.5

,vean

tinurno

atistronlinridri

nseO

T

PHYSICAL REVIEW B 1 SEPTEMBER 2000-IIVOLUME 62, NUMBER 102~ab !

where Aa ,b is the decay rates of the a and b levels, and T2(ab)is an effective relaxation rate associated with phonon broad-ening, which is also known as the dephasing time.1 In ahomogeneously broadened spectrum, due to the uniformityof the field around the rare-earth ions, a single frequencypump will interact with all the dopant ions with equal prob-ability. The pump, therefore, produces a gain spectrum inwhich any signal wavelength can saturate the entire gainband. The uniformity of field around the rare-earth ions in aglass host will also enable signals at different wavelengths tointeract with each other, which may be advantageous in ahigh-power laser and amplifiers. However, this feature willprove detrimental in a wavelength-division-multiplexing~WDM! system in which the power amplifiers may sufferfrom cross-saturation and cross-talk. On the other hand, thevariations in the nearest-neighbor environment, local sitesymmetry, and ligand field strengths produce an inhomoge-neously broadened spectrum. As a result of the site-to-sitevariations, there occurs a difference between the nonradiativeand radiative rates, and individual sections of a gain spec-trum respond independently, should these differences be-come large. In a WDM system, the cross-channel interaction

dependence of the emission oscillator strength of a transitiona-b, f ab on host glass is understood for designing a broad-band amplifier. The emission oscillator strength f ab is depen-dent on local field correction, (E loc /E)25x2, which is ahost-dependent property. x may be either due to electric di-pole or magnetic dipole; in each case it is proportional to thesquare of the refractive index, n i.e., n2. By including theemission cross section (sab) and local-field correction ~x!,f ab for a given transition a-b is proportional to the spectralintegral of the cross section, which can be expressed by Eq.~2!.2~a! In Eq. ~2!, the summation sign defines all the subleveltransitions, of which the transition between the levels a and bis an example:

f ab5Z0n

x (i51i5 j

Dnabs~ab !peak

. ~2!

Z0 is a constant in Eq. ~2!. Upon rearranging Eq. ~2! yieldsthat the line boadening (Dneff) is dependent upon two mainfactors: the local field effect determined by the ion-host in-teraction and the emission cross section. The ion-host inter-action also determines the extent of Stark splitting. TheStructural origin of spectral broadening of 1

A. Jha,* S. ShenDepartment of Materials, Clarendon Road, Uni

~Received 20 July 1999; revised m

The emission spectra and the lifetimes of the lasingternary glasses have been studied. The investigationRaman spectroscopic analysis of the host glass structEr31-ion photoluminescence near 1500 nm on OH2-ioEr2O3 and the glass structure modifiers ~ZnO, Na2O!emission spectra has been analyzed. It was observed thspectral broadening as a consequence of Er31 being dof 4I13/2 emission vary between 4.5 and 7.8 ms and storigin of the spectral broadening of the Er emissioncomponents in the tellurite, modified silicate, and fluoroom temperature in Er-doped TeO2, silicate, and fluostrengths and site-to-site variation of the rare-earth iounusually large broadening of the OH2 band in the Ttellurite glass structure.

I. INTRODUCTION

A rare-earth ion, for example, Er31 doped in a glass host,when excited using a pump photon usually exhibits an inho-mogeneously broadened emission line for a transition be-tween the levels a and b. The amplifier channel gain is de-fined by the homogeneous spectral width, Dnab of a lasingtransition, which is dependent on the characteristics of theline-shape broadening. The spectral width Dnab can be ex-pressed as

Dyab5Aa1Ab12

, ~1!PRB 620163-1829/2000/62~10!/6215~13!/$15.00-mm emission in Er3-doped tellurite glasses

and M. Naftalyrsity of Leeds, Leeds LS2 9JT, United Kingdomuscript received 13 January 2000!

ransition 4I13/2-4I15/2 in Er31-doped TeO2-ZnO-Na2Ocludes two main studies: The first part discusses thee. In the second part, we explain the dependence ofconcentration. The influence of the concentrations ofn the lifetimes of 4I13/2 level and line shape of thethe increasing concentrations of Er31 ions resulted in

ributed in the glass structure. The measured lifetimesgly depend on OH-ion concentrations. The structurale is particularly discussed by analyzing the spectrale glass hosts. The number of spectral components atde glasses are compared, and the dependence of lineon the structure of each glass is also explained. The2 family of glasses is analyzed in the context of the

must be minimized, but not at the expense of a large varia-tion in the power out of each channel. Gain excursion mustbe minimized. The design of a rare earth ~RE!-doped glassthat yields an inhomogeneously broadened spectrum withuniform output power for each channel will prove beneficialfor WDM applications.

The low-loss window at 1.5 mm in the installed silicaglass optical fiber networks is undergoing a major upgradingfor a WDM operation. For an efficient WDM operation, theEr31-doped amplifiers with broadband and flat gain charac-teristics are required to compensate and maintain the gain ateach channel, so that the gain excursion over the whole spec-tral width can be minimized. It is, therefore, essential that the6215 2000 The American Physical Society

31

teed

ea

~

~

~

45

ANtrations of Er ions without causing concentration quench-ing, and at the same time should exhibit a broader FWHMthan Al/P silica and ZBLAN. In 1997, Mori andco-workers12 and Yamada and co workers13 reported the re-sults from an Er31-doped TeO2 fiber with flat gain over 70nm covering the range between 1530 and 1600 nm. The de-vice was pumped at 1480 nm and the gain was larger than 20dB.

Before the Er31-doped TeO2 amplifiers were invented,these glasses have been known to have a good glass-formingtendency with much better resistance to crystallization andsusceptibility to moisture than ZBLAN. The glass has a

temperature, the oxide melt was homogenized for 30 minand then quenched into a preheated brass mold to form glass.The quenched glass samples were annealed for several hoursat 260 C in a muffle furnace, and then allowed to coolslowly inside the furnace by turning the power supply off.The glass samples were doped with Er2O3 in the compositionrange 50050 000 ppm. Samples of OH2 ion doped glasseswere also prepared for spectroscopic analysis. In order tocarry out a reliable measurement of emission and absorptionspectra, glass samples were polished carefully using a 1-mmdiamond paste.

The analysis of Er31 and OH2 ion absorption was carriedemission cross-section together with the lifetime ~t! of meta-stable level b determines the gain ~G! of an amplifier, i.e.,G}(st).

