Erbium-doped glasses for fiber amplifiers at 1500 nm...

234 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 9. NO. 2, FEBRUARY 1991

Erbium-Doped Glasses for Fiber Amplifiers at 1500 nm

William J . Miniscalco (Invited Paper)

Abstract-The importance of the host glass composition has been recognized from the earliest days of fiber lasers and amplifiers. Material- dependent properties influencing the performance of fiber amplifiers are reviewed together with the available data for Er3+. The major glass types potentially useful in this application are considered and compared to silica. The topics addressed include quenching processes and the solubility of rare-earth ions, transition strengths and bandwidths at the 1500-nm gain transition, and the characteristics at the 800-, 980-, and 1480-nm pump bands. Aluminum is shown to he an extremely useful codopant for silica, improving its ability to dissolve rare- earth ions and providing desirable spectroscopic properties for ErJ+. For some of the attributes considered, other glasses have advantages over AI silica, but only with respect to gain bandwidth and pumping at 800 nm is significantly better performance expected from other glass compositions.

I . INTRODUCTION HE importance of glass composition has been recognized T from the very beginning of the rather venerable study of

gain in doped optical fibers. The first lasers [ I ] and amplifiers [ 2 ] were demonstrated in the early 1960's using Nd3+-doped barium crown glasses which were chosen for the long lifetime of the metastable state and their immunity to concentration quenching [ 11. About a decade later another major advance was achieved with the operation of end-pumped glass-clad Nd3+- doped silica fiber lasers [ 3 ] . These investigators recognized that AI silica behaves like a multicomponent glass host for Nd3+ while pure silica is anomalous [3]. Still another decade later, a major impetus was provided by reports of procedures for doping rare earths into low-loss single-mode silica fiber [4] and the demonstration of low threshold, high-slope-efficiency Nd3+ fiber lasers [5]. In a further acceleration of the development of this technology, there has been a veritable explosion of activity following the announcement of high-gain low-noise E$' amplifiers at a principal optical communications window [6], [7].

One reason for this interest is found in Fig. 1 which presents the E?' luminescence spectrum for a Ge/P-silica fiber. Gain has been demonstrated over practically the entire emission spectrum, with values over 20 dB readily obtainable between 1530 and 1560 nm. A critical factor in the success of E?+- doped fiber amplifiers in optical communications is the long lifetime of the metastable state which permits the required high- population inversions to be obtained under steady-state conditions using modest pump powers. An optically pumped device must have suitable pump bands to be practical. Fig. 2 illustrates the absorption spectrum for an Ef'+-doped silicate glass. All five of the absorption bands between 500 and 1500 nm have

Manuscript received October 15, 1990; revised November 1, 1990. The author is with GTE Laboratories Incorporated, Waltham, MA 02254. IEEE Log Number 9041474.

WAVELENGTH (nm)

Fig. I . Emission spectrum of Er'+-doped Ge/P-silica fiber.

0.5 - - U)

C 3

2 0.4

-

L e s Y 0

.-

I- 0.3 z

U U

0

z

g 0.2

k 0.1

2

0 II:

m a n

111 i n I I I I I I I I I I

50

300 500 700 900 1100 1300 1500

WAVELENGTH (nm)

Fig. 2. Absorption spectrum of Er3+-doped L22 silicate glass. The peaks of the strong bands at 380 and 520 nm are 1.5 and 1 .O, respectively.

been successfully used to pump amplifiers, although those which match the output of diode lasers are to be preferred. The absorption bands of Er3+ are sparse throughout the visible and near infrared, in addition to being relatively weak except at wavelengths I 540 nm. Although this has historically limited the utility of flashlamp-pumped bulk Ef" lasers [8], 191, the high-pump intensity realized by coupling a laser into single- mode fibers provides good inversion and high gain at all the pump bands.

0733-872419110200-0234$01 .OO 0 1991 IEEE

Authorized licensed use limited to: Hitesh Mehta. Downloaded on March 08,2010 at 02:23:23 EST from IEEE Xplore. Restrictions apply.

MINISCALCO: ERBIUM-DOPED GLASSES FOR FIBER AMPLIFlERS AT 1500 nni

24-

20-

16- - 'E E,12- > 0 5 w

235

4

5 1 2 450 nm 630 nm

'b I2 490 nm 715 nm

4 1 1 2 520 nm 790 nm - 4s312 550 nm 850 nm

512

2

650 nm 1140 nm 4

5 1 2 -

41912 800 nm 1680 nm

- 41r112 980 nm 2700 nm

8-

4 ~ , 3 1 2 - 1530 nm -

The performance of solid-state lasers and amplifiers is governed by the relevant electronic and optical characteristics of the active ion, such as cross sections, spectral shapes of the emission and absorption bands, excited state lifetimes, ion-ion interactions, as well as the static and dynamic ion-lattice interactions. The host material has an important influence on all these properties, but only for Nd3+-doped glasses has this role been extensively investigated [ 101. Much less information is available for E?+-doped multicomponent glasses, and practically all the work reported on E?+-doped fiber lasers and amplifiers has used silica glass codoped with various combinations of Ge, P, and AI. From the extensive body of research on Nd3+-doped glasses it is known that pure silica is an anomalous host for rare- earth ions, being able to dissolve only low concentrations and, in the case of Nd3+, providing low transition strengths [ l l ] . Other glasses have properties leading to superior performance in specific applications. For example, phosphates and fluorophosphates have higher cross sections while germanates, tellurites, and fluorides have lower nonradiative relaxation rates.

The objective of the present work is to identify the material- dependent properties that influence the performance of optical amplifiers and review the information available on E?+-doped glasses, including silica, that is relevant to assessing the relative merits of these glasses in this application. The section that follows addresses the issues affecting the efficiency of amplifiers, principally nonradiative relaxation and concentration quenching. Section I11 discusses the cross sections and spectra at the signal wavelength and how these are affected by the host glass. Since the performance and utility of an amplifier is strongly influenced by the pump band used, Section IV treats the composition-dependent properties of the 800-, 980-, and 1480-nm pump bands. Rather than providing an exhaustive survey, it is our intent to convey an understanding of the trends in properties among the various glass compositions as well as their general suitability as optical amplifiers. The emphasis is placed on silica and the most promising alternative glasses for this application: fluorophosphates and fluorozirconates.

