April 1987 / Vol. 12, No. 4 / OPTICS LETTERS 269

Brillouin-gain spectra for single-mode fibers having pure-silica,GeO2-doped, and P2 05-doped cores

Nori Shibata

NTT Electrical Communications Laboratories, Nippon Telegraph & Telephone Corporation, Tokai, Ibaraki-ken, 319-11, Japan

Robert G. Waarts

Weizmann Institute of Science, Rehovot, Israel

Ralf P. Braun

Heinrich-Hertz-lnstitut fUr Nachrichtentechnik Berlin GmbH, Einsteinufer 37, D-1000 Berlin 10, Federal Republic of Germany

Received August 20, 1986; accepted January 7, 1987

Brillouin-gain spectra are measured for pure-silica core, GeO 2 -doped core, and P2 05-doped core single-mode fibers

with different index profiles. A narrow-linewidth semiconductor laser operating at 828 nm is used as the pumplight source. The spectral shape of the Brillouin gain is found to be strongly related to the refractive-index profiles.The Brillouin linewidths evaluated experimentally are 90 and 215 MHz for a step-index single-mode fiber and a

graded-index single-mode fiber, respectively. The Brillouin Stokes shifts depend on the core/clad-ding dopantmaterials and their concentrations and range from 20 GHz for the P2 05 -doped core fiber to 21.6 GHz for the pure-silica core fiber in the 0.8-,um-wavelength region.

As an optical nonlinearity in fibers, stimulated Bril-louin scattering (SBS) has a significant influence onthe operation of optical transmission systems usingnarrow-linewidth single-frequency lasers.1 Waartsand Braun2 pointed out a negative feature of SBS as across-talk factor in frequency-multiplexed coherentcommunication systems.3'4 Accordingly, carrier-channel frequency separation should be carefully cho-sen for coherent communication systems operating atwavelengths of 0.8, 1.3, and 1.55 Aim. Bit-error-ratedegradation due to SBS was also observed in the co-herent single-mode transmission system.3 On theother hand, as an active feature of SBS, Olsson andVan der Ziel5 demonstrated the first operation of afiber Brillouin amplifier pumped by a semiconductorlaser operating at 1.5 jm and reported cancellation ofthe fiber loss plus 5 dB of net gain in a 37.5-km-longsingle-mode transmission line. Inhomogeneities inthe five fiber spools were applied to broaden the Bril-louin-gain linewidth artificially. Hence precise mea-surements of the Brillouin-gain spectra and the fre-quency shift from the pump frequency are of greatimportance from the viewpoint of system design inconstructing frequency-multiplexed coherent commu-nication networks and semiconductor-laser-pumpedfiber Brillouin amplifiers.

In this Letter we show Brillouin-gain spectra forsingle-mode optical fibers having various core dopantmaterials. The measurements were carried out at828-nm wavelength by a highly sensitive techniqueutilizing a heterodyne receiver and lock-in detection.2

A similar technique has been applied for measuringoptical powers generated through the three-wave mix-ing process in a coherent single-mode transmissionline.6

We prepared three kinds of test fiber for measuringBrillouin-gain spectra. Waveguide parameters andtransmission characteristics of the test fibers areshown in Table 1. Fibers A, B, and C have a pure-silica core, a 3.2 mol % GeO 2 -doped core, and a 9.9 mol %P205-doped core, respectively. Refractive-index pro-files of the fibers are shown in Fig. 1. Fibers A and B,with a length L of 2.1 km, have graded-index and step-index profiles without core center dip, and fiber C,with L = 1.48 km, has a step-index profile with corecenter dip.

The experimental setup for measurement of theBrillouin-gain spectrum is shown in Fig. 2. The pumpand the probe lasers operating at 828 nm are chan-neled-substrate planar lasers emitting light with alinewidth of approximately 20 MHz. After passingthrough a mechanical chopper, the pump source candeliver approximately 1 mW of optical power into sin-gle-mode fibers A, B, and C with respective cutoffwavelengths of 780, 760, and 850 nm. The two lasers

Table 1. Waveguide Parameters and TransmissionCharacteristics of Test Fibers A-C

Items Fiber A Fiber B Fiber C

Core diameter 5.8 gm 5.1 ,m 4.2 AmIndex difference 0.30% 0.32% 0.50%Cutoff wavelength 780 nm 760 nm 850 nmFiber length 2.1 km 2.1 km 1.48 kmCore dopant Pure silica GeO2 P2 05

(3.2 mol %) (9.9 mol %)Cladding dopant F Pure silica P2 05 -FOptical loss (dB/km)a 2.1 2.3 8.1

a Measured values at X = 828 nm.