Two main methods have already been adopted for broad-ening and flattening the gain curve in Er-doped silica de-vices. The first approach is based on modifying the structureof host glass by incorporating Al2O3 and P2O5 codoping ofsilica glass ~known as Al/P silica!,3 which is known to dis-perse Er31 ions in the silica glass matrix and thereby providemultiple sites. The modification of glass structure withAl2O3P2O5 addition leads to an increase in the value of thefull width at half maximum ~FWHM! from 8 nm in puresilica to 44 nm in Al/P silica.2~b! A significant spectral flat-tening has been also achieved via filtering reported byGeorges.4 From Eq. ~2! the emission cross section increaseswith the increasing refractive index of the host glass. Thevalue of the FWHM is 65 nm for the 1500 nm transition inEr31-doped fluoroziconate family of glasses.5,6 ConsequentlyEr31-doped fluorozirconate glass fiber amplifiers have beenconsidered as an alternative to Al/P silica-based devices. Al-though the emission and the gain curves are much broader inEr-doped zirconium-barium-lanthanum-aluminium-sodium-fluoride ~ZBLAN! glass, it suffers from three main disadvan-tages. The first disadvantage is related with the low phononenergy of ZBLAN host ~580 cm21!, which makes the deviceunsuitable for pumping at 980 nm due to an enhanced ex-cited state absorption at pump wavelength.7 As a result, thedevice is only pumped at 1480 nm. The second major prob-lem is that the glass cannot be heavily doped with Er31, asthe concentration quenching starts dominating above 1000ppm concentration level.811 The third problem is the com-paratively poor stability and durability of ZBLAN fibers. It ismuch weaker than Al/P silica. The unsuitability ofEr31-doped ZBLAN and limitation of Al/P silica led to asearch for better glass hosts that will dissolve higher concen-

TABLE I. The identification of Raman peaks and their relative inof vibrational modes in TeO2 glasses are based on the results report

Glass composition ~mol %!Urbach edge (E0),

~eV! P

80 TeO2, 10 ZnO, 10 Na2O (RR2) 5.44 44090 TeO2, 10 ZnO (RR14) 5.08 420

60 TeO2, 10 ZnO, 20 Na2O (RR15) 6.25 450a-TeO2

6216 A. JHA, S. SHEN,higher phonon energy ~780 cm21! and refractive index~.1.9! than ZBLAN.1416 The solubility of RE ions in theseglasses is also known to be larger than 5000 ppm.911 In thispaper, we have designed a series of ternary tellurite glassesbased on TeO4, ZnO, and Na2O oxides as hosts for Er31ions, and analyzed the structural origin of the emission line-broadening in Er-doped TeO2 glasses. The spectroscopicproperties of tellurite glasses are also compared with a modi-fied silicate and a fluorozirconate glass, called ZBLAN. Thestructural analysis has been carried out by using Raman andUV-visible spectroscopy for analyzing the structural unitsand Urbach tail, respectively, of the TeO2 glasses as a func-tion of the composition. The room-temperature emissionspectra in tellurite, ZBLAN, and silicate glasses have beenanalyzed for their spectral components, which enabled us todetermine the spectral width of Stark sublevels and site-to-site variation in each host. A structural model for broadeningis also presented. The effect of OH2 ions on concentrationquenching and the dependence of spectral broadening onEr2O3 concentration is also explained on the basis of thestructural model discussed herein.

II. EXPERIMENT

The composition of Er31-doped tellurite glass was chosenafter a systematic search for stable glass compositions. Mul-ticomponent tellurite glasses are known to be more stablethan pure TeO2 glasses,1416 the latter requires an extremequenching rate to form. The compositions chosen in this in-vestigation fall in the following composition range: (1002x2y) mol % TeO2 , x mol %, ZnO, and y mol % Na2O~see Table I below!. The glasses were prepared from 99.99%pure powders of Na2CO3, ZnO, Er2O3, and TeO2. Beforemelting the powders in a gold crucible, the molar masseswere calculated, weighted, and mixed thoroughly, using amortar and pestle. The mixed oxides were melted in a goldcrucible at 750 C in a stream of dry air. At the melting

nsities determined from Gaussian peak fit program. The assignmentin Refs. 1922.

Raman peak (nR) cm21 and relative peak intensity ~%!k A Peak B Peak C Peak D Peak E

48%! 591 ~19%! 663 ~91%! 730 ~42%! 778 ~49%!53%! 601 ~19%! 656 ~77%! 717 ~30%! 768 ~38%!17%! 583 ~9%! 681 ~51%! 765 ~64%! 793 ~31%0 611 659 716 773

PRB 62D M. NAFTALYout using a Perkin-Elmer Lamda-19 UV-visible-NIR and aFourier transform infrared spectrophotometers, respectively.

ECThe absorption peak for OH2 ion was normalized with re-spect to the thickness of the glass. Emission in bulk glasswas measured using a Ti-sapphire laser to pump at 980 nm.The emission spectra were recorded by scanning spectrom-eter ~Macam Photometrics! equipped with an In-Ga-As de-tector. The resolution was 1 nm. The 980-nm excitationwavelength was also used to measure the lifetimes of theupper (4I11/2) and the lasing (4I13/2) levels in Er31-dopedglass samples. A mechanical chopper was used to modulatethe pump, and fluorescence decay curves were recorded by adigital storage adaptor. The lifetimes were calculated by fit-ting a single exponential function to the measured data. Thestructure of glass was analyzed using a Raman spectrometerin the 50900-cm21 wave-number range in 90 geometryusing the Ar1-ion laser ~514.5 nm! as excitation source. Thespectra were observed at an angle of 90 to the propagationdirection of the pump beam. The emission spectra ofEr31-doped tellurite and other host glasses used for compara-tive purposes reported in this paper were analyzed using apeak fit analysis software ~SPSS Science!. The analyzedVoigtian shapes are described below; for each fitting thevalue of r2 was maximized to reduce the standard error. Thespectral analysis was carried out in wavelength domain in-stead of frequency for clarity and convenience. Since thespectral range is so narrow, 200 nm, the error due to conver-sion is negligible. A Voigtian shape defines both the homo-geneous ~Gaussian! and the inhomogeneous ~Lorentzian!parts. Examples of the Voigtian amplitude are given below,and in some cases, the magnitude of the fitted amplitudes andwidths are also given. The thermal characterization of eachglass was carried out using a differential scanning calorim-eter ~DSC-7, Perkin Elmer!, in which small fragments ~,20mg! of each glass were heated isochronally at a rate of 10K min21 to determine the glass transition, crystallization, andpeak crystallization temperatures.

III. RESULTS

A. Analysis of vibrational spectroscopic data

Figure 1~a! shows the absorption spectra of OH2 ions inthe RR2 glass composition. The OH2 absorption band inRR2 is broader, with a FWHM ;900 cm21, when comparedwith fluoride ~FWHM 410 cm21! and silica ~FWHM 130cm21! glasses. The fundamental OH2 band in TeO2 glasspeaks approximately at 2980 cm21 ~3.356 mm!. The Voigtiancomponents of high-OH peak between 2250 and 3750 cm21are fitted shown in Fig. 1~b!. Each spectral component ap-pears to be at least as broad as observed in the silicate andfluoride glasses. By comparison the fundamental absorptionpeaks in SiO2 and fluorozirconate glasses are at 2.72 mm and2.87 mm, respectively.17 The broadening of OH2 band inTeO2-based glass appears to have a relationship with theglass structure, as, for instance, has been shown for aNa2O-modified silicate glass, in which the broadening ofOH2 fundamental is associated with the presence of non-bridging oxygen. The nonbridging oxygen sites, according toAdams,18 contribute more to the hydrogen bonding in silica,which is associated with the OH2 absorption.

