670 OPTICS LETTERS / Vol. 30, No. 6 / March 15, 2005

40-GHz, 100-fs stimulated-Brillouin-scattering-freepulse generation by combining a mode-lockedlaser diode and a dispersion-decreasing fiber

Ken-ichi Hagiuda, Toshihiko Hirooka, and Masataka NakazawaResearch Institute of Electrical Communication, Tohoku University, 2-1-1 Katahira, Aoba-ku,

Sendai 980-8577, Japan

Shin Arahira and Yoh OgawaOki Electric Industry Co., Ltd., 550-1 Higashiasakawa-cho, Hachioji-shi, Tokyo 193-8550, Japan

Received October 26, 2004

A 40-GHz, 100-fs pulse train was successfully generated by soliton compression of a mode-locked laser diode(MLLD) pulse with a dispersion-decreasing fiber. The MLLD had a longitudinal mode linewidth as broad as60 MHz, which made it possible to suppress stimulated Brillouin scattering and achieve stable, ultrahigh-speed pulse compression without applying external frequency modulation. © 2005 Optical Society of America

OCIS codes: 320.5520, 190.5530, 190.5890.

Adiabatic soliton compression using a dispersion-decreasing fiber (DDF) is a useful and easy method ofgenerating a femtosecond optical pulse train with agigahertz repetition rate.1,2 Such an ultrashort pulsetrain is becoming increasingly important with a viewto realizing ultrahigh-speed optical time-division-multiplexed transmission, all-optical sampling,and all-optical switching. Recently, 10-GHz, 54-fspulses and 50-GHz, 280-fs pulses were successfullygenerated by use of a polarization-maintainingdispersion-flattened dispersion-decreasing fiber(PM-DF-DDF).3,4 However, as we increase the repeti-tion rate, stimulated Brillouin scattering5 (SBS)starts to play an important role, because the opticalpower of each longitudinal mode increases in propor-tion to the square of the repetition rate and easily ex-ceeds the SBS threshold.6 Since SBS is initiated byamplification of spontaneous scattering, the back-scattered power varies randomly and the compres-sion process becomes unstable. Tamura and Sato4 ob-served SBS in pulse compression at 50 GHz andovercame the power limitation by frequency modulat-ing the injection current of a mode-locked laser diode(MLLD) to increase the linewidth.4 The SBS effect inDDF with adiabatic soliton compression has alreadybeen described in detail.6

In this Letter we report SBS-free generation of a40-GHz, 100-fs pulse train, which we achieved bycompressing the output of a MLLD with PM-DF-DDF. The SBS was sufficiently suppressed and stablepulse compression was achieved without employingfrequency modulation, because the MLLD that weused had a longitudinal linewidth as broad as60 MHz.

To determine the SBS effect on soliton compressionat a high repetition rate, we first derive the opticalpower of each longitudinal mode of a soliton pulsetrain. The average power of a fundamental soliton isgiven by

0146-9592/05/060670-3/$15.00 ©

PavgN=1 = 0.776

l3

p2cn2

uDu

DtBAeff, s1d

where D is the group-velocity dispersion (GVD), l isthe wavelength, n2 is the nonlinear refractive index,Dt is the FWHM, B is the repetition rate, and Aeff isthe effective area. By roughly assuming that thepower is equally divided into M=Dn /B modes withinthe FWHM of the optical spectrum, Dn, we obtain thepower per mode,

P0 =Pavg

N=1

M= 2.463

l3Aeff

p2cn2uDuB2, s2d

where we use DnDt=0.315. Equation (2) indicatesthat P0, which increases in proportion to B2, easilyexceeds the SBS threshold and that SBS has a sig-nificant disadvantageous effect when the repetitionrate is increased.

Figure 1 shows the optical spectrum of backscat-tered light from a DDF. In this case we launched a40-GHz, 1.7-ps pulse train generated by a mode-

Fig. 1. Optical spectrum of the backscattered light when a40-GHz MLFL output is launched into DDF. The inset is anexpansion of one mode.

