530 OPTICS LETTERS / Vol. 34, No. 4 / February 15, 2009

Excited state absorption measurement in the900–1250 nm wavelength range for bismuth-doped

silicate fibers

Seongwoo Yoo,* Mridu P. Kalita, Johan Nilsson, and Jayanta SahuOptoelectronics Research Centre, University of Southampton, Southampton SO17 1BJ, UK

*Corresponding author: sey@orc.soton.ac.uk

Received December 22, 2008; accepted December 23, 2008;posted January 16, 2009 (Doc. ID 105660); published February 13, 2009

The feasibility of direct laser diode pumping of Bi-doped fiber lasers at the wavelengths of 915 and 975 nmwas examined by measuring excited state absorption in Bi-doped silicate fibers for the wavelength range of900–1250 nm. When the Bi-doped fibers were pumped at 1047 nm a strong excited state absorption wasfound at 915 and 975 nm, whereas no significant excited state absorption was observed in the 1080 nmpumping band nor in the emission band, �1160 nm, of Bi-doped fiber lasers. © 2009 Optical Society ofAmerica

OCIS codes: 060.2270, 060.2300, 060.2320.

Current Bi-doped fiber lasers (BiDFLs) utilize laserdiode (LD)-pumped Yb-doped fiber lasers for pump-ing at �1080 nm [1–3]. However, silicate-based Bi-doped fibers (BiDFs) exhibit broad and intense ab-sorption bands, which suggest the use of variousother pumping wavelengths, in addition to 1080 nm.In particular, high-power LDs emitting at 915 and975 nm have good overlap with the absorption bandsof BiDFs. Direct LD pumping would make BiDFLsmore compact and cost effective, which makes thesewavelengths particularly attractive for pumping. Atthese wavelengths, LDs are commercially availablewith more than 100 W of output power and with un-matched brightness.

The performance of fiber lasers and amplifiers canbe considerably impaired by undesired effects such asunsaturable absorption and excited state absorption(ESA). In fact, the unsaturable absorption at thepump wavelength of 1080 nm has been revealed as amain limitation of the efficiency of BiDFLs [1,2]. TheESA at pump wavelengths is also detrimental, be-cause it typically absorbs the pump power withoutexciting any ions into the metastable state. Hence,pump ESA should be avoided (as should ESA at theemission wavelength). The possibility of the ESA at808 nm was ruled out in Li2O–Al2O3–ZnO–SiO2glass [4], but the measurement was performed onlyat relatively low pump power, which may create onlya few percent of population excitation. Significantfurther excitation from such a low population inver-sion level is unlikely, in our opinion. Moreover, themeasurement captured all possible excitation mecha-nisms, not only ESA but also energy-transfer upcon-version [5].

Here we report our measurements of the ESA inthe 900–1250 nm wavelength range for BiDFs under1047 nm pumping. We further use the measured datato investigate the feasibility of pumping at 915 and975 nm.

We fabricated a BiDF preform by modified chemi-cal vapor deposition and solution doping. The

core composition was roughly estimated as

0146-9592/09/040530-3/$15.00 ©

0.6 GeO2:1.2 Al2O3:98.2 SiO2 (mol. %). We note herethat the properties of a BiDF critically depend on thecompositions or fabrication processes, and a BiDFwith different host materials by different fabricationprocess would show different properties. However,the compositions in our fibers are close to otherBiDFs used for laser operations. The preform wasdrawn into a fiber with 125 �m outer diameter. Thefiber lased at 1160 nm with 10% of slope efficiencywith respect to the launched pump power in a linearcavity with 100% and 4% feedback in its two ends [2].We measured ESA in the BiDF under 1047 nmpumping. The experimental arrangement is shown inFig. 1(a). We determined the change of transmissionin the BiDF in the range of 900–1250 nm as the1047 nm pump was switched on and off. For this, wefollowed the experimental procedure described in [2].A fiber-coupled Nd-doped yttrium lithium fluoride(Nd:YLF) laser was used as a pump source. When theBiDF pumped, the power of the transmitted probechanged, with a larger change for a larger pump

Fig. 1. (Color online) Measured ground state absorptionand changes in transmission in Bi-doped fiber. Inset, ex-perimental setup for ESA measurement. WLS, white light

source.