PRB 62 STRUCTURAL ORIGIN OF SPThe structure of TeO2 glass, described above, was ana-lyzed by Raman spectroscopic technique, and the vibrationalspectra for all compositions are compared in Fig. 2. In thisfigure, the symmetric vibrational modes in TeO2 glasses havebeen analyzed and assigned. These assignments are based onthe data reported in the literature16,1922 for other types ofTeO2 glasses containing modifying oxides, and a-TeO2crystals. The assigned vibrations for a-TeO2 crystals arecompared with those determined in glasses, and their relativeintensities are compared in Table I. The relative intensitieswere derived from the peak fit program, which enabled de-convolution of Raman peaks and yielded the relative magni-tude of amplitude of each peak. The assigned vibrationalmodes for peaks in a-TeO2 and glass compositions are com-pared in Table II.16,1922

The analysis of Raman peaks points out that the vibra-tional modes and their intensities depend on the composi-tions of divalent and monovalent modifying ions. Since theRaman vibrations are sensitive to the polarizability of thenetwork modifying ions, the larger and more polarizable ionscontribute more to intensify the high-energy peaks than themonovalent ions. The high-energy Raman peaks ~E and D!are due to the stretching mode contributions not only fromTeO4 trigonal bipyramid ~tbp! structural units, but also fromTeO31d polyhedra and TeO3 trigonal pyramid ~tp! units inthe environment of nonbridging oxygen.20,22 Peak C is as-signed to a convolution of antisymmetric vibration modes,

FIG. 1. A comparison of OH2 absorption bands in high and lowOH2 ion containing RR2 tellurite glasses. The FWHM in highOH2RR2 glass is 900 cm21 approximately. ~b! A deconvolvedGaussian amplitude peaks fitted to the broad OH2 absorption bandin high OH2 containing RR2 glass.

6217TRAL BROADENING OF . . .arising from the stretching of two dissimilar Te-O bondlengths. Sekiya and co-workers22 proposed that this mode

a6218 PRB 62determines the connectivity between TeO4, TeO31d andTeO3 structures. According to Komatsu and co-workers,19and Himei and co-workers,21 two tbp structures can connectwith each other with an equatorial oxygen. In the same way,two tp (TeO3) and two TeO31d polyhedra can be connectedwith each other via an equatorial oxygen. Other combina-tions of TeO31d and TeO4 connectivities are also possible,resulting in a range of antisymmetric vibrations. The anti-symmetric vibrations of equatorial oxygens arise due to thepresence of the lone-pair electrons ~LPE!, which leads to theformation of short and long equatorial Te-O bonds,17,20 thechemical nature and structure of such bonds are discussedbelow.

The Raman peaks at lower wave number A and B are dueto the bending ~wing mode! and stretching modes of thenetwork as described in Table II. Therefore, the presence of

FIG. 2. Relative intensities of Raman peaks in RR2, RR14, andRR15 compositions. Excitation wavelength is Ar1 laser at 514 nm.The curves with broken lines are deconvoluted peaks, A, B, C, D,and E, which correspond to 450, 611, 659, 716, and 773 cm21,respectively, in a-TeO2 ~Refs. 1822!.

TABLE II. Vibrational mode assignments for Ram~Refs. 1922!.

Assigned vibrational modes in TeO2 glass

Stretching vibrations of Te-O in TeO31d andTeO3

Vibrations of TeO4 tbps, Te-Onbo in TeO31dand TevOnbo in TeO3

Coupled symmetric vibrations alongTe-O-Te axes in TeO31d /TeO4, TeO4 /TeO3

pairs and similar combinationsVibration of continuous TeO3 /TeO4 networks

Symmetric and bending vibrations of

Te-O-Te linkages at corner sharing sitesA and B peaks define the network connectivity of TeO2glasses.19,20,22 On the basis of Raman spectroscopic analysispresented in this paper, which is consistent with the pub-lished results,14,1922 it can be concluded that a networkmodified tellurium oxide glass, for example RR2 used fordoping erbium ions, has more than one type of structuralunits, namely, TeO4, TeO31d , and TeO3 structures. The ex-tent of deviation between the two oxygen bond distances isstrongly dependent on the nature of the modifying oxides.22Each of these structural units, which depend on the concen-trations of modifying oxides,20,22 has a unique ligand fieldand these will vary depending on the value of d. From theliterature,22 it is apparent that the Te-O bond distances arealso dependent on the concentration of modifying ions.

B. UV-visible spectroscopy of TeO2 glass

The UV absorption in samples of RR2, RR14, and RR15glasses were measured by a lamda-19 UV-visible spectro-photometer. The measured Urbach tail for each glass is com-pared in Fig. 3. The position of tail changes with the com-position of the glass, indicating that as the ratio of Te41 to

FIG. 3. The measured and fitted electronic absorption edges inRR2, RR14, and RR15 tellurite glasses. The derived values of E0from fitted Eq. ~3! are 5.44, 5.08, and 6.25 eV, respectively.

n peaks and their comparison with a-TeO2 structure

a-TeO2,nR (cm21)

Glass nR (cm21)RR2 RR14 RR15

773 778 768 793

716 730 717 765

659 664 657 681

611 591 601 582450 440 421 450A. JHA, S. SHEN, AND M. NAFTALY

ECmodifying ions (M n15Na1, Zn21) increases, the position oftail shifts to longer wavelength. The measured electronicedges fit UV energy-gap equation ~3! satisfactorily andyields the value of E0 ~eV!, which are compared in Table I.For RR15, the Te41/M n1 ratio is 2:1, and the correspondingvalue of E0 is 6.25 eV, which decreases to 5.08 eV in RR14composition having Te41:M n1 ratio equals to 9:1,

auv5a0 expF2 A~E2E0!kT G5a0 exp@2a~E2E0!# . ~3!Using the UV-visiblenear IR spectrophotometer, the ab-sorption cross sections at 1480 nm in Er31-ion doped glasseswere also measured, and an example is shown in Fig. 4.Evidently, TeO2 glasses have a much larger cross sectionthan ZBLAN and Al/P silica glasses. A similar trend wasalso observed for the absorption at 980 nm wavelength.

C. Er3 emission line-shape analysis at room temperature

In order to study the effect of Er2O3 concentration on4I13/2-4I15/2 transition and lifetime of the lasing level, RR2glass was doped with various amounts of Er2O3. Fig. 5~a!shows the fluorescence spectra at different Er31-ion concen-trations; the fitted spectral components for 0.5 and 2 mol %Er2O3 are shown in Figs. 5~b! and 5~c!, respectively. It illus-trates that the higher the concentration of Er2O3 doped in theglasses, the broader its emission, while the peak positionremains unchanged. Because of the differences in the shapeof the emission spectrum in different hosts, FWHM is usedas a semiquantitative indication of increase in the bandwidth.Figure 6 compares the FWHM and the wavelength at halfmaximum fluorescence intensity ~HFMI! at different Er31concentration. The parameter HFMI, plotted in Fig. 6, is de-fined as the blue (xi) and red (yi) wavelengths at FWHM.The relationships plotted as a function of Er2O3 concentra-tion in a RR2 glass shows that the FWHM increases gentlyup to 11 000 ppm, whereas the points xi and yi virtuallyremain unchanged. Above 11 000 ppm concentration, there is

FIG. 4. Comparison of measured absorption cross sections ofRR2 and ZBLAN glasses for 4I15/24I13/2 transition.

