March 15, 2006 / Vol. 31, No. 6 / OPTICS LETTERS 751

Passively Q-switched fiber laser that usessaturable Raman gain

Yucheng Zhao and Stuart D. JacksonOptical Fibre Technology Centre, Australian Photonics Cooperative Research Centre, University of Sydney,

206 National Innovation Centre, Australian Technology Park, Eveleigh, Sydney, New South Wales 1430, Australia

Received September 13, 2005; revised November 22, 2005; accepted November 30, 2005; posted December 6, 2005 (Doc. ID 64781)

Operation of a short all-fiber passively Q-switched Raman laser pumped by a continuous-wave laser diode isexperimentally demonstrated. The passively switched fiber laser consists simply of a double-clad ytterbium-doped silica fiber that is spliced directly to a moderately germanium-doped silica fiber. The placement of theGe-doped silica fiber within the fundamental (Raman pump) cavity allows interplay between fundamentaland Stokes fields to take place, which leads to saturation of the Raman gain as a result of pump depletion.Pulse widths of 70 and 60 ns at the first and second Stokes wavelengths of 1168 and 1232 nm, respectively,are produced at a stable 588 kHz repetition rate. © 2006 Optical Society of America

OCIS codes: 140.3510, 140.3540, 140.3550.

Fiber lasers have emerged as a practical alternativeto conventional bulk solid-state lasers in manyapplications.1 The main advantages relative to con-ventional bulk solid-state lasers include high beamquality, high efficiency, compactness, robustness, andexcellent heat dissipation. Raman fiber lasers are ofgreat importance to the laser field because they gen-erate a large number of emission wavelengths thatare not offered by rare earth ions when they aredoped into a silica host. Passive Q-switching tech-niques provide a convenient way to generate intenselight pulses at these new wavelengths, and there area variety of ways to achieve passive Q switching, e.g.,by incorporating saturable absorber materials intothe laser cavity.2 Rayleigh scattering (RS) and stimu-lated Brillouin scattering (SBS) backscattering pro-cesses have been recently exploited as passiveQ-switching mechanisms to generate short pulses.3

The stochastic nature of RS and SBS, however, leadsto unstable repetition rates, amplitudes, and pulsewidths.4 A stable and short all-silica fiber-based pas-sively Q-switched Raman laser therefore is impor-tant, particularly to the large variety of applicationsthat require laser pulses emitted over a wide wave-length range.

In this Letter we report the experimental demon-stration of a passively Q-switched fiber laser that em-ploys intracavity saturable Raman gain as the pro-cess that enables switching of both the fundamentaland the Stokes fields. The mechanism underlying thecurrent passive Q-switching method differs in an im-portant way from existing Q-switched Ramanlasers.3,4 Stimulated Raman scattering (SRS) in thepresent case acts as the nonlinear scattering processfor the fundamental field while at the same time act-ing as the gain mechanism for the Stokes emissions.Unlike the previous demonstration of a cw Raman fi-ber laser5 that had the Ge-doped silica fiber spliced tothe Yb3+-doped double-clad fiber laser (YDCFL) buthad resonators separated by a Bragg grating, in thepresent case, we place the Raman gain fiberdiscretely within the fundamental lasing cavity. As aresult, a dynamic occurs between the fundamental

0146-9592/06/060751-3/$15.00 ©

and the Stokes fields that causes pulsing of both out-puts in a systematic fashion.

A schematic diagram of the experimental setup isshown in Fig. 1. The passively Q-switched fiber laserconsisted of a 975 nm diode pump laser, a high reflec-tivity (�99% at 1040 to 1190 nm) broadband dielec-tric mirror (labeled M) butted against the input endto the double-clad Yb3+-doped fiber, which was�36 m long, tapered at the output end, and splicedwith a loss of �0.1 dB to a moderate length of highRaman gain Ge-doped fiber. The YDCFL was not op-timized for high efficiency before splicing it onto theGe-doped fiber. The YDCFL can operate in a broadspectral range, i.e., between 1040 and 1120 nm, us-ing a fiber end reflector (FER) as one cavity mirror6;however, in the present case, the Q-switched fiber la-ser uses FER from the facet of the cleaved Ge-dopedsilica fiber.

