26 OPTICS LETTERS / Vol. 32, No. 1 / January 1, 2007

10-W-level diode-pumped compact 2.78 �mZBLAN fiber laser

Xiushan Zhu and Ravi JainCenter for High Technology Materials, The University of New Mexico, 1313 Goddard SE, Albuquerque,

New Mexico 87106, USA

Received July 26, 2006; revised October 4, 2006; accepted October 9, 2006;posted October 11, 2006 (Doc. ID 73481); published December 13, 2006

We report on �9 W transverse-fundamental-mode CW output near 3 �m from a 4 m heavily erbium-dopedZBLAN double-clad fiber laser pumped by a collimated 100 W 975 nm laser diode array. The pump thresholdof the fiber laser was about 1 W, and the slope efficiency was 21.3%. The peak wavelength of free runningwas about 2708 nm at low pump power and moved to around 2785 nm at high pump power. Output of 9 Wwas obtained when the launched pump power was 42.8 W. The output, however, fluctuated intensively likea pulsed laser, and the operation broke down with optical damage of the pumping end facet when the pumpwas increased beyond 42.8 W. Therefore, alleviation of the operation fluctuation, heat management, andstrengthening the pumping fiber are crucial considerations for the stable operation of 10-W-level mid-IRZBLAN fiber lasers. © 2006 Optical Society of America

OCIS codes: 060.2320, 140.3070, 140.3510.

Erbium- (Er-) doped ZBLAN mid-infrared (mid-IR) fi-ber lasers, which have broad emissions from 2.6 to2.9 �m (Ref. 1) through the 4I11/2→ 4I13/2 transition ofEr ions, recently have been paid much attention be-cause of their potential applications to spectroscopy,2

laser surgery,3,4 and optical pump sources for longer-wavelength mid-IR oscillators. 10-W-level diode-pumped 2.7 �m Er-doped ZBLAN fiber lasers withdiffraction-limited output are promising candidatesfor laser scalpels with high ablation efficiency and ul-trafine resolution (several micrometers)4 and for highefficiency optical pump sources for mid-IR Ramanlasers,5 mid-IR optical parametric oscillators,6

longer-wavelength mid-IR gas lasers, and THz localoscillators.7 Although various developments of Er-doped ZBLAN fiber lasers8–14 have been demon-strated, their maximum outputs were not beyond2 W. In 2004, we reported on a 5.4 W Er/Pr co-dopedZBLAN fiber laser in which the power scaling was re-alized through a dual end pumping configuration.15

In this Letter, we report on a �9 W diode-pumpedheavily Er-doped ZBLAN fiber laser and investigatethe obstacles of output scaling of 2.7 �m fiber laserswith recent, commercially available high-power diodelasers and custom-designed Er-doped ZBLAN fibers.

Physically, the longer natural lifetime of the lowerlaser level (4I13/2, �1=9 ms) relative to that of the up-per laser level (4I11/2, �2=6.9 ms) of Er ions in ZBLANglass results in a population bottleneck that inhibitsefficient steady-state lasing and limits the outputpower in Er-doped ZBLAN fiber lasers.16 Cascadelasing,8 the energy-transfer (ET) process between Erions and co-doped praseodymium (Pr) ions,9–12 andthe energy-transfer upconversion (ETU) process be-tween Er and Er ions,13,14 which are depicted in Fig.1, have been demonstrated as being able to overcomethe population bottleneck. Cascade lasing is efficientonly in low-doped, core-pumped fiber lasers8 and can-not be used for low-brightness diode- pumped high-power (watt-level) mid-IR fiber lasers, because thepopulation inversion is established due to the large

0146-9592/07/010026-3/$15.00 ©

branch ratio for the transition from 4I11/2 to 4I15/2��20=0.63�, and the continuous lasing is supported bythe excited-state absorption (ESA) of Er ions at the4I13/2 level and strong 1.7 �m transition from 4S3/2 to4I9/2. Er-Pr ET and enhanced Er-Er ETU from 4I13/2(ET1 and ETU1) are larger than their correspondingprocesses from 4I11/2 (ET2 and ETU2), so they can beused to solve the population bottleneck efficientlyeven with low-brightness pumping and can be usedfor high-power diode-pumped Er-doped ZBLANdouble-clad fiber lasers. Consequently, high-concentration singly Er-doped ��1 mol. % � ZBLANdouble-clad fibers and Er/Pr co-doped ZBLANdouble-clad fibers are two alternative gain media forwatt-level or 10-W-level mid-IR lasers operating at2.7 �m. Technically, increasing the absorbed pumppower and improving the fiber laser efficiency aretwo direct methods to scale up the output of fiber la-sers. Although �10 mol.% singly Er-doped ZBLANfiber lasers are predicted to possibly reach a slope ef-ficiency close to or even larger than the Stokeslimits,7 much work is still needed to determine ex-

Fig. 1. Partial energy-level scheme of Er and Pr indicatingthe processes of cascade lasing, energy transfer (ET1�ET2), and energy-transfer upconversion (ETU1�ETU2)to deplete the population of 4I13/2 and solve the population

bottleneck.