PRB 62 STRUCTURAL ORIGIN OF SPa rapid increase in FWHM and in the values of xi and yi .Above 20 000 ppm, there is no significant change in theFIG. 5. ~a! Comparison of emission cross sections as a functionof Er31 ion concentrations in RR2 tellurite glasses for a 4I13/24I15/2 transition. ~b! Voigtian spectral components of the overallemission peak for a RR2 glass with 0.5 mol % Er2O3. The widths~w! and amplitudes ~a! of 7 Voigtian peaks are ~i! 1494.4 nm, ~w!31.57 nm, ~a! 0.2935; ~ii! 1535.3 nm, ~w! 10.35 nm, ~a! 0.4906; ~iii!1554.7 nm, ~w! 29.14 nm, ~a! 0.5198; ~iv! 1511.9 nm, ~w! 28.73nm, ~a! 0.2889; ~v! 1528.7 nm, ~w! 18.13 nm, ~a! 0.3714; ~vi!1543.9 nm, ~w! 9.06 nm, ~a! 0.2101; ~vii! 1588.9 nm, ~w! 50.85 nm,

6219TRAL BROADENING OF . . .~a! 0.1814. ~c! Voigtian spectral components of the overall emis-sion peak for a RR2 glass with 2.0 mol % Er2O3.

ANshape of the spectrum as function of erbium concentration inthe glass. The relationships in Fig. 6 also help in explainingthe dependence of the 4I13/2 lifetimes on Er31 and OH2 ionconcentrations described in Fig. 12.

Figures 7~a!, 7~b!, and 5~b! compare the normalized emis-sion spectra of Er31 ions doped in three different glass hostswith their fitted spectral components, namely, a modified

FIG. 6. The dependence of FWHM and HMFI on Er31 concen-tration in RR2 glasses.

FIG. 7. ~a! The emission line for a modified silicate glass ~seeTables III and IV! and its eight Voigtian spectral components. ~b!

6220 A. JHA, S. SHEN,The emission line for a ZBLAN glass ~see Tables III and IV! and itssix Voigtian spectral components.silicate, fluoride ~ZBLAN!, and a ternary tellurite glass, re-spectively. The compositions of these glasses are given inTable III. The concentration of Er31 ions in each host was 1mol %. The values of emission peak (lp nm), FWHM ~nm!,and emission cross section of Er31-doped glasses are alsocompared in this table, from which it is evident that the TeO2glasses studied in the present investigation have the largestvalue of emission cross section (8.5310221 cm2). Unlike themodified silicate and ZBLAN glass hosts studied in this in-vestigation, the spectral broadening in RR2 tellurite glass ismuch larger with the values of FWHM ranging between 50and 77 nm, compared to 65 nm in ZBLAN, 44 nm in Al/Psilica, and 8 nm in the silicate glass. The main reason for thebroadening of emission line shape in tellurite glass is thelocal ligand field environment of the dopant ions.

The dependence of the emission line broadening of Er31on the composition of the glass is described below, in whichthe overall line shape at room temperature has been decon-volved to identify the spectral components, see Figs. 5~b!,5~c!, 8~a!, and 8~b! for ternary and binary TeO2 glasses, andFigs. 7~a! and 7~b! for modified silicate and ZBLAN glasses,respectively. Figures 9~a!, 9~b!, and 9~c! are the examples ofthe absorption line broadening in two binary tellurite andmodified silicate glasses, respectively. The analyzed absorp-tion data for other compositions are summarized in Table IV.On the basis of the peak fit analysis for emission and absorp-tion curves at room temperature, the overall extent of Starklevel and sublevel splitting can be determined. In Table IVbelow, from the peak-fit analysis, the values of the sum ofthe energy differences between the 4I13/2 and 4I15/2 levels,DE total ~cm21! can be found, and compared with the valuesfor DE1 , which is for the ground-state absorption. The dif-ference (DE total2DE1) is the value of the extent of the Starksplitting of the ground state, 4I15/2 . The analysis unam-biguously demonstrates that the ground state encounters agreater degree of Stark splitting than the 4I13/2 level, whichis consistent with the earlier findings in the Er-doped silicateglasses.23 The table also shows that the number of compo-nent peaks in a binary tellurite glass is higher than in a ter-nary glass, thereby proving that as the molar concentrationsof modifying oxides increase, the TeO4 ~Raman peak C! toTeO31d ~peaks D and C! and TeO31d ~peaks D and C! toTeO3 ~peak E! structural transformation increase. The pres-ence of larger number of TeO3 units in RR15, as evidencedin the Raman spectra, Fig. 2, confirms that the continuity ofthe glass network is substantially decreased, leading to alesser variation in the environment of Er31 ions from one siteto other. As a result, in ternary glasses, there are only sevenspectral components @see Figs. 5~b! and 5~c!# compared to 10in a binary TeO2-ZnO glass, see Fig. 8~a!. There is also astrong evidence from the spectral analysis of the ternaryglasses in Figs. 5~a! and 5~b! that with increasing dopantconcentrations the Stark sublevel splitting increases, therebyindicating that the Er31 ions occupy new sites at higher con-centrations, which leads to a further line-shape modification.

In Table IV, we have also compared the room-temperature spectral components for three other types ofglasses besides binary and ternary tellurite compositions,e.g., a modified silicate and a ZBLAN glass. From the com-

PRB 62D M. NAFTALYparison of DE total and DE1 values, we find that the ZBLANglass has the smallest value of DE total and DE1 with only six

raction mayow Z values.een the Er-

ide glasses.ission crosse large, andorozirconateis clear thatglasses.

RR2 glassFIG. 8. Tglasses. ~a! 70mol % TeO2, 9

ione,

2)

i

ECwith larger atoms ~i.e., high Z!, the spin-orbit intedominate more than where the host atoms have lThis may be also an important distinction betwdoped silicate, fluorozirconate, and tellurium ox

For a broadband amplifier, the product, emsection (sab) times FWHM (lFWHM) should bthe results for RR2, modified silicate, and fluglasses are compared in Table III, from which itthe product sDl has the largest value for TeO2

D. Lasing level lifetimes

The decay rate from the 4I13/2 lasing level inhe fitted Voigtian peaks for two binary telluritespectral components. By comparison, the modified silicatehas eight spectral components for both absorption and emis-sion. At room temperature, the 4I13/2 Stark level and its sub-components in ZBLAN and TeO2 glasses have high popula-tion density fraction ~.0.5!, which is determined from theexp(2DE1 /kT) term. On the other hand, the same level in amodified silicate is less populated due to a larger Stark split-ting. The Stark splitting increases with increasing host glasscovalency, which is consistent with the comparison of DE1and DEground values in three different glass hosts discussed in

TABLE III. A comparison of emission cross-sectand FWHM ~nm! in tellurite (RR2), modified silicat

Glass composition ~mol %! s (10221 cm80 TeO2 , 9.5 ZnO, 9.5 Na2O, 1 Er2O3

(RR2)8.5

61 SiO2 , 11 Na2O, 3 Al2O3 , 12 LaF312 PbF2 1 ErF3 ~Modified silicate 193!

7.5

52 ZrF4 , 20 BaF2 , 3 LaF3 , 4 AlF3,20 NaF and 1 ErF3 ~ZBLAN!

5.1

a1 mol % Er2O3, FWHM for 2 mol % Er2O3 is 90 nm

PRB 62 STRUCTURAL ORIGIN OF SPmol % TeO2, 29 mol % ZnO, 1 mol % Er2O3 ; ~b! 90.5 mol % ZnO, 0.5 mol % Er2O3.this work. The bond covalency decreases in the followingorder: silicate.TeO2.ZBLAN. Based on the absorption andemission data for a binary tellurite glass ~69 mol % TeO2, 30mol % ZnO, 1 mol % Er2O3! the transitions between theStark subcomponents are shown in Fig. 10, and the corre-sponding data for the spread of energy sublevels are alsoshown in this figure. Due to the lack of low-temperaturephotoluminescence data, it is not possible to determine pre-cisely the energy sublevels.

