348 OPTICS LETTERS / Vol. 21, No. 5 / March 1, 1996

Photodegradation of near-infrared-pumped Tm31-dopedZBLAN fiber upconversion lasers

Ian J. Booth, Jean-Luc Archambault, and Brian F. Ventrudo

SDL Optics, Inc., 6703 Rajpur Place, Saanichton, British Columbia V8M 1Z5, Canada

Received July 21, 1995

Photodegradation has been observed in Tm31-doped ZBLAN fiber lasers pumped with laser diodes at 1135 nm.After upconversion lasing at 482 nm, the fiber develops color centers that absorb strongly at wavelengths below,650 nm, affecting further upconversion lasing. The rate of damage formation is strongly dependent on thepump power level and on the thulium concentration. The color centers are bleached by intense blue lightbut recover with thermal excitation and can be removed by thermal annealing at temperature near 100 ±C. 1996 Optical Society of America

Upconversion lasing in rare-earth-doped ZBLANoptical fiber offers the potential for compact blue-greenlaser sources pumped by infrared laser diodes. Apromising candidate for such a source is the thuliumsystem first demonstrated by Grubb et al.1 Lasing at482 nm was demonstrated in Tm31:ZBLAN fiber withsingle laser diodes operating at 1135 nm.2 However, aconcern with any upconversion laser is the possibilityof long-term deterioration in the glass f iber. Photo-sensitivity and rapid darkening have been observed,for example, in thulium-doped silica fiber pumped at476 nm (Ref. 3) and at 1064 nm.4 Our results suggestthat similar effects occur in Tm31:ZBLAN fiber, albeiton a time scale that makes it more difficult to observe.

Figure 1 illustrates the relevant energy levels ofTm31 and the upconversion mechanism that leads tolasing at 482 nm. Ions excited to the short-lived 3H5and 3F2,3 levels decay rapidly by phonon emission tothe 3H4 and 3F4 levels, which are metastable in a low-phonon-energy material such as ZBLAN. The centerwavelengths of the three transitions are 1220 nm (3H6to 3H5), 1130 nm (3H4 to 3F2), and 1150 nm (3F4 to1G4), but because of the broadening of the transitionsin glass it is possible to excite all three with a single-wavelength pump in the range 1110–1160 nm. Thenext energy level above 1G4 in thulium is 1D2, with atransition of 1480 nm separating the two, so that thereis a very small cross section for further upconversionabove 1G4 for pumping in the 1100-nm range.

Thulium-doped ZBLAN fiber (Le Verre Fluore, Vern/Seiche, France) with a core diameter of 3 mm and anumerical aperture of 0.21 was used in these experi-ments. Fibers with 1000, 2500, and 1250 parts in 106

(ppm) Tm31 were tested; the 1250-ppm fiber also had5000-ppm Yb31. Upconversion lasing at 482 nm wassuccessfully demonstrated in all three fibers with laserdiode pumping near 1135 nm. The pump sources forthese measurements were 80-mW laser diodes, of whichapproximately 50 mW could be launched into the f iberwith aspheric lenses. Pump light absorption per unitlength of f iber was consistent with the thulium con-centration in each fiber, with the 1000-ppm fiberabsorbing approximately 3 dBym. Pump saturationwas not observed. Multilayer dielectric mirrors re-f lective at 480 nm were butted to the cleaved fiber

0146-9592/96/050348-03$6.00/0

ends, with a high ref lector at the pump input end anda 90% ref lector at the output to couple the blue laserpower out of the fiber. Blue-light outputs of severalmilliwatts were obtained. A 0.5-W tunable diode laser(SDL Model 8630) was also used as a pump sourcein some experiments. Approximately 230 mW of in-frared power at 1130 nm could be launched, producingan output of as much as 25 mW of blue light from a 1-m1000-ppm fiber.

After a piece of f iber was lased several times, wenoticed that the threshold pump power required forlasing to start would increase by 10–20%. Once las-ing was established, the threshold returned to normallevels. This transient increase in start-up thresholdwas larger the longer the fiber had been left inactive.Turning the laser off and on again quickly produced noeffect. When the 0.5-W tunable diode laser was usedas a pump source the effect in exposed fibers becamemuch greater, with the transient threshold rising to5–10 times the normal operating threshold level. Adecrease in blue output power was also observed afterlong periods of operation at high power. A 1000-ppmfiber laser that initially produced 25 mW of blue power

Fig. 1. Energy levels of Tm31 ions in ZBLAN glass, indi-cating the steps involved in pumping to achieve upconver-sion lasing at 482 nm.