11. AMPLIFIER EFFICIENCY Although much attention has been focused on the effects of

waveguide design and pump wavelength on amplifier efficiency, the host glass and doping concentration also play a significant role. Ion-dynamic lattice and ion-ion interactions fall into this category and are generally deleterious, converting pump photons into heat. The mechanisms which have been identified are thermal quenching due to multiphonon emission by individual ions, excited state absorption, and two cooperative processes involving interacting ions: cross relaxation and upconversion. In addition, the issue of clustering in silica is considered since it is sensitive to composition and exacerbates the cooperative processes.

A . Dissipative Processes

Fig. 3 is an energy diagram for Er3+ indicating the levels up to 24 000 cm-'. The 41,3/2 metastable state is the initial level for the transition producing gain at 1500 nm and any process removing E?' ions from this state, other than stimulated emission into the signal mode, necessarily decreases the efficiency of an amplifier. One possible mechanism is nonradiative relaxation directly to the 41,5/2 ground state as a result of the interaction between the electrons and the dynamic lattice. This oc-

" L 44s12-

Fig. 3 . Energy levels of Er'+ labeled with the dominant Russell-Saunders 'L, term. For each state the GSA column lists the wavelength of the ground- state absorption transition terminating on it, while the ESA column con- tains the wavelength of the excited-state absorption transition to that level originating on the 4/,3,2 metastable state.

curs through multiphonon emission and is inversely proportional to the exponential of the energy gap separating the two levels [ 121-[ 141. The higher the energy of the phonons coupling to the rare-earth ion, the fewer required to bridge the gap and the faster the nonradiative rate. The general rule of thumb is that the lifetime of the excited state will be temperature dependent if the gap is less than ten times the effective phonon fre: quency and completely quenched if less than four times [ 151. The 6500 cm-' gap between the 41,3/2 and 4115/2 states is large enough that nonradiative relaxation is only significant at room temperature for borates [8], [ 161. Phosphate glasses also have effective phonon frequencies larger than silicates and there have been reports of slightly reduced quantum efficiencies in some compositions [9]. In contrast, germanate [ 171, [ 181, tellurite [18], [19], and fluoride glasses [15] have low-effective phonon frequencies which lead to additional metastable states. This leaves open for consideration most of the practical glass compositions with the caveat that phosphates should also be examined for radiative efficiency.

Another loss mechanism is excited-state absorption (ESA), the excitation of an ion in the metastable 41,3/2 state to a still higher level through the absorption of a pump or signal photon. Even if the ion ultimately returns to the 41,3/2 level, a photon will be lost either to heat or spontaneous emission at an un- wanted wavelength. Excited ions have an ESA spectrum which differs from their GSA spectrum due to the new energy gaps to the higher lying states. This is illustrated in Fig. 3 where each state above the 41,3/2 is labeled with the wavelength of the ESA transition terminating on it. All have been identified in either absorption or emission. Fig. 4 compares the GSA absorption coefficient of G e / P silica to the change in absorption coefficient induced by exciting most of the ions into the ' I I3 l2 state. A negative change corresponds to bleaching and where it mirrors the GSA band (e.g., 980 nm) there is no ESA. A positive band indicates a region in which ESA dominates since the increase in absorption due to populating the metastable state more than offsets the decrease from depleting the ground state. Fig. 4 indicates that ESA transitions from the 4113/2 state of Er3+ coin-


236

v) .- C

3

c

IOURNAL OF LIGHTWAVE TECHNOLOGY. VOL. 9. NO 2. FEBRUARY 1991

~ l l l l l l l

--... Ground State Absorption Coefficient

- Pump-Induced Change in Absorption Coefficient

cidentally overlap many GSA bands. The 41,3/2 -+ 2Hl , /2 ESA transition at 790 nm has particularly important implications for the 4 1 , 5 / 2 + 419/2 pump band at 800 nm and will be examined in Section IV.

The remaining dissipative processes to be considered involve ion-ion interactions leading to concentration quenching and cooperative upconversion. The greatest amount of information is available for Nd3+-doped glasses, and this may be drawn upon because of the chemical similarity between rare-earth ions. However, its applicability is limited since the dominant relaxation mechanism differs for Nd3+ and E?’ [20]. Concentration quenching for Nd3’ is believed to occur through the cross relaxation mechanism [21] illustrated in Fig. 5. If an ion excited into the metastable 4F3/2 level interacts with a nearby ion in the ground state (Fig. 5(a)), the first ion can transfer part of its energy to the second, leaving both in the intermediate 4115/2 state (Fig. 5(b)). Since the energy gaps to the lower lying states are small, both ions quickly decay nonradiatively to the ground state. The net result is the conversion of the original excitation into heat in a time short compared to the radiative lifetime. A luminescence decay measurement will sample a large number of ions, each in a different environment with a different cross relaxation rate. Accordingly, this quenching process will man- ifest itself as nonexponential decay which is independent of pump power since only one excited ion is required for cross relaxation. This behavior has been observed for both Nd3+- doped bulk glasses [ 101 and fibers [20].

The major cause of inefficiency for E?+-doped glasses is believed to be the cooperative upconversion process [20] illustrated in Fig. 6. Auzel [22] and Wright 1231 have comprehensively reviewed cooperative upconversion, although only recently has this phenomenon been extensively investigated in glass hosts. Unlike Nd3+, E?’ has no states between the metastable and ground state, thus cross relaxation between an excited ion and one in the ground state cannot occur. How- ever, if two excited ions interact (Fig. 6(a)), one can transfer its energy to the other, leaving itself in the ground state and the other in the higher 419/2 state (Fig. 6(b)). In oxide glasses the

‘F3 ,2

4 1 i 5 ! 2

4 1 i 3 ~ z

411112

(912

(b)

Fig. 5 . The cross relaxation process for Nd’+. In (a), one of a pair of interacting ions is excited to the ‘F , l2 metastable state. In (b), it transfers part of its energy to the second ion, exciting it to the state while decaying to the same level. Both ions then undergo nonradiative relaxation to the ground state.

Fig. 6 . The upconversion process for Er”. In (a), both interacting ions are excited to the metastable ‘ f , , I 2 level. In (b), the donor ion transfers all its energy to the acceptor, leaving itself in the ground state and the acceptor in the ‘ I 9 / > state. For oxide glasses the acceptor ion quickly decays nonradiatively back to the ‘ f , , I 2 level.

