SEMICONDUCTOR optical amplifiers (SOAs) have been. Two-Wave Competition in Ultralong Semiconductor Optical Amplifiers

Transcription

1 1260 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 41, NO. 10, OCTOBER 2005 Two-Wave Competition in Ultralong Semiconductor Optical Amplifiers Gero Bramann, Hans-Jürgen Wünsche, Ulrike Busolt, Christian Schmidt, Michael Schlak, Bernd Sartorius, and Hans-Peter Nolting Abstract The copropagation of two waves in an ultralong semiconductor optical amplifier (SOA) is considered in theory and experiment. One wave is a modulated signal, whereas the other one is unmodulated (continuous wave). The theory bases on a comprehensive traveling-wave model and predicts an exponential improvement of the signal extinction ratio (ER) of the modulated signal, caused by the presence of the unmodulated signal. Conditions for achieving this two-wave competition (TWC) effect are as follows. The SOA is operated under saturation, both waves are copolarized, they have comparable gain and their spectral correlation is between certain limits. The TWC effect is due to nondegenerate four-wave mixing (FWM) in the saturated part of a long SOA and is expected to have a high-speed potential. In order to check the theoretical predictions, 4-mm-long SOAs are developed and experimentally investigated under the given conditions. The measured ER improves by 1.3 db for a 5-GHz sinusoidal signal, which compares well with the 2 db theoretically predicted for this configuration. FWM is identified also experimentally as the basic mechanism. Variation of wavelength detuning, pump current, modulation frequency and ER of the injected signal are used to determine optimum conditions for the given device. Index Terms Four-wave mixing (FWM), nonlinear wave propagation, optical saturation, optical signal processing, semiconductor optical amplifiers (SOA),. I. INTRODUCTION SEMICONDUCTOR optical amplifiers (SOAs) have been used for multiple purposes in optical signal processing, exploiting mostly cross-gain modulation (XGM) or cross-phase modulation (XPM). Important examples are wavelength conversion [1], [2] and switching in interferometer arrangements [3] [5]. Most applications use SOAs shorter than 1 mm because amplification, XGM, or XPM saturate in longer devices [cf. Fig. 1(a)]. Only recently, an ultralong SOA (ULSOA) has been used as nonlinear element in a 3R-regeneration circuit at 40 Gb/s [6]. This application exploits the saturation of power and carrier Manuscript received February 14, 2005; revised June 14, This work was supported in part by the Bundesministerium für Bildung und Forschung (BMBF) in the frame of the MultiTeraNet project. G. Bramann, C. Schmidt, M. Schlak, B. Sartorius and H.-P. Nolting are with the Fraunhofer Institute for Telecommunications, Heinrich-Hertz-Institut, Einsteinufer 37, Berlin, Germany ( gero@bramann.com; Ulrike.Busolt@fh-furtwangen.de; christian.schmidt@hhi.de; schlak@hhi.fhg.de; sartorius@hhi.fhg.de; nolting@hhi.de). U. Busolt is with the Fachhochschule Furtwangen, Hochschule für Technik und Wirtschaft, Villingen-Schwenningen, Germany ( Ulrike.Busolt@fh-furtwangen.de). H.-J. Wünsche is with the Humboldt University of Berlin, Physik Photonik, Berlin, Germany ( ede@physik.hu-berlin.de). Digital Object Identifier /JQE density beyond 1 mm for equalizing, reshaping, and jitter reduction of degraded signals by properly superposing them with high quality pulse trains of an optical clock. This paper considers a different and new effect which takes place especially in the saturated part of an SOA [7] [9]. Configurations are regarded, where a modulated signal is injected together with a continuous wave (CW). It will be demonstrated in theory and for the first time also in experiment that, under appropriate conditions, the extinction ratio (ER) of the modulated signal improves during the copropagation along the device [Section II-B, cf. Fig. 1(b)]. This effect occurs on the modulated input wavelength and is not a XGM. Our theoretical analysis (Section II) attributes it to two-wave competition (TWC) caused by four-wave mixing (FWM). This type of competition between two waves leads to a suppression of always the weaker wave in favor of the stronger one, in analogy to an electrical differential amplifier. The TWC in ULSOAs has a high-speed potential because the FWM behind is based to a large extend on fast intraband effects such as spectral hole burning or carrier heating [2], [10], [11]. Aiming at a proof of principle, we restricted our experiments to sinusoidal modulation of the signal wave. It should be noted however, that modeling has already demonstrated that TWC acts in a similar manner with realistic bit sequences, if the ULSOA is additionally stabilized by an optically clocked scheme [7], [8]. In particular, the 1 bits increase to the well defined level of the saturation power, and the noise and crosstalk on the 0 position is suppressed. Especially this second functionality can not be achieved easily with other effects or devices, and that makes TWC a very interesting and promising technique. Several applications of TWC for signal processing are proposed in [7] and [8]. Section III presents