Spectroscopic properties and upconversion mechanisms in Er31-doped fluoroindate glasses T. Catunda, L. A. O. Nunes, and A. Florez Instituto de Fı´sica de Sa˜o Carlos (IFSC), Universidade de Sa˜o Paulo, Caixa Postal 369, 13560-970 Sa˜o Carlos, Sa˜o Paulo, Brazil Y. Messaddeq Instituto de Quı´mica, Universidade Estadual de Sa˜o Paulo, Araraquara, Sa˜o Paulo, Brazil M. A. Aegerter Instituto de Fı´sica de Sa˜o Carlos (IFSC), Universidade de Sa˜o Paulo, Caixa Postal 369, 13560-970 Sa˜o Carlos, Sa˜o Paulo, Brazil ~Received 24 July 1995! Fluorindate glasses containing 1,2,3,4 ErF3 mol % were prepared in a dry box under argon atmosphere. Absorption, Stokes luminescence~under visible and infrared excitation!, the dependence of4S3/2, 4I 11/2, and 4I 13/2 lifetimes with Er concentration, and upconversion under Ti-saphire laser excitation atl5790 nm were measured, mostly atT577 and 300 K. The upconversion results in a strong green emission and weaker blue and red emissions whose intensity obeys a power-law behaviorI;Pn, whereP is the infrared excitation power andn51.6, 2.1, and 2.9 for the red, green, and blue emissions, respectively. The red emission exponentn51.5 can be explained by a cross relaxation process. The green and blue emissions are due to excited state absorp- tion ~ESA! and energy transfer~ET! processes that predict a factorn52 andn53 for the green and blue emissions, respectively. From transient measurements we concluded that for lightly doped samples the green upconverted emission is originated due to both processes ESA and ET. However, for heavily doped samples ET is the dominant process. I. INTRODUCTION There is a great interest in the study of rare-earth doped heavy metal fluoride glasses. These materials present a high transparency from UV to IR region; they can be easily pre- pared and a relative high concentration of transition metal and rare earth can be incorporated into the matrix.1 Due to low multiphonon emission rates, rare-earth doped fluoride glasses present large upconversion efficiencies and fluorozir- conate fiber lasers have been reported.2 Besides the well-known zirconates glasses several compo- sitions based on indium fluoride have been studied. Com- pared to fluorozirconate glasses, these compositions present higher transparency in the mid-infrared range~up to 8mm!- lower multiphonon emission rates and are also more stable against atmospheric moisture.3 The spectroscopy of Eu31 and Gd31 doped fluoroindate was studied showing evidence of Eu-Eu and Gd-Eu energy transfer.4 Upconversion of flu- oroindate glasses doped with Pr31 ~Ref. 5! and Er31 ~Ref. 6! have been reported. In a previous work,6 some of us studied upconversion using a red laser~l5647 nm! to pump the Er31 ions from the ground state to4F9/2. Er 31 doped fluoride glasses conventional spectroscopy7–9and upconversion6,10–12 have been studied very much lately. Upconversion have also been studied in many other glass types doped with Er31, like silicate,13 fluoride,10–13 tellurite,14–15 oxide,16 chloride17 and others. A comparative study between fluoride and many oth- ers glasses shows that the superior upconversion emission is caused by their lower phonon energies.18 In the present work, we report results of absorption, fluo- rescence, lifetimes measurements and upconversion mea- surements in Er31 doped fluoroindate glasses. The upconver- sion was studied by using a Ti-saphire laser pumping the4I 9/2 level. The results obtained in this work are substantially