419/2 level quickly relaxes through multiphonon emission back to the 4113/2 state, the net result of the process being to convert one excitation into heat. For fluoride glasses radiative relaxation of the ‘1912 or 41, , l2 levels to the ground state may dominate, in which case both excitations will be lost. Since cooperative upconversion requires two interacting ions in the excited state it will not be evident at low pumping levels, but at high pump powers it will appear as accelerated and nonexponential


MINISCALCO: ERBIUM-DOPED GLASSES FOR FIBER AMPLIFIERS AT 1500 nm 231

decay, the latter due to variation in the coupling strengths between excited ions. This has been observed [20], [24] and further evidence has been the report of upconverted 411 l /* emission at 980 nm when pumping at 1480 nm [25]. Even more con- vincingly, when pumped at 1500 nm the decay kinetics of the upconverted luminescence at 800 nm from the ' 1 9 , ~ state was found to be proportional to the square of the 1500-nm emission as is expected for a process involving two ions in the 41,3/2 state [20], [24]. The pump-power dependence of the upconversion mechanism has important device consequences since the quenching process will be most deleterious at the high population inversions required for amplifiers to achieve high single- pass gain and good signal-to-noise ratio. This is reflected in the amplifier experiments of Kagi et al . which show a large difference between the gains of high and low concentration fibers at high pump powers although the gain thresholds for the two fibers are close [25].

Even when the interaction mechanisms are known it is generally not possible to perform ab initio calculations, and concentrations at which ion-ion interactions become important are determined empirically. Since little information is available for E?+-doped glasses, we may obtain some guidance for multicomponent glasses by considering Nd3+. Table I is drawn from the data of Stokowski et al . [ lo] and illustrates for the major glass types the range of ion concentrations which reduces the lifetime of the 4F3/2 state by a factor of 2. The 4F3/2 level is the metastable state for Nd3+ lasers and amplifiers. The empirical formula relating the observed lifetime robs to the ion concentration p is [ lo]

where ro is the lifetime in the limit of zero concentration and Q is the quenching concentration given in Table I . From (1) one expects less than a 1 % reduction in efficiency if p is kept to less than 10% of the values in the table. Of the glasses for which data are available, phosphates are the most resistant to quenching while fluorophosphates are the least; the survey of the latter includes mostly high-fluorine-content compositions. Overall, no multicomponent glasses should exhibit efficiency problems at the Nd3+ concentrations of 5 lOI9 cm-3 normally used in fiber devices. Qualitatively similar behavior is expected for E?+ and this has been confirmed by power amplifier experiments in which differential conversion efficiencies approaching the quantum limit have been achieved using Al- and Al/Ge-silica fibers with concentrations in the 1018-1019 range [26]-[28]. As is discussed below, Al- and Al/Ge-silica behave in many respects like multicomponent silicate glasses.

The OH- complex is extremely effective at quenching excited rare-earth ions [8], and an alternative mechanism for efficiency reduction involves fast energy transfer between interacting ions until the excitation reaches one near an OH- impurity. If the OH- concentration is low, the energy transfer process and hence the rare-earth concentration will control the quenching rate. Since this process should apply to both Nd3+ and E?+, the differences in behavior reported by Wyatt [20] appear to exclude this possibility.

B. Solubility of Rare Earths in Silica

The above concentration limits for maintaining the efficiency of an amplifier or laser apply only to materials in which the rare- earth dopant is uniformly dispersed, as is the case for multicom-

TABLE I Nd" CONCENTRATIONS FOR FACTOR OF Two

REDUCTION IN 4F3,z LIFETIME

Quenchin Concentration Glass Type (180 cm-3)

Silicate Phosphate Fluorophosphate Fluorozirconate Fluoroberyllate

3.9-6.0 3.9-8.6

4.2 3.0-4.0

3.8-5.3

Source: Stokowski er al. [lo]

ponent glasses. Pure silica, however, is an exception to these guidelines because it can incorporate only very small amounts of rare earths before microscopic clustering appears and quenches the active ions. At still higher concentrations crystalline phases may appear [29]. Rare-earth ions require large coordination numbers, and Arai et al . have explained this phenomenon with a structural model in which the absence of a sufficient number of nonbridging oxygens to coordinate isolated rare earths in the rigid silica network causes them to cluster in order to share nonbridging oxygens 1211. The Nd3+ concentration for the onset of clustering in pure silica was estimated to be = l O I 9 cm-3 ( = 1000 ppm-wt) from luminescence decay data [30], while fiber amplifier experiments have revealed significant E?' ion-ion interactions at levels = 10" cm13 [31]. Power amplifier [27], [28] and small-signal amplifier [25] investigations have indicated that quenching also appears at = 10" cm-3 for Ge-doped silica. This is consistent with the expectation that four-coordinated Ge does not alter the tetrahedral silica network sufficiently to significantly increase the solubility of rare-earth ions. Further evidence has been provided by Ainslie et al . who reported the emission spectrum of Nd3+-doped pure silica and Ge silica to be identical [32]. Since Nd3+ and Er3+ are expected to be equally insoluble in silica, the order of magnitude difference in the reported onset of quenching is likely due to either a stronger interaction between E?' ions or a greater sensitivity to nonradiative processes. in the amplifier investigations.

The addition of A1 has been found to eliminate the clustering problem and Nd3+-doped Al-silica has been reported to follow the same concentration quenching curve as the multicomponent ED2 silicate glass 1211. For ED2, Q = 3.9 X loZo cm-3 and no significant quenching is expected for Nd3+ concentrations of = 1000 ppm-wt in A1 silica. Extensive measurements are not available for E?+-doped silica codoped with Al, but power amplifier studies found a level of = 800 ppm-wt provided approx- imately the same efficiency as very low concentration Ge silica [26]. By comparison, a similar level in Ge silica reduced the efficiency by nearly an order of magnitude from the low concentration limit [26]-[28]. Arai et al. have explained the role of A1 in terms of solution chemistry: Nd203 dissolves in A1203 but not in S O 2 , while A1203 dissolves in S i02 [21]. Thus A1203 forms a solvation shell around the rare-earth ion and the result- ant complex is readily incorporated into the silica network [2 11. The correlation between the positions of the rare earth and A1 ions implied by this model is consistent with the small Al-to- Nd ratio ( = 10) required to prevent clustering and produce the multicomponent silicate glass emission spectrum [2 11. In studies of E?' amplifiers where silica was codoped with Ge and Al, a Ge-to-A1 ratio of 10 was found sufficient to eliminate


238 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. Y, NO. 2. FEBRUARY 1991

quenching effects [27], [28] and provide an emission spectrum similar to that of AI silica [27]. Using typical fiber parameters, this translates into an AI-to-Er ratio of = 50. Thus the properties of the rare-earth dopant and the waveguide can be indepen- dently controlled using AI doping for the former and Ge doping for the latter, thereby avoiding the crystallization and high background-loss problems often encountered when using AI doping alone to produce high NA ( -0 .2 ) fibers. Although AI silica behaves like a multicomponent glass when doped with Nd3+, this is not true in every respect for E?'. As is discussed in Section 111, the emission spectrum of E?+-doped AI silica is not typical of a silicate glass.

