Simultaneous Amplification and Compression of Ultrashort Solitons in an Erbium-Doped Nonlinear Amplifying Fiber Loop Mirror

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 39, NO. 4, APRIL 2003 555 Simultaneous Amplification and Compression of Ultrashort Solitons in an Erbium-Doped Nonlinear Amplifying Fiber Loop Mirror Ping Kong A. Wai, Senior Member, IEEE, and Wen-hua Cao Abstract A simple technique for simultaneous amplification and compression of ultrashort fundamental solitons is proposed. It is based on an erbium-doped nonlinear amplifying fiber loop mirror. Numerical simulations show that, unlike conventional erbium-doped fiber amplifiers in which nonlinear effects lead to serious degradation of pulse quality, the proposed device performs efficient high-quality amplification and compression of ultrashort solitons while nearly preserving the soliton nature of the input pulses. We have also studied the effects of loop characteristics, nonsoliton input pulses, and higher order fiber effects on the device performance and show that the proposed scheme is fairly insensitive to small variations in both the loop and input pulse parameters. Index Terms Optical fiber amplifiers, optical pulse amplification, optical pulse compression, optical solitons, optical switching. I. INTRODUCTION ULTRASHORT pulse amplification is required in many fields such as ultrafast spectroscopy, optical signal processing, and soliton-based communication systems, etc., because of the relatively low output power levels of commonly used short pulse sources, such as mode-locked semiconductor and fiber lasers. Erbium-doped fiber amplifiers (EDFAs) are widely used for pulse amplification owing to their broad bandwidths ( 50 nm), high gains ( 40- db), and high pulse-saturation energies ( 1 J). For most applications of ultrashort pulses, the pulse quality is a key factor. It is, therefore, not sufficient to achieve a high gain, but the amplification process must also preserve the pulse quality, which is especially important for soliton-based communication systems. However, distortionless amplification of ultrashort soliton pulses in EDFAs is difficult if fiber nonlinearities such as self-phase modulation (SPM) and Raman self-scattering (RSS) are large [1] [5]. These nonlinear effects lead to undesirable pulse shaping and serious degradation of the pulse quality. Although adiabatic amplification (small gain) gives high-quality pulses, Manuscript received September 30, 2002. This work was supported by the Research Grant Council of the Hong Kong Special Administrative Region of China under Project PolyU5096/98E, by the National Natural Science Foundation of China under Project 60277016, and by the Guandong Natural Science Foundation of China under Project 021357. P. K. A. Wai is with the Photonics Research Center and Department of Electronic and Information Engineering, The Hong Kong Polytechnic University, Hung Hom, Hong Kong. W. Cao is with the Photonics Research Center and Department of Electronic and Information Engineering, The Hong Kong Polytechnic University, Hung Hom, Hong Kong, and also with the School of Information, Wuyi University, Guandong 529020, China. Digital Object Identifier 10.1109/JQE.2003.809327 the resulting energy gains are small [5], [6] and the amplifier length must increase exponentially with the input pulsewidth in order to satisfy the adiabatic condition. The difficulty can be overcome using the chirped pulse amplification (CPA) technique [7] [11] in which an input pulse is first stretched by a dispersive delay line to ensure linear amplification and then the stretched and amplified pulse is recompressed through another delay line having the opposite sign of dispersion. Since the nonlinearities in the amplifier are suppressed, efficient and distortionless amplification of femtosecond pulses can be achieved. However, the CPA technique does not compress the pulses because the amplifier nonlinearities are suppressed. For femtosecond pulse applications, it is desirable to find a technique capable of achieving both high-quality pulse amplification and a large compression factor, especially when the input pulsewidth is in the picosecond range. In this paper, we show that an erbium-doped nonlinear amplifying fiber loop mirror (hereafter, we call it ED-NALM) can perform such a function. The technique simultaneously utilize the gain, the multisoliton pulse compression effect, and the switching characteristics of the ED-NALM, so that the output pulses nearly preserve the fundamental soliton nature of the input pulses. We note that