dif- ferent from the ones obtained by pumping the4F9/2 level and provide a much better understanding of the upconversion mechanisms involved. II. EXPERIMENT Fluoride glasses with bath compositions 20ZnF2-20SrF2-2NaF-16BaF2-6GaF3-~362x!InF3-xEr3 with x51, 2, 3, and 4 mol % were prepared. The concentration range ~1–4 %! corresponds to 2.0–8.031020 Er ions/cm3. The mixture was heated in a platinum crucible at 800 °C during one hour for melting and at 850 °C for fining, both treatments were performed in a dry box under argon atmo- sphere. The melt was cast into a preheated mold at 260 °C and slowly cooled down to room temperature. The samples were cut and polished into a parallel piped shape. The ab- sorption paths of the samples were 2.07 mm, 1.77 mm, 2.01 mm, and 1.43 mm for concentrations 1,2,3, and 4 mol %, respectively. The emission spectra~both conventional or Stoke lumi- nescence and upconversion or anti-Stoke luminescence! were analyzed using a SPEX 1403 double monocromator equipped with RCA 31034 photomultiplier, connected to a PAR-128 lock-in amplifier. The infrared radiation~at 1500 and 970 nm! was detected by using a nitrogen cooled Judson 516-D detector. The Stokes luminescences were obtained by pumping the Er31 with a Coherent Innova 400 Ar ion laser with a mirror for the UV lines~351 and 364 nm!. Upconver- sion spectra were obtained by using a Coherent Mira-Basic Ti-saphire laser atlp5790 nm~used in cw mode!, resonant with the 4I 15/2- 4I 9/2 transition. Lifetimes were measured by chopping the cw pumping beam with a mechanical chopper, PHYSICAL REVIEW B 1 MARCH 1996-IIVOLUME 53, NUMBER 10 530163-1829/96/53~10!/6065~6!/$10.00 6065 © 1996 The American Physical Society and by signal averaging the resulting fluorescence decays on a Hewlet Packard digital oscilloscope. III. RESULTS AND DISCUSSION A. Stokes emission and lifetimes measurements By using the absorption data9 the energy level diagram for Er31 fluoroindate glass was determined. Table I shows the energy~cm21!, peak cross section and radiative lifetime of the most important energy levels. These calculations were carried out by using the Judd-Ofelt theory; the complete data with Judd-Ofelt parameters, transitions probabilities, branch- ing ratios, radiative lifetimes, and peak cross sections for stimulated emission can be found in Ref. 9. The Stokes emission~conventional luminescence! spectra for the 3 mol % sample are shown in Fig. 1 and 3. Figures 1~a! and 1~b! show the visible Stokes emission obtained by pumping the sample with UV radiation from the Ar laser. The main laser lines are at 351 and 364 nm, the second one is in resonance with the4G9/2 level. These spectra show three main groups of lines~blue, green, and red! that appear also in the absorption spectrum and correspond to the following transitions: 2G9/2→4I 15/2; 2H11/2, 4S3/2→4I 15/2, and 4F9/2→4I 15/2. The line centered at;12 000 cm21 corre- sponds to transition4S3/2→4I 13/2. The 300 K spectrum@Fig. 1~a!