We report remarkably fast and strongly wavelength-dependent gain recovery in a single SOA without the aid of an offset filter. Full gain recovery times as short as 9 ps were observed in pump-probe measurements when pumping to the blue wavelength side of a continuous wave probe, in contrast to times of 25 to 30 ps when pumping to the red wavelength side. Experimental and numerical analysis indicate that the long effective length and high gain led to deep saturation of the second half of the SOA by the probe. The consequent absorption of blue-shifted pump pulses in this region resulted in device dynamics analogous to those of the Turbo-Switch.
©2010 Optical Society of America
Over the course of recent years, extensive research has been carried out on semiconductor optical amplifiers (SOAs) for use as active switching devices in future high-speed optical communication networks . Gain and phase recovery times limit the rate at which SOAs can switch high-speed signals. In general, there is an ultrafast component dependent on intraband effects such as spectral hole burning, carrier heating and subsequent carrier cooling, and a slower, interband, band-filling process [1–3]. The speed of the band-filling process is ultimately determined by the current injection.
Many experimental and theoretical investigations of the physics underpinning both the gain and phase dynamics of SOAs have been carried out e.g [4–8]. Despite ultrafast processes in the carrier recovery, the full gain recovery in most SOAs is limited by the slow band-filling process. Various techniques, such as narrow bandpass filtering, the use of a holding beam or employing two SOAs in either a Turbo-Switch arrangement or a “push-pull” interferometer, can improve the switching capability of SOAs [9–15].
Narrow bandpass filtering is a technique that exploits the ultrafast chirp dynamics in an SOA with the aid of a narrow optical bandpass filter [10,15]. The filter selects the blue-shifted sideband of the output probe spectrum from the SOA. An effective recovery time of 1.8 ps or less can be attained and wavelength conversion has been demonstrated at 320 Gb.s−1  using this method, proving that it can achieve high signal processing rates. One disadvantage of this method is that the signal magnitude following the narrow-band filter is reduced, which compromises the optical signal-to-noise ratio (OSNR).
A holding beam is a saturating cw beam at a wavelength other than that of the pump or the probe that is incident on the SOA and increases its effective carrier recovery rate [11,17,18]. The holding beam depresses the SOA gain, which reduces the gain and phase modulation and this leads to faster recovery times for the gain and phase. This method reduces the band-filling recovery process rather than eliminating it, so patterning effects can become apparent at high data rates. The SOA bias current must be large to allow fast recovery rates to be observed .
The Turbo-Switch is an arrangement consisting of two identical or similar SOAs in series separated by a broad optical bandpass filter . A cw probe beam and pump pulses are introduced into the first SOA and cross-gain modulation (XGM) takes place in this SOA. The filter prevents transmission of pump pulses to the second SOA. The signal at the probe wavelength proceeds to the second SOA, where self-gain modulation (SGM) occurs, negating much of the slow tail of the gain and phase response of the first SOA. This technique also gives rise to an overshoot in the gain recovery, which can mitigate patterning effects. Wavelength conversion at 170.4 Gb.s−1 has been demonstrated using the Turbo-Switch .
SOAs can be placed within push-pull interferometers using a differential switching scheme to increase their switching capability as the switching “window” of the interferometer can be much shorter than the natural carrier recovery time . Mach-Zehnder and Sagnac loop interferometers are frequently employed, although other configurations such as the ultrafast nonlinear interferometer (UNI) are also commonly used .
The recovery time in a single MQW SOA can be shortened by increasing the differential gain by means of p-doping the SOA active region. For instance, 1/e phase recovery times as short as 11 ps in Zn-doping of the barriers within a MQW SOA have been demonstrated [22,23].
In this paper we discuss a multiple quantum well (MQW) SOA that exhibits a highly wavelength-dependent gain response. We have demonstrated predominantly ultrafast gain recovery in a single MQW SOA without employing specialized filters, additional SOAs, p-doping of the MQW barriers or push-pull interferometry. Pump–probe measurements using a 2.5ps pump pulse and a continuous wave (cw) probe showed a gain recovery consisting almost entirely of an ultrafast component followed by a slow overshoot, when the pumping wavelength was to the blue of (shorter than) the probe. When the pump wavelength was to the red of (longer than) the probe, slower recovery times without a gain overshoot were observed. We used a simple delay interferometer arrangement to extract gain and phase responses, rather than heterodyne detection [24,25], and pump and probe wavelengths were varied independently.
