron. 12, 1303–1321 (2006). [CrossRef]