Aluminum codoping has an additional benefit in controlling the radial doping profile of rare earths. Confining the E?' to the center of the core improves the pump efficiency of amplifiers [33], but the MCVD preform fabrication process depletes rare- earth dopants at the center of the core for Ge and Ge /P silica [34]. This appears to be related to the "bum-out" of Ge and P in MCVD preforms since the rare-earth concentration profiles have the same shape [34], but the mechanism by which it occurs is not presently known. The addition of AI, even when codoped with Ge or P, prevents the depletion of the rare earth [34] and is a straightforward way to improve the efficiency of amplifiers made from MCVD fibers.

The effects of P doping are not so clear. For a P-to-Nd ratio of = 15, Arai et al . reported spectroscopic changes and solubility improvements similar to those found for AI doping [21]. In contrast, investigations of Nd3'-doped Ge/P-silica fibers found that to prevent clustering the P-to-Nd ratio needed to be an order of magnitude larger 1321, [34]. Doping P at < 1 mol % , levels which are sufficient to derive full benefit from AI doping, offers no improvement over Ge silica: rare-earth concentrations < 10" cm-3 are still required to prevent quenching of both Nd3+ [29] and Er3+ [27]. At higher rare-earth concentrations P doping is reported to improve amplifier efficiency somewhat [27]. For very high rare-earth levels ( 1 loi9 ~ m - ~ ) a P- and rare- earth-rich crystalline precipitate appears [29]. An additional drawback to P doping is that it does not prevent the depletion of rare-earth ions at the center of the core for MCVD preforms [34]. Adding P has been shown to increase the amount of A1 that can be incorporated before phase separation occurs [35], permitting higher NA fibers to be made without the use of Ge.

The only other codopant characterized in any depth has been F. Imai et al . found it has no effect upon the optical spectra of Ce-doped silica, in contrast to AI and P codoping [36]. This is because F bonds preferentially to Si and does not affect the li- gands in the vicinity of the rare-earth ion [36]. They did not determine if F can play a role in improving solubility [36], but their structural model for the incorporation of F into the silica network implies it will not have a beneficial effect. Although only Ce was investigated, these conclusions should apply equally well to other rare-earth dopants. Based on the available evidence it is apparent that AI is the most beneficial codopant for rare-earth ions in silica fiber.

111. CROSS SECTIONS AND SPECTRA AT 1500 NM

Glass composition has an important effect on the absorption and stimulated emission cross sections of rare-earth ions as well as the shapes of these bands. Although this has been extensively documented for Nd3' [lo], less information is available for E?'. Most of the published data for the latter include only os-

and less on optical bandwidths. Alekseev et al. present a com- prehensive discussion of bulk E?+-doped glass lasers but include detailed information for only a few glass compositions. Sandoe et al . provide data on compositional trends [37], and silica fiber has been investigated by several authors [38]-[43]. Although multicomponent silicate glasses have been used for Nd3' fiber lasers and amplifiers for some time [l], an E?'- doped fiber amplifier made from such a glass has only just been reported [44]. In recent years phosphate and fluorophosphate glasses have been identified as excellent bulk hosts for Nd3+ [ 101, and lasers have been demonstrated for phosphate glass single-mode fibers doped with Nd3+ [45], [46] and E?' [45]. Because of their lower nonradiative decay rates, extended infrared transmission, and reduced excited state absorption for Nd3+ at 1300 nm, fluorozirconates have assumed an important role in fiber applications [47]. Nevertheless, spectra and cross sections have only recently been determined for Er3+-doped fluorophosphate and fluorozirconate glasses [40], [41], [43]. The sections that follow consider the principles underlying the shape and strength of rare-earth optical transitions, and discuss the observed magnitudes and bandwidths of cross sections at 1500 nm for Er'+-doped glasses.

A. Electronic Structure and Transition Rates

The electronic structure of rare-earth ions is dominated by electron-electron interaction, with spin-orbit next in importance. These produce the free-ion J multiplets illustrated in Fig. 3 which are usually labeled by their dominant Russell-Saunders SLJ term. The host has the least influence on the electronic structure and changes the relative positions of these levels only slightly. The effect of the host upon the rare-earth dopant is customarily treated by replacing the host by an effective crystal field potential at the ion site. The potential is usually expanded in a power series of tensor operator components b_kk that trans- form as spherical harmonics

where Bi are the crystal field components (k 5 6 for f electrons) and the summation i is over the f electrons of the ion. The B", are often found by replacing the other ions in the glass, by their point-charge equivalents and considering only near neighbors. The even-k terms in the expansion split the free-ion J multiplets into Stark components generally separated by 10- 100 cm- ' . The rare-earth site symmetry in glasses is usually low enough to lift all but the Kramers degeneracy which persists for ions like Er3+ with an odd number of electrons. The odd-k terms admix higher lying states of opposite parity (e .g . , 5d ) into the (4f )" configuration, thereby introducing a degree of electric-dipole strength into the otherwise parity forbidden in- tra-f-shell transitions. In addition to this host-dependent forced electric dipole contribution, the 41,3 /2 * 41,5/2 transition of E?' at 1500 nm has considerable magnetic dipole strength which is relatively insensitive to host material. Individual dopant ions occupy very dissimilar sites in a glass and experience different crystal field parameters B",. This disorder in the even-k terms leads to a distribution of energies for a given Stark component and produces inhomogeneous broadening of the emission and absorption spectra. Although seldom discussed, the corresponding disorder in the odd-k terms will produce variations in transition strengths between a given pair of components. The

cillator strengths; little has been reported on peak cross sections transitions are also homogeneotky broadened through dephas-


MINISCALCO: ERBIUM-DOPED GLASSES FOR FIBER AMPLIFIERS AT 1500 nm

~

239

ing processes. This is dominated by the electron-phonon interaction and is therefore strongly temperature dependent.