2006 Optical Society of America

January 1, 2007 / Vol. 32, No. 1 / OPTICS LETTERS 27

perimentally the optimum concentration for maxi-mum efficiency and to fabricate low-loss ZBLAN fi-bers with such high concentration. In this Letter, wereport on our investigation of the output scalability of2.7 �m fiber lasers by using an advanced, singly Er-doped ZBLAN double-clad fiber.

The experimental setup for our heavily Er-dopedZBLAN fiber laser is depicted in Fig. 2. A 100 W975 nm collimated laser diode (Apollo Instruments,C100-975-0) with a beam size of 20 mm�16 mm andwith a divergence of 4 mrad�4 mrad was used as thepump source. The collimated laser diode had dimen-sions of 164 mm�164 mm�92 mm, which made ourfiber laser very compact and simple. The custom-designed active fiber (fabricated by FiberLabs, Inc.)was a piece of 4 m singly Er-doped ZBLAN double-clad fiber. The core [diameter 15 �m, numerical aper-ture (NA) 0.12] had an Er3+ concentration of 6 mol.%.The inner rectangle cladding of the fiber had dimen-sions of 200 �m�250 �m and a NA of 0.55. The outercircular cladding had a diameter of 370 �m. The ab-sorption coefficient of this singly Er-doped ZBLAN fi-ber was about 2.3 dB/m �0.54 m−1� as measured by acutback experiment. The collimated pump beam wascoupled to the inner cladding of the fiber by aspherical-aberration-corrected silica lens assemblythat had a diameter of 32 mm and a focal length of30 mm. Every side of the lens assembly was antire-flection coated at the pump wavelength. The spot sizeat the lens focal point was about 250 �m. In front ofthe pumping end of the fiber, a dichroic mirror (highreflection �99.5% at 2.7 �m, high transmission�95% at 975 nm) was placed with an angle of inci-dence of 45° to couple out the laser beam. Behind theother end of the fiber, a sapphire lens was used to col-limate the laser beam, and a highly reflective mirrorwith an angle of incidence of 0° was placed after thesapphire lens to provide high reflection at laser andpump wavelengths. To purify the output, a long-wavelength pass filter (transmission �95% at ��2.5 �m) was placed before the detector. The outputspectrum of the fiber laser was measured by a mono-chromator of focal length 50 cm. To reduce the ther-mal damage and the unstable performance of the fi-ber laser (output fluctuation) induced by thermal-optical effects in the pumping end of the fiber, we

Fig. 2. Experimental setup for the high-power heavily Er-doped ZBLAN double-clad fiber laser.

used an actively water-cooled fiber chunk to hold the

double-clad fiber, with �1 cm stripped fiber protrud-ing. We found that active water cooling significantlyprevents softening and bending of the pumping fiberend and somewhat improves the stability of the fiberlaser.

The output power of the singly Er-doped ZBLANdouble-clad fiber laser as a function of the launchedpump power is shown in Fig. 3. The threshold pumppower was about 1 W. Output of 9 W was obtainedwhen the launched pump power was 42.8 W, corre-sponding to a slope efficiency of 21%, which is smallerthan half of the simulation value.17 The big differencebetween our experimental result and Pollnau’s theo-retical value is caused mainly by a larger backgroundloss of the used fiber �500 dB/km� compared withthat of the simulation fiber [100 dB/km (Ref. 17)] andby an approximation of dividing the fiber cavity intoonly eight cascaded short cavities used in the simu-lation. Comparing this experiment with our previousexperiment,13 the slope efficiency of singly Er-dopedZBLAN fiber lasers, however, was increased greatlyby using higher dopant density. This fact indicatesthat, as predicted by the simulation, higher slope ef-ficiency may be obtained when the fiber has a higherEr concentration. However, the background loss ofZBLAN fiber and the effects of processes that impairthe laser efficiency (such as the ESA of the upper la-ser level 4I11/2 and the cross-relaxation of the lowerlaser level 4I13/2) will also increase at a higher dopantdensity. Thererfore, it is most likely that there is anoptimum dopant density for singly Er-doped ZBLANfiber lasers, similar to the case of Er:LiYF4 crystallasers.18 The output spectra at different pump pow-ers are shown in Fig. 4. When the pump power was5 W, the laser peak was at 2708 nm and broad ampli-fied spontaneous emission (ASE) also was observedaround 2780 nm. The peak wavelength moved toaround 2785 nm and strong ASE was observed whenthe pump power was lager than 10 W. The strongASE of the free-running heavily Er-doped ZBLAN fi-ber laser may result from a large population in thesublevels of 4I11/2. The strong ASE can be restrainedeffectively and the output spectra can be improvedgreatly by using a bulk grating or a fiber Bragggrating19 in the laser cavity.

Fig. 3. Output power as a function of launched pump

power.