2009 Optical Society of America

February 15, 2009 / Vol. 34, No. 4 / OPTICS LETTERS 531

power. Above 100 mW of launched pump power weobserved saturation in the transmitted probe power.The measurement resolution was 2 nm. The smallsignal ground state absorption (GSA) spectrum wasalso determined by cutting back the fiber and mea-suring the transmission with the pump off. The GSAand change in the transmission are shown in Fig.1(b). The change in the transmission �T can be ex-pressed in logarithmic units per unit length as [6]

�T = −1

Lln�Ton

Toff� = N���ESA − ��GSA + �e��, �1�

where L is a fiber length; Ton and Toff are white lighttransmission (linear units) with pump on and off, re-spectively; N� is a constant accounting for the overlapbetween the signal and the population in the excitedstate; and �ESA, �GSA, and �e are cross sections ofESA, GSA, and emission, respectively. Equation (1)suggests that the ESA is dominant over the sum of�GSA and �e when the change is negative, i.e., whenthe transmission decreases when the pump is turnedon. Strong ESA is observed in the 900–1000 nmrange, growing toward shorter wavelengths, asshown in Fig. 1(b). On the contrary, the pumpingband at �1080 nm as well as the Bi emission band,�1160 nm, show a negative change; i.e., the trans-mission increases with larger pump powers.

To obtain �ESA, we determined �GSA and �e fromthe BiDF. The fluorescence was measured from theside of the BiDF, to determine the fluorescence spec-trum and decay time. Then, the Füchtbauer–Ladenburg relation was used to evaluate emissioncross-section spectrum [7]. The emission cross-section spectra determined are shown in Fig. 2 forpump wavelengths of 915, 975, and 1090 nm. We seethat the emission cross sections strongly depend onthe pump wavelength. The peak shifts to shorterwavelength, and the linewidth becomes broader withshorter pump wavelength. Moreover, the peak of thecross sections becomes lower with shorter pumpwavelengths. The observed fluorescence linewidth

Fig. 2. (Color online) Measured stimulated emission cross-section spectra for Bi-doped fiber. Inset, measured GSA and

ESA cross-section spectra.

dependence on the pump wavelength indicates awide site distribution of the active Bi ions. It is worthmentioning that such dependence was not observedin the pump region in 1064–1090 nm [3]. The mea-surement resolution was 1 nm. The fluorescence de-cay time under 1090 nm pumping was 750 �s, and itreduced to 670 �s under 915 and 975 nm pumping.

To determine �GSA we measured saturation inten-sity, Isat, in the BiDF under 1090 nm pumping. A uni-form pump intensity and uniform distribution of ac-tive Bi ion across the fiber core was assumed. Thesaturation in the fluorescence was observed at Isat of�36 kW/cm2. Note that we assumed 30% of ratio ofunsaturable absorption to the GSA in the wholewavelength range for the �GSA calculation, based onour measurement at 1080 nm [2]. In fact, the unsat-urable absorption is wavelength dependent [2,3].Nonetheless, we believe it would still be instructiveto determine �GSA despite the uncertainty. The ab-sorption cross-section spectrum was determined byscaling the GSA spectrum to fit the value found for�GSA at 1090 nm. The results are shown in the insetof Fig. 2. For �ESA we compared the spectral shapes of���GSA+�e�−�ESA� and ��GSA+�e�, and we found nosignificant distortion around the fluorescence peak,which indicates negligible ESA at the peak emissionwavelength. The N� in Eq. (1) was then calculated atthe peak wavelength, and finally the cross-sectionspectrum for ESA (inset of Fig. 2) was determined.The ratio of the ESA to GSA cross sections are 7.97and 2.48 at 915 nm and 975 nm, respectively, whilethe ESA is nearly negligible at 1080 nm and in thelaser emission band of the BiDFL.

We next estimate the influence of the pump ESA at915 and 975 nm on the performance of the 1160 nmBiDFL. We assume the energy transitions betweenpump and signal as those of Yb3+ in a glass [1], withan additional upper excited state level accounting forthe ESA. We also assume that the population in theupper excited state level is negligible owing to rapiddecaying to the metastable level. We ignore signalESA and any other pump loss mechanism such as co-operative upconversion [5] in our calculation. Therate equations become

dn2

dt= −

dn1

dt= ��GSA

p Ip

hvp+

�GSAs Is

hvs�n1

− ��epIp

hvp+

�esIs

hvs+

1

��n2, �2�

where n1 and n2 are the populations in the groundand metastable states and the total population, n, isequivalent to �n1+n2�. Furthermore, �GSA

p and �ep are

the GSA and emission cross sections for the pumpwavelength, �GSA

s and �es are the corresponding val-

ues for the signal wavelength. In addition, Ip and Isare pump and signal intensities, h is Planck’s con-stant, vp and vs are pump and signal frequencies, and� is the lifetime of the metastable state. For simplic-ity, we assume that the pump and signal propagate in

the fundamental mode of the fiber. Furthermore, Ip

532 OPTICS LETTERS / Vol. 34, No. 4 / February 15, 2009

and Is are simplified as Ip=�p�Pp /A�, Is=�s�Ps /A�,where Pp and Ps are pump and signal powers, A isthe area of doped region in the core, and �p and �s arepump and signal overlap factors.