For E?+-doped glasses most of the expected seven Stark levels of the metastable 41i3 /2 state and eight components of the 'II5l2 ground state are resolved at low temperature. At room temperature even the highest components have some thermal occupation and the emission and absorption are composites of 56 largely unresolved Stark transitions. The magnitude and shape of the absorption and emission cross-section spectra will be determined by the crystal field coefficients B", together with the homogeneous and inhomogeneous broadening. The relative significance of the two broadening mechanisms has important implications for the gain saturation properties of an amplifier and is a continuing subject of investigation for silica. The situation is complicated by several of the important parameters being of comparable size. For Al-and Al/Ge-silica the lower- lying Stark components of the two manifolds are split by = 20- 50 cm-' [48], [49], the inhomogeneous broadening is =40-50 cm- [48], [49], while the room temperature homogeneous width is = 50 cm-l [50]. Accordingly, mixed homogeneous and inhomogeneous behavior is expected with the former dominat- ing. A similar situation occurs for Nd3+-doped glasses. For the latter, Hall and coworkers modeled energy extraction from bulk power amplifiers and found a relatively minor departure from homogeneous behavior [51], [52]. The saturation of Er3+ fiber amplifiers has been modeled for AI silica by Desurvire er al. who also report essentially homogeneous behavior [53]. Re- ports of inhomogeneous saturation for Ge silica [54] have been explained in terms of a smaller homogeneous width for this glass compared to A1 silica [55].

Attempts to calculate spectra from first principles are seldom undertaken because of their complexity and the lack of information regarding the environment of the rare earth. More often, low temperature measurements of the Stark splittings are used to deduce the nature of the ion site. This approach was applied to E?+-doped glasses by Robinson using absorption spectra [56], [57]. In a tour de force, Brecher and Riseberg used site- selection emission spectroscopy to determine not only the site symmetry but also the distribution of sites for Eu3+-doped silicate [58] and fluoroberyllate [59] glasses. Weber and Brawer have pointed out, however, that although optical spectroscopy can indicate possible rare-earth sites, it cannot be used to prove which of the possibilities are actually present [60]. They sought to overcome these shortcomings using molecular dynamics simulations of the glass structure to provide the rare-earth environment to be used as inputs to (2). For rare-earth-doped fluoroberyllate glasses they successfully reproduced the Stark splittings and their dependence on composition [60]. Neverthe- less, molecular dynamics simulations require both the assumption of ionic species and the use of simple interaction potentials; these can lead to errors when applied to covalent glasses such as silicates and phosphates. A useful semiempirical technique for the analysis of rare-earth transitions was developed inde- pendently by Judd [61] and Ofelt [62]. It relies upon the ap- proximation, usually well satisfied, that the energy range occupied by opposite parity states admixed by the odd-k terms of (2) is small compared to its separation from the (4 f ) " multiplets. The spectrally integrated strength of the transition between a given pair of J multiplets is then determined by three empirical parameters and the appropriate reduced matrix elements. The latter are independent of material and have been tabulated [63]. All the host dependence is contained in the three intensity parameters which are found for a given dopant/mate-

i r, -..- Ai/P-Silica

..... Silicate L22

I\

1500 1550

WAVELENGTH (nm)

1600

Fig. 7 . 1500-nm absorption cross-section spectra for different glass types illustrating the variation in magnitude and shape.

rial combination from a least-squares fit to the absorption spectrum. Judd-Ofelt analysis is accurate to about 10-15% and is particularly valuable for providing linestrengths of transitions for which direct measurements are difficult. It applies only to the integrated transition strengths, however, and provides no spectral information.

B. Absorption and Stimulated Emission Cross Sections

For a three-level gain system such as E?+ at 1500 nm, not only the stimulated emission but also the absorption cross section play an important role in determining the performance of a device. Fig. 7 illustrates the strong effect of the host upon the magnitude and shape of the absorption cross section. Table I1 provides compositions for the glasses which will be discussed in greatest detail. The stimulated emission cross sections in Fig. 8 exhibit no less variation. As discussed in the previous section, the differences in spectra between glasses are caused by changes both in the crystal field splittings and in the intensities of the Stark transitions. Table Il l lists integrated absorption cross sections, K, = { ( T , ~ ( U ) du, for compositions representing several important glass types. KO is used instead of the more conven- tional oscillator strength so that the extensive survey by Sandoe et al . [37] can be included in the table. It is seen that the host can affect the integrated absorption strength by slightly more than a factor of 2. Silicates have the lowest values while phosphates, fluorophosphates, and borates can achieve significantly larger values. Germanate glasses also have weak transition strengths [17], [18] as might be expected from their structural similarity to silicates. Within the silicates, transition strength decreases with increasing atomic number of the alkaline earth, S6-S8 containing Mg, Ca, and Sr, respectively [37]. For simple phosphates the opposite holds true with K, increasing from Mg (S l ) through Ba (S4) [37]. A high silica content tends to reduce K,: L22 differs from S7 primarily in having 7 1 versus 57 mol % silica, respectively. In contrast, increasing the B203 content from 74 to 88 mol% causes the increase in absorption strength from borate S23 to S 18 [37]. Values for the fluoride glasses fall midway in the large range spanned by the phosphates. Among fluorophosphates, K , is sensitive to the relative amounts of oxygen and fluorine. It varies from that of a good phosphate to a value nearer the fluoride glasses as the oxygen-to-fluorine ratio


240 JOURNAL OF LIGHTWAVE TECHNOLOGY. VOL. 9. NO. 2. FEBRUARY 1991

TABLE I1 GLASS COMPOSITIONS

Glass Compositions

TABLEJIII INTEGRATED 4i,5/2 + ii3,? ABSORPTION

FOR Er3+-DOPED GLASSES

Component L22 ED2 L28 L11 L14 F88 Glass

p205 Si02 ZrF, A1203 AIF, MgF2 CaO CaF, BaO BaF2 Li20 LiF Na20 NaF K2O InF,

Er@, LaF,

ErF,

50.1 20.8 70.7 59.2

1.5 2.6 3.5 6.9

6.2

54.0 2.1

20.2 3.5 19.1

12.0 10.2 19.1

15.5 27.7

1.5 28.1 8.7 35.9

10.5

19.1 20.0

12.4 15.5 3.9 31.0

2.5 1.9 4.4

0.005 0.051 0.025 0.022 0.048 0.100

- AIIP-Silica ' t

WAVELENGTH (nm)

Fig. 8. 1500-nm stimulated emission cross-section spectra shown for different glass types.

goes from 1.5 for L11 to 0.15 for L14. As mentioned in Section 11, the A l / P silica is unusual and stands out for the value of its transition strength which is large even for a multicomponent silicate and more typical of a fluorophosphate.