28 OPTICS LETTERS / Vol. 32, No. 1 / January 1, 2007

The damage threshold of rare-earth-doped ZBLANglass is about 25 MW/cm2, as measured by using10 ms long pulses at 2.8 �m, so the damage power ofour active ZBLAN fiber with a diameter of 15 �mshould be about 44 W. Theoretically, �10 W output isachievable just by increasing the pump power. Ourexperimental results, however, showed that when thepump power was increased the output began to fluc-tuate chaotically, as shown in Fig. 5. When the out-put power was �9 W (shown in Fig. 3), the fluctua-tion became periodic like a pulsed laser, andconsequently the pumping fiber end was likely toburn. Nonlinear effects and the effect of ESA inZBLAN fiber lasers20 may be taken into account asan explanation for the generation of harmful fluctua-tion and final pulsed operation. We found that activecooling can prevent softening and bending of thepumping fiber21 and somewhat improves the stabilityof the laser performance; however, in large part, itcannot overcome the output fluctuation. Therefore,reducing or eliminating the fluctuant operation of theEr-doped ZBLAN fiber laser is very important for ob-taining 10 W or tens of watts output. In addition, thefragility, low damage threshold, and low meltingpoint �265°C� of ZBLAN fibers are obstacles for ob-taining 10 W or tens of watts mid-IR fiber lasers.Therefore, increasing the damage threshold of the fi-

Fig. 4. Output spectra of the singly Er-doped ZBLAN fiberlaser at different pump powers.

Fig. 5. Output of the singly Er-doped ZBLAN fiber laser atdifferent pump powers.

ber (e.g., large-mode-area fibers), good heat manage-

ment (e.g., efficient cooling), and strengthening thestripped pumping fiber end (e.g., metallizing the fi-ber) should be considerations for the power scaling ofmid-IR ZBLAN fiber lasers.

In conclusion, we have produced nearly 10 W out-put near 3 �m from an advanced heavily Er-dopedZBLAN double-clad fiber laser. Our experimentsdemonstrated that, using a dopant density of6 mol.%, the slope efficiency of the singly Er-dopedZBLAN fiber laser was increased to 21%, and evengreater efficiency is likely to be obtained at a higherEr concentration. To find the optimum dopant den-sity that can provide maximum efficiency, however,experimental investigations of higher Er concentra-tions are necessary.

This work was supported financially by NASA, theAir Force Office of Scientific Research, and the Opto-electronics Industry Development Association. Theauthors thank IR Photonics, Inc., for providing thedamage threshold of rare-earth-doped ZBLAN fibers.X. Zhu’s e-mail address is xshan@email.arizona.edu.References

1. F. Auzel, D. Meichenin, and H. Poignant, Electron.Lett. 24, 909 (1988).

2. P. Werle, F. Slemr, K. Maurer, R. Kormann, R. Mucke,and B. Janker, Opt. Lasers Eng. 37, 104 (2002).

3. R. Kaufmann, A. Hartmann, and R. Hibst, J.Dermatol. Surg. Oncol. 20, 112 (1994).

4. S. D. Jackson and A. Lauto, Lasers Surg. Med. 30, 184(2002).

5. P. A. Thielen, L. B. Shaw, J. S. Sanghera, and I. D.Aggarwal, Opt. Express 11, 3248 (2003).

6. T. H. Allik, S. Chandra, D. M. Rines, P. G.Schunemann, J. A. Hutchinson, and R. Utano, Opt.Lett. 22, 597 (1997).

7. W. Benedict, M. Pollack, and W. Tomlinson, IEEE J.Quantum Electron. 5, 108 (1969).

8. M. Pollnau, Ch. Ghisler, G. Bunea, M. Bunea, W.Luthy, and H. P. Weber, Appl. Phys. Lett. 66, 3564(1995).

9. J. Y. Allan, M. Monerie, and H. Poigant, Electron. Lett.27, 445 (1991).

10. B. Srinivasan, J. Tafoya, and R. K. Jain, Opt. Express4, 490 (1999).

11. S. D. Jackson, T. A. King, and M. Pollnau, Opt. Lett.24, 1133 (1999).

12. S. D. Jackson, T. A. King, and M. Pollnau, Electron.Lett. 36, 223 (2000).

13. E. Poppe, B. Srinivasan, and R. K. Jain, Electron. Lett.34, 2331 (1998).

14. B. Srinivasan, E. Poppe, J. Tafoya, and R. K. Jain,Electron. Lett. 35, 1338 (1999).

15. X. Zhu and R. Jain, in IEEE/LEOS 2004 AnnualMeeting (IEEE, 2004), paper ThP5.

16. S. Bedo, M. Pollnau, W. Luthy, and H. P. Weber, Opt.Commun. 116, 81 (1995).

17. M. Pollnau and S. D. Jackson, IEEE J. QuantumElectron. 38, 162 (2002).

18. T. Jensen, A. Diening, and G. Huber, Opt. Lett. 21, 585(1996).

19. N. J. C. Libatique, J. Tafoya, N. K. Viswanathan, R. K.Jain, and A. Cable, Electron. Lett. 36, 791 (2000).

20. F. Z. Qamar and T. A. King, Appl. Phys. B 81, 821(2005).

21. D. J. Coleman and T. A. King, Meas. Sci. Technol. 14,

998 (2003).