I March 1995 OPTICS COMMUNICATIONS ELSEVlER Optics Communications 115 (1995) 105-109 Time resolved spectral development of ultrashort pulse solitons in erbium fiber loop lasers D.U. Noske, N. Pandit, J.R. Taylor Femtosecond Optics Group, Physics Department, Imperial College, Prince Consort Road, London SW7 282. UK Received 13 October 1994 Abstract Using a gated pumping technique, the time resolved spectral development of ultrashort pulses in an erbium fibre figure of eight laser have been examined in the instability regime where the fibre laser length is comparable to the characteristic soliton length. The associated spectral sideband development has been resolved in addition to the development of other lines not predicted by the simple theoretical model of the system. Solitons exist in optical fibre as a result of the continuous balance of nonlinear refraction and anomalous group velocity dispersion [ I]. To overcome transmission losses, pulse amplification must be provided. The solitons are stable with respect to two principle extremes of perturbation in the pulse energy (or other system parameters such as dispersion, etc.). The first is the adiabatic regime, where variations in the pulse energy take place on a length scale which is long compared to the characteristic soliton period [ 2,3]. In the second extreme, the perturbations take place on a length scale significantly shorter than the soliton period [ 4,5 ] and the theoretical description of the soliton is that of an average for the system [ 61. Over the past few years considerable interest has been directed towards erbium doped fibre lasers as sources of ultrashort pulse solitons [ 7-91. The bandwidth of erbium doped silicate glass is capable of supporting sub 100 fs pulses, however, for pulses of this duration the associated soliton length, under normal operating conditions, is generally comparable to the optical length of conventional fibre laser geometries. In such a situation, short pulse solitons are affected by the periodicity of the amplification process and deviations from the average soliton behaviour are expected. In particular, the dispersive wave which is shed off from the perturbed soliton plays a significant role in the dynamics of the pulse formation/reformation. Spectral beating occurs, enhanced by the periodicity of the amplification process and this phase matched process gives rise to characteristic spectral sidebands [ 10-131, of order n, the separation (Ah,) of which from the central wavelength (A) of the soliton isgivenby [ll] J h* 8nZ, 1 Ah,=*- -- 27Tct() z, where Z,, and Z, are the soliton period and periodic amplification length respectively, of a soliton of pulse width 1.763t,,. Good agreement between theoretical prediction [ 111 and experimental measurement [ 141 has been demonstrated for the time integrated spectra of solitons generated in various cavity configurations [ 151 for pulsewidths in the picosecond and femtosecond regimes. In this letter we describe measurements on the time resolved spectral development of ultrashort pulse soli- 0030-4018/95/$09.50 0 1995 Elsevier Science B.V. All rights reserved SSDIOO30-4018(94)00699-7
106 D.U. Node etal. /Optics Communications 11.5 (1995) 105-109 980 nm cw pump ) Ti:Al203 1 Electrical BOXC?X output ) Integrator I 20 Hz Sqw-Wave Generator 1 Time-resolved output Fig. I. Schematic of the experimental arrangement used to measure the time resolved soliton spectral development in a figure of eight fibre laser. tons in a fibre laser loop geometry and the associated sideband evolution which is a result of the associated soliton length being comparable to the fibre laser 2ms A- -1,,,1,,,,11 1.560 1 S65 1.560 1.565 wavelength ( w ) wavelength ( pm ) Fig. 2. Time resolved (as indicated) spectral evolution of the figure of eight Er fibre laser. length. In addition, other spectra1 features were observed which are not explained by the simplified theoretical models above. A conventional figure of eight fibre laser geometry [ 71 was employed in the laser under investigation, the components, configuration and characterisation of which has been described in detail previously [ 141. In this series of measurements an overall fibre cavity length of 158 m was used with an average dispersion of 5 ps/nm km. Minimum transform limited pulse lengths of 1.5-2.0 ps were routinely obtained and the time integrated spectra exhibited the now characteristic sideband structure. Fig. 1 shows the experimental arrangement configured to examine the time-resolved spectra1 evolution. A square wave generator provided the drive signal to an acousto-optic modulator. The 980 nm pump beam from a Ti:sapphire laser was modulated at 20 Hz with a 1: 1 mark-space ratio and a rise time of approximately 10 ns, generating 100% modulated pulsetrains from the fibre laser which followed the pump pulse modulation. The spectral output from the figure of eight laser was monitored via a scanning optical spectrograph in conjunction with a germanium diode and time resolved using a box car integrator driven in synchronism by the square wave generator. A sampling gate width of 15 ps was employed which corresponded to an integration over 19 round trips of the fibre laser cavity and the output was sampled sequentially through the 25 ms lasing time. For optimised polarization controller settings and pump power, the fibre laser was self starting and reproducible in terms of generated pulsewidth, spectrum and output power.