0146-9592/87/040269-03$02.00/0 © 1987, Optical Society of America

270 OPTICS LETTERS / Vol. 12, No. 4 / April 1987

S0 2 Core

Fiber A

GO-Doperd

C-,

Fiber B

P2F5 'pmdCo,

Fiber C

Fig. 1. Refractive-index profiles of test fibers A-C.

ProbeLaser

Isolat

F[1lLosol LoserSingle-Mode LUt |

FiberIsolator IAFC

Coupler I

Coupler

Chopper Amp___ _

F Isolator | Rrder|

m Pump Laser

Fig. 2. Experimental arrangement for measuring Brillouingain spectrum. AFC, autofrequency control.

serve for the two counterpropagating waves in thesingle-mode fiber transmission line constructed by fi-bers A, B, and C. Optical isolators, with 27 dB ofisolation, are inserted between the lasers and the fiber.The probe laser launches 10-100,uW of power into thefront end of the transmission line. At the receiver endthe transmitted light from the probe laser is hetero-dyned with the light of the local laser. The intermedi-ate frequency is maintained at 1.5 GHz. The com-bined spectrum resulting from the probe and thepump lasers was observed on a grating-spectrum ana-lyzer. The frequency difference Af between thepump-laser frequency fpump and the probe-laser fre-quency fprobe was adjusted to be approximately 20GHz. A power transfer from the chopped light of thepump laser into the optical frequency fprobe would ap-pear as a chopped signal at the heterodyne receiver,and this signal component was detected with a lock-inamplifier. The frequencyr difference Af was sweptover a range of about 3 GHz. The output waveformfrom the lock-in amplifier'as a function of the frequen-cy fprobe gives the Brillouin-gain profile.

To measure the Brillouin Stokes shifts for fibers A-C, the fibers were spliced with an arc-fusion splicemachine. Figure 3 shows the Brillouin-gain spectrafor the sequentially spliced fiber by fibers A and C andfor that by fibers A-C. The vertical scale gives theBrillouin gain G in arbitrary units. The Brillouinshifts for the peak gain were found to be 21.6,21.2, and20.0 GHz for fibers A, B, and C, respectively. Themaximum Brillouin shift of 21.6 GHz was observed forthe pure-silica core fiber (fiber A), and the Brillouinshift was made to decrease by doping GeO2 and P20 5into the silica.

The Brillouin-gain spectra of the test fibers for dif-ferent values of pump power are shown in Fig. 4. Inthis figure the vertical scale was calibrated to give theBrillouin gain G (=APprobe/Pprobe). Here Pprobe andAPprobe are the probe power and the power differencebetween the Brillouin-amplified probe power andPprobe at the heterodyne receiver, respectively. Thegain G increases linearly with the input pump powerPpump. The measured linewidth A.vB, which is definedby the full width at half-maximum (FWHM) of thegain spectrum, is 275, 150, and 180 MHz for fibers A,B, and C, respectively. The measured gain spectra donot correspond to intrinsic Brillouin-gain spectra be-cause of the finite spectral width of the lasers used inthe experiment. Assuming that the laser line is aLorentzian profile of width AvL, after heterodyning atthe receiver the actual measured Brillouin linewidthcorresponds to a Lorentzian of width AVB + 3AVL re-sulting from convolving the laser linewidth with theBrillouin-gain spectrum.' As a result of the convolu-tion correction, we obtained the intrinsic Brillouinlinewidth AvB of 215, 90, and 120 MHz for fibers A, B,and C, respectively. The evaluated value AVB for fiberA is significantly different from the predictedlinewidth of 53 MHz for bulk fused silica.' We believethat the discrepancy between them is due to the inho-mogeneous structure of the optical fibers in the senseof cross-sectional index-profile variation. For a grad-ed-index single-mode fiber such as fiber A, the densitydistribution changes gradually over the core and clad-ding. Hence the Brillouin shift due to the contribu-tion from each part in the core-cladding cross sectionwill take different values. In fact, the linewidth forgraded-index fiber A is much greater than that forstep-index fiber B. Furthermore, the effective inter-action length in the optical fiber is much'longer thanthat of the bulk sample. As a result, waveguide fluc-tuation along the fiber will reflect the Brillouin-gainspectra. Inhomogeneities in the fiber cross-sectionalarea and along the fiber length broaden the linewidth.

is

06

Frequency Difference Af (GHz)-22 -21 -20

Optical Frequency fprobe

Fig. 3. Brillouin gain spectra obtained for the sequentiallyspliced transmission line composing test fibers.