nonlinear optical loop mirrors (NOLMs) have been widely used for pulse compression [12], [13] and pulse-pedestal suppression [14], [15], but NOLMs do not provide amplification because they are passive devices. To obtain high-quality pulse amplification in an NOLM, one can insert a lump amplifier at one end of the loop [16], [17]. In this work, we place a distributed gain in the loop instead of a lump gain. As a result, the new device has different features than those in [16], [17]. In [16] group-velocity dispersion (GVD) was suppressed using a low-dispersion fiber loop in order to achieve high-quality amplification. As a result, pulse compression was inefficient; the typical compression factor is less than two. Whereas in the present design, we utilize the soliton effect, i.e., the interplay between GVD and SPM, to efficiently compress the input solitons. The proposed technique also differs from that in [17] in that no adiabatic condition is required, which not only permits more efficient pulse amplification and compression, but also allows one to amplify and compress long pulses with short loop lengths. II. ULTRASHORT SOLITON AMPLIFICATION IN EDFAS Before we investigate soliton amplification in an ED-NALM, it is useful to discuss soliton amplification in a conventional EDFA. With the inclusion of GVD, SPM, gain, gain dispersion, 0018-9197/03$17.00 2003 IEEE

556 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 39, NO. 4, APRIL 2003 and higher order fiber effects, ultrashort pulse amplification in an EDFA can be described by [3], [9], [18] (1) where,, and are the normalized distance, time, and pulse envelope in soliton units respectively. The parameters,,, and are the normalized gain, gain dispersion, Raman time constant, and third-order dispersion (TOD) coefficient respectively. In real units where is the half-width (at -intensity point) of the input pulse, is the group velocity, is the GVD coefficient, is the TOD coefficient, is the Raman resonant time constant, is the fiber loss, is the dipole relaxation time, is the unsaturated gain, and is the dispersion length. We do not include self-steepening and two-photon absorption effects since they play much smaller roles when compared with the other effects. The effect of gain saturation is also neglected in (1), which is justified since we will only be concerned with the amplification of a single pulse with a typical energy much lower than the saturation energy for most EDFAs, which is on the order of 1 J [3]. The input to the EDFA is assumed to be a fundamental soliton of the for (2) (3) Fig. 1. Local compression factor, normalized peak intensity, and corresponding pedestal energy for an input 2-ps (FWHM) fundamental soliton as a function of the EDFA length. The simulation parameters are =2:3, d =0:01144, =0, and =0. with a peak power determined by (4) Fig. 2. Temporal pulse shapes in (a) linear and (b) logarithmic scale at an EDFA length of 75.5 m. (c) Spectrum and (d) frequency chirp of the pulse. The input pulse and simulation parameters are identical to those used for Fig. 1. (5) perbolic-secant pulse having the same peak intensity and width (FWHM) as those of the amplified pulse, i.e., where is the nonlinearity coefficient. Equation (1) is solved numerically with initial condition (4) by using the split-step Fourier method. For illustration, we consider the amplification of a fundamental soliton with a full-width at half-maximum (FWHM) of ps ( ). Using typical EDFA parameters of ps km, ps km, fs, and fs near 1.55 m, and assume that the EDFA has a 10-dB gain per dispersion length,wehave, and. The parameters and for such a pulse are so small that their effect on pulse amplification is negligible. Fig. 1 shows, respectively, the normalized peak intensity, the compression factor, and the corresponding pedestal energy of the amplified pulse as a function of the amplifier length. The peak intensity of the amplified pulse is normalized to that of the input pulse, the compression factor is defined as the ratio of the FWHM of the input pulse to that of the amplified pulse, and the pedestal energy is defined as the relative difference between the total energy of the amplified pulse and the energy of a hy- Pedestal energy (6) Note that the energy of a hyperbolic-secant pulse with peak power and pulsewidth is given by Fig. 1 shows that efficient soliton amplification and compression occur when the propagation distance is longer than 50 m. However, the quality of the amplified pulse declines seriously beyond this distance. For example, at a distance of 75.5 m, the peak intensity is amplified by a factor of 142 with a compression factor of 12.7, but nearly 20% of the pulse energy is contained in the pedestal component. Fig. 2(a) and (b) shows, respectively, the temporal shapes of the amplified pulse in linear and logarithmic scale at this distance (75.5 m), while Fig. 2(c) shows the spectrum and Fig. 2(d) shows the frequency chirp of the amplified pulse. The spectrum intensity is normalized to (7)