The peak-fit analysis of room-temperature emission spec-tra shows that the individual Stark sublevels have both ho-mogeneous and inhomogeneous broadening. Homogeneousbroadening is described by Eq. ~1!, whereas the inhomoge-neous width of each Stark sublevel is due to site-to-sitevariation and depends on the composition of the glass host.As an example, the amplitude and width of each Stark sub-level is given in Figs. 5~b! and 10. Another important featureof the Stark component analysis is that the short-wavelength~i.e., the bottom sublevels! and the long-wavelength ~topsublevels! peaks are much broader than those in the centralwavelength regions of both the emission and absorptionspectra. This feature appears more dominant in the telluriteglasses than either a silicate or a ZBLAN glass. The com-parison of emission line shape and Stark sublevels points outthat in fluoride glass hosts, Er31 ions appear to have a muchless diversified range of environment due to its ionic charac-ter compared to those observed in the covalent TeO2 andmodified silicate glasses. The emission cross section in TeO2glasses is higher than ZBLAN and silicates due to theirhigher refractive index and a larger variation in the Er31sites. The effect of host glass electric dipole on the electro-static field and spin-orbit interactions is not distinguishablequantitatively from our present result. However, for glasses

(s310221 cm2), metastable lifetime of 4I13/2 leveland ZBLAN glasses.

t ~ms! FWHM, Dl ~nm! st sDl (cm2 nm)7.5 76a 64a 645

11.0 53 82 398

9.5 65 49 331

n RR2 compositions.

6221TRAL BROADENING OF . . .was found to be single exponential in all samples, regardlessof Er31-ion concentration. We have shown that the tellurite

ANglasses have multiple dopant sites for Er31 ions, and that the4I13/2 emission exhibits a considerable inhomogeneousbroadening due to a significant variation between the dopantsites. Therefore, the single exponential decay character of

FIG. 9. The fitted Voigtian peaks for the ground-state absorptionin two binary tellurite and modified silicate ~a! 90 mol % TeO2, 9.5mol % ZnO, 0.5 mol % Er2O3 ; ~b! 90 mol % TeO2, 9.5 mol %Na2O, 0.5 mol % Er2O3. ~c! A modified silicate, see Tables III andIV for composition.

6222 A. JHA, S. SHEN,fluorescence decay indicates the presence of fast energytransfer among Er31 ions, which takes place on time scalesmuch shorter than the decay lifetime, shown in Fig. 11. Thegoodness of fitted single-exponential equation confirms thatthe contributions from 2/e and 3/e folding times are virtuallyabsent. More detailed data analysis confirms that there is anabsence of Er-ion clustering even at high doping levels ~upto 12 000 ppm!, as can be seen in Fig. 12 in which the life-time changes with Er2O3 concentrations from 500 to 45 000ppm in the RR2 glass with two different OH2 concentra-tions. The corresponding OH2-ion absorption spectra for thelifetime versus Er2O3 curves shown in Fig. 12 are comparedin Fig. 1. These samples were made by melting the RR2composition at the same temperature for a fixed duration oftime, but with two different concentrations of OH2 contain-ing air ~high and low OH2!. The low-OH concentrationsamples were melted in dry air, while the high-OH sampleswere prepared in the laboratory air atmosphere. The lifetimeversus Er-ion concentration curves in Fig. 12 peak and thendrop gently, indicating that the lifetime of the lasing levelincreases with erbium ion concentration in theOH2-containing glass. The observed trend in this figure isalso consistent with the corresponding change in the value ofFWHM plotted as a function of concentration in Fig. 6. Atconcentrations of erbium ions less than 10 000 ppm, thesharp increase in the lifetime of the 4I13/2 level confirms thatneither the OH-activated quenching nor the ion-ion cross re-laxation is a dominating factor. Over and above 15 000 ppm,the lifetime drops more gently in low-OH glass than in ahigh-OH glass, indicating that the Er-Er ion cross-relaxationprocess dominates only after 15 000 ppm. The lower valuesof the measured lifetimes in high-OH glass are not unex-pected, which is due to an increase in OH2-ion inducedquenching.

For lower OH2 ion concentration, maximum lifetime of;7.6 ms was observed at around 11 000 ppm Er2O3. Withthe increase in OH2 ion concentration in host glass, the peakshifts to lower concentration of Er31 ions at around 5 000ppm having lifetime of ;4.9 ms. The values of measuredlifetimes reported in this work are at least 1.5 times longerthan the values ~,4 ms! reported by the investigators in Ref.12. Devices with large dopant concentrations and short pathlengths are possible without the effect of ion-ion cross relax-ation and/or OH-induced quenching. By implementing eitherdry oxygen or air atmosphere processing condition, glasseswith low OH2 and large ~sDl! product can therefore bedesigned for a broadband amplifier application.

E. Relaxation behavior of Er2O3-doped TeO2 glasses

As a part of the structural investigation on TeO2 glasses,thermal relaxation properties were measured using a differ-ential scanning calorimeter. The glass transition (Tg), onsetof devitrification (Tx), and the peak crystallization (Tp) tem-peratures were measured using a differential scanning calo-rimeter at an isochronal rate of 10 K mm21. The values ofTg , Tx , and Tp are plotted as a function of Er2O3 concen-tration in ppm in glass. The three characteristic temperaturesincrease moderately with Er31-ion concentration and thenlevel off, see Fig. 13. The increase in Tg and Tx is an indi-cation of strong bonding between Er ion and the host glass,

PRB 62D M. NAFTALYwhich might be dependent on the concentrations of Er31ions. The broadening of emission line with increasing erbium

The line sstates u ja& anand magnet^ jauMuib&, reportional to nThe line strental lattice fielspin-orbit ~LSnant with incever, the LSsis of hostenvironmentemission lineof Er31-ion ebasis of the sUV-visible spon the effecdynamic hostmodel for Temission lineferent types omodified tellu

a binary zincspectral widthw! 39.83 nm,i! 1517.5 nm,m, ~a! 0.446,m, ~w! 14.960.303, ~viii!

om1

10.6ii!99

ECstructural units

trength and the shape of a transition between thed uib& is dependent on the electric dipole ~D!ic dipole matrix elements ^ jauDuib& andspectively, both of (x (D) ,x (M )) which are pro-2, where n is the refractive index of the host.

gth also depends on the spin-orbit and the crys-d effect, also known as the ion-ligand field. The! mixing is also known to become more domi-reasing value of Z of rare-earth ions.2~a! How-mixing is also host glass dependent. The analy-glass structural units, which determines theof dopant ions, is essential in understanding thebroadening. The origin of spectral broadening

mission in TeO2 glasses can be explained on thetructural information deduced from Raman andectroscopic techniques. The discussion focuses

t of glass host structure ~ion-static host/ion-! interaction in Er-doped glasses. The followingeO2 glass network helps us in explaining thebroadening in Er-doped glasses. The three dif-