1996 Optical Society of America

March 1, 1996 / Vol. 21, No. 5 / OPTICS LETTERS 349

Fig. 2. Absorption versus wavelength of color centersformed in Tm31:ZBLAN fiber used in an upconversion f iberlaser. The absorption is the change in transmission of thefiber over 96 h after the centers were bleached by lasing.Inset: Increase in the turn-on lasing threshold versus thetime during which the laser was switched off.

had a reduced output power of 18 mW after 8 h of oper-ation. No change in the throughput of the pump lightwas observed.

To explain the observed degradation in laseroperation, we conjectured that color centers were beingproduced in the core of the fiber. These centers absorbonly at shorter wavelengths, thereby increasing theblue lasing threshold without noticeably affecting theinfrared pump throughout. Under intense blue lightin the fiber core, the absorption of these centers isbleached, so that the laser threshold is reduced oncelasing starts. The recovery of the defects from theirbleached state occurs on a time scale of several hours.

We obtained the absorption spectra of the color cen-ters in several f ibers by taking a white-light trans-mission spectrum on the same piece of f iber imme-diately after lasing at 482 nm to bleach the defects,and again some time later when some of the defectshad unbleached. The ratio of these spectra gives theincrease in absorption caused by the color centers.Figure 2 shows the spectrum obtained with the 1-m1000-ppm fiber that had been run at high power for8 h with 230 mW of pump light. Similar spectra wereobtained with 2500-ppm fibers.

We measured recovery time of the centers frombleaching by leaving a fiber inactive for varying peri-ods of time between lasing and measuring the turn-onthreshold (Fig. 2, inset). The increase in threshold isapproximately proportional to the increase in absorp-tion at 480 nm and provides a measure of the num-ber of unbleached defects in the fiber. Attempts tofit the defect recovery to an exponential function in-dicated that the centers had no well-defined time con-stant; rather, the recovery rate appeared to slow downat longer times. This suggests that the defects havea range of activation energies.5 Some fraction of thecenters always unbleach quickly, effectively producinga permanent increase in f iber loss; this explains the de-crease in lasing power observed after several hours ofoperation at high power.

Further investigation indicated that the color cen-ters were produced by the infrared pump light withoutthe presence of the blue lasing light. Fiber pumpedwith 1130-nm light without allowing blue lasing sub-sequently demonstrated the transient threshold effect,with approximately the same magnitude as the f iberthat had been lased. The corollary to this experiment,pumping a fiber with high-power 480-nm light to seewhether this caused color centers to form, was not donefor lack of an appropriate laser source. Thresholdhysteresis was not present in fresh fiber, but it ap-peared and increased with exposure to pump light.The increase was much faster with high infrared powerlevels, but the exact relation between pump power anddeterioration rate has been diff icult to measure, ow-ing to the transient nature of the phenomenon. Directmeasurement of the white-light absorption caused bythe defects was possible only in heavily exposed fibersbecause of the poor signal-to-noise ratio.

Deterioration was observed to occur more quicklyin fiber with 2500-ppm Tm31 doping and in theYb31-codoped fiber than in the 1000-ppm fiber. A sys-tematic test was initiated with prototype blue laserpigtails made with 1000- and 2500-ppm fiber. The1000-ppm fiber was 38 cm long, whereas the 2500-ppmfiber was only 13 cm long, the higher doping levelcompensating for the reduced length. The units wererun for approximately 8 h each day and turned off for16 h. The turn-on threshold at the start of each dayas well as the operating threshold was monitored. Thefiber lasers were operated with only 26 mW of launchedpump power in both units as a result of poor launch ef-ficiency in the 1000-ppm pigtail. Figure 3 shows howthe transient thresholds increased with use, both lasersstarting with some transient effects owing to exposureof the f ibers during testing and construction. Theturn-on threshold in the 2500-ppm pigtail increasedapproximately 10 times faster than in the 1000-ppmunit, based on a straight-line f it to the data up to150 h, and soon exceeded the maximum availablepump power. Lasing could still be started by pump-ing the fiber for several minutes, indicating that theblue f luorescence generated in the f iber was suff icientto cause some bleaching of the defects. The results for

Fig. 3. Turn-on threshold of two Tm31:ZBLAN upcon-version fiber lasers versus the operating time. Squares,2500-ppm laser; circles; 1000-ppm laser.

350 OPTICS LETTERS / Vol. 21, No. 5 / March 1, 1996

Fig. 4. Upconversion f luorescence spectra from 1000-ppm(dashed curve) and 2500-ppm (dotted curve) doped fiberspumped at 1135 nm.

the 1000-ppm fiber suggest that the transient thresh-old levels off after 150 h of operation. This may in-dicate that the number of color centers reaches anequilibrium, with some process acting to remove cen-ters at the same rate as they are formed.