We fabricated the Raman fiber with �18 mol.%germanium doping to increase the Raman gain perunit length of fiber because germinate glass has oneof the the highest Raman cross sections among thewidely used vitreous Raman materials SiO2, GeO2,B2O3, and P2O5.7 This fiber has been used in severalrecent demonstrations of compact Raman fiberlasers5 and had a mode-field diameter (MFD) of3.87 �m and an effective area Aeff=12 �m2 at1300 nm. The estimated loss and Raman gain coeffi-cient were measured to be �3.3 dB/km and�10.1 dB/km W, respectively, at 1230 nm. The well-known Stokes shift of the Ge-doped silica fiber was�13.2 THz.

The spectral characteristics of the Q-switched fiberlaser output were measured with an Ando AQ6317optical spectrum analyzer. Figure 2 shows a typicalspectrum of the output. The fundamental Q-switchedlasing wavelength was 1112 nm, the first Stokeswavelength was 1168 nm, and the second Stokeswavelength was 1232 nm. The 56 and 64 nm wave-length separations between the laser lines and theStokes lines are in good agreement with the 13.2 THzRaman frequency shift. (The long operating wave-

length of the YDCFL output suggests that the �36 m

2006 Optical Society of America

752 OPTICS LETTERS / Vol. 31, No. 6 / March 15, 2006

long fiber was longer than the optimal length.) Sincefiber Bragg gratings are not used to select the funda-mental and Stokes wavelengths, the 3 dB band-widths were wider than 2 nm.

The average output power for each Stokes emissionas a function of the launched diode power is shown inFig. 3. The measured overall slope efficiency, i.e., thetotal Raman output measured with respect to thelaunched diode power, was �24%. Compared withthe �72% slope efficiency measured for the recentlydemonstrated Raman fiber laser that also usesbroadband pumping,6 the comparatively lower slopeefficiency in the present case, which occurs as a re-sult of splicing the Ge-doped silica fiber directly tothe YDCFL, was due to the high cavity loss for thefundamental wavelength of 1112 nm. Initially, whenthe launched diode power exceeded the fundamentallasing threshold of 2.5 W, the fundamental wave-length operated in cw mode and the output power in-creased linearly until the power of the fundamentalreached the first Stokes threshold of 1.6 W (which

Fig. 1. Experimental setup of a passively Q-switchedthree-wavelength Raman fiber laser: M, dielectric mirror;FER, fiber end reflector; Ps, prisms; OSA, optical spectrumanalyzer; Ds, optical detectors; CFL, fundamental laser cav-ity; CRS, Raman Stokes cavity; �FL, fundamental laserwavelength; �RS, Raman Stokes wavelengths.

Fig. 2. Measured optical spectrum for the passivelyQ-switched three-wavelength Raman fiber laser.

corresponds to a launched diode power of 7.2 W).

Upon increasing the launched diode power further,all three lasing wavelengths (fundamental, firstStokes, and second Stokes) operated in passivelyQ-switched modes. The maximum average outputpower for the fundamental, first Stokes, and secondStokes was 1.6, 1.3, and 1.25 W, respectively.

A Tektronix TDS 794D digital oscilloscope moni-tored the pulses that were emitted from the passivelyQ-switched laser. A synchronized comparison of theshapes of the pulses at the three wavelengths isshown in Fig. 4. The duration of the Q-switchedpulses varied for the fundamental and the Stokesemissions: the pulse widths for the fundamental, firstStokes, and second Stokes were 150, 70, and 60 ns,respectively. The pulse widths were found to increaseas the length of Ge-doped silica fiber increased. Themaximum pulse energies for the fundamental, firstStokes, and second Stokes emissions were �2.7,�2.2, and �2.0 �J, respectively, at the maximum di-ode power of 12 W. Because SRS is a nonlinear pro-cess, we observe pulse narrowing of the subsequentStokes emissions. The time delay between the funda-mental pulse and the first Stokes pulse, and the sec-ond Stokes pulse and the first Stokes pulse shown inFig. 4 was 180 and 20 ns, respectively. When thelength of Ge-doped fiber was reduced from134 to 54 m for the constant Yb3+-doped fiber lengthof �36 m, the time delay was reduced from180 to 50 ns. (Note that the threshold for pulsation isalso increased when the shorter Ge-doped fiber isused). Likewise when the length of the Yb3+-doped fi-ber was reduced from 36 to 19 m and the Ge-dopedfiber length was kept at �54 m, the time delay be-tween the fundamental pulse and first Stokes pulsechanged from 50 to 20 ns. These results suggest thatthe time delay between the first Stokes pulse and thefundamental pulse is related to the length of both theYb3+-doped and Ge-doped fibers.

