Ultrashort soliton source, with very high repetition rate, is a necessity for next generation optical fiber communication. A very simple passively mode-locked fiber ring laser centered at 1566 nm, which is based on the modulation instability (MI) theory, is proposed. Using high Erbium-doped (6470ppm) Bismuth fiber, a soliton pulse train, with 660 GHz repetition rate and pulse width of 420 fs, has been observed in a 11.5m short cavity for the first time.
©2003 Optical Society of America
Ultra-short solitons have attracted much attention due to their potential application in high-speed optical fiber communication. Actively and passively mode-locked fiber lasers are the two potential ways to achieve this objective. The actively mode-locked fiber laser uses one modulator to realize mode-locking. To date, the highest repetition rate is limited to 40 GHz due to the bandwidth of the modulator and signal generator . Various multiplexing technologies, such as rational harmonic , Talbot effect , intracavity Fabry-Perot etalon  and 3-couplers method, have been used to obtain high repetition rates. But they lead to other problems such as uneven pulse and poor stability. The output pulse is usually in the order of picoseconds with a Gaussian profile. Actively mode-locked fiber laser is also very expensive due to the usage of high-frequency RF components in it.
In the past years, nonlinear polarization rotation (NPR) technology , stretched pulse laser , additive pulse mode locking and Figure-8 fiber laser employing nonlinear amplifying loop mirror techniques  have been reportedly used for a passively mode locked fiber laser. Single solitons, multi-solitons and bound-solitons have been observed. The present passively mode locked fiber ring laser can produce very short hyperbolic pulses but with very low repetition rate, which is usually from 1MHz to 100MHz. Some of the designs that combined the merits of both actively and passively mode-locked fiber lasers have met with only partial success. In order to realize higher repetition rate, closely-spaced two-bound-solitons was reported by our group in 2001 for the first time . Two-bound-solitons from NPR fiber ring laser has a separation of 938 fs and pulsewidth of 326 fs. But normally, there are several pairs of bound solitons or several bunches of solitons in the cavity which have a separation of about tens of picoseconds between pairs.
Masataka et al. [9,10,11] first proposed another kind of passively fiber soliton laser which was based on MI, which can produce soliton trains at 115GHZ higher repetition rate. They used a longer cavity that was composed of a 1.5km Polarization Maintaining Fiber in order to produce enough nonlinearity. But this will lead to instability of operation, and self-induced Raman effect may stop the MI effect . In this paper, we report a simpler design for a passively mode-locked fiber ring laser which is based on the MI theory. Compared to the former report, a shorter and non-polarization maintaining fiber was used and the filter and Etalon have also been removed. A 660GHz repetition rate with pulse width of 420fs was observed for the first time to the author’s knowledge.
When the light power is very high, as a result of an interplay between nonlinear and dispersive effects, a nonlinear system will exhibit an instability that leads to modulation of the steady state. This MI can lead to spontaneous breakup of the continuous wave (CW) beam into a periodic pulse train. Self-phase modulation (SPM) induces spectral broadening which acts as a probe in this situation and is amplified by the gain provided by the MI. This MI exhibits a symmetrical gain spectrum centered at around the central frequency, with the maximum gain occurring at a shifted frequency given by :
where γ is the nonlinearity coefficient, P 0 is the peak power of incident light and β 2 is the group velocity dispersion coefficient. Since the largest gain occurs for frequencies ω 0±Ωmax, these frequency components have the maximum amplification gain of g=2γP 0. Thus, a clear-cut evidence of spontaneous MI is the creation of spectral side lobes located symmetrically at ±Ωmax on either side of the central wavelength ω 0. In the time domain, the CW beam is converted into a periodic pulse train whose period is given by , where
If the cavity round trip period is a harmonic of the period of the pulse generated by the MI, then mode locking can be achieved. A stable output of soliton pulse train can be obtained. In the cavity of a fiber soliton laser with high power, P 0 is actually the soliton peak power, and it can be expressed as:
This means that a higher repetition frequency pulse can be obtained, which is inversely proportional to the pulse width. The duty cycle is around 0.4. The key factor of this kind of fiber soliton laser is its need of very high power, which should be far above the threshold of MI, in the fiber laser cavity.
Figure. 1 shows the configuration of our MI mode-locked fiber ring laser (MIMLFRL). The loop length of the laser cavity is about 11.5 m, which comprises of a 41.5 cm highly Erbium-doped Bismuth fiber (Er+ concentration ~6470 ppm). A polarization-insensitive isolator is used to force unidirectional operation in the laser cavity. A polarization controller is used to adjust the polarization of the light in the cavity to achieve mode locking. A 1480/1550 nm wavelength-division-multiplexing (WDM) coupler has been used to couple the pump power into the ring cavity. A pigtailed Raman laser operating at 1480 nm, with the maximum pump power of 1.2 W, is used as the pump source for the system. Finally, two outputs are taken from 10/90 ratio coupler which is made of dispersion shifted fiber (DSF).