The structure of this paper is as follows: Section 2 gives details of the experiments, the results of which are presented in Section 3. Section 4 gives a description of the modelling results which explain the observed effects .The main findings of this work are summarised in Section 5.
2. Experimental details
In order to measure the gain and phase impulse responses of the SOA, the pump-probe experiment shown schematically in Fig. 1 was used. The SOA employed was a commercially available (CIP Technologies) InGaAsP/InP buried heterostructure MQW SOA with a confinement factor of 0.18, an effective length of 2.2mm and a small signal gain peak at 1565 nm. It had a −3 dB bandwidth of 125nm, a small signal fibre-to-fibre gain of 36 dB and, under typical experimental conditions described here, a fibre-to-fibre saturated gain of 21 dB. It had a saturation input power of −17 dBm. These gain values include input and output coupling losses of 2.3dB and 1.4 dB respectively. A 500 mA bias was used for all the measurements described in this paper.
The pump was a 2.5 ps pulse clock stream, with −3 dB bandwidth of 1.68 nm, from an actively mode-locked tunable laser,, whose repetition rate was reduced from 10.65 GHz to 665 MHz using an optical modulator. This pump pulse stream was combined with a tunable cw probe signal using a 50:50 coupler and both signals were injected into the SOA. The pump and probe wavelengths were varied in steps of 5 nm within the ranges 1535≤ λpump≤1570 nm and 1535≤ λprobe≤1580 nm respectively and the gain and phase evolution curves were measured. The input probe power was fixed at −3.5 dBm and the pump energy per pulse was varied from 3 to 130 fJ in order to maintain a constant amplitude modulation depth of 50% for the range of wavelengths used. A 4nm band-pass filter after the SOA removed both ASE and pump signals. The output probe signal was passed through an erbium-doped fibre amplifier (EDFA) to an optical sampling oscilloscope having ~1 ps resolution, and to a spectrum analyser.
The gain response could be taken directly using the set-up in Fig. 1, without the highlighted section. To determine the phase evolution, a delay interferometer (DI) with a differential delay of 1bit period at 10.65 GHz was inserted after the SOA as shown in Fig. 1 (highlighted section). The phase response was calculated using both the gain response and output temporal waveform from the DI.
In order to extract the phase, it is necessary to employ both the gain data and the delayed interference setup data. The DI has two outputs, only one of which was used. The electric field E1 of the signal from the constructive interference port is given by.Eq. (2).Eq. (3). The factor of 4 in Eq. (3) ensures that T = 1 when G(t) = G0 and ∆φ(t) = 0.26]:
3. Experimental results
Gain and phase data were measured within the wavelength ranges allowed by the available filters and the EDFA. The points in the pump-probe measurements were separated by 0.42 ps. Figure 2(a) and Fig. 2(b) show the gain and phase evolution for five pump wavelengths with a fixed probe at 1550 nm. The dashed line in Fig. 2(a) marks the steady-state gain level. For this data, the pump pulse energy was within the range of 3 to 60 fJ, and was varied to maintain a constant amplitude modulation depth.
It can be seen from Fig. 2(a) that there was a marked increase in the recovery rate and an overshoot when the pump was to the blue of the probe (1535 nm and 1540 nm) compared to when the pump was to the red (1560 to 1570 nm). Measurements were taken for various fixed probe wavelengths (1545 to 1560 nm), always resulting in the same observed qualitative difference in recovery rates between pumping to the blue and to the red. Figure 2(b) shows the phase change recovery for a 1550nm probe and it shows that there was a larger ultrafast contribution with a blue-shifted pump compared to a red-shifted pump, although the total recovery time remained approximately the same.
The gain recovery times shown in Fig. 3 and Fig. 4 are the times for the gain to change from 10% to 90% of the full recovery (10/90 recovery) when compared to the steady state. This method of recovery time estimation avoids errors caused by fluctuations in the gain curve as full recovery is approached. Similarly, the trend in the phase recovery is followed by calculating a 10/90 recovery time, based on the time for the phase to change from 90% to 10% of the maximum phase shift. For all of the data in Fig. 3 and Fig. 4, we estimate an associated error of approximately ± 2.5 ps, as displayed in the figures. However, error bars are not attached to each point to avoid obscuring the trend in the data.