FIG. 10. An illustration of Stark subcomponents intellurite glass @see Table IV; Figs. 8~a! and 9~a!#. Theof each subcomponent is also given. ~i! 1498.9 nm, ~~a! 0.391, ~ii! 1558.0 nm, ~w! 17.42 nm, ~a! 0.659, ~ii~w! 29.75 nm, ~a! 0.217, ~iv! 1531.5 nm, ~w! 22.32 n~v! 1534.1 nm, ~w! 11.83 nm, ~a! 0.286, ~vi! 1544.6 nnm, ~a! 0.503, ~vii! 1569.6 nm, ~w! 16.90 nm, ~a!ion concentration strongly suggests that the Er31 ions may beoccupying sites as the modifying oxides and stabilizing non-bridging oxygen bonds, which is why the thermal relaxationresistance of glass increases nearly by 20 C with the addi-tion of Er2O3 in the glass. The increase in the characteristictemperatures may also be due to a strong bonding of erbiumions with nonbridging oxygen. This may suggest that Er31ions due to their high coordination number ~which variesbetween 6 and 7! cannot participate in the formation of glassnetwork.14,24 The contribution of 6/7-fold erbium to networkformation may be further restricted due to less than fivefoldcoordination of Te in TeO2 glass.

IV. DISCUSSION

A. Broadening contribution due to the glass-host

TABLE IV. The analyzed Stark subcomponent frfit data for a binary zinc tellurite glass 69 TeO2 30 ZnO16.59, ~ii! 1558.0 nm, ~w! 17.42 nm, ~a! 0.659, area6.88, ~iv! 1531.5 nm, ~w! 22.32 nm, ~a! 0.446, area 1~vi! 1544.6 nm, ~w! 14.96 nm, ~a! 0.503, area 8.02, ~v1582.3 nm, ~w! 21.92 nm, ~a! 0.197, area 4.73, ~ix! 15nm, ~w! 56.51 nm, ~a! 0.164, area 9.88.

Composition of glass~mol %!

DE total~cm21!

DEground~cm21!

69 TeO2 30 ZnO 1 Er2O3 472 41290 TeO2 9.5 Na2O

0.5 Er2O3443 371

RR2 with 0.5 mol %Er2O3

397 278

RR2 with 2.0 mol %Er2O3

425 303

ZBLAN 1.0 mol % ErF3 333 267Modified silicate-193,

1 mol % Er2O3570 357

PRB 62 STRUCTURAL ORIGIN OF SPf structural units TeO4, TeO31d , and TeO3 inrium oxide glasses exist, and their relative pro-portions are dependent on the chemical nature and the con-centrations of the network modifiers used. By comparison,the main structural units in modified silicate and ZBLANglasses are SiO4

42 tetrahedron and ZrF622 octahedron. These

three different types of structural units must therefore deter-mine the ligand field environment ~ion-static, ion dynamicfield, and ion-ion cross relaxation, which is also dependenton the structure! of the dopant ion and the host glass. In apure TeO2 glass, the TeO4 structure dominates, and the evo-lution of TeO31d and TeO3 commences with the addition ofthe network modifiers, as has been suggested by several au-thors in the past.1922 By referring to the previous structuralresearch in the literature,1922 the following models for struc-tural changes are proposed, which differ from the previous

the emission and absorption peak fit analysis. PeakEr2O3. ~i! 1498.9 nm, ~w! 39.83 nm, ~a! 0.391, area

2.23, ~iii! 1517.5 nm, ~w! 29.75 nm, ~a! 0.217, area, ~v! 1534.1 nm, ~w! 11.83 nm, ~a! 0.286, area 4.24,1569.6 nm, ~w! 16.90 nm, ~a! 0.303, area 5.45, ~viii!.3 nm, ~w! 39.79 nm, ~a! 0.179, area 7.61, ~x! 1610.2

DE1~cm21!

Fittedpeaks Experiment (2DE1 /kT)

60 10 0.7572 8 0.71

120 7 0.56

122 7 0.55

66 6 0.72212 8 0.36

6223TRAL BROADENING OF . . .1582.3 nm, ~w! 21.92 nm, ~a! 0.197, ~ix! 1599.3 nm, ~w! 39.79 nm,~a! 0.179, ~x! 1610.2 nm, ~w! 56.51 nm, ~a! 0.164.

structural analysis. The structural model proposed is consis-tent with the spectroscopic analysis reported in this paper.The approach is based on the understanding of the structuralchanges observed in the modified silicate and phosphateglasses.

~a! Two fourfold coordinated TeO4 units combine with adivalent oxide, e.g., ZnO and produce a new unit with Zn21as a bridging ion between a TeO3 unit and a TeO4 unit. Thebridging zinc ion (Zn21) can be considered as a part of aTe2O7 ~or TeO31d! structure. The zinc ion can also bridgebetween two TeO4 tbps via the bridging oxygen, as shownin Eq. ~5!. It should be noted that a TeO3 structure, which isa pyramid with Te at the apex and a double-bonded oxygen,favors an energetically favorable tetrahedron structure by al-lowing the Te atom at the apex to a new site similar to Te ina tbp structure.

~b! The structural modification to a TeO31d structurealso takes place in the presence of a monovalent sodiumion, as exemplified in Eq. ~5!, except in the case of sodiumion as a modifier where the zinc ion site is replaced bya bridging oxygen, as shown in Eq. ~4!. As a result,three nonbridging oxygens are produced, which are distrib-uted over two TeO4 tbps. Two such nonbridging oxygen

FIG. 11. The measured and fitted single exponential decay ratefrom 4I13/2 level to 4I15/2 in an RR2 glass. Excitation wavelengthwas 980 nm, and the concentration of Er31 was 0.2 mol %.sites may be similar to those sites, which are shared by, say,Na1 ions, in Eq. ~5!; the third one may also become part ofthe erbium ion environment at a second nonbridging oxygensite.

~c! The third possibility can also be exemplified via Eq.~5!, in which a TeO3 unit can replace a TeO4 unit by forminga structural unit:

Indeed such combinations of structural units with TeO3,TeO31d , and TeO4 will have a major influence on the ion-host ligand field interaction and contribute to a greater varia-tion between various rare-earth ion-point group sites. Bytransforming the glass structure from a tbp to a polyhedronTeO31d , which is the most dominant vibrational mode, i.e.,peak C, in the Raman spectra, the site for a double-oxygenbonded Te must change to a single-oxygen bonded Te. Thepresence of multiple structural units is the first contributoryfactor affecting the ion-host interactions both in the staticand dynamic fields.

~4!

~5!

