August 15, 1999 / Vol. 24, No. 16 / OPTICS LETTERS 1133

Diode-pumped 1.7-W erbium 3-mm fiber laser

Stuart D. Jackson, Terence A. King, and Markus Pollnau*

Laser Photonics Group, Schuster Laboratory, Department of Physics and Astronomy,University of Manchester, Manchester M13 9PL, UK

Received May 14, 1999

We report what is to our knowledge the first 3-mm fiber laser of the 1-W class. 1.7 W of output power and17.3% slope eff iciency (with respect to the launched pump power) at a wavelength of 2.71 mm are demonstratedfrom a double-clad erbium-doped ZBLAN fiber diode pumped at 790 nm. Energy transfer from the Er31 lowerlaser level to a Pr31 codopant decreases ground-state bleaching and excited-state absorption, thus avoidingoutput-power saturation. This result represents more than an order-of-magnitude improvement over previouswork of which we are aware. Advantages over current crystal-laser designs include nearly transverse-fundamental-mode operation, reduced thermal effects, and ease of use, e.g., in medical endoscopy. 1999Optical Society of America

OCIS codes: 140.3460, 140.3500, 140.3480, 140.3510, 140.3570, 140.3070.

In recent years there have been enormous researchefforts to improve the performance of lasers emittingat 3 mm, mainly because of their potential applicationsin laser surgery.1 Because of the high absorptionof 3-mm radiation in water, high-quality cutting orablation has been demonstrated in biological tissueby use of erbium-doped solid-state lasers.2 From bulklaser sources, cw output power of 1.1 W at 2.8 mm wasdemonstrated from a diode-pumped Er31:LiYF4 laser,3

and a cw output of 1.15 W was recently obtained froma diode-pumped Er31:Y3Al5O12 laser.4 Unfortunately,further power scaling that is required for surgicalapplications has been hindered by thermal lensing anddegradation in output-beam quality as well as thefrequent occurrence of rod fracture.

The erbium-doped f luoride fiber is a promising can-didate for use in the construction of a compact andeff icient all-solid-state laser emitting on the tran-sition at 2.7 mm (Fig. 1). Because of its geometrythe fiber provides signif icant f lexibility and poten-tially high laser intensity without the drawbacks ofthermal and thermo-optic effects. A high excitationdensity, with the consequences of pump excited-stateabsorption5 (ESA) and uncontrolled redistribution ofthe upconverted energy, can, however, lead to output-power saturation in the fiber laser.6 This satura-tion was overcome by recycling of the upconvertedenergy in a cascade lasing regime, and 150-mW out-put power was achieved under Ti:sapphire pumping at791 nm.7 However, stringent demands on the pumpintensity prevent this system from being pumped withthe double-clad configuration because of the compara-tively lower brightness of diode laser sources. Diodepumping of an erbium-doped double-clad fiber resultedin similar output powers of �140 mW,8 but with fur-ther power scaling one might encounter the same prob-lems as those observed earlier.6

In this Letter we follow a theoretical proposal toscale the output power of the erbium 3-mm fiberlaser to the 1-W region.9 Ground-state bleaching andconsequent ESA losses are avoided by the combination

0146-9592/99/161133-03$15.00/0

of a highly erbium-doped fiber and the relatively lowpump intensity that is present in a cladding-pumpedfiber with an active reduction of the excitation densityby energy transfer to a Pr31 codopant.10,11 By useof the rectangular double-clad geometry of a designthat was successful in scaling a Tm31 silica-fiber laserto high output powers,12 we demonstrate 1.7 W oftransverse-fundamental-mode output power and 17.3%slope efficiency at 2.71 mm from a diode-pumped Er31,Pr31:ZBLAN laser.

The core of the double-clad fiber (KDD, Japan) con-tained a 35,000-parts-in-106 �ppm� molar concentrationof ErF3 and a 3000-ppm molar concentration of PrF3.With these doping levels, the Er31 and Pr31 ion densi-ties were 6.3 3 1026 and 5.4 3 1025 m23, respectively.The core had a diameter of 15 mm and a numerical

Fig. 1. Partial energy-level diagram of Er31 and Pr31 inZBLAN glass, indicating the processes that are relevantfor operation of the high-power diode-pumped Pr31-codopedEr31 fiber laser at the transition 4I11/2 ! 4I13/2. GSA,ground-state absorption.