a first experimental verification of the TWC effect with modulated sinusoidal signals using an ULSOA with a length of 4 mm. Fabrication and basic characterization of these specially designed devices are outlined in Section III-A, followed by a brief description of the experimental setup in Section III-B. To our knowledge, TWC has not been proven in dynamic experiments yet. Therefore, the existence and behavior of the TWC effect are analyzed in Sections III-C and III-D, respectively. The aim of the dynamic measurements of the TWC is to validate the simulations and to characterize the effect in experiments. Another aspect is to find out the optimal conditions for the TWC. The impact of the driving SOA current and of the spectral correlation between the two injected waves are analyzed in Section III-E. The influence of the input ER and the /$ IEEE

2 BRAMANN et al.: TWC IN ULSOAs 1261 TABLE I PARAMETER VALUES Fig. 1. Calculated copropagation of a signal with a CW in a ULSOA. (a) Mean power and carrier density versus position. (b) Evolution of the ER along the device for sinusoidal signals with different frequencies. Input: 1 and 1.8 mw mean power in signal and CW, respectively. The values at position z equal those at output of a SOA with length L = z. used frequency are analyzed as well. All together, the results provide a clear picture of the performance of the TWC. II. THEORY A. Model Equations The features of TWC are first studied in theoretical calculations. For this purpose, the laser model of [12] is adapted to the present SOA configuration. It bases on the following traveling-wave equations (TWE) for the slowly varying amplitudes of linearly polarized optical waves, which counterpropagate with group velocity forward and backward along the longitudinal axis of the device The waves are generally driven by spontaneous emission as well as by optical injection. Spontaneous emission is treated by the Langevin forces as described in [12]. Optical injection will be described by boundary conditions in (5). Although only unidirectional operation is considered, the reverse wave is taken into account to properly describe the amplified spontaneous emission (ASE) which travels in both directions. Properties of the bulk active waveguide enter (1) by the complex wave number where is the linewidth enhancement factor, is the spectral maximum of material gain, is with nonlinear saturation, is the internal optical losses, and is the dispersion contribution as specified in [12]. The dependence of the gain on the carrier density and its nonlinear saturation are described by the simple standard model (1) (2) (3) where is the gain slope coefficient, is the transparency carrier concentration, and is the nonlinear gain saturation coefficient due to fast intraband processes like carrier heating and spectral-hole burning [10], [11]. The quantities and include the optical confinement factor. The system of equations is completed by the carrier rate equation where is the injection current per SOA length, is the cross section of the active zone, and is the polynomial recombination law. By assuming proper normalization, is the local photon density. B. Results of Simulation Calculations In order to numerically solve the described model equations, we use the program suite LDSL [13]. The optical injection into the SOA at facet is described by the boundary condition The indexes 0 and 1 represent the CW and the signal, respectively. Harmonic modulation of the signal intensity with period and ER is assumed. At the opposite facet it holds. Before regarding TWC, we have checked whether the model yields basic characteristic of the devices measured in regimes close to the targeted power saturation. With the parameter set given in Table I, we could reasonably well reproduce measured spectra of ASE, small signal gain, and FWM, as well as their dependence on current. In particular, this procedure yielded a reasonable estimate of, which is the ratio between spontaneous emission rate into a guided wave and recombination rate. This quantity is the only free parameter of the noise sources in (1) [12]. (4) (5)

3 1262 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 41, NO. 10, OCTOBER 2005 Fig. 1 displays spatially resolved results obtained with these parameters for the TWC in an ULSOA. Mean carrier densities and power saturate after about 1 mm. Everywhere in the saturated part, the calculated intensity of the backward propagating wave (not depicted) is negligibly small. The SOA operates unidirectionally here and the forward amplitude at position does not depend on the total length of the device. The plotted curves are thus representative for SOAs of any length beyond 1 mm and the values at are equal to those at the output of a SOA with length. ASE noise is known to limit the performance of SOAs. This effect may become strong in very long devices, because each part of the device contributes spontaneous emission. However, our simulations show that the contribution of the saturated part of the SOA to the forward ASE is negligible up to several 10-mm lengths. This result is easily understood when regarding the device as cascade of two different SOAs, a