FIG. 12. The dependence of the measured lifetimes of 4I13/2level on the concentrations of Er31 at two different OH2 concen-trations in RR2 glass.6224 PRB 62A. JHA, S. SHEN, AND M. NAFTALY

ECIt has been reported1820 that when modifying oxides,such as BaO, ZnO, and Na2O, are added in a TeO2 glass, theTeO4 network gradually transforms from 4-oxygencoordinated-Te41-LPE~O ! units ~tbp! to TeO31d polyhedra.The polyhedron structure is the result of a corner-sharingarrangement of TeO3 ~tp! and TeO4 ~tbp! structures, and thisis consistent with the model proposed in Eq. ~4!. Since aTeO3 and a TeO4 structure chemically differ, the structuraldistortion at the oxygen bridging sites shown in Eq. ~4! isexpected. The evolution of TeO31d structure is manifestedby changes in the intensities of Raman vibration peak C forantisymmetric structural breathing mode and peak A for thebending mode of the TeO2 glass structure. As a result of thecorner sharing of TeO4 and TeO3 units, the bending modeand antisymmetric vibrations of the Te2O7 polyhedron struc-tures intensify peak A at 450 cm21 and peak C at ;660 cm21in RR2 glass compared to RR14 and RR15 compositions.The emergence of Te2O7 ([TeO31d) polyhedron unit,therefore, gives rise to an additional set of weak and stronghost ligand Er31 ion interactions. The increased tendency forthe formation of TeO31d structure shifts the band edge of theTeO2 glasses to higher energies from 5.08 eV in RR14 ~aTeO2-rich glass! to 5.44 eV in RR2 ~see Fig. 3!, in which thestructure is more modified due to the presence of larger con-centrations of Na2O and ZnO. By comparison, the RR15composition has the highest concentrations of modifying ox-ides, Na2O and ZnO, resulting in the value of band edge at6.25 eV. The shift in the band-edge energy is also an indi-cation of the change in the cation field strength around theLPE, which may also affect the refractive index of the glass.It is well known that when an alkali ~or alkaline-earth! oxideis incorporated in germanate glasses,25 the refractive indexincreases up to 25 mol % of the modifying oxides, whichmeans that the band edge shifts to lower energy. Because inheavily modified tellurite glasses, the band edge shifts tohigher frequency, it can be concluded that the effect of alkaliand alkaline-earth oxide additions produces an opposite ef-fect in TeO2 glasses, which is an important factor in under-standing the emission line broadening, as it depends on the

FIG. 13. The dependence of the characteristic glass transition(Tg) and crystallization (Tx) temperatures on Er31 concentrationsin RR2 glass.

PRB 62 STRUCTURAL ORIGIN OF SPvalue of x (5E loc /E). On the basis of the analysis of Ramandata and the measured UV-edge results, it can be concludedthat the formation of Te2O7 structure with different modify-ing cations produces additional environments for Er31 ionsin TeO2 glasses, around which the possible bond configura-tions for LPE increase. The combination of LPE and Te2O7structural units is the second contributory in producing awide variety of ligand-ion environment and field strengthsaround Er ions. It should be noted that the lone-pair electronsare completely absent in fluoride and silicate, phosphate, andborate glasses.

The modifying cations, as expected, also create nonbridg-ing oxygens in the tellurite glass structures, as they do in thesilicate, phosphate, borate, and germanate glasses. In silicaglass, the solubility of RE ions is low and extensive ionclustering takes place. The clustering of RE ions is signifi-cantly reduced by the addition of Al31 and P51 ions, whichmodify the silica structure and disperse the RE ions in theenvironment of nonbridging oxygens.2628 Arai andco-workers26 showed that the dispersion of Nd31 ions in Al/P-containing silica is much better than in pure silica; thesame glass also exhibits a much better solubility and ion-ionseparation between erbium ions. As a result, the 1500 nmemission in Er-doped Al/P is much broader than in puresilica.2~a!,27,28 A similar effect is seen in the TeO2 glass due tothe addition of modifying oxides. Wells,29 on the basis of thecrystal structures of lanthanide sesquioxides, describeshigher than six coordination of Er in the Er2O3 oxide crys-tals. Such possibilities may also exist, which cannot be sub-stantiated in this paper without a detailed glass structure,particularly with reference to the sites for RE ions. From thecrystal coordination chemistry of sesquioxides, an erbiumion therefore may assume either a sixfold or sevenfold coor-dination by sharing the LPE, as shown in Fig. 14. Alterna-tively it may choose to share the nonbridging oxygen sites,as does a sodium ion, shown in Eq. ~5!. The Er31 ions atsuch sites will be then surrounded by six or seven TeO4 unitsby forming an octahedral-like cage. The possibility of morethan one type of coordination for an Er31 ion in a telluriteglass host is the third contributory factor.

B. Broadening of OH absorption peak

The multiplicity of LPE sites and nonbridging oxygens isresponsible for the extensive broadening of OH2 peak in aTeO2 glass, which is consistent with the model proposed byAdams.18 A wide range of structural sites, which are presentin a TeO2 glass, is not available in the SiO2, fluoride, andgermanate glasses. As a result OH2 ions disperse and formhydroxyl bonds over range of sites. Arnaudov andco-workers30 proposed the sites in barium-tellurite glassesare due to TeO3-OflHOTeO4 units. The authors suggested

FIG. 14. The schematic diagram of sixfold Erbium coordinationvia LPE in a TeO2 glass structure.

6225TRAL BROADENING OF . . .that the oxygen atom-electrodonors are strongly polarizedand become nonbridging sites.

ide anion in OH acts as an oxygen donor to TeOn structures2

refractive indices of silicates and ZBLAN glasses range be-

at the LPE sites at lower concentrations of OH ~see effectof OH2 on lifetimes of 4I13/2 level in Fig. 11!. By consider-ing the approach of Arnaudov and co-workers, we can con-sider that when an LPE site transforms into a nonbridgingdonated oxygen site ~NDOS!, by accepting the oxide anionfrom OH2, and then forms a bond with the neighboring Er31ion, there would be a net negative charge deficiency in theenvironment of Er31 ion. The donation of oxygen will ne-cessitate that the protons ~H1 ion! must move away at leastinto a second coordination cell of Er31 ion. This would alsomean that the new O-H2 bond due to the release of a protonwill not be available in the immediate vicinity of Er31 toinduce the quenching of luminescence. The modifying ionsin the same way donate oxygen to LPE sites to form NDOSand cause a major shift in the position of Urbach tail to shortwavelengths ~see Table I, E0 values!. It is expected that inthe presence of Er31 ions coupled with NDOS, the excessprotons will contribute to more nonbridging oxygen sitesaway from the original LPE sites. These may become evi-dent from the increase in the intensity of peaks C and D inRaman spectrum of the glass. Once all the LPE sites havetransformed into NDOSs, the ion-ion cross relaxation be-comes dominant. There is an additional complementary in-formation on the effect of OH2 ions on quenching of Erluminescence. The initial slope of lifetime versus Er2O3 con-centration is steeper in glasses with higher concentrations ofOH2 than glasses with lower OH2. This difference in theinitial slopes between high and low OH2 glasses suggeststhat there are much fewer LPE sites, which have not trans-formed to NDOSs in a high-OH glass than in a low-OH2glass. It is, therefore, possible to control LPEs and NDOSsby selecting a gaseous atmosphere together with the modify-ing cations in tellurite glass.