1999 Optical Society of America

1134 OPTICS LETTERS / Vol. 24, No. 16 / August 15, 1999

aperture of 0.16, providing nearly single-transverse-mode output at the laser wavelength of 2.71 mm. Thenominal intrinsic loss was less than 50 dB�km at awavelength of 2.7 mm. The introduction of scatteringcenters arising from the relatively high combinedconcentration of the dopant ions was the main con-tributing factor to the overall intrinsic loss. Thepump cladding of the fiber had a 100 mm 3 200 mmrectangular cross section and was composed of approxi-mately 53% HfF4, 18% BaF2, 22% NaF, 4% LaF3, and3% AlF3. Surrounding the pump cladding was atransparent f luororesin second cladding layer (300-mmdiameter with a refractive index of 1.38) that provideda numerical aperture of 0.55 for the pump cladding.The diode-laser pump source was a Diomed concen-trator unit (Diomed, Ltd., Cambridge, UK) originallydesigned for use as a high-power pump sourcefor pumping a Tm31-doped silica double-clad fiberlaser.12 This system optically multiplexed the outputfrom sixteen 2-W-rated AlGaAs diodes, each operatingat a wavelength of 790 nm. The focus of the pumplight (which had a 200 mm 3 50 mm spot size) waslocated in the interior of the pump-source housing, andhence for the experiments the focus was translatedwith a 1:1 magnif ication triplet lens pair and a totalincident pump power of 22.4 W was available. Withthis pump configuration, 45–50% of the pump lightthat was incident upon the end of the fiber waslaunched into the pump cladding.

In the design of the fiber-laser cavity we made al-lowance for the fact that it is diff icult to design mir-rors that have simultaneously a high transmission at790 nm and a high ref lection at 2700 nm. To mini-mize backref lection onto the diode lasers of our pumpsource, one should ensure that the maximum ref lec-tion at the pump wavelength of a dichroic mirror (thatwould ordinarily be butted against the input end tothe fiber) is less than approximately 4%, a mirrorcharacteristic that we have found difficult to find.As a consequence, a mirror that was highly ��99%�ref lecting at 790 nm and was placed at 45± to thepump-light direction steered the pump light onto theinput end of the fiber while also effectively transmit-ting �.99.5%� the fiber-laser output; see Fig. 2. Ina readily arranged fiber laser cavity, Fresnel ref lec-tion at each end provides feedback (denoted resonatorA). For a second cavity arrangement, a mirror thatwas highly ref lecting at both the pump (.99.9% re-f lecting) and the laser (.99.7% ref lecting) wavelengthswas butted against the distal end of the fiber (denotedresonator B). The laser output was measured with aCoherent 210 powermeter after it was collimated witha CaF2 lens that was antiref lection coated for 2700–2800 nm. The temporal dynamics of the fiber laseroutput was detected with a liquid-nitrogen-cooled InAsphotodiode.

The fiber-laser output as a function of launchedpump power for both resonators for a 10.5-m fiberlength is shown in Fig. 3. In the case of resonator Athe counterpropagating output was determined to havea slope efficiency of 12% with respect to the launchedpump power, and for the copropagating output, a slopeeff iciency of 5.2%. The overall threshold launched

pump power for this fiber-laser configuration wasmeasured to be 0.52 W. For resonator B the slopeeff iciency was determined to be 17.3%, with a slightlyreduced threshold pump power of 0.47 W. The slopeeff iciency of the output from each resonator is �60% ofthe Stokes efficiency limit of 29%. For a fiber lengthof 10.5 m the effects of Beer’s law absorption leadto a significant difference between the gain at theproximal end of the fiber and the gain at the distalend. Since the ref lectivity at each end of the fiberis the same for resonator A, the difference betweenthe counterpropagating and copropagating output istherefore a result of gain asymmetry. In a separateexperiment, in which 4.96-m length of fiber was usedin a resonator A configuration, we measured thecounterpropagating and the copropagating outputs tohave similar values, as a consequence of a reductionin the gain difference. The overall optical-to-opticaleff iciency for resonator B therefore is calculated to be�7.6%, a value that can be signif icantly increased byoptimization of the launch conditions. Experimentsconcerning the optimization of the launch of the pumplight are ongoing.

From measurements of the transmitted pump powerduring cutback experiments, we determined that theeffective absorption coefficient for the fiber is approxi-mately 0.15 m21. Taking into account the core-to-cladding area ratio, we calculate that the effective

Fig. 2. Experimental setup of the high-power diode-pumped 3-mm ZBLAN fiber laser: L1, pump-focusinglens; L2, laser output collimating lens.