short one with high gain and a long, saturated one with zero gain. It is well known that the highly amplifying initial component of such an amplifier cascade dominates the noise figure [14]. This argumentation holds because the SOA is saturated by the injected waves but not by the ASE itself. Irrespective of the saturation beyond 1 mm, the ER of the signal still increases with the length of the saturated part [Fig. 1(b)]. This increase is nearly exponential with only slight effects of the modulation frequency. After 9 mm, the improvement is about 13, 11, and 8.5 db for signal frequencies 5, 40, and 80 GHz, respectively. Possible applications of this high-speed effect have been discussed elsewhere [8] on base of analogous simulations with realistic bit sequences, which yielded similar results. In the following we focus, therefore, on a deeper physical understanding and on experimental evidence. C. Analysis of Wave Competition Under Complete Saturation Now we present a closer inspection of the two-wave copropagation in the saturated part of an SOA. Our approximate treatment of this situation will show that nondegenerate FWM is the physical effect behind the observed improvement of the ER. Nondegenerate FWM in general is the creation of secondary waves from the superposition of two primary waves with different frequencies. The frequencies of the additional waves are separated from those of the primary ones by integer multiples of the detuning [Fig. 6(c)]. The well developed theory of FWM in SOAs (cf. e.g., [11] and [15]) is applied now in the framework of our TW model. Ideal conditions are assumed, which account for the situation obtained in the simulation: complete saturation, negligible noise contributions, negligible backward traveling-waves, equal gain for both primary waves, detuning large compared to the signal modulation frequency. The forward propagating optical field is expressed as a superposition of equally spaced partial waves with slowly varying amplitudes. The intensity of such a multiwave mixture contains beating contributions oscillating (6) with and multiples of it. They drive carrier density pulsations [11], [15], which by (2) transform into according beating oscillations of the wave number. The latter one couple the different partial waves and cause the FWM. In addition, the intensity beating in the nonlinear gain saturation (3) contributes directly to the same effect. FWM theory usually considers the interaction of a weak probe wave with a strong pump wave and focusses to the question how much power is transferred by FWM from the probe wave to the FWM products. In contrast, here we ask how the amplitudes of two primary waves evolve along the SOA, which have comparable magnitude at the input. To our best knowledge, this problem has not been theoretically investigated, yet. In order to get equations for the primary amplitudes, expression (6) is inserted into the TW model (1) (4). Standard steps like linearization with respect to the beating terms, averaging over the short beating period, and restricting to leading order in the primary amplitudes yield and the abbreviations is the photon density in wave (7) (8) (9) (10) are introduced with being and the carrier densities averaged over the beating period. The second term in describes the contributions of induced carrier density modulations to FWM [11], [15]. It decreases rapidly with increasing detuning. Although with present parameters, this term still contributes and gives rise to asymmetries between the two primary waves, which also appear in the optical spectra (cf. Fig. 6). In order to keep the analysis simple, we neglect this contribution now and confine ourselves to the limit of large detuning. By transforming to a frame of coordinates moving with the waves,, and taking the real part of the difference between (7), we get (11) This equation depends only parametrically on the time at which the waves entered the saturated SOA part at. Thus, the ratios of intensities entering at different times evolve independently of each other. Whether they increase or decrease depends only on the intensity ratio at the input. The ratio increases when being positive at the input, because the right-hand side is positive and keeps positive. Reversely, negative input ratios decrease. Thus, that wave injected with smaller intensity is further suppressed relative to the stronger

4 BRAMANN et al.: TWC IN ULSOAs 1263 Fig. 2. Extinction ratio of the signal wave versus relative position. Solid line: calculated from analytic approximation (12) (14), supposing signal and CW waves of equal mean intensities. The symbols illustrate the case of Fig. 1: an improvement of the ER from 10 to 20 db requires a propagation over 2.4 L. one. These conclusions hold irrespective of the concrete spatiotemporal variation of. Only is required, i.e. pumping above transparency. Under complete saturation, it holds, where is the saturation photon density with. Furthermore, the gain equals the losses,. Using these relations, (11) is easily transformed into an equation for the relative intensity difference. Integration by separation of variables yields with and (12) (13) (14) Beyond, approaches, i.e., the