Compared with the former report , our configuration has the following merits: 1) It has a very short cavity of around 11.5m rather than the former hundreds of meters. Hence, it is more tolerant to the environmental perturbation and it can stop the occurrence of Raman effect. 2) We used non-polarization-maintaining fiber which is cost effective. 3) We removed the filter in the cavity in order to obtain narrower pulses, although it increased the difficulty in tuning the polarization.
Increasing the pump power above the self-start threshold and with careful tuning of the polarization, MI occurs and soliton trains are generated. Fine-tuning of the polarization controller leads to the appearance of two symmetrical wavelength peaks. The CW background can be greatly suppressed. The typical optical spectrum of the output with a pump power of 800mw is shown in Fig. 2, which gives six clear spectral peaks that are due to the consequence of modulation of very close pulses. The interval of the neighboring peaks is around 6.0 nm. Fig.3 gives the autocorrelation trace of the output in different scan ranges. Fig. 3(a) is in the full scan range of 50ps, and (b) is in the 5ps scanning range. From Fig. 3(b), we can see that the period of the pulse is around 1.51 ps, which corresponds to 660 GHz. The 3dB width of the autocorrelation trace of the pulse is 648 fs, and it corresponds to a pulse width of 420 fs if Sech2 profile is assumed. According to Eq. (4), if the repetition frequency f is 660 GHz for our laser, then τ should be around 600 fs. There is a 30% error between our calculation and measurement. We think that this is due to some overlap of neighboring pulses. The estimated 3dB spectral width of the spectral envelope is around 5.0 nm and the time-bandwidth production is 0.38. This means that it is nearly transform-limited.
With a suitable polarization bias, the self-start threshold is around 44 mw. Following the CW operation, mode-locking is obtained suddenly. In the mean time, the oscilloscope shows a very stable, periodic, wide-pulse train. Increasing the pump will broaden the width of the pulse on the oscilloscope. More periodic wide-pulses appeared with the increase pump power. The period of the wide-pulse was automatically adjusted and this lead to the increase of the repetition rate of wide-pulse. When the pump power exceeded 750mw, no pulse output was observed on the oscilloscope. The explanation is as follows:
With the increase of pump power, the part exceeding the threshold of MI first forms the soliton train (hundreds of periodic, closely spaced solitons, can be regarded as a bunch) suddenly because of MI effect. The rise and fall time of our detector is around 10ps (45GHz bandwidth), which is much slower than the soliton repetition period of 1.5ps. Hence, a single soliton pulse can’t be identified. Under such pump power, the population inversion of Er ions is still not strong enough to prevent gain depletion. It only supports a periodic soliton train. Hence it is only observed as periodic wide-pulses on the oscilloscope, this actually indicates the length of the soliton train (bunch), as shown in Fig. 4, whose period is dependent on the pump level. The higher the pump power, the shorter the period of wide-pulses, until it finally fills up the entire cavity. In our experiment, when the pump power is around 750 mw, the repetition rate of the wide-pulses is about 12GHz, which is higher than the frequency limit of our oscilloscope’s trigger signal. Hence, a further increase of pump power will lead to no pulses output on the oscilloscope.
Our experiment showed that there were no exact rules to follow for the increase of either the width of the wide-pulse or its frequency in every pump sweep. Sometimes the width increased faster and sometimes the frequency. But, the total energy should be decided by the pump power. According to our estimation, if it is assumed that the Er-doped fiber is not saturated, when the pump power reaches 4W, it may have a continuous soliton train output. If the pump power is even much higher than this threshold, higher order soliton train may be formed, or the Raman effect may even stop the MI effect.
The soliton output is very stable once self-started. The soliton separation and soliton width doesn’t change for hours even if the pump power varies. If the polarization controllers are kept in the same position, soliton output can be restarted at every pump sweep. And in this laser, we don’t find the obvious phenomenon of pump power hysteresis, which is typical in the previous passively mode-locked fiber ring laser.
There exist several discrete polarization settings to realize the soliton output. But, they show very little change in output results, with discrete pulse period varying from 1.5 ps to 1.7 ps. We believe that this is because, under different polarizations, the light travels different paths and this leads to a small dispersion fluctuation in the cavity. Hence the soliton period has to be self-adjusted accordingly in a certain small range in order to achieve mode-locking.