In Fig. 3(a), the 10/90 gain recovery times are plotted as a function of the pump-probe separation for the four fixed probe wavelengths (1545 to 1560 nm) for which there was data for pumping both to the red and to the blue. The zero position on the x-axis in Fig. 3(a) corresponds to equal pump and probe wavelengths, with the positive separation representing pumping red and the negative separation representing pumping blue. It is clear from Fig. 3a that the gain recovery time became longer as the pump moved further to the red. In Fig. 3(b), the graph of the 10/90 recovery times for both the gain and phase with a 1550 nm fixed probe demonstrates that the phase recovery rate approximately followed the same trend as that of the gain, within the limits set by the error bars.
In addition, the recovery time as a function of probe wavelength with the pump at a fixed wavelength was studied. In Fig. 4(a), the 10/90 gain response time as a function of the separation between the probe and the small signal gain peak (1565 nm) is shown, with positive and negative separation representing probing at longer and shorter wavelengths respectively. Note that in Fig. 4, the pump was always blue-shifted relative to the probe. An approximate parabolic shape to the curves was observed when the pump was between 1535 and 1545 nm, with a minimum located near the 0 nm separation. This demonstrates that the fastest response was achieved when the probe wavelength was close to the gain peak. When the pump was within 1550 to 1555 nm, the curves do not appear to be parabolic and the data was inconclusive.
In Fig. 4(b), the pump wavelength was kept to the blue of the probe and remained fixed for each separate curve while the probe wavelength was incrementally changed. An approximate parabolic shape to the curves was observed in when the pump was within the range 1535 to 1545 nm and indicates that there was an optimum pump-probe separation (20 to 30 nm) for obtaining a fast gain response at a given pump wavelength. However, this was not the case when the pump is within 1550 to 1555nm, where the data was inconclusive.
For several wavelengths of the pump and probe, having the optimum wavelength separation of ~30nm, the gain response was almost completely dominated by a large ultrafast component with just a small band-filling component manifested as an overshoot. The magnitude of the overshoot was low but it persisted for a long time (>60 ps). Some of these gain and phase evolution diagrams are displayed in Fig. 5 . As may be seen in Fig. 6(a) , the gain made a 10/90 recovery within 6 to 9 ps and a full recovery within 9 to 20 ps. These were remarkably short recovery times and were substantially shorter than both the recovery time of 100ps for the same SOA when the cw probe was non-saturating and the typical full recovery time of ~60 ps for fast SOAs [5,6,27]. The phase recovery time, however, was longer than that of the corresponding gain recovery, although there was a significant ultrafast component to the phase (Fig. 5(b)). Also, since the phase recovery took longer than the differential delay of the AMZI (94 ps), the full phase recovery could not be shown. The pump pulse energies for these results ranged from 60 to 130 fJ.
Using a 1535nm pump and a 1560 nm probe, and we observed the same very large fast responses of Fig. 5. Increasing modulation depth further gave the same response as in Fig. 5(a), but the 10/90 times were larger simply due to the increased depth from which the recovery started.
The SOA gain spectra when there was a cw input of −3.5 dBm within the wavelength range 1565 to 1575 nm along with the spectrum when there was no input are displayed in Fig. 6. The gain at the cw wavelength could not be measured, leading to discontinuities in the spectra. It is clear from Fig. 6 that the cw input strongly suppresses the gain and that the gain at wavelengths to the blue of the cw is more strongly suppressed.
4. Theory and modelling
The SOA model used was a multi-section time-domain model of the carrier dynamics within the device, similar to that described in , but with some additional features. The temporal and longitudinal variations of both carrier concentration and carrier temperature in response to the pump, probe and ASE powers were calculated using rate equations [28,29]. The approximation to the gain spectrum introduced in  was employed to enable the gain and ASE spectra to be represented efficiently as functions of both time and distance along the SOA.