wave with larger input intensity approaches the saturation level, whereas the other one dies out exponentially with the characteristic decay length. The time, at which the waves have been injected into the saturated SOA, enters the right-hand side of (12) only via the input value. Thus, the evolution of the signal extinction along the device can be obtained by applying (12) to the extrema of. Fig. 2 displays the resulting increase of the signal extinction in case of waves with equal mean intensities. With present parameters we find mm. Thus, the improvement of the extinction by 10 db indicated in the figure requires about 8.7 mm propagation through a completely saturated SOA, in very good agreement with the full numerical simulation presented in Fig. 1. We conclude this section with remarks on the speed potential of these effects. The derived analytic formulae are independent of the modulation rate of the injected signal. Of course, their validity is limited by the applied approximations. In particular, the wave amplitudes must change slowly compared to the period of beating between both waves. In other words, the frequency detuning between the waves must be much larger than the modulation rate. On the other hand, the possible detuning is limited by the known decrease of the gain saturation coefficient in the Fig. 3. Photograph of the fabricated ULSOA module. terahertz range of detuning (see, e.g., [11]). Nevertheless, the improvement of the ER by about 1 db/mm remains valid up to modulation frequencies of some hundred gigahertz. This estimate underlines the high-speed potential of the TWC effect. III. EXPERIMENT A. Fabrication of ULSOA Device and Module SOAs are commercially available, but only with lengths up to 1 mm. In conventional operation, longer devices are not reasonable due to the saturation of the amplification [16] and due to strong patterning effects that degrade signals traveling through such long SOAs [6]. However, for the present investigation of the new TWC effect we need ULSOAs, as was shown in the previous chapter. Consequently, devices with a length of several millimeters have to be fabricated. The according wafer was realized in the InGaAsP material system using a buried laser structure. A strained bulk heterostructure was applied for achieving polarization insensitive operation in the 1550-nm range. The suppression of reflections from the facets is a very critical issue in long SOAs. We combined a two layer TiO SiO antireflection coating with the tilted facet technique. The ASE ripple thus was suppressed to less than 1 db with the waveguide having an angle of 12 relative to the normal of the facet. The light emits with an angle of 42 relative to the straight case. Spot size converters were not applied for the present wafer since we wanted to have the option to adjust various device lengths just by cleaving. The ULSOA used in this paper had a length of 4 mm, which appeared as a good compromise between device preparation and operation on one hand and detectable TWC effect on the other hand. Multiple bonding is used for achieving a uniform pumping and a homogeneous carrier density distribution. The ULSOA is packaged into a module for the following investigations. A photograph of the module is shown in Fig. 3. The fiber to fiber small signal gain is about 18 db, which seems to be relatively low for such a long device. We attribute this fact to saturation and to the high coupling losses caused by the large tilt and the lack of mode converters. B. Experimental Setup The TWC in the ULSOA was investigated by injecting simultaneously a sinusoidal modulated signal and a constant signal into the device (Fig. 4). The mean power of the modulated signal and the power of the constant signal were both about 2 dbm

5 1264 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 41, NO. 10, OCTOBER 2005 Fig. 4. Experimental setup. A CW and a modulated signal travel through an ULSOA. The effect of polarization and wavelength correlation on the ER is analyzed using a scope. (in the fiber). The maximum of the modulated signal thus had higher power than the constant signal, the minimum had lower power. External cavity lasers (ECL) were used as sources and both wavelengths thus could be varied. For most of the following investigations the modulated signal was at 1569 nm and the constant signal at 1573 nm, both close to the gain maximum (1570 nm) of the SOA and within the flat part of the gain spectrum. The effect of the TWC on shape and ER of the modulated signal traveling through the ULSOA was analyzed using a sampling oscilloscope and comparing input and output pulse shape. The modulation used in the experiments was sinusoidal with variable ER. It was generated using an electro-absorption modulator, driven by a synthesizer and operated at matched bias voltage. Though we propose an ultrahigh-speed potential of the TWC effect we used a moderate modulation frequency of 5 GHz during the experiments. Target of the measurements was the analysis of small improvements in the ER of the sinusoidal signal, and toward this it was important to be not limited by the bandwidth of the detection system. In our