The results from output 2 are almost the same as that from output 1. The estimated average dispersion in our fiber laser is around nd 2 ps/nm/km and hence, the soliton peak power should be around 1W. The output power is around 16 dBm under full pump power. If we consider that the output coupler has a coupling ratio of 10:90, then the peak power in the cavity is 26dBm. This is a little lower than what was calculated. We think that this is so because the wide-pulses still do not fully fill the cavity although the solitons have been very closely spaced. We believe that the usage of a much stronger pump and higher Erbium-doped fiber can fill solitons in the entire cavity.
A polarizer was inserted in the fiber bench when the MI laser was running, and its polarization was adjusted with all other parameters remaining constant. At a suitable bias position, the laser still gave the same results. This verifies that the light in the cavity has linear polarization.
The output from the current MI fiber laser can be regarded as hundreds of solitons bound together. We can consider this as a good progress in comparison to our previous report of bound solitons from NPR fiber laser [7,8], where there were only 2 solitons, or up to 4 solitons that were closely bound together, and its whole repetition rate is still lower. There was pump hysteresis and in the present case we don’t observe this phenomenon. And according to our experience, to achieve polarization tuning is more difficult in NPR laser for bound-soliton output than that in a MI fiber laser. But MI laser needs a much higher pump power, and the modulation depth of the output signal from the MI laser is much lower than that from the NPR laser.
A simple passively mode-locked fiber ring laser centered at 1566nm, which is based on the MI theory, has been established. With the use of a highly Erbium-doped (6470ppm) Bismuth fiber and a 1.2 W pump source, soliton pulses having 660 GHz repetition rate and pulse width of 420fs have been observed for the first time. A short cavity helps us to stop the occurrence of self-induced Raman effect and make the system more stable. We think that if the pump power and Er-doped fiber gain are strong enough, solitons can fill the entire cavity and a real 660GHz soliton source can be obtained.
I would like to acknowledge Varghese Paulose for his help in proofreading and editing this manuscript.
References and Links
1. Baoxi Xu, Gaoqiang Yuan, Chong Wei Chuah, Honggang Wang, and Tow Chong Chong, “40GHz pulse generation and performance analysis of actively mode-locked fiber ring laser,” ICOCN, 14H2, Singapore (2002)
2. Min-Yong Jeon, Hak Kyu Lee, Joon Tae Ahn, Kyong Hon Kim, Dong Sung Lim, and Ei-Hang Lee, “Pulse-amplitude-equalized output from a rational harmonic mode-locked fiber laser,” Opt. Lett. 23, .855–857 (1998) [CrossRef]
3. J. Azana and M.A. Muriel, “Temporal self-imaging effects: Theory and application for multiplying pulse repetition rates,” IEEE J. Sel. Quantum Electron. 37, 728–744 (2001) [CrossRef]
4. Kamal K. Gupta, Noriaki Onodera, Kazi S. Abedin, and Masaharu Hyodo, “Pulse repetition frequency multiplication via intracavity optical filtering in AM mode-locked fiber ring lasers,” IEEE Photo. Tech. Lett. 14, 284–286 (2002) [CrossRef]
5. G. Yandong, S. Ping, and T. Dingyuan, “298fs Passively Mode Locked Ring Fiber Soliton Laser,” Micro. & Opt. Tech. Lett. 32, 320–333 (2002) [CrossRef]
6. I. N. Duling, III, “Subpicosecond all fiber erbium lasers,” Elec. Lett. 27, 544–545 (1991) [CrossRef]
7. Y. D. Gong, P. Shum, D. Y. Tang, C. Lu, Z. W. Qi, W. J. Lai, W. S. Man, and H. Y. Tam, “Closed spaced ultra-short bound solitons from DI-NOLM figure 8 fiber laser,” Opt. Commun. 220, 297–302 (2003) [CrossRef]
8. Y.D. Gong, P. Shum, T.H. Cheng, Q. Wen, and D.Y. Tang, “Bound soliton pulses in passively mode locked fiber laser,” Opt. Commun. 200, 389–399 (2002) [CrossRef]
9. Masataka Nakazawa, Kazunri uzuki, and Hermann A Haus, “The modulational instability laser—Part I: Experiment,” IEEE J. Quantum Electron. 25, 2036–2044 (1989) [CrossRef]
10. Eiji Yoshida and Masataka Nakazawa, “Low-threshold 115GHz Continuous-wave modulational-instability erbium-doped fiber laser,” Opt. Lett 22, 1409–1411 (1997) [CrossRef]
12. Govind P. Agrawal. Nonlinear fiber optics, (New York: Academic Press, 1995)
13. Masataka Nakazawa, Kazunri uzuki, Hirokazu Kubota, and Hermann A Haus, “The modulational instability laser—Part II: Theory,” IEEE J. Quantum Electron. 25, 2045–2052 (1989) [CrossRef]