C. Prolonged metastable lifetime

There are two main reasons for a prolonged lifetime of7.8 ms in 11 000 ppm doped TeO2 glass structure. The firstone is associated with an increased ion-ion separation dis-tance, which may be at least of the order of 0.390 nm ~twotimes the shortest Te-O bond!, if not more, which can beestimated by considering the structural model in Fig. 14. Thesecond one is due to the increased Er31/OH2 ion distance.As proposed above, the LPE/NDOS transformation in theTeO2 glass structure ~and hence Er31/NDOS interaction!causes the H1 protons to move away in the second or highercoordination shells. As a result, on average the Er-OH sepa-ration distance is at least going to increase by a factor of 4.With the increasing distance, the probability of impurity-induced quenching is greatly reduced. In the present investi-

*Corresponding author. Email address: a.jha@leeds.ac.uk1 S. Sudo, Optical Fibre AmplifiersMaterials, Devices, and Ap-

plications, 1st ed. ~Arctec House, Boston, 1997!, pp. 178181.

2 ~a! W. J. Miniscalco, in Optical and Electronic Properties of

Rare Earth Ions in Glasses in Rare Earth Doped Fibre Laserstween 1.45 and 1.51, which are significantly lower than theaverage index of a tellurite glass composition ~;2.1! dis-cussed above. The radiative rate increases with the square ofrefractive index of the host, which means that a factor ofnearly 2 is required to compare the measured lifetimes oftellurite glasses with those of silicates. The measured life-times of 4I13/2 level for Er-doped silicates and ZBLAN areapproximately 12 ms, which by comparing the measuredlifetime and refractive index of an Er-doped tellurim oxideglass refractive index, should be of the order 1516 ms,thereby indicating that in the silicate and ZBLAN glass hostsdue to the structural constraints, the erbium ions are neitheruniformly dispersed nor they can be dissolved in higher con-centrations compared to a TeO2 glass. The measured longlifetime at higher concentrations of Er31 ions in the telluriteglasses may not be just due to the nonradiative component,but also due to the structural factors, which we are unable todistinguish in the present investigation.

V. CONCLUSIONS

~a! There are three major structural factors due to the hostglass, which contribute to line-shape broadening in Er-dopedTeO2 glasses. The first contribution is due to the presence ofthree different types of structural units in the TeO4 structure,which causes the Te-O bond length to change in equatorialand axial directions. As a result, the ion-host field strengthschange. The second contribution arises due to a change in theinteraction between an LPE and the modifying cations, e.g.,Zn21, Na1. The third contribution arises due to more thanone type of coordination of Er31 in a tellurium oxide glass.The combination of these factors contribute to the line-shapebroadening.

~b! The broad OH2 ion absorption band in TeO2 glass isdue to the multiplicity of sites, where a nonbridging oxygenforms with the TeO3, TeO4, and Te2O7 structures.

~c! The OH2 ions donate oxygens to LPE sites and re-lease protons, which occupy sites away from Er31-ion sites.The increased Er-OH ion separation distance reduces theprobability of OH-induced quenching of the metastable4I13/2 level in Er ions. The release of proton increases theEr-Er ion distance. As a result ion-ion clustering diminishesrapidly. These two factors contribute in prolonging the life-times of 4I13/2 level to 7.8 ms in 11 000 ppm doped glassesbesides the radiative components.

ACKNOWLEDGMENTS

The authors acknowledge the help and guidance given byProfessor Rui Almeida and Dr. L. Santos for carrying out theRaman spectroscopic experiments at INESC, Lisbon.

and Amplifier, edited by M. J. F. Digonnet ~Dekker, New York,1993!, pp. 2530; ~b! ibid., pp. 8284.

3 J. N. Sandoe, P. H. Sarkies, and S. Parke, J. Phys. D 5, 1788In the present investigation, from the spectroscopic obser-vations of Er31- and OH2 ions, the behavior of hydroxylions appears more complex and is described below. The ox-

2

6226 A. JHA, S. SHEN, Agation, we have also measured the lifetimes of 4I13/2 level insilicates and fluorozirconate glasses ~ZBLAN!. In each casethe measured total lifetimes vary between 10 and 11 ms. The

PRB 62ND M. NAFTALY~1972!.4 E. S. Georges ~unpublished!.

5 W. J. Miniscalco, J. Lightwave Technol. LT-9, 234 ~1991!.6 M. Semenkoff, M. Guibert, D. Ronarch, Y. Sorel, and J. F. Ker-

diles, J. Non-Cryst. Solids 184, 240 ~1995!.7 T. Whitley, C. A. Miller, R. Wyatt, M. C. Brierly, and D. Sze-

berta, Electron. Lett. 27, 1785 ~1991!.8 M. C. Brierly, C. A. Miller, and P. W. France ~unpublished!.9 R. Reisfeld and Y. Eckstein, Solid State Commun. 13, 741

~1973!.10 R. Reisfeld and Y. Eckstein, J. Non-Cryst. Solids 15, 125 ~1974!.11 R. Reisfeld and Y. Eckstein, J. Non-Cryst. Solids 12, 357 ~1973!.12 A. Mori, Y. Ohishi, and M. Yamada ~unpublished!.13 M. Yamada, A. Mori, K. Kolayashi, H. Ono, T. Kanamosi, K.

Oikavia, Y. Nishida, and Y. Ohishi, IEEE Photonics Technol.Lett. 10, 1244 ~1998!.

14 J. E. Stanworth, Nature ~London! 169, 581 ~1952!.15 G. W. Brady, J. Chem. Phys. 24, 477 ~1956!; ibid. 27, 300 ~1957!.16 W. Vogel, in Chemistry of Glasses, 2nd ed., edited by N. Kreidl

~The American Ceramic Society, Westerville, OH, 1985!, pp.167174.

17 P. W. France, Fluoride Glass Optical Fibres ~Blackie, London,1990!, pp. 135142.

18 R. V. Adams, Phys. Chem. Glasses 2, 50 ~1961!.19 T. Komatsu, H. G. Kuin, and H. Mohri, J. Mater. Sci. Lett. 15,

2026 ~1991!.20 J. C. Sabadel, P. Armand, D. Cachan-Herreillat, P. Baldeck, O.

Doclot, A. Ibanez, and E. Philippot, J. Solid State Chem. 132,911 ~1997!.

21 Y. Himei, A. Osaka, T. Nanba, and Miura, J. Non-Cryst. Solids177, 164 ~1994!.

22 T. Sekiya, N. Mochida, and A. Ohtsuka, J. Non-Cryst. Solids 168,106 ~1994!.

23 E. Desurvire and J. R. Simpson, Opt. Lett. 15, 547 ~1990!.24 J. H. Simmons, C. J. Simmons, R. Ochoa, and A. C. Wright, in

Fluoride Glass Fiber Optics, 1st ed., edited by G. Lu and I. D.Aggarwal ~Academic, New York, 1991!, pp. 40 and 41.

25 A. Paul, Chemistry of Glasses, 2nd ed. ~Chapman and Hall, Lon-don, 1990!, p. 104.

26 K. Arai, H. Namikawa, K. Kumata, and T. Honda, J. Appl. Phys.59, 3430 ~1986!.

27 A. Nakazawa and Y. Kimnsa, Electron. Lett. 27, 1065 ~1991!.28 M. Ohashi and K. Shiraki, Electron. Lett. 27, 2143 ~1991!.29 A. F. Wells, Structural Inorganic Chemistry ~Clarendon, Oxford,

1962!, pp. 464467.30 M. Arnaudov, Y. Dimitriev, V. Dimitrov, and M. Dimitrova-

Pankova, Phys. Chem. Glasses 27, 48 ~1986!.

PRB 62 6227STRUCTURAL ORIGIN OF SPECTRAL BROADENING OF . . .