setup we applied photodiodes XPDV 2020R of u2t and a sampling oscilloscope A of Agilent, both with a bandwidth of about 40 GHz and thus not broadening the pulses of our 5-GHz signals. The relative polarization of modulated and constant signal could be adjusted by means of polarization controllers. The different relative polarization states could be detected easily using the optical spectrum analyzer behind the ULSOA. The parallel state of polarization was identified by the strongest FWM components in the spectrum [2], while the orthogonal state was obtained by minimization of the FWM lines. C. Demonstration of TWC in Dynamic Measurements In the following series of measurements, a modulated signal with an ER of 8 db, left panel of Fig. 5, was injected into the ULSOA. First the constant wave was switched off. The according output trace is also shown in Fig. 5. The output modulation is flat with a few weak structures in it. One can attribute the flattening to saturation: the saturation power (8.5 dbm, measured in the fiber) is approached independently of the input power, not only for maxima but also for the minima of the modulation. The remaining structures are related to the finite gain recovery time in the ULSOA. Obviously, and in accordance with the previous knowledge, a long SOA makes no sense for amplifying or processing of signals in conventional operation. The situation changed significantly when injecting additionally a wavelength-detuned CW. First we investigated the Fig. 5. Conventional operation of the 4-mm ULSOA (injection current 1.2 A) for one signal. Left: 5-GHz sinus input signal at 1569 nm, ER 8 db. Right: output signal, flattening, and degradation due to gain saturation and patterning effects can be observed. Powers are normalized to the respective mean power. Fig. 6. Impact of an additional CW injection at 1573 nm onto the 1569-nm sinus signal traveling through the ULSOA. (a) (b) Orthogonal polarization. (c) (d) Parallel polarization. (a) and (c) Optical spectra. (b) and (d) Pulse trace at input (thin) and output (thick). The significant improvements from (b) to (d) are attributed to the TWC effect that is caused by FWM, detectable in panel (c). characteristics for orthogonal polarized waves. That state can be noticed by the minimized weak FWM lines in the optical spectrum [Fig. 6(a)]. FWM is known as a strongly polarization sensitive effect, but detailed analysis identifies weak contributions also for orthogonal polarized waves [17]. The latter becomes visible in the optical spectrum for long SOAs, like in Fig. 6(a). However, the efficiency of FWM is extremely different for parallel and orthogonal polarization. According to our theory TWC is based on FWM, and thus no significant effect of TWC is expected for the case of orthogonal polarized waves. The additional constant wave here has only the function of a holding beam, which is known to improve the gain recovery time and to reduce patterning effects [18]. The result with our 4-mm ULSOA is shown in Fig. 6(b). Due to the holding beam, a distinctly modulated signal at the ULSOA output was obtained now. However, compared to the input signal, the ER was decreased and the shape was degraded. Again this degradation can be attributed to gain saturation and gain recovery processes. The device is still not useful. At last, the two signals were adjusted for parallel polarization. Strong FWM components in the optical spectrum [Fig. 6(c)] indicate that, and one can expect impacts of the TWC effect now. The output signal measured in this configuration is shown in Fig. 6(d). A clear

6 BRAMANN et al.: TWC IN ULSOAs 1265 Fig. 7. Normalized input and output signals in log scale, allowing for a direct graphic determination of improvement of the ERs. (a) Experimental data of Figs. 5(a) and 6(d). (b) Simulation results of Part II. sinusoidal pulse trace with a good ER can be observed. The additional constant wave injected at parallel polarization had obviously a strong and positive impact on the performance of the modulated signal traveling through the ULSOA. These findings can be attributed only to the effect of the TWC. The significantly different output pulse shapes obtained with an orthogonal polarized holding beam [Fig. 6(b)] on one hand and with a parallel polarized FWM beam on the other hand [Fig. 6(d)] thus represent the first dynamic verification for the existence and function of the TWC effect. D. Analysis of the Regenerative Feature of TWC An important intended application of the TWC effect is optical signal regeneration. In particular, the improvement of the eye opening is the target. One may assume an eye with the 1 bits having the amplitude of the maxima of the sinus signal used here, and the 0 bits having noise and crosstalk equivalent to the power in the minima of the sinus. The improvement in the ER of the sinus test signal applied in our experiment thus can be taken as a measure for the eye opening function of TWC, the suppression of the noise level relative to the 1 bit level. The ER improvement thus has to be determined in detail. For this purpose the measured input and the output signals were normalized to their maxima and both were plotted in