,-pumpMeosurmd Io a, <

MecsurerJ for Sn +

prob I -*

SiO25i02 ts0

P20515.02

I

tor

April 1987 / Vol. 12, No. 4 / OPTICS LETTERS 271

CL

-0o

a.0

30

20

10

0

36

20

10

Frequency Difference Af (GHz)

-22 -21.5 -21

Optical Frequency fpobe(al

Frequency Difference Af (GHz)

-21.5 -21 -20.5

Fiber B Pm:Pr-20S94iWPp,6be.S.6PW / - 0.47WiWL.21 ke,

| . 024 W

.1 50 M H.

Opt ical Frequency fox

(bi

Frequency Difference Af (GHz)

.0

c

a(12

a.0

1 5

10

5

0

Optical Frequency f,,be.(c)

Fig. 4. Brillouin-gain spectra obtained for (a) fiber A, (b)fiber B, and (c) fiber C.

In Fig. 4(c) two resonances for fiber C with corecenter dip in the index profile are clearly visible. Forfibers A-C, the field power of the HE1, mode propa-gating in the cladding is approximately 20% of thetotal power at the operating V values, 2.2-2.5.7 Richand Pinnow8 reported observation of a Brillouin linedue to scattering from longitudinal acoustic phononsin the cladding for a silica-core-borosilicate-clad mul-timode fiber, and Thomas et al.9 and Waarts andBraun2 also reported such a Brillouin line for a single-mode fiber. Unfortunately, the index profiles of thefibers were not shown in Refs. 2, 8, and 9. In our case,a small-resonance Brillouin peak was observed onlyfor fiber C, and such a resonance peak could not beobserved for fibers A and B under approximately the

same V value. This indicates that the second small-resonance peak is due to scattering from the acousticphonons in the core center dip, where the field powerof the HE11 mode is mostly concentrated. A similartheory about the second resonance has been reportedby Tkach et al.,10 and our work supports theirs. As-suming a linear relation between dopant concentra-tion and the Brillouin-shift deviation from the shiftfor the pure silica, the Brillouin-shift coefficients, de-fined by the frequency shift per unit mol % concentra-tion, for GeO2 and P205 are evaluated from Fig. 3 to be125 and 162 MHz/mol %, respectively. The valueevaluated for GeO2 at an 828-nm wavelength coincideswith that of 89 MHz/wt % (= 154 MHz/mol %) (Ref.10) measured at 1.525 Am.

In conclusion, the Brillouin gain profiles were mea-sured for three types of fiber with different index pro-files. The spectral shape of the Brillouin gain wasfound to reflect the waveguide structure, especiallythe index profiles of the fiber. The experimental re-sults suggest that the Brillouin linewidth can be artifi-cially broadened by selecting the core-cladding dop-ant materials and by controlling the index profile.This work provides useful information for design andperformance evaluation of coherent communicationsystems employing frequency-division multiplexingand active transmission systems utilizing SBS.

The authors would like to express their sincerethanks to S. Seikai and M. Ohashi of NTT ElectricalCommunications Laboratories for providing the fibersused in our experiments and to C. Baack and B. Stre-bel of Heinrich-Hertz-Institut for their kind sugges-tions and continuous encouragement. We also wouldlike to thank the German Bundespost and the GermanMinistry for Research & Technology for partiallyfunding this work.

This research was carried out at Heinrich-Hertz-Institute fur Nachrichtentechnik Berlin GmbH.

Note that we have revised our Letter to include Ref.10 after we learned about it from one of the reviewers.Reference 10 was received for publication one weekbefore ours was received.

References

1. D. Cotter, J. Opt. Commun. 1, 10 (1983).2. R. G. Waarts and R. P. Braun, Electron. Lett. 21, 1114

(1985).3. E.-J. Bachus, R. P. Braun, W. Eutin, H. Foisel, E. Gross-

mann, K. Heims, and B. Strebel, Electron. Lett. 21,1203(1985).

4. T. G. Hodgkinson, D. W. Smith, R. Wyatt, and D. J.Malyon, Br. Telecommun. Technol. J. 3, 5 (1985).

5. N. A. Ollson and J. P. van der Ziel, Appl. Phys. Lett. 22,1329 (1986).

6. N. Shibata, R. P. Braun, and R. Waarts, Electron. Lett.22,675 (1986).

7. D. Gloge, Appl. Opt. 10, 2252 (1971).8. T. C. Rich and D. A. Pinnow, Appl. Opt. 13,1376 (1974).9. P. T. Thomas, N. L. Rockwell, H. M. van Driel, and G. I.

Stegeman, Phys. Rev. B 19,4986 (1979).10. R. W. Tkach, A. R. Charaplyvy, and R. M. Derosier,

Electron. Lett. 22, 1011 (1986).