# also shows transitions4G11/2→4I 13/2, 2G9/2→4I 11/2, and 2H11/2, 4S3/2→4I 13/2. These transitions are indicated in pro- cess~a! in Fig. 2. As the2H11/2 level is thermally populated its fluorescence disappears at low temperatures. The peculiar feature of these spectra, both at 300 and 77 K, is that the blue emission is much more intense than the red one. The oppo- site is found, for instance, in ZBLA~ZrF4-BaF2-LaF3-AlF3! glass.7 It is also remarkable that the green emission increases by about one order of magnitude when the sample is cooled to 77 K. Figures 3~a! and 3~b! show the infrared Stokes emission spectra from 13 000–6000 cm21 ~corresponding to 0.8–1.7mm!, obtained by pumping with the 488 nm radia- tion from the Ar laser@see also process~b! in Fig. 2#. It is interesting to observe that at 77 K, the4S3/2→4I 13/2 lumines- cence is more intense than that of4I 13/2→4I 15/2 transition and the transition4S3/2→4I 11/2 cannot be seen at 300 K. This is a consequence of the growth of the4S3/2 population when the temperature decreases as shown by relative increase of the 547 nm peak in Figs. 1. The peak observed in Fig. 3 indi- cated by an asterisk was not identified as an Er31 transition FIG. 1. Visible emission spectra of the 3 mol % sample at~a! 300 K and~b! 77 K. In part~b! the vertical scale is 12 times larger than part~a!. The spectra were obtained by pumping withl;355 nm radiation. Unless when indicated, all the peaks are related to transitions from excited states to the ground state4I 15/2. FIG. 2. Energy level diagram and Stokes emission processes under 364 nm~a!, 488 nm~b!, and 790 nm~c! laser pumping. TABLE I. Energy, peak cross sectionrP and calculated radiative lifetime tR at 300 K~all the transitions are from the indicated levels to the ground state4I 15/2!. Energy level Energy~cm21! rP ~10220 cm2! tR ~msec! 4I 13/2 6 602 0.4879 9.54 4I 11/2 10 279 0.2563 10.83 4I 9/2 12 612 0.2328 5.58 4F9/2 15 379 1.1547 0.623 4S3/2 18 552 0.7218 1.06 2H11/2 19 267 2.1767 0.211 4F7/2 20 592 1.3233 0.316 4F3/2, 4F5/2 22 362 0.7227 0.50 2G9/2 24 692 0.3265 0.70 4G11/2 26 457 3.0054 0.048 4G9/2 27 451 1.2264 0.17 6066 53CATUNDA, NUNES, FLOREZ, MESSADDEQ, AND AEGERTER and is possibly due to an impurity. Both absorption9 and Stokes emission spectra obtained from samples with differ- ent concentrations show identical characteristics and no sig- nificant effect of the doping level on the band structure has been observed. The lifetimes were obtained by fitting the luminescence time decay curves. The4S3/2 lifetimes were measured by pumping the4F7/2 level with 488 nm radiation from the Ar ion laser. For Er doped materials, usually all the levels ex- cited above4S3/2 decay very fast to this level, by nonradia- tive cascade processes, and emit a green luminescence at 547 nm. Consequently, the nonradiative decay to the4S3/2 level is much faster than the4S3/2 lifetime and it can be disregarded in the lifetime analysis. The measurements were taken at 300 and 77 K and the results are shown in Table II. The fluorescence decays originated from the4I 11/2 and 4I 13/2 levels, were obtained by pumping the4I 9/2 Er 31 level by using infrared radiation~lp5790 nm! from a Ti-sapphire laser. Level4I 9/2 has a very short lifetime~of the order of tens of microseconds7! due to fast nonradiative decay to 4I 11/2. Thus the4I 9/2 lifetime is negligible compared with 4I 11/2 radiative lifetime which is about 8 msec as shown in Table II. Consequently, by taking the same arguments used for 4S3/2 fluorescence, 4I 11/2 lifetime can be determined by analyzing its fluorescence decay underlp5790 nm excita- tion @process~c! in Fig. 2#. However, we cannot use this argument to explain the4I 13/2 fluorescence~using lp5790 nm! because the decay times4I 11/2→4I 13/2 and 4I 13/2→4I 15/2 are comparable. The fluorescence temporal behavior can be obtained from the rate equations for populationsn2 andn1 of level 2 ~4I 11/2! and level 1~4I 13/2!, as indicated in Fig. 2, given by dn3 /dt5s03Fn32A2n2 , ~1a! dn2 /dt5A32n32A2n2 , ~1b! dn1 /dt5A21n21A31n32A1n1 , ~1c! wheresi j is the absorption cross section for the transition from level i to j , F the incident pump flux forlp5790 nm, Ai j is the partial relaxation rate from leveli to j andAi the total relaxation rate from leveli given by Ai5( jAi j . We should remark that Eqs.