a logarithmic scale in Fig. 7(a). Comparing the minima of both one can notice a suppression of 1.3 db for the output relative to the input. For comparison with theoretical predictions the TWC for a device with similar parameters and with similar operating conditions was calculated. Results are also summarized in Fig. 7(b). The improvement obtained by TWC here is 2 db, slightly better than measured in the experiments, but close enough to support the experimental results and to give confidence in the theoretical calculations. Most important, however, is the verification of an ER enhancement of the signal traveling through the SOA. To the best of our knowledge, TWC is the only effect that predicts such a behavior. An eye opening improvement of db may not appear as sufficient for a good signal regenerator. However, the TWC effect increases significantly with the length of the ULSOA. Looking to the theoretical curve in Fig. 1, doubling the length Fig. 8. Impact of different operating conditions on the performance (ER improvement) of the TWC in the 4 mm ULSOA. The investigated parameters are: (a) detuning of the wavelength of the CW wave relative to the signal, (b) pump current, (c) signal modulation frequency, and (d) ER of the signal at input. of the ULSOA to 8 mm will result in an eye opening improvement of more than 10 db (from 10 to 20 db) at 5 GHz and of 8 db (from 10 to 18 db) at 80 GHz. Assuming a similar difference between modeling and experiment as found for the 4-mm device one can estimate an ER improvement of 5 db at 80 GHz, which represents a promising regenerative function at this high speed. E. Analysis of Operating Conditions for TWC Finally, the impact of further operation conditions on TWC is studied. The results are summarized in Fig. 8. The ER improvement is taken here as a measure for the efficiency of the effect. Most important besides polarization is a proper spectral correlation between signal and CW wave. No significant ER improvement can be detected if the CW wavelength is shorter than that of the modulated signal [Fig. 8(a)]. This experimental observation can be attributed to the asymmetry of the FWM efficiency for up- and down-conversion [11], [15]. Thus, it is a prerequisite for achieving an ER improvement, to set the CW beam to the long wavelength side of the signal. The efficiency of TWC then decreases with the wavelength difference, similar as known for the originating FWM effect. The TWC effect requires a sufficiently high pumping level of the ULSOA. It increases with pumping and saturates at moderate current densities of about 300 ma/mm [Fig. 8(b)]. The modulation frequency of the signal wave has no significant impact, within the accessible range of frequencies up to 5 GHz [Fig. 8(c)]. This observation is in concordance with the high-speed potential, theoretically predicted in Section II. An important issue regards the range of signal ERs that can be improved by TWC. Results on that are summarized in Fig. 8(d). The ER improvement shows no significant dependency on the ER of the input signal. It keeps closely above 1 db within the large range from 4 to 14 db at input. We conclude the TWC operates even in strongly degraded signals with bad eye opening,

7 1266 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 41, NO. 10, OCTOBER 2005 and it has also the potential for further improving the quality of relative good signals. These features are highly favorable for possible practical applications. IV. SUMMARY We have studied theoretically and experimentally the novel effect of TWC that occurs if two wavelength-detuned waves copropagate through saturated SOAs. TWC results in the suppression of the weaker wave by the stronger one. The ER of a modulated wave thus can be improved by the presence of a CW with a power higher than the minimum and lower than the maximum of the modulation. Our theoretical calculations based on the LDSL program suite clearly identified FWM as origin of TWC. High speed potential was proposed, copolarized waves and ULSOAs were predicted as requirements for exploitation of the TWC. ULSOAs (4 mm) thus were fabricated for the experiments. A 5-GHz sinusoidal modulated test signal and a constant wave were simultaneously injected into the ULSOA and the ERs of the sinus before and behind the ULSOA were measured and used to characterize the TWC. Improvement of the ER could be detected in presence of a parallel polarized constant wave which means TWC to be effective, while degradation of the ER was observed in the case of an orthogonal polarized constant wave acting purely as a holding beam. That result represented the first direct and dynamic experimental verification on the existence and the regenerative characteristic of the TWC effect. In a series of further measurements, the characteristics of TWC were studied in detail. The constant wave has to be set on the long wavelength side to get the TWC effect on the modulated signal. The ER improvements by TWC were observed on strongly degraded input signals (ER 4 db) as well as for input signals with high ER of 14 db. These features together with the high-speed potential proposed by modeling to be similar to FWM makes TWC