For four-level and highly inverted three-level lasers and amplifiers, the maximum gain occurs at the peak of the stimulated emission cross-section spectrum. The peak cross section is determined by the spectral shape as well as the integrated cross section K . In Table IV are peak values of the stimulated emission cross section uZ1 and absorption cross section uI2 for rep- resentative glasses as well as the wavelengths at which they occur. The absorption and emission oscillator strengths, f i 2 and

f Z 1 , respectively, are also included. These are common mea- sures of the integrated strength of a transition and are proportional to K . The absolute values of the absorption cross sections for the bulk glass were obtained using standard spectrophoto- metric techniques on samples of known. Er3+ concentration. Determining the absorption cross sections for the A1 /P-silica

Silicate L22 Silicate S8 Silicate S7 Silicate S6 AI/P silica Borate S23 Borate S 18 Phosphate SI Phosphate S4 Phosphate L28 Fluorophosphate L14 Fluorophosphate L11 Fluorozirconate F88 Ba-Zn-Lu-Th fluoride

2.6 3.0 3.7 4.0 5.1 4.7 5.9 3.2 4.4 5.5 5.1 5.5 4.6 4.6

Sources: Sample numbers beginning with S are from Sandoe er ul. [37] and are the same numbers used therein with an S appended; all others this work.

single-mode fiber required a more complicated procedure using high resolution attenuation measurements performed using the cut-back method, knowledge of the Er3+ concentration and radial distribution obtained by secondary-ion mass spectrometry (SJMS), and the profile of the optical mode which was found from far-field measurements [64]. Stimulated emission cross sections were found from measurements of the 411,/z radiative lifetime T and the 41,3 /2 + 411s/z emission spectrum corrected for the wavelength dependence of instrumental sensitivity. These quantities are then related by [65]

1 - = S(u) du = 7 V ~ U ~ ~ ( U ) du

where U is the photon frequency, n is the index of refraction for the host material, and c is the speed of light. The luminescence spectrum is proportional to S( U ) which is defined as the spontaneous emission rate (photons /second) per hertz integrated over all solid angles and summed over both polarizations. The measured lifetimes are included in Table IV.

The peak of absorption and emission generally lie within 1 nm of each other for a given glass, indicating that the same pair of Stark components dominates both emission and absorption. Low temperature measurements have revealed these to be the lowest level of each manifold [9], [48], [49]. With the exception of the silicate and G e / P silica (not shown), the peak emission wavelengths for all compositions fall in a narrow range around 1531.5 nm. The longer emission wavelengths observed for silicates, including pure and Ge-codoped silica, is usually attributed to their greater covalency and is also observed for Nd3' transitions [ll]. Again AI/P silica is unusual and has characteristics more in keeping with a fluorophosphate or fluoride glass. The oscillator strengths of the L22 silicate in Table IV are unusually low but the very narrow bandwidths lead to peak cross sections comparable to the other glasses. For comparison, a lifetime of 8.8 ms and peak uZ1 of 7.3 X cm2 have been reported for an E?+-doped phosphate glass [45]. Alekseev et al. provide peak cross sections for a number of phosphate glasses, all of which fall into the range 6-8 X lo-'' cm2 [9]. This confirms the expectation that phosphate and high-


MINISCALCO: ERBIUM-DOPED GLASSES FOR FIBER AMPLIFIERS AT 1500 nm 24 I

TABLE IV ‘Il3l2 +, 41,siz TRANSITION STRENGTHS FOR E ~ + - D o P E D GLASSES

Emission Absorption

Lifetime fil Peak A Peak u21 flz Peak X Peak uI2 Glass ( m s ) (IO-‘) ( n m ) ( I O - * ’ cm’) ( n m ) (IO-’’ cm’)

AI/P silica 10.8 1.16 1530.8 5.70 1.48 1530.1 6.60 Silicate L22 14.5 0.821 1535.8 7.27 0.737 1536.4 5.80 Fluorophosphate

L11 8.25 1.38 1532.6 7.16 1.53 1532.6 6.99 Fluorophosphate

L14 9.5 1.33 1532.0 5.79 1.48 1531.4 5.76 Fluorozirconate

F88 9.4 1.26 1530.6 4.95 1.29 1530.4 4.98 1530.6 - 1.27 1529.4 4.82 Ba-Zn-Lu-Th fluoride - -

Sources: Lifetimes for Al /P silica this work, all other lifetimes R . S . Quimby, private communication; absorption uIZ for AI/P silica from Singh er al. 1641; all other data this work.

phosphate-content fluorophosphate (e.g., L1 1) glasses provide the highest cross sections. Ignoring L22, the emission and absorption oscillator strengths vary by only & 10% while the peak cross sections fluctuate by + 2 0 % . This additional factor of two arises from differences in bandwidth, an issue that will be discussed in the next section. The oscillator strengths in Table IV include a magnetic dipole contribution of 0.49 X so that the relative variations of the material-dependent electric dipole contribution are actually much larger than the changes in the total oscillator strength.

The success of the E?+-doped silica fiber amplifier is due in large measure to the long lifetime for the metastable state. This permits the large population inversions needed for high gain and low noise to be achieved using CW pumping. It also leads to the absence of distortion and cross talk for modulation frequencies above = 100 kHz in saturated amplifiers [54], [66], [67]. Table IV shows that other glass compositions offer similarly long radiative lifetimes, the shortest being ~ 8 0 % of that for Al /P silica. Thus excited state lifetime will not be a factor in selecting appropriate hosts.

In the analysis of rare-earth transitions it is common to measure only one of the cross-section spectra, usually absorption, and calculate the other. Since procedures for obtaining absorption cross sections in fibers can be complex, it is more conve- nient to measure the stimulated emission cross section for fiber samples [38]-[40]. The integrated cross sections are connected by the relation between the Einstein A and B coefficients gen- eralized to account for the finite linewidth

g1 s u 2 u 1 2 ( u ) du = gz s U2U21(.) du ( 3 )

where gi = 2 J + 1 are the degeneracies of the J multiplets involved. This expression is commonly seen in a more approximate form, sometimes referred to as the Ladenburg-Fucht- bauer relationship [37], with U’ replaced by the square of an effective frequency and taken outside the integral. For the ‘Il5/* ground state g, = 16, and for the ‘II3l2 metastable state g, = 14. Table V compares measured emission oscillator strengths to those calculated from (3) with the discrepancy evaluated relative to the calculated value. In all cases the measured oscillator strengths are smaller, usually by 20% or more. This is consistent with reports of significant differences between measured cross sections and those obtained using the Einstein relation for this transition of E?+ [42], [68], [69]. The problem stems from