Fig. 3. Output power at 2.7 mm versus launched pumppower at 790 nm of the diode-pumped ZBLAN fiber laserfor resonators A and B. The pump threshold is �0.5 W,and the slope eff iciency is 17.3% with respect to thelaunched pump power for resonator B.

August 15, 1999 / Vol. 24, No. 16 / OPTICS LETTERS 1135

Fig. 4. Measured rf spectrum of the laser output forresonator B, in which 3.6 W of pump power was launchedand a 10.5-m length of fiber was used. The inset showspump-induced modulations at 10 and 125 kHz and thecorresponding beats.

absorption coefficient of the core is �17 m21, whichtranslates to an overall absorption cross section withour pump source of 2.74 3 10226 m 2. The �6-nmFWHM bandwidth of the pump source, therefore,has the effect of reducing the absolute absorptioncross section at 790 nm of 4 3 10226 m 2 (Ref. 5) by�30%. Note that in making this calculation we as-sumed that almost all the propagation rays within thepump cladding are absorbable and that the absorptioneff iciency is therefore related to the core-to-claddingarea ratio and the dopant concentration only.

The temporal dynamics of the output were measuredwith a Tektronix 492P rf spectrum analyzer, and theresults of the measurements are displayed in Fig. 4.It can be observed that there exist a number of oscil-lations in the output from the fiber laser that haveperiods that are approximately 1�n (where n is aninteger) times the cavity round-trip time. The rela-tive strengths of the oscillations were not constantwith time, and the results presented in Fig. 4 areonly representative. The origin of these oscillationsis currently under investigation. The inset of Fig. 4displays the pump-induced modulations in the out-put at �10 and at �125 kHz. The strength of the125-kHz modulation in the output is �10 times moreintense than the 10-kHz modulation, and the beat be-tween the two pump-induced modulations produces anadditional sideband (at 115 and 135 kHz) on each sideof the 125-kHz modulation. The incorporation of aquieter current source to the diode lasers would im-prove the overall temporal stability on the microsecondtime scale.

In conclusion, we have demonstrated an outputpower of 1.7 W from an erbium 3-mm fiber laser. Fur-ther power scaling seems possible by optimization ofthe resonator design and the launch efficiency. Sinceground-state bleaching and a high excitation densityof the laser levels could be avoided by energy transferto the Pr31 codopant, pumping at 980 nm directly intothe upper laser level is expected to increase further thequantum efficiency of this laser without introducingsignif icant ESA losses from the upper laser level. Atthe present power level, we are able to perform laser–tissue interactions and investigate the potential of thislaser in microsurgery.

The authors appreciate helpful discussions with AlexCable of Thorlabs, Inc. This work was financially sup-ported by Engineering and Physical Sciences ResearchCouncil grant GA744NP. M. Pollnau is indebted toHans-Ulrich Gudel of the Department of Chemistryand Biochemistry, University of Bern, Bern, Switzer-land for his support. S. D. Jackson’s e-mail address issj@fs4.ph.man.ac.uk.

*On leave from Department of Chemistry and Bio-chemistry, University of Bern, Freiestrasse 3, CH-3012Bern, Switzerland.

References

1. L. Esterowitz and R. Allen, Proc. SPIE 1048, 129(1989).

2. M. Ith, H. Pratisto, H. J. Altermatt, M. Frenz, andH. P. Weber, Appl. Phys. B 59, 621 (1994).

3. T. Jensen, A. Diening, G. Huber, and B. H. T. Chai, Opt.Lett. 21, 585 (1996).

4. D. W. Chen, C. L. Fincher, T. S. Rose, F. L. Vernon, andR. A. Fields, Opt. Lett. 24, 385 (1999).

5. M. Pollnau, Ch. Ghisler, W. Luthy, and H. P. Weber,Appl. Phys. B 67, 23 (1998).

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

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

8. X. Zhao, B. Srinivasan, P. Pulaski, S. Gupta, andR. K. Jain, in Digest of 1998 Conference on Lasersand Electro-Optics/Europe (Optical Society of America,Washington, D.C., 1998), postdeadline paper CPD1.7.

9. M. Pollnau, IEEE J. Quantum Electron. 33, 1982(1997).

10. L. Wetenkamp, G. F. West, and H. Tobben, J. Non-Cryst. Solids 140, 25 (1992).

11. J. Schneider, D. Hauschild, C. Frerichs, and L.Wetenkamp, Int. J. Infrared Millim. Waves 15, 1907(1994).

12. S. D. Jackson and T. A. King, Opt. Lett. 23, 1462 (1998).