WAI AND CAO: SIMULTANEOUS AMPLIFICATION AND COMPRESSION OF ULTRASHORT SOLITONS 557 the peak intensity of the input pulse spectrum. Because of the broad pulse pedestal, the spectrum exhibits a three-fold structure and the pulse is seriously chirped. This pulse cannot be used as long-distance information carriers because the pedestal will lead to intersymbol interference. Note that the pulse quality can be improved by using small amplifier gain [5], [6] (i.e., adiabatic amplification), but the amplifier length for achieving the same amplification factor will be excessively long. In the next section, we will show that this difficulty can be overcome using an ED-NALM. III. ULTRASHORT SOLITON AMPLIFICATION IN ED-NALM A. Demonstration of the Technique Fig. 3 shows the configuration of the ED-NALM which is identical to that of the NOLM [19] except that the loop is constructed from an uniform erbium-doped fiber. In order to provide an uniform gain along the entire loop length, the loop can be pumped simultaneously in both clockwise and counterclockwise directions using two semiconductor lasers located at the two ends of the loop. The input pulse is split at the coupler according to [19] Fig. 3. Schematic of the ED-NALM. (8) (9) where and are the pulse amplitudes right after the coupler and is the coupler power-splitting ratio. To compare this device to the conventional EDFA, we assume that the input soliton to the ED-NALM is the same as that used for Figs. 1 and 2. All of the loop parameters,,,, and are also identical to those assumed before. Fig. 4(a) (b) shows the temporal shapes of the output (transmitted) pulse in linear and logarithmic scale respectively when higher order effects are neglected. The coupler has a power-splitting ratio of. The loop length is optimized to 92.6 m so that the pedestal of the transmitted pulse is minimal. At this optimum loop length, the peak intensity is amplified by a factor of 165, the compression factor is 10.48, and the pedestal energy accounts for only 3.8% of the total energy of the transmitted pulse which means that the pulse shape is very close to hyperbolic-secant. Fig. 4(c) shows the pulse spectrum which is also close to a hyperbolic-secant shape except a small notch in the central frequency region caused by the nonadiabatic compression of the counterpropagating pulses in the loop. Fig. 4(d) shows the frequency chirp of the transmitted pulse. We see that the chirp across the main pulse is very small and is almost linear. The time-bandwidth product of the transmitted pulse is 0.303 which is very close to the transform-limited value 0.315 of a hyperbolic-secant pulse. From its peak intensity and width, the pulse is close to a soliton with a soliton order of 1.23. To verify the soliton nature of the transmitted pulse, we couple it into a passive (undoped) lossless fiber having the same parameters and as those of the loop. Fig. 5 shows the evolution over a fiber length of 101 m, which corresponds to 110 soliton periods in terms of the initial pulsewidth (0.19 ps of FWHM). The pulse narrows initially since its peak intensity is higher than that of a fundamental soliton. It then broadens Fig. 4. Temporal shapes of the transmitted pulse from the ED-NALM in (a) linear and (b) logarithmic scale. (c) Spectrum and (d) frequency chirp of the pulse. The coupler power-splitting ratio is 0.55 and the optimum loop length is 92.6 m. The input pulse and other parameters, d,, and are identical to those used for Fig. 2. Fig. 5. Propagation of the transmitted pulse shown in Fig. 4 in a passive (undoped) lossless fiber line with GVD coefficient and nonlinearity coefficient identical to those of the ED-NALM. because of the initial frequency chirp induced dispersive wave around the pulse wings. However, as propagation continues, the pulse becomes stable and approaches a fundamental soliton with a FWHM of 0.26 ps. The pedestal suppression and soliton-like pulse formation shown in Fig. 4 result from the switching characteristics of the ED-NALM. Since the coupler is asymmetric, the counterpropagating pulses in the loop are amplified differently and they acquire different phase shifts when they recombine at the coupler. At the optimum loop length, switching condition is satisfied for the central peak but not for the rest of the pulse leading to pedestal-free transmitted pulse. Fig. 6(a) shows the clockwise (solid line) and counterclockwise (dashed line) pulse shapes after they travel around the loop but before they