~1! are valid only at low excitation power so that upconversion terms can be neglected. By re- minding thatA32@A2 , theA32 term in Eq.~1b! can be ne- glected for longer times (t@1/A32). Also the radiative decay termA31 in Eq. ~1c! can be neglected because almost all the ions decay nonradiatively to level 2 (A32@A31). With these simplifications then2(t) population decay~ 4I 11/2! is given by a single exponential andn1(t) ~4I 13/2! by a double exponen- tial: n2~ t !5N2e 2A2t, ~2a! n1~ t !5 A2N1 A22A1 S e2A1t2 A1 A2 e2A2tD , ~2b! whereN35s03FN0/A3 , N25A32N3/A2;s03FN0/A2 , and N15A21N2/A1 . The 4I 11/2 lifetime t251/A2 was obtained by fitting the 980 nm fluorescence decay with an exponential curve. By using thist2 value as a fixed parameter, the4I 13/2 fluorescence decay atl51.5 mm was fitted by a double ex- ponential in order to determine the lifetimet151/A1. Figure 4 shows the experimental fluorescence decays of levels4I 11/2 and4I 13/2 with fits which agree well with Eq.~2b!. The ratio (A1/A2)50.81 is in good agreement with the factor 0.80 obtained in the fit. In solids doped with rare-earth ions, the experimentally observed decay rate is a sum of three rates: a radiative decay ~WR5t R 21 whose theoretically calculated values are given in Table I!, a multiphonon emission rateWph and a rate due to luminescence quenchingWq ~also called concentration quenching!. Wph depends strongly on the matrix phonon spectra and decreases exponentially with the effective pho- non number~p5E/hveff!, which is the well-known energy gap law. Raman spectra for ZBLAN~ZrF4-BaF2-LaF3-AlF3! FIG. 3. Infrared emission spectra of the 3 mol % sample at~a! 300 K and~b! 77 K. The spectra were obtained by pumping with l5488 nm radiation. The peak indicated by asterisk is probably due to an impurity. TABLE II. Lifetimes measurements. Energy level Er concentration x ~mol %! T577 K t ~msec! T5300 K t ~msec! 4S3/2 1 0.73 0.43 4S3/2 3 0.68 0.17 4S3/2 4 0.68 0.16 4I 11/2 1 10.7 8.5 4I 11/2 4 10.3 8.1 4I 13/2 1 11.8 10.3 4I 13/2 4 11.7 10.0 53 6067SPECTROSCOPIC PROPERTIES AND UPCONVERSION . . . show a polarized band at 580 cm21.19 In fluoroindate glasses Raman spectra show a strong polarized band at;507 cm21 and a broad depolarized band centered at 203 cm21.20 IR reflection spectra show modes at;484 and 225 cm21. In Eu31 and Gd31 doped fluoroindate glasses only vibronic bands associated with a vibrational mode at around 329 cm21 could be observed.4 The temperature behavior of the emission rate in Er31 doped ZBLA was studied in the range 4–500 K.7 For the 4S3/2 level it was observed thatWph is constant at low tem- peratures and increases aboveT;100 K. In Er doped crys- tals and glasses the4S3/2 lifetime decreases with the concen- tration by two experimentally indistinguishable cross- relaxation processes~Refs. 14 and 21!: (4S3/2, 4I 15/2) →(4I 9/2, 4I 13/2) and (4S3/2, 4I 15/2)→(4I 13/2, 4I 9/2). In the tem- perature range 77–300 K, for Er31 doped ZBLA, the4S3/2 experimental lifetime decreases by a factor 1.3 and 3.0 for 0.5 and 2 mol % Er concentration, respectively. For tempera- tures lower than 100 K both samples, with 0.5 and 2 mol % Er have