D.U. Nnske et al. /Optics Communications 115 (1995) 105-109 107 Fig. 2 shows the time resolved spectral evolution in the figure of eight laser in the first two milliseconds from pump switch on. After 50 ps cw laser action alone occurred around 1.5625 pm. This continued to grow in intensity until after 200 ps, the passive mode locking mechanism of the NALM permitted the formation of solitons, which resulted in the appearance of a broad spectral feature around 1.5615 pm. In time the number of solitons in the cavity increased and the intensity of the soliton spectra increased at the expense of the cw component which correspondingly decreased. The short pulse solitons which were formed experienced the spectral instability described above as the soliton length approached that of the amplifier (overall fibre laser loop length) and energy was lost from the soliton shed into the sidebands. It can be seen that after 200 ps, on the formation of the soliton, sideband structure appeared, around 1.5595 pm and 1.5642 pm, which was very apparent after 400 ps. The sideband generation took place immediately on formation of the ultrashort pulse soliton within the resolution limit of our measurement, i.e. the appearance of a soliton feature in the specrum was always accompanied by sideband structure within the limiting 15 /.u temporal resolution. Higher spectral resolution revealed that all orders n of the sidebands appeared simultaneously, as theory predicts [ 111. It should also be noted that splitting of the individual sidebands was observed which is related to interaction via birefringence within the nonlinear amplifying loop mirror [ 181. On the formation of the sidebands an additional spectral feature, labelled for convenience the zero order sideband appeared and in Fig. 2 this can be clearly seen after 400 ps at 1.563 pm. It should be noted however, that this n = 0 label does not infer any relationship to the simple theoretical models of sideband generation in periodically amplified soliton systems [ 10-131 which predict symmetrical behaviour. A single asymmetric spectral feature associated with the effects of third order dispersion on soliton spectra has been theoretically derived and is well established [ 16,171, however, in most fibre lasers second order dispersion is non-negligible, dominating the third order contribution and the resonance feature due to the third order dispersion is absent. The component labeled it = 0 in this work arises from birefringent effects as described below. In Fig. 2, the spectrum taken after 900 ps, labels all the relevant spectral features refered to in the text.. 000 B.MM. oooo 0 0 0.. 0 Soilton,. 1st ridehand 0. I 5 IO 15 20 dclny (ms) Fig. 3. Temporal evolution of the spectral intensity of the soliton and n = 1 sideband. It should be noted that the evolution shown in Fig. 2 is not that of a single soliton. At the pump powers used many tens of solitons can exist simultaneously in the cavity, with the actual number increasing during the evolution period shown. Hence, Fig. 2 shows the superimposed spectra of an increasing number of solitons which are also at differing stages of their evolution. Theory predicts [ 111 that the intensity of the sidebands should increase linearly with the number of amplification periods. However, in the laser system, passive mode-locking based on the operation of the nonlinear optical loop mirror (NALM) is taking place and this acts to limit the energy build-up in the sidebands. In the steady state each soliton sheds radiation at a rate equal to that at which the NALM removes it. Thus each soliton will have a fixed amount of dispersive radiation associated with it and the energy in the sidebands should be proportional to the number of solitons circulating in the cavity and hence the intensity of the soliton spectrum. Fig. 3 shows the time resolved spectral intensity growth of the soliton and first order sideband. In the first millisecond both exhibited a linear growth. Higher temporal resolution than that of Fig. 3 showed that some periodicity was present due to Q-switching during the start up process. For the experimental parameters used solitons appeared approximately 80 ps from the start of the pump pulse, although this is a function of the pump power etc. As expected, the sidebands appeared at the same time as the soliton and the first order side band intensity was shown to be directly proportional to the soliton intensity. After approximately 24 ms (see Fig. 3) both the soliton and sideband energies clamp to a constant level, indicating that the