558 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 39, NO. 4, APRIL 2003 Fig. 6. (a) Temporal shapes and (b) spectra of the clockwise (solid lines) and counterclockwise (dashed lines) traveling pulses before recombination. The dashed-dotted lines represent the transmitted pulse. The input pulse and the ED-NALM are identical to those used for Fig. 4. recombine at the coupler under conditions identical to those of Fig. 4. The transmitted pulse shape is also shown by the dashed dotted line. Fig. 6(b) shows the spectra corresponding to Fig. 6(a). We observed that although the shapes and spectra of both the clockwise and counterclockwise pulses deviate significantly from a hyperbolic-secant shape, the transmitted pulse and its spectrum are close to hyperbolic-secant. B. Effects of Loop Characteristics and Nonsoliton Input Pulses The above results demonstrated the amplification of a particular fundamental soliton in a fixed ED-NALM. To study the robustness of the device, we study in this subsection the effect of varying the loop parameters and input pulse parameters on ultrashort pulse amplification. The varied parameters include the loop length, loop gain, coupler power-splitting ratio, input peak power, input pulse shape, and frequency chirp of the input pulse. Fig. 7(a) shows the normalized peak intensity, compression factor, and corresponding pedestal energy of the transmitted pulse as a function of the loop length. In all cases, the input pulse and the loop parameters,,,, and are identical to those used for Fig. 4 except that the loop length is varied. We see that variation in has a relatively large effect on the compression factor and the normalized peak intensity, but has little effect on the pedestal energy. The pedestal energy remains below 4% over a loop length range of 17 m, indicating that soliton-like transmitted pulses can be obtained over such a loop length range. Fig. 7(b) and (c) shows, respectively, the effects of varying the coupler power-splitting ratio and the loop gain parameter on the normalized peak intensity, compression factor, and pedestal energy of the transmitted pulses, under conditions identical to those used for Fig. 4 except that is varied for Fig. 7(b) and that is varied for Fig. 7(c). Again, we see that efficient high-quality pulse amplification and compression can be achieved despite the relatively large variations in and. Comparing Fig. 7(c) to Fig. 7(a), we see that varying the gain parameter has a similar effects on pulse amplification as that of varying the loop length. This is because the same gain can be obtained using either a long loop with small or a short loop with large. Varying the coupler power-splitting ratio, however, affects pulse amplification differently as shown in Fig. 7(b). In this case, both the compression factor and the pedestal energy increase almost linearly with increasing, which can be understood as follows: As increases, the difference between the clockwise and counterclockwise pulses increases, thus the interference between the pulses when they recombine at the coupler is reduced. We now consider how pulse amplification is affected when the input pulse deviates from an ideal fundamental soliton. Fig. 8(a) shows the amplified results under conditions identical to those for Fig. 4 except that the input soliton order is varied from 0.85 to 1.1, which corresponds to a variation of the peak power of the input pulse from 3.74 to 6.27 W assuming that the nonlinearity coefficient of the loop has a value of km W. As expected, the main feature shown here is similar to that shown in Fig. 7(a) and (c) because variation of the input pulse energy is similar to variation of the loop gain. Fig. 8(b) shows the amplified results under conditions identical to those for Fig. 4 except that the input pulse is linearly chirped, which can be expressed as (10) where is the chirp parameter. As compared to the cases shown in Figs. 7 and 8(a), initial frequency chirp affects pulse amplification in a complicated way. Both the normalized peak intensity and the compression factor oscillate rapidly as changes from negative to positive values. The positive chirp appears to enhance the compression factor because the anomalous GVD compresses positively chirped pulses