All three pulsed laser emissions operated with thesame repetition rate of 588 kHz. When the 54 m longGe-doped silica fiber was used, thus providing a 90 mlong cavity, a pump-insensitive pulse repetition rateof 1.124 MHz was obtained. Overall, the repetition

Fig. 3. Measured average output power of the fundamen-tal and the Stokes emissions versus the 975 nm launched

diode power.

March 15, 2006 / Vol. 31, No. 6 / OPTICS LETTERS 753

rate of the output, f, was related to the cavity round-trip time in which f=c /2nL, where c is the light ve-locity in vacuum, L is the length of the laser cavity,and n is the refractive index of the laser medium.

The inset to Fig. 4 shows the pulse characteristicsthat lead up to the emission of the Q-switched Stokespulses. Prior to the emission of the Q-switched firstStokes pulse, we observe that the fundamental emis-sion comprises a number of small pulses, which maybe considered as relaxation oscillations that becomelimited with the onset of the Q-switched first Stokespulse. As expected, as the diode power was increased,the period between these relaxation oscillations de-creased.

Based on the measured experimental results andobserved phenomena, we can describe the passiveswitching process by the following simplified model.Before the fundamental intensity reaches the SRSthreshold, the fundamental emission operates in cwmode with Raman scattering acting as a loss processwithin the YDCFL cavity as shown in Fig. 1. In thiscase, the threshold is high and the slope efficiency islow compared with previous demonstrations of simi-lar Raman fiber laser arrangements.6 After pulsedemission is observed, the fundamental field builds upas series of relaxation oscillations until one of the os-cillations exceeds the SRS threshold. Note that in theinset to Fig. 4 some first Stokes emission is observedprior to the emission of the Q-switched first Stokespulse. The first Stokes pulse builds up relatively

Fig. 4. Pulse characteristics of the passively Q-switchedthree-wavelength fiber laser.

quickly; however, as the intracavity first Stokespower is increasing, the threshold for the secondStokes emission is reached, and a Q-switched secondStokes pulse is emitted. The Stokes pulses depletethe power within the fundamental laser field, an ef-fect that ultimately causes the cessation of theStokes pulses. Some residual fundamental emissionis observed after the cessation of the Stokes pulsesbecause of the reduced Stokes losses acting on thefundamental field. This entire process is repeated ac-cording to the round-trip time of the cavity. Since theinterplay between the Stokes and the fundamentalfields is intricate, to fully elucidate the complex be-havior of this laser will require a numerical model.

In conclusion, we have experimentally demon-strated, for the first time to our knowledge, a pas-sively Q-switched Raman fiber laser that uses pump-depleted SRS to provide the saturation mechanismnecessary for passive switching. With an overallslope efficiency of �24%, we have generated averageoutput powers of 1.6, 1.3, and 1.25 W at the funda-mental wavelength of 1112 nm, the first Stokeswavelength of 1168 nm, and the second Stokes wave-length of 1232 nm, respectively. The Q-switched Ra-man fiber laser produced pulse widths of 150, 70, and60 ns, respectively, for the three wavelengths at astable repetition rate of 588 kHz. This demonstrationhas opened a new pathway toward the generation ofmultiwavelength pulsed output.

The authors gratefully acknowledge financial sup-port from the Australian Research Council and theAustralian Photonics Cooperative Research Centre.Y. Zhao’s e-mail address is y.zhao@oftc.usyd.edu.au.

References

1. J. A. Alvarez-Chavez, H. L. Offerhaus, J. Nilsson, P. W.Turner, W. A. Clarkson, and D. J. Richardson, Opt.Lett. 25, 37 (2000).

2. U. Keller, Nature 424, 831 (2003).3. S. V. Chernikov, Y. Zhu, J. R. Taylor, and V. P.

Gapontsev, Opt. Lett. 22, 298 (1997).4. G. Ravet, A. A. Fotiadi, M. Blondel, and P. Megret,

Electron. Lett. 40, 528 (2004).5. Y. Zhao, Y. H. Li, and S. D. Jackson, Opt. Express 12,

4053 (2004).6. Y. Zhao and S. D. Jackson, Opt. Express 13, 4731

(2005).7. F. L. Galeener, J. C. Mikkelsen, Jr., R. H. Geils, and W.

J. Mosby, Appl. Phys. Lett. 32, 34 (1978).