2005 Optical Society of America

March 15, 2005 / Vol. 30, No. 6 / OPTICS LETTERS 671

locked fiber laser (MLFL). The power level of thelaunched pulse was set at the fundamental solitonpower. The linewidth of the MLFL was approxi-mately 1 kHz,7 which easily exceeded the SBSthreshold. The backscattered power was monitoredthrough an optical circulator at the DDF input. Fivedistinct Brillouin modes, denoted in the figure by thearrows, were clearly seen around the central wave-length region, and the power of each mode reached ashigh as 7 dBm. Other longitudinal modes observed ata level lower than −20 dBm were the direct output ofthe source that leaked through the circulator. The ex-panded view in the inset shows that the backscat-tered modes were frequency shifted by 0.09 nms11 GHzd from the source wavelength toward alonger wavelength. This result indicates that thesereflected modes are caused by SBS in the DDF. SBSoccurs for longitudinal modes with a high power levelthat exceeds the SBS threshold, namely, around thecenter of the spectrum. In this case it was difficult toobtain stable short pulses at the DDF output.

It is well known that one can overcome the SBS-induced power limitation by increasing the sourcelinewidth.8 SBS gain is dependent on the signal line-width, Dnp, and the bandwidth of the SBS gain spec-trum, DnB (,16 MHz in silica glass), which is ex-pressed as

gB = S DnB

DnB + DnpDgB0, s3d

where gB0=5310−11 m/W is the SBS gain when Dnp=0. Since the MLFL has a linewidth as narrow as1 kHz, frequency modulation must be employed tosuppress the SBS and stabilize the pulse compres-sion. In the study reported in Ref. 6 frequency modu-lation of 30 MHz was applied to compress the outputof a 40-GHz MLFL without SBS. In contrast, aMLLD has a linewidth much broader than that of aMLFL because of the low cavity Q value resultingfrom the short cavity length. The linewidth-enhancement factor caused by the variation of the re-fractive index with carrier density fluctuation, whichis commonly denoted by a, is also responsible for thebroader linewidth.9,10 These properties are beneficialin terms of suppressing SBS, which makes ultrahigh-speed femtosecond pulse compression possible with-out the need for frequency modulation.

As the next step, we carried out an adiabatic soli-ton compression experiment at 40 GHz with aMLLD. The experimental setup is shown in Fig. 2.We used a 40-GHz hybrid mode-locked monolithicdistributed Bragg reflector laser diode with a satu-rable absorber as a mode locker.11 A 40-GHz, 2.4-pspulse train was generated from a MLLD operating at1553 nm. The input pulse had a time–bandwidthproduct of 0.35. The pulses were amplified to the soli-ton power level with a high-power erbium-doped fiberamplifier (HP-EDFA). The amplified pulses werethen launched into a 950-m-2 long PM-DF-DDF. TheGVD in the PM-DF-DDF decreased linearly from

2

11 to 0.2 ps/ snm kmd and Aeff was 28 mm . The aver-

age power of the fundamental soliton was 250 mW.The linewidth of the MLLD was measured with aheterodyne detection method. Figure 3 shows the rfspectrum of the heterodyne signal, which we ob-tained by extracting one mode from the MLLD by useof a narrowband optical filter and by detection of thebeat note with a MLFL (linewidth, ,1 kHz). Fromthis result, the FWHM of the rf beat signal, whichcorresponds exactly to the actual laser linewidth, wasapproximately 60 MHz.

Figure 4(a) shows the optical spectrum of the back-scattered light from the DDF monitored through anoptical circulator. Unlike in Fig. 1, with the MLLD,the SBS was sufficiently suppressed to as little as−40 dBm. This was due to the fact that, from Eq. (3),gB for the MLLD was reduced to approximately 21%of gB for the MLFL. Figure 4(b) shows the transmit-ted average power measured at the DDF output andthe backscattered power for various input power lev-els. The backscattered power was less than 0.1 mWfor all the input power levels. The launched powerwas successfully transmitted through the DDF with-out pump depletion, indicating that stable pulse com-pression was achieved.