and thus expedites soliton-effect compression of the counterpropagating pulses in the loop. Conversely, negatively chirped pulse broadens initially, which acts against the soliton-effect compression. Nevertheless, as shown in Fig. 8(b), efficient high-quality pulse amplification and compression can be achieved when the chirp parameter changes from 0.7 to 0.7. Thus far we have assumed that the input pulses to the ED-NALM have an ideal hyperbolic-secant shape. The effect of a nonhyperbolic-secant pulse shape is also an important consideration since in practice most lasers produce Gaussian shape pulses. Fig. 9 compares the amplified result shown in Fig. 4 (where the input pulse is with hyperbolic-secant shape) to the amplified result with a Gaussian input pulse given by

WAI AND CAO: SIMULTANEOUS AMPLIFICATION AND COMPRESSION OF ULTRASHORT SOLITONS 559 Fig. 7. Variation of the compression factor, normalized peak intensity, and pedestal energy with (a) loop length, (b) coupler power-splitting ratio, and (c) gain parameter. In all cases, the input pulse and the loop parameters are identical to those used for Fig. 4 except that the loop length, coupler power-splitting ratio, and gain parameter are varied, respectively, for (a)-(c).. The solid and dashed lines represent, respectively, the results with hyperbolic-secant and Gaussian input. The loop is the same for both cases and is identical to that used for Fig. 4. We see that the effect of input pulse shape on pulse amplification is very small. Similar peak intensity and pulse width are obtained in both cases. However, a smaller pedestal is created with the Gaussian input, as shown in Fig. 9(b), because the input Gaussian pulse has steeper leading and trailing edges than those of the input hyperbolic-secant pulse. As a result, the frequency chirp around the edges of the transmitted pulse is a slightly larger in the case of Gaussian input as shown in Fig. 9(d). We conclude that the performance of the ED-NALM is fairly insensitive to variations in both the loop parameters and input pulse parameters because for pulse amplification in an erbiumdoped fiber, gain dispersion tends to stabilize the pulse when its peak power and width approach those of the so called autosoliton [18]. When the counterpropagating pulses are amplified to some extent, they will be stabilized leading to robustness of the device performance to small variations in the parameters. C. Influence of Higher Order Effects The results presented so far concentrated on a fixed input pulsewidth ( ps). Simulations not shown here indicated that the qualitative behavior of the ED-NALM is the same for longer input pulses except that the optimum loop length increases. As a result, both the compression factor and the amplified pulse energy increase with input pulsewidth. As mentioned in Section II, higher order fiber effects such as RSS and TOD can be neglected when the input pulsewidth is in the picosecond range. For femtosecond input pulses, however, these effects become important. It was shown [3], [20] that RSS causes time delay of a propagating pulse and TOD causes oscillations near the pulse edges. Since the ED-NALM contains an asymmetric coupler, mismatched higher order effects of the counterpropagating pulses will lead to relative time delay and cause poor pulse overlap when the two pulses recombine at the coupler. This final subsection discusses the effects of RSS and TOD on ultrashort pulse amplification in ED-NALM. We consider the amplification of a fundamental soliton with initial width of ps ( ps). The ED-NALM is assumed to have a 10-dB gain per dispersion length ( ) with a coupler power-splitting ratio of. Note that the dispersion length for ps is approximately one eighth of that for ps, thus, for the same value of the normalized gain parameter, the unsaturated loop gain assumed here is eight times larger than that assumed before. The reason why we choose such a large is that a much shorter optimum loop length than before can be used to reduce the influence of higher order effects. Using the same parame-