The effect of moderate heating on fibers with colorcenters was measured. Temperatures of approxi-mately 50 ±C greatly speeded recovery of the centersfrom bleaching. However, baking at higher tempera-tures of 70–100 ±C for 24 h appeared to remove mostof the defects, as the subsequent threshold hysteresiswas smaller, and in heavily deteriorated f ibers thewhite-light spectrum recovered to its bleached level.

In Tm31-doped silica glass fiber pumped at 1064 nmthe cause of photodarkening was hypothesized to be anupconversion ladder leading to a highly excited state ofthe Tm31 ion, creating an electronic defect in the sur-rounding glass.4 The photodarkening rate followedapproximately a fifth-power dependence on the pumpintensity, suggesting a five-step upconversion, pastthe 1G4 and 1D2 levels to the P states of thulium. Asimilar process may be responsible for the defects ob-served in our lasers. If so there should be observablef luorescence from the Tm31 energy levels above 1G4and a correlation between this f luorescence and therate of fiber deterioration. We tested this by pumpingTm31:ZBLAN fiber with an 1135-nm laser diode withno end mirrors on the f iber; the f luorescence emittedfrom the fiber end was measured with an opticalspectrum analyzer. Spectra obtained from equivalentlengths of 1000- and 2500-ppm doped fiber (71 and30 cm, respectively) are compared in Fig. 4. Thepresence of f luorescence at 360, 455, and 515 nmfrom, respectively, the 1D2 ! 3H6, 1D2 ! 3H4, and1D2 ! 3H5 transitions indicates that some upconver-sion is occurring past 1G4. It is not clear to whatextent direct pumping from 1G4 to 1D2 by the 1135-nmlight is occurring; however, the transfer of energybetween thulium ions provides another upconversionpath, and such ion–ion interactions become stronger

with higher doping concentration. It is evident thatthe f luorescence spectra from the 2500-ppm dopedfiber shows increased 1D2 f luorescence relative tothe 1000-ppm fiber, which strongly suggests thation–ion interactions are indeed contributing to thisupconversion. Spectra taken from the thulium–ytterbium codoped fiber also exhibited enhanced 1D2f luorescence. From the 1D2 level resonant pump-ing by 1135 nm is possible to the 3P1 level, whichrelaxes to the 1I6 state. This level is thought to beresponsible for the severe photodarkening observedin Tm31-doped silica glass fiber. Fluorescence fromthe 1I6 state is expected at 310 nm, which we cannotmonitor with our equipment. It was also observedthat the 1D2 f luorescence peaks increased with pumppower faster than the 480-nm peak. The dependencewas less than the fourth-power law expected for afour-level upconversion and tended to f latten out athigher powers, indicating the saturation of some of theupconversion transitions. In fact, the transitions3H4 –3F2,3 and 3F4 –1G4 are expected to saturate near10-mW pump power.

In conclusion, photodarkening has been observed inTm31-doped ZBLAN fiber pumped with diode lasersat 1135 nm. This is probably caused by color cen-ters induced by ultraviolet f luorescence generated bymultistep upconversion from thulium to a highly ex-cited state. Such a damage process has been observedpreviously in Tm31-doped silica glass f iber. The rateof damage formation is strongly dependent on thepump power level and on the thulium concentration.The centers absorb strongly only at wavelengths be-low ,650 nm, affecting the performance of blue up-conversion lasing. The color centers are bleached byintense blue light but recover with thermal excitation.They can be removed by thermal annealing at tempera-tures near 100 ±C. Stronger f luorescence from the 1D2state is observed in fiber with higher levels of thuliumdoping and in fiber with thulium–ytterbium codoping;these f ibers also develop photodarkening faster. Anobvious step toward reducing the photosensitivity is touse fiber with lower levels of thulium doping. It is notclear, however, how effective this will be in prevent-ing damage, since part of the 1G4 –1D2 upconversionmay be directly pumped as a result of broadening of thetransition in glass. It is also possible that the suscep-tibility of the glass to damage depends on the compo-sition and the doping of other elements that are dopedinto the glass to tailor the refractive index. These ele-ments may provide sites for color centers to form.

References

1. S. G. Grubb, K. W. Bennett, R. S. Cannon, and W. H.Humer, Electron. Lett. 28, 1243 (1992).

2. I. J. Booth, C. J. Mackechnie, and B. F. Ventrudo, IEEEJ. Quantum Electron. 32, 118 (1996).

3. W. S. Brocklesby, A. Mathieu, R. S. Brown, and J. R.Lincoln, Opt. Lett. 18, 2105 (1993).

4. M. M. Broer, D. M. Krol, and D. J. DiGiovanni, Opt. Lett.18, 799 (1993).

5. T. Erdogan, V. Mizrahi, P. J. Lemaire, and D. Monroe,J. Appl. Phys. 76, 73 (1994).