The experimental device was modelled with CW probe having a power of −3.5 dBm and wavelength 1550 nm and with a pump either to the blue (1535 nm) or to the red (1565 nm). The pump pulse energies at the SOA input facet were 17 fJ and 0.7 fJ respectively, chosen to achieve the same gain compression as the corresponding experimental results (50%). Figure 7 shows the carrier density variation along the amplifier length with and without the probe. With no input, the forward and backward travelling ASE reached their maximum levels at each end of the amplifier where they caused symmetrical reductions in the carrier density. With the input probe present, the forward travelling power was enhanced and the carrier density beyond the first quarter of the SOA was further progressively reduced, until the modal gain per unit length at the probe wavelength was depressed to near zero by the end of the device (Fig. 8(a) ). The backward travelling ASE was also reduced but still caused a small dip in the carrier density close to the SOA input.
As a result of the saturation induced by the strong probe, the pump pulses experienced decreasing gain as they travelled along the SOA (Fig. 8(a)). With the blue-shifted pump, the gain became negative beyond the midpoint of the SOA and the pulses were attenuated. With the red-shifted pump, on the other hand, the pulses continued to grow up to the end of the SOA, albeit at a reduced rate (Fig. 8(b)). In the latter case, smaller pulse energies were sufficient to obtain the same gain compression.
With the blue-shifted pump, the behaviour of the device was similar to that of a Turbo-Switch , which employs a filter between two SOAs to block the onward transmission of the pump. Here, the pump was effectively filtered out in the second section of the SOA because the carrier density was below the transparency point for the pump wavelength. In the first section, where the pump experienced gain, cross-gain modulation (XGM) of the probe by the pump occurred. In the second section, where the pump was absorbed, self-gain modulation (SGM) of the probe signal took place without XGM. Thus, the gain compression in the first section was partially opposed by a gain increase in the second that compensated for the slow band-filling tail in the recovery.
When the pump was red-shifted, however, it was not absorbed and therefore gave rise to XGM throughout the entire length of the SOA. Although SGM took place to some extent, it was not sufficient to eliminate the slow tail of the response. This is illustrated in Fig. 9 which shows the evolution of the change in carrier density as a function of time at several points along the active region length. When the pump was blue-shifted, the carrier density change quickly became positive from the centre of the SOA onwards, whereas when the pump was red-shifted, the carrier density change was always negative except at the input facet (where the influence of backward-travelling ASE predominated).
The normalized probe gain was plotted for both pump wavelengths and compared with the corresponding experimental measurements in Fig. 10 . With the blue-shifted pump, the response resembled that of the Turbo-Switch in showing a rapid recovery succeeded by an overshoot. It is interesting to note that, although the probe amplitude had returned to its original level following the overshoot by t = 200 ps, 125 ps after reaching its minimum value, the carrier density was still far from equilibrium at that time (Fig. 9). Clearly the increase in carrier density near the midpoint of the SOA was successfully compensating for the reduction in density in the preceding section.
Finally, the good agreement between the modelled and measured probe gain gives confidence that the foregoing explanation of the observed behaviour of the device is correct. Further numerical analysis of the variation of the SOA recovery time as a function of pump-probe and probe-gain peak separation is ongoing.
In summary, the gain and phase dynamics of a long (2.2 mm) non-linear SOA with high gain have been investigated. The recovery rates were found to be wavelength dependent and there was a marked difference in the recovery profiles between pumping to the blue and pumping to the red of the cw probe wavelength.
When the pump was blue-shifted and the pump-probe separation was optimized (~20 to 30 nm), we observed gain recovery curves almost entirely dominated by an ultrafast component, accompanied by an overshoot. Although phase recovery times were slower, they followed the same trend as a function of pump-probe wavelength separation. When the pump was red-shifted, the gain and phase recovery rates were slower, the ultrafast component was reduced and there was no appreciable gain overshoot. Computational modelling of the SOA explained these observations in terms of behaviour similar to that of a Turbo-Switch when pumping to the blue, where the pump was effectively filtered out within the SOA. We believe this is the first time such an ultrafast gain response from a single SOA has been reported, without the use of a filter or an additional SOA. Interestingly, the optimum behaviour of this device was similar to that of an ideal quantum dot (QD) SOA regarding the dominance of the ultrafast component in the gain recovery , although a QD SOA generally requires a higher bias current than did this MQW SOA. As in the case of the QD SOA, the phase response was slower than that of the gain. This does mean that an interferometer based switch will not demonstrate the ultrafast response we observe in the gain.