the same lifetime. The fact that the 2 mol % sample has a shorter lifetime~for T.100 K! indicates that cross- relaxation processes play an important role. We observed similar results, i.e., a decrease by a factor 1.1 and 2.7 for the 1 mol % and 3 mol % samples, respectively, when the tem- perature increases from 77 to 300 K~Table II!. A similar behavior was also observed in tellurite glasses.14 The lifetime decreases strongly at high temperature but is almost constant at low temperature. The decreasing of the lifetime due to the increasing of the Er concentration (Wq) was also studied. At T5300 K the4S3/2 lifetime decreases by a factor 2.3 when Er31 concentration increases from 2.0 to 8.031020 ions/cm3. In the same concentration range, we observed a decrease by a factor of 2.8~Table II!. As observed in ZBLA,7 tellurite,14 aluminate and gallate18 glasses, the4I 11/2 and 4I 13/2 lifetimes have a much less pronounced dependence with both concen- tration and temperature than the4S3/2 lifetime. This is a con- sequence of the small number of phonons involved. B. Upconversion process Upconversion spectra have been measured by pumping Er ions with a Ti-saphire laser atlp5790 nm. Figures 5~a! and 5~b! show the spectrum for the 3 mol % sample atT5300 K andT577 K, respectively. We observed a very intense green luminescence atl;547 nm corresponding to the thermally coupled2H11/2, 4S3/2→4I 15/2 transition, a red luminescence at l;650 nm due to the4F9/2→4I 15/2 transition and a blue luminescence atl;407 nm from the2G9/2→4I 15/2 transition. At 77 K @Fig. 5~b!# the green upconversion emission in- creases about one order of magnitude similar to the behavior observed in the visible Stokes emission spectra~Fig. 1!. We also remind that in both visible and infrared Stokes emission spectra~Figs. 1 and 3! the intensity of all transitions coming from the4S3/2 level increase at 77 K indicating an increase of its population. In barium-thorium fluoride glasses, Yehet al.7 observed an increase of the upconversion green emission by a factor of;2.8 when the temperature is decreased from 300 to 77 K. The green and blue spectrum shapes are very similar to the ones obtained by pumping the4F9/2 level with a Kr ion laser~lp5647 nm!.6 However, the integrated blue emission is much smaller for pumping atlp5790 nm than for pump- ing atlp5647 nm. To obtain more insight into the upconversion mechanisms the dependence of the upconversion intensityI was mea- sured as a function of the incident pump powerP atlp5790 nm. All the experimental results can be fitted to a power-law behavior I;Pn, with n52.1 for the green luminescence, n51.6 for the red luminescence, andn52.9 for the blue luminescence as shown in Fig. 6. A similar behavior for the green and blue luminiscences was observed by Harriset al.12 in ZBLAN glasses and for the red emission in Te based glasses by Oomenet al.10 FIG. 4. Fluorescence decays from the4I 13/2 level ~l51.5 mm! and4I 11/2 level ~l50.973mm! of the 1 mol % sample at 300 K. The line are the fits for the4I 11/2 and 4I 11/2 fluorescences, a single expo- nential decay ~t258.1 msec! and a double exponential n1;(e2t/1020.80e2t/8.1), respectively. These fits are in good agreement with Eq.