to a promising approach for applications in high-speed all-optical signal processing and regeneration. REFERENCES [1] A. D Ottavi, F. Girardin, L. Graziani, F. Martelli, P. Spano, A. Mecozzi, S. Scotti, R. Dall Ara, J. Eckner, and G. Gueckos, Four-wave mixing in semiconductor optical amplifiers: a practical tool for wavelength conversion, IEEE J. Sel. Topics Quantum Electron., vol. 3, no. 3, pp , Apr [2] S. Diez, C. Schmidt, R. Ludwig, H. G. Weber, K. Obermann, S. Kindt, I. Koltchanov, and K. Petermann, Four-wave mixing in semiconductor optical amplifiers for frequency conversion and fast optical switching, IEEE J. Sel. Topics Quantum Electron., vol. 3, no. 5, pp , Oct [3] S. Diez, E. Hilliger, M. Kroh, C. Schmidt, C. Schubert, H. G. Weber, L. Occhi, L. Schares, G. Guekos, and L. K. Oxenloewe, Optimization of SOA-based Sagnac-interferometer switches for demultiplexing to 10 and 40 Gb/s, Opt. Commun., vol. 189, pp , [4] Y. Ueno, S. Nakamura, and K. Tajima, Nonlinear phase shifts induced by semiconductor optical amplifiers with control pulses at repetition frequencies in the GHz range for use in ultrahigh-speed all-optical signal processing, J. Opt. Soc. Amer., vol. 19, no. 11, pp , [5] H. J. S. Dorren, G. D. Khoe, and D. Lenstra, All-optical switching of an ultrashort pulse using a semiconductor optical amplifier in a Sagnacinterferometric arrangement, Opt. Commun., vol. 205, pp , [6] J. Slovak, C. Bornholdt, U. Busolt, G. Bramann, H. P. Nolting, and B. Sartorius, Dynamic behavior of optically clocked 4 mm UL-SOA at 40 Gbit/s, in Proc. SPIE Integrated Optics and Photonic Integrated Circuits, vol. 5451, Strasbourg, France, 2004, pp [7] H.-P. Nolting and H.-J. Wünsche, Novel ultra-high speed switching technique for all-optical signal processing, in Proc. Integrated Photonics Research Conf., Monterey, Ca, Jun. 2001, Paper PDP 1. [8] H.-P. Nolting and B. Sartorius, Fast clocked decision a novel ultrahigh speed switching technique, in Proc. ECOC, 2002, Paper [9] U. Busolt, G. Bramann, H. J. Wünsche, C. Schmidt, B. Sartorius, H.-P. Nolting, and T. Rosin, Two-wave competition for all-optical signal processing: first experimental verification in ultra-long SOAs, in Proc. ECOC, vol. 4, 2003, Paper Th3.5.1, pp [10] A. Uskov, J. Mørk, and J. Mark, Wave mixing in semiconductor laser amplifiers due to carrier heating and spectral-hole burning, IEEE J. Quantum Electron., vol. 30, no. 8, pp , Aug [11] M. A. Summerfield and R. S. Tucker, Frequency-domain model of multiwave mixing in bulk semiconductor optical amplifiers, IEEE J. Sel. Topics Quantum Electron., vol. 5, no. 3, pp , May/Jun [12] H.-J. Wünsche, M. Radziunas, S. Bauer, O. Brox, and B. Sartorius, Modeling of mode control and noise in self-pulsating PhaseCOMB lasers, IEEE J. Sel. Topics Quantum Electron., vol. 9, no. 3, pp , May/Jun [13] M. Radziunas et al.. (2005, Apr.) LDSL: Tool for the simulation of longitudinal dynamics in semiconductor lasers. [Online]. Available: [14] D. M. Baney, P. Gallion, and R. S. Tucker, Theory and measurement techniques for the noise figure of optical amplifiers, Opt. Fiber Technol., vol. 6, pp , [15] G. P. Agrawal, Population pulsations and nondegenerate four-wave mixing in semiconductor lasers and amplifiers, J. Opt. Soc. Amer. B, vol. 5, no. 1, pp , [16] K. J. Blow, R. J. Manning, and A. J. Poustie, Model of longitudinal effects in semiconductor optical amplifiers in a nonlinear loop mirror configuration, Opt. Commun., vol. 148, pp , [17] R. Paiella, G. Hunziker, J. H. Zhou, K. J. Vahala, U. Koren, and B. I. Miller, Polarization properties of four-wave mixing in strained semiconductor optical amplifiers, IEEE Photon. Technol. Lett., vol. 8, no. 6, pp , Jun [18] R. J. Mannings and G. Sherlock, Recovery of a pi phase shift in 12.5 ps in a semiconductor laser amplifier, Electron. Lett., vol. 31, pp , Gero Bramann was born in Schotten, Germany, in He received the Diploma in electrical engineering from the University of Applied Science, Frankfurt, Main, Germany, in He received the MSc. degree in data communications from the University of Central England, Birmingham, U.K., in 2002 In 1998, he joined the telecommunication provider O2 and worked in the field of intelligent networks until After receiving the M.Sc. degree, he joined the Fraunhofer-Institut für Nachrichtentechnik Heinrich-Hertz-Institut (HHI) Berlin, Germany. In 2003, he started his research in special applications of SOAs. Since July 2004, he has been working for the German optical fiber manufacturer fiberware GmbH in research and development technology. Hans-Jürgen Wünsche was born in Nossen, Germany, in He received the Diploma, the Dr.rer.nat, and the Dr.sc.nat. degrees in physics from Humboldt University, Berlin, Germany, in 1972, 1975, and 1983, respectively. He is currently with the Humboldt University and has carried out theoretical research on tunneling and high excitation phenomena in semiconductors. Between 2001 and 2004, he also was with Fraunhofer Institute for Telecommunications, Heinrich-Hertz-Institute, Berlin. His current research interests concentrate on the nonlinear physics of photonic materials and devices. Dr. Wünsche is member of the German Physical Society.