Figure 5(a) shows the autocorrelation waveformsobtained at the input and output of the DDF. The in-put power was Pin=216 mW, corresponding to N= sPin/Pavg

N=1d1/2=0.93. The output durations were104 fs assuming a sech shape, and the pedestals ofthe compressed pulses were less than −20 dB fromthe peak. The output spectrum is shown in Fig. 5(b).The spectral FWHM of the sech fit of the output spec-trum, which is shown by the dashed curve, was4.1 THz s33 nmd, corresponding to a time–bandwidth

Fig. 2. Experimental setup (abbreviations are defined intext).

Fig. 3. Longitudinal linewidth of MLLD, obtained by de-tecting the beat note between a mode extracted from a

40-GHz MLLD and a MLFL.

power versus input power.

672 OPTICS LETTERS / Vol. 30, No. 6 / March 15, 2005

product of 0.43. However, the actual FWHM, denotedby the arrows in Fig. 5, is approximately 3.3 THzs26 nmd. This corresponds to a time–bandwidth prod-uct of 0.34, which is close to 0.32 for a sech pulse. Foran input power level of 333 mW sN=1.15d, the pulsewas further compressed to 82 fs; however, the outputpulse had broad and rippled pedestals in the tail.This is because the wings of the spectrum had al-ready broadened into the normal GVD regime as aresult of fourth-order dispersion in the PM-DF-DDF.The optimum pulse compression with the DDF is de-termined by the broadened spectrum remaining inthe anomalous GVD region. This can be confirmed inthe pulse compression experiment by combining theMLFL with frequency modulation and the sameDDF.6 In this case we also obtained the shortestpulse width of 104 fs with no pedestal, which indi-cates that regardless of the input pulse condition theoptimum output is the same because of the fixedanomalous GVD bandwidth.

In conclusion, we have successfully obtained a40-GHz, 100-fs pulse train with low pedestalsthrough the use of the adiabatic soliton compressioneffect in a PM-DF-DDF. The input optical source wasa mode-locked laser diode that had a longitudinallinewidth as broad as 60 MHz. This broad linewidthmade it possible to achieve stable, ultrahigh-speedpulse compression without any input power limita-tion caused by SBS. In this case there was no need toemploy frequency modulation.

Part of this work was undertaken under the man-agement of the Femtosecond Technology Association,supported by the New Energy and Industrial Devel-opment Organization. T. Hagiuda’s e-mail address ishagiuda@riec.tohoku.ac.jp.

References

1. S. V. Chernikov and P. V. Mamyshev, J. Opt. Soc. Am. B8, 1633 (1991).

2. M. Nakazawa, E. Yoshida, H. Kubota, and Y. Kimura,Electron. Lett. 30, 2038 (1994).

3. K. R. Tamura and M. Nakazawa, Opt. Lett. 26, 762(2001).

4. K. R. Tamura and K. Sato, Opt. Lett. 27, 1268 (2002).5. E. P. Ippen and R. H. Stolen, Appl. Phys. Lett. 21, 539

(1972).6. T. Hirooka, S. Ono, K. Hagiuda, and M. Nakazawa,

Opt. Lett. 30, 364 (2005).7. S. Choi, M. Yoshida, and M. Nakazawa, IEICE Trans.

Electron. J86-C, 1054 (2003).8. M. Denariez and G. Bret, Phys. Rev. 171, 160 (1968).9. C. H. Henry, IEEE J. Quantum Electron. 18, 259

(1982).10. K. Haneda, M. Nakazawa, H. Yokoyama, and Y.

Ogawa, paper TuE39 presented at the 14thInternational Conference on Ultrafast Phenomena,July 25–30, Niigata, Japan (2004).

11. S. Arahira, K. Mineo, K. Tachibana, and Y. Ogawa,

Fig. 5. Pulse compression characteristics: (a) Input andoutput autocorrelation waveforms. Inset, logarithmic ex-pression of the center of the waveforms. (b) Output spec-trum (solid curve) and its sech fit (dotted curve).

Fig. 4. (a) Optical spectrum of the backscattered lightwith a 40-GHz MLLD. (b) Transmitted and backscattered

Electron. Lett. 39, 287 (2003).