This work was supported by Science Foundation Ireland under grant 06/IN/I969.
References and links
1. D. Cotter, R. J. Manning, K. J. Blow, A. D. Ellis, A. E. Kelly, D. Nesset, I. D. Phillips, A. J. Poustie, and D. C. Rogers, “Nonlinear optics for high- speed digital information processing,” Science 286(5444), 1523–1528 (1999). [CrossRef] [PubMed]
2. W. Mathlouthi, F. Vacondio, P. Lemieux, and L. A. Rusch, “SOA gain recovery wavelength dependence: simulation and measurement using a single-color pump-probe technique,” Opt. Express 16(25), 20656–20665 (2008). [CrossRef] [PubMed]
3. L. Occhi, Y. Ito, H. Kawaguchi, L. Schares, J. Eckner, and G. Guekos, “Intraband gain dynamics in bulk semiconductor optical amplifiers: measurements and simulations,” IEEE J. Quantum Electron. 38(1), 54–60 (2002). [CrossRef]
4. R. Giller, R. Manning, and D. Cotter, “Gain and phase recovery of optically excited semiconductor optical amplifiers,” IEEE Photon. Technol. Lett. 18(9), 1061–1063 (2006). [CrossRef]
5. F. Girardin, G. Guekos, and A. Houbavlis, “Gain recovery of bulk semiconductor optical amplifiers,” IEEE Photon. Technol. Lett. 10(6), 784–786 (1998). [CrossRef]
6. R. Gutierrez-Castrejon, L. Schares, L. Occhi, and G. Guekos, “Modeling and measurement of longitudinal gain dynamics in saturated semiconductor optical amplifiers of different length,” IEEE J. Quantum Electron. 36(12), 12 (2000). [CrossRef]
7. L. Schares, C. Schubert, C. Schmidt, H. Weber, L. Occhi, and G. Guekos, “Phase dynamics of semiconductor optical amplifiers at 10-40 GHz,” IEEE J. Quantum Electron. 39(11), 1394–1408 (2003). [CrossRef]
8. A. Uskov, J. Mork, and J. Mark, “Theory of short-pulse gain saturation in semiconductor laser amplifiers,” IEEE Photon. Technol. Lett. 4(5), 443–446 (1992). [CrossRef]
9. R. Giller, R. Manning, and D. Cotter, “Recovery Dynamics of the'Turbo-Switch',” Optical Amplifiers and Their Applications (Optical Society of America), paper OTuC, Whistler (2006).
10. J. Leuthold, R. Ryf, D. Maywar, S. Cabot, J. Jaques, and S. Patel, “Nonblocking all-optical cross connect based on regenerative all-optical wavelength converter in a transparent demonstration over 42 nodes and 16800 km,” J. Lightwave Technol. 21(11), 2863–2870 (2003). [CrossRef]
12. R. Runser, D. Zhou, C. Coldwell, B. Wang, P. Toliver, K. Deng, I. Glesk, and P. Prucnal, “Interferometric ultrafast SOA-based optical switches: From devices to applications,” Opt. Quantum Electron. 33(7/10), 841–874 (2001). [CrossRef]
13. J. Sokoloff, P. Prucnal, I. Glesk, and M. Kane, “A terahertz optical asymmetric demultiplexer (TOAD),” IEEE Photon. Technol. Lett. 5(7), 787–790 (1993). [CrossRef]
14. G. Talli and M. Adams, “Gain recovery acceleration in semiconductor optical amplifiers employing a holding beam,” Opt. Commun. 245(1-6), 363–370 (2005). [CrossRef]
15. H. Yu, D. Mahgerefteh, P. Cho, and J. Goldhar, ““Optimization of the frequency response of a semiconductor optical amplifier wavelength converter using a fiber Bragg grating,” Lightwave Technology,” Journalism 17, 308–315 (2002).