560 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 39, NO. 4, APRIL 2003 Fig. 10. Influence of higher order effects on the amplification of a 0.7 ps (FWHM) fundamental soliton. (a)-(b) Pulse shapes in linear and logarithmic scale, respectively. (c) Spectra and (d) frequency chirps of the transmitted pulses. The solid and dashed lines represent, respectively, the results without and with higher order effects. The gain parameter ( =2:3) and the gain dispersion parameter (d =0:092) are the same for both cases. The higher order parameters are = 0:0075 and = 0:0021. Fig. 8. Variation of the compression factor, normalized peak intensity, and pedestal energy with (a) input soliton order and (b) initial frequency chirp. In all cases, the ED-NALM and the input pulsewidth are fixed and are identical to those used for Fig. 4. clearly shows that higher order effects have little influence on the device performance even though the amplified pulse is as short as 113 fs (FWHM). The compression factor is the same for both cases, the normalized peak intensity decreases from 61.69 (the ideal case) to 58.32, and the pedestal energy changes from 1.5% to 1.7%. The only difference is that higher order effects cause a slight time delay of the output pulse and a little red-shifting of the spectrum. The robustness of the ED-NALM to higher order effects is due the switching characteristics of the device. Since higher order effects have nearly the same effects on the counterpropagating pulses, the RSS generated pulse tails and the TOD induced low-intensity oscillations near the pulse edges are reflected when the two pulses recombine at the coupler, leading to a soliton-like transmitted pulse. IV. CONCLUSION Fig. 9. Comparison of the amplification of Gaussian (dashed lines) and soliton (solid lines) pulses in the ED-NALM. (a) (b) Pulse shapes in linear and logarithmic scale, respectively. (c) Spectra and (d) frequency chirps of the transmitted pulses. The ED-NALM is the same for both cases and is identical to those used for Fig. 4. ters,,, and as assumed before, we have,, and. Fig. 10 shows the amplified results when RSS and TOD are included (dashed lines) and neglected (solid lines) respectively. The optimum loop length with and without higher order effects are 10.66 and 10.75 m, respectively, which are much shorter than those used before. Fig. 10 We have proposed and demonstrated a technique for simultaneous amplification and compression of ultrashort solitons using an erbium-doped nonlinear amplifying fiber loop mirror. Numerical simulations show that, in contrast to conventional erbium-doped fiber amplifiers in which nonlinear effects lead to serious degradation of pulse quality, the proposed device performs high-quality pulse amplification and compression such that the amplified pulse nearly retains its soliton nature. We have also shown that the scheme is quite tolerant of small variations in both the device parameters and input pulse parameters such as the loop length, loop gain, coupler power-splitting ratio, input peak power, input pulse shape, initial frequency chirp, and higher order fiber effects. Although coherent effects [4] and the frequency dependent phase change [7] induced by the amplifier which are neglected in the numerical model may play a role for ultrashort pulse amplification in a conventional EDFA, we believe that these effects would not drastically affect the performance of the ED-NALM because they have similar effects on the counter propagating pulses in the loop and should not drastically affect the switching characteristics of the device.

WAI AND CAO: SIMULTANEOUS AMPLIFICATION AND COMPRESSION OF ULTRASHORT SOLITONS 561 REFERENCES [1] K. Kurokawa and M. Nakazawa, Wavelength-dependent amplification characteristics of femtosecond erbium-doped optical fiber amplifiers, Appl. Phys. Lett., vol. 58, no. 25, pp. 2871 2873, June 1991. [2] I. Y. Khrushchev, A. B. Grudinin, E. M. Dianov, D. V. Korobkin, V. A. Semenov, and A. M. Prokhorov, Amplification of femtosecond pulses in Er -doped single-mode optical fibers, Electron. Lett., vol. 26, no. 7, pp. 456 458, Mar. 1990. [3] G. P. Agrawal, Effect of gain dispersion and stimulated Raman scattering on soliton amplification in fiber amplifiers, Opt. Lett., vol. 16, no. 4, pp. 226 228, Feb. 1991. [4] B. Gross and J. T. Manassah, Numerical solutions of the Maxwell- Bloch equations for a fiber amplifier, Opt. Lett., vol. 17, no. 5, pp. 340 342, Mar. 1992. [5] W. Hodel, J. Schutz, and H. P. Weber, Limits to the amplification efficiency of ultrashort fundamental solitons using Er-doped fibers, Opt. Commun., vol. 88, no. 2-3, pp. 173 179, Mar. 1992. [6] M. Nakazawa, K. Kurokawa, H. Kubota, K. Suzuki, and Y. Kimura, Femtosecond erbium-doped optical fiber amplifier, Appl. Phys. Lett., vol. 57, no. 7, pp. 653 655, Aug. 1990. [7] W. Hodel, D. S. Peter, and H. P. Weber, Chirped pulse amplification in Er-doped fibers, Opt. Commun., vol. 97, no. 3-4, pp. 233 238, Mar. 1993. [8] M. E. Fermann, A. Galvanauskas, and D. Harter, All-fiber source of 100-nJ subpicosecond pulses, Appl. Phys. Lett., vol. 