The signal and pump power propagation throughthe fiber in the presence of the pump ESA will begiven by

±dP±

p

dz= �p���e

p − �ESAp �n2 − �a

pn1�P±p − �pP±

p,

±dP±

s

dz= �s��e

sn2 − �asn1�P±

s − �sP±

s, �3�

where �p and �s are the loss at pump and signalwavelengths. The signs depend on the direction ofpropagation. A linear laser cavity with feedback of100% in one end and 4% in the other, outcouplingend, was assumed. The pump was launched to the 4%feedback end. The unsaturable absorption wastreated as a background loss [1]. The concentration ofactive Bi ions was calculated to �8.3�1024 m−3 fromthe measured GSA and cross section. The fiberlengths were optimized for the maximum output sig-nal power. For 1080 nm pumping, we used 25 m tomatch our experiments [2], however, the 25 m fiberlength is close to the optimized length �22 m�. Theoutput signal power was calculated for up to 2.1 W ofpump power according to our laser experiments. Inaddition, 1 dB of signal and 2 dB of pump losses wereincluded for the calculation of 1080 nm pumping tomatch the experimental condition. These additionallosses arose from imperfection of the WDM coupler.Figure 3 shows the calculated laser performances.

Fig. 3. (Color online) Influence of pump ESA on Bi-dopedfiber laser performance. Numbers in parentheses are fiber

lengths used for the calculation. OC, output coupling.

The simple model well describes the experimental re-sults by 1080 nm pumping. This indicates the ESA ata specific wavelength is independent of the excitationwavelength as the relative population of the excitelevel does not depend on the pump wavelength. InFig. 3 the laser efficiency is indeed under the influ-ence of the pump ESA ��ESA/�GSA�. Under 975 nmpumping, the signal output power, limited to�0.02 W owing to the pump ESA, could otherwise bemore than 0.5 W. The high ESA strength togetherwith low emission cross section did not allow theBiDF to oscillate under 915 nm pumping within thegiven pump power level. We note that the impact ofthe pump ESA can be mitigated by reducing a popu-lation inversion factor. An average population inver-sion factor for the 975 nm pumping is �0.6 at 2.1 Wof pump power, but it reduces to �0.3 by replacingthe 4% feedback end with 50%. Consequently, theoutput power can reach �0.13 W (see Fig. 3). Itshould be noted that cladding pumped lasers usuallyoperate in a lower inversion factor than core pumpedlasers.

In summary, the ESA in BiDFs was measured inthe 900–1250 nm range under 1047 nm pumping. Astrong ESA appeared at 915 and 975 nm, whereas nosignificant ESA was found in the 1080 nm pumpingband nor in the primary emission band of Bi-dopedsilicate fiber lasers. The spectral shape of the emis-sion cross sections depended on the pump wave-length, and the emission cross-section peak reducedunder 915 and 975 nm pumping. Thus, the pumpESA together with the weak emission cross sectionwould limit the efficiency of the BiDFL under 915and 975 nm pumping. It was also suggested to reducethe population inversion factor to mitigate the impactof the pump ESA.

References

1. E. M. Dianov, A. V. Shubin, M. A. Melkumov, O. I.Medvedkov, and I. A. Bufetov, J. Opt. Soc. Am. B 24,1749 (2007).

2. M. P. Kalita, S. Yoo, and J. K. Sahu, Opt. Express 16,21032 (2008).

3. V. V. Dvoyrin, V. M. Mashinsky, and E. M. Dianov,IEEE J. Quantum Electron. 44, 834 (2008).

4. M. Peng, J. Qiu, D. Chen, X. Meng, and C. Zhu, Opt.Express 13, 6892 (2005).

5. I. A. Bufetov, S. V. Firstov, V. F. Khopin, A. N.Guryanov, and E. M. Dianov, in Proceedings of the 34thEuropean Conference on Optical Communication(ECOC 2008) (IEEE, 2008), paper Tu.3.B.4.

6. J. E. Román, M. Hempstead, C. Ye, S. Nouh, P. Camy,P. Laborde, and C. Lerminiaux, Appl. Phys. Lett. 67,470 (1995).

7. W. L. Barnes, R. I. Laming, E. J. Tarbox, and P. R.

Morkel, IEEE J. Quantum Electron. 24, 1004 (1991).