TABLE V COMPARISON OF CALCULATED A N D MEASURED OSCILLATOR

STRENGTHS FOR THE 411312 + 41,s12 TRANSITION

Calculated Measured Discrepancy f i l f i l

Glass (10 -6 ) ( %

Al/P silica 1.75 1.16 - 34 Silicate L22 0.873 0.821 -6 Fluorophosphate L11 1.81 1.38 - 24 Fluorophosphate L14 1.75 1.33 - 24 Fluorozirconate F88 1.52 1.26 - I7

the validity of (3) depending upon at least one of two conditions being met. Either all Stark components of the two multiplets must be equally populated, or all the transitions must have the same strength regardless of the components involved. Either of these conditions insures that the observed oscillator strength is identical to that of the degenerate (unsplit) two-level case to which the Einstein relation applies. Since the manifold widths for the 4115/2 and 4113/2 states of E?+-doped glasses are typi- cally 300-400 cm-I [9], [48], [49], the first condition is not satisfied at room temperature ( k T = 200 cm-’ ). Moreover, low temperature absorption and emission measurements indicate that for all glasses the transition strength is quite sensitive to the Stark levels involved [9], [48], [49]. Since neither of the above requirements is satisfied, it is not surprising to find that (3) leads to significant errors.

Recently an alternative analysis technique has been presented [70], [71] using the more general theory of McCumber [65] which was developed for the study of phonon-terminated lasers based upon transition metal ions. The only assumption required by McCumber’s theory is that the time required to establish a thermal distribution within each manifold be short compared to the lifetime of that manifold. From consideration of detailed balance, the absorption and emission cross sections are then related by [65]

u2, (u ) = u12(~)e‘‘-hu’/kT (4 )

where h is Planck’s constant. The cross-section spectra are scaled with respect to each other by E which is the net free energy required to excite one Er3+ ion from the 4115/2 to the 4113,2 state at temperature T . In addition to requiring less restrictive


242 IOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 9. NO. 2 . FEBRUARY 1991

assumptions than the Einstein relation, (4) has the added advantage of providing spectral information and may be used to generate a relative cross-section spectrum if the spectrum of the inverse process is known. The offset between emission and absorption is readily seen by noting that at only one frequency U = ~ / h are the two spectra equal. At higher frequencies the absorption cross section is larger while at lower the emission cross section is larger. The principal disadvantage of the McCumber analysis is that E can only be calculated if the positions of all the Stark levels are known for both manifolds. This has been done for A l / P silica using the electronic structure reported for similar glasses [9], [49] and the calculated peak emission cross section was found to be within 3% of the value in Table IV [70], [71]. To eliminate the need for detailed measurements of the electronic structure, an approximate procedure has been developed for determining t and usually produces more accurate cross sections than an Einstein analysis [70], [71].

C. Bandwidth

Optical amplifiers are generally more useful if they provide gain which is relatively independent of signal wavelength. This relaxes the wavelength tolerances on transmitters in a single channel system and increases the number of optical channels which can be multiplexed without compromising performance. The spectra of the 4115/2 -+ ‘II3l2 absorbing and 4 1 1 3 / 2 + 411s,2 emitting transitions are important considerations in determining the tunability of a laser and the gain bandwidth of an amplifier. Since the spectrum of the gain coefficient depends upon the lo- cal inversion which is in tum a function of position within the fiber, there is no procedure for quantitatively predicting the bandwidth short of using a fiber amplifier model. If the population inversion is complete over the entire length of the fiber, the problem is greatly simplified and the logarithm of the gain spectrum will be proportional to the linear spectrum of the stimulated emission cross section. This condition is automatically satisfied in four-level systems but requires very high pump levels for three-level devices. For the 1500-nm transition of Er3+, incomplete inversion leads to reabsorption of the signal by ions in the ground state. Since the absorption spectrum is offset from the emission spectrum, the reabsorption process will not only alter the value of the gain but change the shape of the gain spectrum. This effect has been discussed by Armitage who points out that gain bandwidths up to 40 nm can be obtained for any pump wavelength through the proper choice of fiber length [72]. While the gain spectrum can be flattened through underpumping or using fiber lengths that are longer than optimum, it exacts a penalty in noise figure and widening the fully inverted bandwidth would be preferable. This requires identifying glasses which provide a broader emission spectrum.

To illustrate the wide range of bandwidths available from glasses, Fig. 9 compares normalized emission spectra for several compositions including those showing the greatest variation. More extensive data are provided in Table VI which lists measured full-widths at half-maximum (FWHM) for compositions representing most of the glass types suitable for optical applications. (It should be noted that extreme care must be exercised in obtaining 4113/2 + 41,5/2 emission spectra since the reabsorption process can significantly distort the observed emission spectrum and lead to an FWHM 50% or more larger than the true value.) Because of differences in the shape of the emission spectrum, particularly the relative strength of the short wavelength peak, FWHM is not necessarily a good indication

WAVELENGTH (nm)

Fig. 9. 1500-nm emission spectra normalized to emphasize variations in shape and bandwidth.

TABLE VI 1500-nm BANDWIDTHS FOR E?’ GLASSES

Gain Bandwidth ( n m )

FWHM 2 0 d B 3 0 d B 4 0 d B Glass ( n m ) Peak Peak Peak

Ge/P silica AI / P silica Silicate ED2 Phosphate L28 Fluorophosphate L74 Fluorophosphate L56 Fluorozirconate F88 Fluorohafnate F114 Ba-Zn-Lu-Th fluoride Borate L87

24.7 43.3 20.4 27.1 32.0 43.3 63.3 63.8 59.2 41.2

5.2 7.9 4.0 4.5 5.3 7.8

16.4 17.3 15.5 7.4

3.8 3.2 6.4 5.3 3.3 2.8 3.6 3.1 3.7 3.3 5.8 5.0

11.7 9.0 13.2 10.4 10. I 8.3 5.6 5.1

of bandwidth in high-gain amplifiers. Accordingly, Table VI also includes gain bandwidths that have been calculated for the fully inverted amplifier which represents the worst case. The columns are labelled by the peak gain since this influences the bandwidth. Experimental values for highly inverted Ge silica [7] and AI/Ge silica [73] amplifiers are in good agreement with those of the similar Ge/P- and Al/P-silica compositions in Ta- ble VI.