108 D.U. No& et al. /Optics Communications 115 (1995) 105-109 available cw radiation had been depleted. It can be seen in Fig. 2 that after approximately 2 ms the cw lasing component of the spectrum has also been totally eliminated. The zero order (n = 0) line also exhibited an identical development saturating at a similar point in time to that of the soliton. It should be noted that for a fixed pump power, the features of the time resolved spectra were highly reproducible. Further insight into the origin of the zero order line was obtained by investigating the polarization dependence of the time resolved spectra. Fig. 4 shows representative polarization resolved data obtained 900 /JS after start-up, while the corresponding integrated spectrum can be seen in Fig. 2. These indicate that the zero order line is strongly polarized perpendicular to that of the soliton. The remaining cw radiation is polarized parallel to the soliton although immediately after start up it is randomly polarized. In the time domain, spectrally filtered and with low temporal resolution, the zero order line appears as a long pulse output in synchronism with the solitons. As the pulses were undetectable using autocorelation techniques the measured duration of the pulses were thus diode limited in our case to 1 ns. Through adjustment of the polarization controllers in the fibre loop, the wavelength of the zero order line could be continuously adjusted relative to the peak wavelength of the soliton, which also varied. However, the zero order line was always to the long wavelength side of the soliton central wavelength. This line simply arises from the complex phase dependent switching characteristic of the NALM. Originating from the dispersive radiation shed from the solitons the zero order line is efficiently switched at a wavelength different from the soliton peak through meeting the phase requirement via polarization rotation. Since autocorrelation measurements of the soliton pulsewidths are polarization dependent, the effect of the zero order component would not normally be apparent on any measurement of the soliton pulsewidth. In fibre lasers where polarisation dependent isolators or polarisation dependent components are incorporated this additional spectral line would not be observed. The theoretical model [ 111 of the sideband generation does not include the affects of polarization. However, the NALM can exhibit transmission at differing wavelengths through meeting the required phase condition by operating in the orthogonal polarization. The so called zero order line is only one example of this. k 1 1.560 Wavelength ( ym ) I.565 Fig. 4. Polarization resolved spectra, 900 + from start-up (a) parallel and (b) perpindicular to the plane of polarization of the soliton. Other examples can occur and numerous additional lines have been observed spectrally distinct from the theoretically predicted sidebands. These additional lines are generally in the orthogonal polarization state and arise from the differing phase matching conditions being satisfied by the birefringence of the fibre system. Clearly, these can be eliminated by the inclusion of an intracavity polarisation selecting element. This will not affect the switching characteristic nor the switching of the additional components, but it will prohibit transmission and regeneration and remove them from the output signals. However, under these conditions, intensity dependent polarisation switching can also possibly contribute to the mode locking process. In conclusion we have time resolved the spectral development of ultrashort pulse soliton generation in optical fibre loop lasers when the cavity size is comparable to the characteristic soliton period. The processes acssociated with the sideband generation have been described. Many of the problems associated with such generation can be avoided by introduction of a bandwidth limiting filter, or defining the polarization of the system or by operating in the non-soliton regime. We are grateful to the Science and Engineering Research Council (SERC) UK for the overall financial support of this work under Research Contract No. GR/ H/64316 References [ 11 A. Hasegawa and F.D. Tappert, Appl. Phys. Lett. 23 ( 1973) 142.
D.U. Noske et al. /Optics Communications 115 (1995) 105-109 109 [2] A. Hasegawa, Optics Lett. 8 (1983) 650. ]3] L.F. Mollenauer, J.P. Gordon and M.N. Islam, IEEE J. Quantum Electron. QE-22 (1986) 157. [4] L.F. Mollenauer, S.G. Evangelides and H.A. Haus, IEEE J. Lightwave Tech. LT-9 (1991) 194. [5] A. Hasegawaand Y. Kodama, Optics Lett. 15 (1990) 1443. [6] S.M.J. Kelly, K. Smith,K.J. Blow and N.J. Doran, OpticsLen 16 (1991) 1337. [7] IN. Duling 111, Annual Meeting, Vol. 15 of 1990 OSA Technical Digest Series (Opt. Sot. Am., Washington, 1990) paper PDP 4. [ IO] L.F. Mollenauer, K. Smith, J.P. Gordon and C.R. Menyuk, Optics L&t. 14 (1989) 1219. [ 1 I] S.M.J. Kelly, Electron. Lett. 28 ( 1992) 806. [ 121 J.P. Gordon, J. Opt. Sot. Am. B 9 (1992) 91. [ 131 N.J. Smith. K.J. Blow and I. Andonovic, IEEE J. Lightwave Tech. 10 (1992) 1329. [ 141 D.U. Noske, N. Panditand J.R. Taylor. OpticsLett. 17 ( 1992) 1515. [ 151 D.U. Noske. N. Pandit and J.R. Taylor, Electron. Lett. 28 (1992) 2185. [16] P.K.A.Wai,C.R.Menyuk,C.LeeandH.H.Chen,OpticsLett. 11 (1986) 464. [8] D.J. Richardson, RI. Laming, D.N. Payne, V.J. Matsas and [ 171 S.M.J. Kelly, T. Brabec, F. Krauszand J.N. Elgin, PaperTuB M.W. Phillips. Electron. Lett. 27 (1991) 1451. Nonlinear Guided Wave Phenomena, Technical Digest 15 [9] V.J. Matsas, T.P. Newson, D.J. Richardson and D.N. Payne, (1993) Opt. Sot. Am. Washington DC, pp 188-191. Electron. Lett. 28 (1992) 1391. [ 181 S.M.J. Kelly and J.R. Taylor, to be published.