16. Y. Liu, E. Tangdiongga, Z. Li, H. de Waardt, A. Koonen, G. Khoe, X. Shu, I. Bennion, and H. Dorren, “Error-free 320-Gb/s all-optical wavelength conversion using a single semiconductor optical amplifier,” J. Lightwave Technol. 25(1), 103–108 (2007). [CrossRef]
17. B. M. K. Stubkjaer, T. Durhuus, G. G. Joergensen, C. Joergensen, T. N. Nielsen, B. Fernier, P. Doussiere, D. Leclerc, and J. Benoit, “Semiconductor optical amplifiers as linear amplifiers, gates and wavelength converters,” in Proc. European Conference on Optical Communications (Zurich, 1993).
18. J. Wiesenfeld, B. Glance, J. Perino, and A. Gnauck, “Wavelength conversion at 10 Gb/s using a semiconductor optical amplifier,” IEEE Photon. Technol. Lett. 5(11), 1300–1303 (1993). [CrossRef]
19. R. Manning, D. Davies, S. Cotter, and J. Lucek, “Enhanced recovery rates in semiconductor laser amplifiers using optical pumping,” Electron. Lett. 30(10), 787–788 (1994). [CrossRef]
20. R. Manning, X. Yang, R. Webb, R. Giller, F. Gunning, and A. Ellis, “The'Turbo-Switch'-a novel technique to increase the high-speed response of SOAs for wavelength conversion,” presented at Optical Fiber Communication Conference (Optical Society of America), paper OWS8, Anaheim (2006).
21. N. S. Patel, K. L. Hall, and K. A. Rauschenbach, “Interferometric all-optical switches for ultrafast signal processing,” Appl. Opt. 37(14), 2831–2842 (1998). [CrossRef]
22. L. Zhang, I. Kang, A. Bhardwaj, N. Sauer, S. Cabot, J. Jaques, and D. Neilson, “Reduced recovery time semiconductor optical amplifier using p-type-doped multiple quantum wells,” IEEE Photon. Technol. Lett. 18(22), 2323–2325 (2006). [CrossRef]
23. L. Zhang, I. Kang, A. Bhardwaj, N. Sauer, S. Cabot, J. Jaques, and D. Neilson, “Significant reduction of recovery time in semiconductor optical amplifier using p type modulation doped MQW,” (IEEE, 2009), pp. 1–2.
24. K. Hall, J. Mark, E. Ippen, and G. Eisenstein, “Femtosecond gain dynamics in InGaAsP optical amplifiers,” Appl. Phys. Lett. 56(18), 1740–1742 (1990). [CrossRef]
25. J. Mark and J. Mørk, “Subpicosecond gain dynamics in InGaAsP optical amplifiers: Experiment and theory,” Appl. Phys. Lett. 61, 2281–2283 (1992). [CrossRef]
26. M. Eiselt, W. Pieper, and H. Weber, ““SLALOM: Semiconductor laser amplifier in a loop mirror,” Lightwave Technology,” Journalism 13, 2099–2112 (1995).
27. J. Dong, X. Zhang, J. Xu, and D. Huang, “Filter-free ultrawideband generation based on semiconductor optical amplifier nonlinearities,” Opt. Commun. 281(4), 808–813 (2008). [CrossRef]
28. G. Talli and M. Adams, “Gain dynamics of semiconductor optical amplifiers and three-wavelength devices,” IEEE J. Quantum Electron. 39(10), 1305–1313 (2003). [CrossRef]
29. G. Talli and M. Adams, “Amplified spontaneous emission in semiconductor optical amplifiers: modelling and experiments,” Opt. Commun. 218(1-3), 161–166 (2003). [CrossRef]
30. J. Leuthold, M. Mayer, J. Eckner, G. Guekos, H. Melchior, and C. Zellweger, “Material gain of bulk 1.55 m InGaAsP/InP semiconductor optical amplifiers approximated by a polynomial model,” J. Appl. Phys. 87(1), 618–620 (2000). [CrossRef]
31. P. Borri, W. Langbein, J. Hvam, F. Heinrichsdorff, M. Mao, and D. Bimberg, “Spectral hole-burning and carrier-heating dynamics in InGaAs quantum-dot amplifiers,” IEEE J. Sel. Top. Quantum Electron. 6(3), 544–551 (2000). [CrossRef]