64, no. 11, pp. 1315 1317, Mar. 1994. [9] A. Galvanauskas, M. E. Fermann, and D. Harter, High-power amplification of femtosecond optical pulses in a diode-pumped fiber system, Opt. Lett., vol. 19, no. 16, pp. 1201 1203, Aug. 1994. [10] J. D. Minelly, A. Galvanauskas, M. E. Fermann, D. Harter, J. E. Caplen, Z. J. Chen, and D. N. Payne, Femtosecond pulse amplification in cladding-pumped fibers, Opt. Lett., vol. 20, no. 17, pp. 1797 1799, Sept. 1995. [11] D. S. Peter, W. Hodel, and H. P. Weber, Efficient distortion-free amplification of 1.3 m femtosecond pulses in a Pr -doped fluoride fiber amplifier, Opt. Commun., vol. 130, no. 1-3, pp. 75 80, Sept. 1996. [12] I. Y. Khrushchev, I. H. White, and R. V. Penty, High-quality laser diode pulse compression in dispersion-imbalanced loop mirror, Electron. Lett., vol. 34, no. 10, pp. 1009 1010, May 1998. [13] J. Wu, Y. Li, C. Lou, and Y. Gao, Optimization of pulse compression with an unbalanced nonlinear optical loop mirror, Opt. Commun., vol. 180, no. 1-3, pp. 43 47, June 2000. [14] M. D. Pelusi, Y. Matsui, and A. Suzuki, Pedestal suppression from compressed femtosecond pulses using a nonlinear fiber loop mirror, IEEE J. Quantum Electron., vol. 35, pp. 867 874, June 1999. [15] K. R. Tamura and M. Nakazawa, A polarization-maintaining pedestal-free femtosecond pulse compressor incorporating an ultrafast dispersion-imbalanced nonlinear optical loop mirror, IEEE Photon. Technol. Lett., vol. 13, pp. 526 528, May 2001. [16] K. Smith, E. J. Greer, N. J. Doran, D. M. Bird, and K. H. Cameron, Pulse amplification and shaping using a nonlinear loop mirror that incorporates a saturable gain, Opt. Lett., vol. 17, no. 6, pp. 408 410, Mar. 1992. [17] M. Matsumoto, A. Hasegawa, and Y. Kodama, Adiabatic amplification of solitons by means of nonlinear amplifying loop mirrors, Opt. Lett., vol. 19, no. 14, pp. 1019 1021, July 1994. [18] G. P. Agrawal, Nonlinear Fiber Optics, 2nd ed. San Diego, CA: Academic, 1995, pp. 506 515. [19] N. J. Doran and D. Wood, Nonlinear-optical loop mirror, Opt. Lett., vol. 13, no. 1, pp. 56 58, Jan. 1988. [20] K.-T. Chan and W.-H. Cao, Improved soliton-effect pulse compression by combined action of negative third-order dispersion and raman selfscattering in optical fibers, J. Opt. Soc. Amer. B, vol. 15, no. 9, pp. 2371 2375, Sept. 1998. Ping Kong A. Wai (SM 96) received the Bachelor of Science (Hons.) degree with first-class honors from the University of Hong Kong in 1981 and the M.S. and Ph.D. degrees from the University of Maryland, College Park, in 1985 and 1988, respectively. In 1988, he joined Science Applications International Corporation, McLean, VA, where he worked as a Research Scientist on the Tethered Satellite System project. In 1990, he became a Research Associate in the Department of Electrical Engineering, University of Maryland. In 1996, he joined the Department of Electronic and Information Engineering, The Hong Kong Polytechnic University, as an Assistant Professor. He became an Associate Professor in 1997, and a Professor and Head of the Department in 2002. His research interests include theory of solitons, modeling of fiber lasers, simulations of integrated optical devices, long-distance optical communications, all-optical packet switching, and network theories. He is an active contributor to the technical field, having over 100 international publications. Dr. Wai is a member of the Optical Society of America. Wen-hua Cao received the B.Sc. degree from Anhui University, Hefei, China, in 1985, the M.Sc. degree from Anhui Institute of Optics and Fine Mechanics, Academia Sinica, Hefei, China, in 1988, and the Ph.D degree from University of Science and Technology of China, Hefei, China, in 1994, all in physics. His Ph.D. work involved the numerical analysis of pulse compression and ultrashort pulse generation in optical fibers. He joined the Department of Electronic and Information Engineering, Wuyi University, Guangdong, China, where he became Vice Professor in 1995 and Professor in 2000. He was a Postdoctoral Fellow with The Department of Electronic Engineering, The Chinese University of Hong Kong, Hong Kong, China, from 1998 to 2000, where he worked on optical pulse compression, generation of ultrashort pulses from continuous-wave light, fiber parameters measurement, and optical switches. Currently, he is with the Department of Electronic and Information Engineering, The Hong Kong Polytechnic University, as a Senior Research Fellow. His research interests include optical solitons, numerical methods, fiber sensors and devices, and nonlinear fiber optics.