~2!. FIG. 5. Upconversion~anti-Stokes emission! spectra of the 3 mol % sample under infrared excitation~l5790 nm!; ~a! 300 K, and ~b! 77 K. 6068 53CATUNDA, NUNES, FLOREZ, MESSADDEQ, AND AEGERTER As discussed previously, thelp5790 nm excitation pumps essentially the Er31 ground state to the level4I 11/2 @process~a! in Fig. 7#. Two possible mechanisms for the green fluorescence can be envisaged: excited state absorption ~ESA! and energy transfer~ET!. As shown by process~b! in Fig. 7, ESA can happen from4I 11/2 or 4I 13/2 states. The 4I 11/2 excited ion absorbs one more infrared photon, goes to4F3/2 and then decays to2H11/2, 4S3/2. In the second case, the 4I 13/2 ion is promoted to the2H11/2 by absorbing another photon and then decay to4S3/2. In both cases the infrared laser is very close to resonance with ESA transitions~the detuning is less than 5%!. However, the ESA absorption cross section of the second case is 2.3 times larger,9 and appears therefore more probable. In ET process, two4I 11/2 excited Er ions ex- change energy so that one of them~the donor ion! decays to ground state4I 15/2 and the other one~the acceptor! is excited to 4F7/2 which decays to4S3/2 @process~c! in Fig. 7#: 2 4I 11/2→4I 15/21 4F7/2. ~3! In both cases~ESA and ET! two infrared photons are needed to reach4S3/2 which emits one green photon. This indicates an n52 exponent in the power-law behavior, that is very close to the experimental valuen52.1. The blue luminescence can also be explained by ESA and ET transfer processes. ESA is possible by the transitions 4I 9/2→2G9/2 or 4F9/2→4G9/2 which decay to2G9/2 @process ~d! in Fig. 7#. ET may happen with the donor ion excited in 4I 11/2 and the acceptor ion excited in 4F9/2. The donor decays to ground state and excites the acceptor to4G11/2 level, which decays to2G9/2 @process~e! in Fig. 7#. In both ESA and ET processes, three infrared photons are needed to excite one ion to 2G9/2 giving one blue photon. In this case, ann53 behav- ior is expected, in good agreement with then52.9 experi- mental result. The blue upconversion luminescence is much weaker than the green one because the excitation mecha- nisms for the blue one involve states of shorter lifetimes and consequently having a lower population than the levels in- volved in the green luminescence. In the case of the red upconversion, the experimentally observedn51.5 dependence can be explained by a cross relaxation process.10 The donor ion in4S3/2 state decays to 4F9/2 while the acceptor ion in4I 9/2 state is excited to4F9/2 @process~f! in Fig. 7#: 4S3/21 4I 9/2→2~4F9/2!, ~4! one infrared photon is needed to pump the4I 9/2 state and two infrared photons are needed to reach4S3/2 producing two red photons. This reasoning gives an51.5 dependence close to the valuen51.6 found experimentally. We should remark that the green and blue upconversion results obtained with a red excitation~647 nm!, close to4F9/2 level, are different from the ones presented in this work. In this case the experiment gave also a power law but with n51.5 for the green emission andn51.6 for the blue one ~results obtained for the 3 and 4 mol % samples!. The dis- crepancy between the experimental values and the expected valuen52 was attributed to saturation effects. The exponent n also varies when the same Er31 energy level is pumped in different types of glasses. Experiments were done in oxide,16 chloride,17 ZBLAN,12 and in the present work, pumping the 4I 9/2 using IR radiation around 800 nm. However in chloride glassesn;2 was obtained for both green and blue lumines- cence and in oxide glassn51.5 was obtained for the green luminescence. This indicates that the upconversion mecha- nisms depend strongly not only on the pumping wavelength, which determines the pumped energy