8 BRAMANN et al.: TWC IN ULSOAs 1267 Society. Ulrike Busolt was born in Bochum, Germany, in She received the diploma in physics and the Ph.D degree in physics at the Free University Berlin, Berlin, Germany, in 1995 and 2000, respectively. From 2001 to 2004, she worked at the Fraunhofer Institute for Telecommunications, Heinrich-Hertz-Institute, Berlin. Since 2004, she has been Professor for medical engineering at Fachhochschule Furtwangen, Hochschule für Technik und Wirtschaft, Villingen-Schwenningen, Germany. Prof. Dr. Busolt is member of the German Physical Bernd Sartorius studied physics in Frankfurt and Berlin. He received the Ph.D. degree from the Technical University of Berlin, Berlin, Germany, in He then joined the Heinrich-Hertz-Institute for Telecommunications, Berlin, Germany, where he first worked on optical techniques for characterization of semiconductors. In 1991, he became head of a technological project developing semiconductor optical amplifiers. He focused the project work on devices for all-optical signal processing. Novel (patented) types of multisection lasers/amplifiers were developed and applied especially for clock recovery and decision in all-optical 3R signal regenerators. Presently, he is head of projects that cover the whole range from design and technology of semiconductor devices up to system experiments on high-speed all-optical signal processing. Christian Schmidt received the Diploma in electrical engineering at the Technical University, Berlin, Germany in He then joined the Heinrich-Hertz-Institut, Berlin. First, he was engaged in the research and development of all-optical switches for OTDM/WDM based on monolithically integrated interferometers. Currently, he is working on the technology and characterization of ultralong semiconductor optical amplifiers. Michael Schlak was born in Berlin, Germany, in He received the M.S. degree in physics from the Technische Universität Berlin, Berlin, Germany, in Since 1980, he has been working as Member of Research Staff, Heinrich-Hertz-Institut, Berlin. He became head of a project group on monolithically integrated all optical demultiplexers in He has diverse publications on epitaxy and integrated optics. He is experienced in crystal growth/epitaxy in the InP-system, InP-technology and measurement technologies, InP integration technology, design and development of InP based devices, measurement technologies for opto-electronic devices. Hans-Peter Nolting was born in Berlin, Germany, in He received the diploma in physics from the Free University of Berlin, Berlin, Germany, and the Ph.D. degree from Technical University of Berlin, Berlin, Germany, in 1967 and 1975, respectively. Since 1967 he is with the Heinrich-Hertz-Institute für Nachrichtentechnik Berlin GmbH. He has worked on microwave techniques, thin- and thick-film technique for high speed electronics, integrated optics components on LiNbO3 and InP, and particularly on monolithic integrated OEICs on InP. He is engaged in device modeling and characterization of guided wave devices like filters, switches and mode converters. He was Head of the Department of Components-Integration and is now Head of the Department of Optical Signal Processing. Work now is dedicated to OTDM- and all-optical signal processing techniques. Here OTDM-signal transmission experiments at 160 Gb/s and loop experiments on 3R-regeneration at 40 Gb/s have been performed successfully. He is currently engaged in modeling of all-optical subsystems like all-optical clock recovery, fast switches (including ULSOA), femtosecond-pulse sources, wavelength converters, and the architecture of 3R-regenerators and other all-optical circuits. He has been head of several national and international projects on optical communication technologies. He has served as an evaluator and reviewer for RACE, ACTS, and IST. He was involved as expert in the Visionary Research Groups Advanced Communications Technologies, Services, and Applications. He was engaged in several COST activities dedicated to devices and the evolution of optical networks. He has organized several workshops on optical devices and subsystems.