level, but also on the glass type. We also studied the transient behavior of the green upcon- version emission. The 1 mol % Er emission decay can be fitted by a double exponential with decay ratesA5 and 2A2, whereA5 andA2 are the total relaxation rates of the levels 4S3/2 and 4I 11/2, respectively. The emission decays from the 3 and 4 mol % samples can be fitted by a single exponential with decay rate 2A2 ~decay time;4 msec!. This lengthening of the 4S3/2 emission decay can be attributed to the ET pro- FIG. 6. Log-log plots of the blue, green, and red upconversion emission intensities as a function of the infrared 790 nm excitation power for the 3 mol % sample. FIG. 7. Er31 energy level diagram and excitation mechanisms. 53 6069SPECTROSCOPIC PROPERTIES AND UPCONVERSION . . . cess given by Eq.~3! as observed in LiYF4:1% Er31.22 Con- sequently, we believe that for lightly doped Er samples the green upconversion is originated from both ESA and ET pro- cess. However, for heavily doped samples ET is the domi- nant process as observed in ZBLAN by Harriset al.12 ACKNOWLEDGMENTS This research was supported by Telebra´s, Fapesp, Capes, CNPq program RHAE-New Materials, Brazil, and the De- partamento de Fı´sica, Universidad Industrial de Santander A. A. 678, Bucaramanga, Columbia. 1Fluoride Glass for Optical Fibers, edited by P. W. France ~Blackie, London, 1990!. 2J. Y. Allain, M. Monerie, and H. Pognant, Electron. Lett.26, 166 ~1990!. 3Y. Messaddeq, A. Delben, M. A. Aegerter, and M. Poulain, J. Mater. Res.8, 885 ~1993!. 4S. J. L. Ribeiro, R. E. O. Diniz, Y. Messaddeq, L. A. O. Nunes, and M. A. Aegerter, Chem. Phys. Lett.220, 214 ~1994!. 5L. E. E. Leite, A. S. Gomes, C. B. Arau´jo, Y. Messaddeq, A. Florez, O. L. Malta, and M. A. Aegerter, Phys. Rev. B50, 16 219~1995!. 6R. Reiche, L. A. O. Nunes, C. C. Carvalho, Y. Messaddeq, and M. A. Aegerter, Solid State Commun.85, 773 ~1993!. 7D. M. Shinn, W. A. Sibley, M. G. Drexhage, and R. N. Brown, Phys. Rev. B27, 6635 ~1983!; D. C. Yeh, W. A. Sibley, M. Suscavage, and M. G. Drexhage, J. Appl. Phys.62, 266 ~1987!. 8R. Reisfeld, G. Katz, C. Jacoboni, R. de Pape, M. G. Drexhage, R. N. Brown, and C. K. Jorgensen, J. Solid State Chem.41, 253 ~1982!; R. Reisfeld, G. Katz, N. Spector, C. K. Jorgensen, C. Jacoboni, and R. de Pape,ibid. 41, 253 ~1982!. 9A. Florez, Y. Messaddeq, O. L. Malta, and M. A. Agaerter, J. Alloys Compounds227, 135 ~1995!. 10E. W. J. L. Oomen, P. M. T. Le Gall, and A. M. A. van Dongen, J. Lumin.46, 353 ~1990!. 11M. Takahashi, R. Kanno, Y. Kawamoto, S. Tanabe, and K. Hirao, J. Non-Cryst. Solids168, 137 ~1994!. 12E. A. Harris, T. M. Searle, J. M. Parker, and D. Furniss~unpub- lished!. 13R. S. Quimby, M. G. Drexhage, and Suscavage, Electron. Lett. 23, 32 ~1987!. 14W. Ryba-Romanowski, J. Lumin.46, 163 ~1990!. 15S. Tanabe, K. Hirao, and N. Soga, J. Non-Cryst. Solids122, 79 ~1990!. 16B. R. Reddy and P. Venkateswarlu, Appl. Phys. Lett.64, 1327 ~1994!. 17M. Shojiyam, M. Takahasi, R. Kanno, and K. Kadono, Appl. Phys. Lett.65, 1874 ~1994!; A. Gharavi and G. L. McPherson, Appl. Phys. Lett.61, 2635~1992!. 18X. Zou and T. Izumitani, J. Non-Cryst. Solids162, 68 ~1993!. 19R. M. Almeida, J. Non Cryst. Solids106, 347 ~1988!. 20R. M. Almeida, J. C. Pereira, Y. Messaddeq, and M. A. Aegerter, J. Non-Cryst. Solids161, 105 ~1993!. 21D. S. Knowles and H. P. Jenssen, IEEE28, 1197~1992!. 22J. Rubin, A. Brenier, R. Moncorge, and C. Pedrini, J. Lumin.36, 39 ~1986!. 6070 53CATUNDA, NUNES, FLOREZ, MESSADDEQ, AND AEGERTER