A definite correlation between glass type and bandwidth emerges from Table VI. Ignoring A1 /P-silica, the silicate and phosphate glasses are seen to have the narrowest emission spectra. The heavy metal fluoride glasses display the largest FWHM’s; this appears to be related to the coordination of Er3+ by fluoride ions. In these glasses rare-earth ions are believed to substitute only for network formers, resulting in less inhomogeneous broadening as well as a more symmetric and lower crystal field strength than for oxide glasses [74]. This is consistent with the observed Stark splittings [48], and on this basis alone narrower emission bands would be expected. The large measured bandwidths lead one to the conclusion that the distribution of strengths among the Stark transitions determines the spectral shape rather than the Stark splittings. Fluorophosphates have intermediate values of bandwidth and show a significant dependence on composition which will be discussed below. Al-


MINISCALCO: ERBIUM-DOPED GLASSES FOR FIBER AMPLIFIERS A T 1500 nm 243

TABLE VI1 1500-nm BANDWIDTHS FOR FLUOROPHOSPHATE GLASSES

0-to-F Ratio FWHM

( n m )

2.3 1 .5 I .2 1.1 I .o 0.94 0.44 0.19 0.17 0.15

32.0 33.6 36.8 34.8 34.6 33.1 41.3 43.8 43.7 43.3

Gain Bandwidth ( n m )

20 dB Peak

5.3 5 . I 5.9 5.5 5.6 5.4 1.2 1.4 7.9 7.8

30 dB Peak.

3.7 3.9 4.6 4.2 4. I 4.1 5.6 5.5 6.2 5.8

~

40 dB Peak

3.3 3.4 4.0 3.1 3.7 3.6 4.8 4.7 5.0 5.0

-

though the borate glass has an FWHM comparable to that of the widest fluorophosphate, only one composition was examined and the range of bandwidths is not known for this type of glass. Once again the AI/P silica is found to be unusual for a silicate glass, having an FWHM more typical of a borate or high-fluorine-content fluorophosphate glass.

With the exception of silica, fluorophosphates show the greatest sensitivity to composition. The clearest correlation is with anion content, and Table VI1 illustrates how the bandwidth depends upon the oxygen-to-fluorine ratio. Compositions containing mostly oxygen have an FWHM of 2 3 4 nm compared to = 43 nm for those with mostly fluorine. The transition occurs between oxygen-to-fluorine ratios of 0.4 and 0.9, but could not be examined in detail because glass compositions in this range were not made. Inhomogeneous broadening is an attractive explanation for large bandwidths in mixed anion glasses, but the relatively discontinuous nature of the change in width argues against this. Another possibility is that the E 8 + site has a symmetry and crystal field strength similar to either that of a phosphate glass or that of a fluoride glass depending upon the anion ratio.

Although the explanation of bandwidths for Er3+-doped glasses must await further investigation, the data clearly indicate the preferred compositions. Using AI/P silica as the benchmark, it is clear that only borates, fluorides, and high- fluorine-content fluorophosphates offer similar or better bandwidths. Borate glasses, however, must be dismissed because the metastable state is quenched by nonradiative relaxation at room temperature (Section 11). Table VI indicates that under fully inverted conditions only fluoride glasses provide wider gain bandwidths than AI/P silica. Recent experiments with fluorozirconate fiber amplifiers bear this out [75], [76].

IV. PUMP WAVELENGTHS Optically pumped laser, amplifier, or superluminescent de-

vices must be driven by another optical source which excites the active dopant in one of its absorption bands. High gains have been demonstrated pumping the four longest wavelength bands in Fig. 2. Nearly 40 dB has been achieved pumping at 654 nm with a dye laser [77]. Using Ti”: sapphire lasers, 37.5 dB has been reported exciting at 820 nm [78] and we have realized 50 dB by pumping the 980-nm band. In addition, 46.5 dB has been demonstrated for an amplifier driven near 1480 nm by four InGaAsP diodes [79]. Although waveguide design plays a major role in determining efficiency [331, the pump band as

well as the particular wavelengths within it will bave a significant effect upon the efficiency of the device and, in the case of a amplifier, the signal-to-noise ratio. The highest small-signal gain efficiencies (in decibels per milliwatts) published are currently 2 0 . 6 at 654 nm [77], 0.43 at 820 nm [go], 10.2 at 980 [81], and 5.9 at 1480 nm [82]. Practical considerations such as reliability, package size, cost, and power consumption dictate that fiber devices be pumped by diode lasers in most applications. The above efficiencies illustrate the trade-offs that must be considered between developing a new diode laser at a good pump band such as 980 nm, or using existing technology such as high-power AlGaAs lasers to excite the relatively poor pump band at 800 nm. Resonantly pumping the 1480-nm band is an intermediate case. The 650-nm absorption band has neither high efficiency nor high-power diode lasers available, and will not be discussed further. The effect of host glass composition on the pump bands must be considered which ever route is taken, and can play a particularly crucial role in enhancing performance if one of the less favorable bands is to be used. The properties relevant to amplifier performance are discussed below for the three principal pump bands of E?+.

A . 980-nm Pump Band

The 4 1 , s / 2 + ‘I,,,* transition of Er3+ corresponds to an absorption band peaking between 970 and 980 nm which has proven to be quite valuable for driving fiber devices. Amplifiers pumped in this band have not only shown the best performance with respect to gain and gain efficiency, but have also achieved noise figures of = 3 dB [83], [84] and signal output power > 500 mW [28] with a differential photon conversion efficiency of 80% [28]. This success is due to the large absorption cross section for this band coupled with the complete absence of stimulated emission and excited state absorption at the pump wavelength. Fig. 10 illustrates that the absorption cross-section spectrum for this transition can vary appreciably with glass composition. The wavelengths of the cross-section peak in Table VI11 show the usual trend with degree of covalency, the silicates and phosphates lying at longer wavelength and the fluorides and fluorophosphates at shorter. The width of the band is an important consideration since it sets the wavelength tolerance for the pump laser. With the exception of the fluorozirconate and the unusual L22 silicate, the bandwidths are all clustered between 17 and 18 nm. Although maintaining the highest level of performance requires that the pump be close to the peak, the FWHM’s imply an adequate tolerance which has been verified for AI silica [73]. [ U ] . Most of the oscillator strengths fall within a 15% range and, due to the relatively narrow distribution of bandwidths, the peak cross sections show a similarly small range with a few exceptions. Pumping at the absorption peak of a glass with a low cross section is equivalent to pumping away from the peak of a high cross-section glass. In either case high gains can be obtained using longer fiber lengths [73] and the major impact of the lower pump cross section will be a reduced inversion and poorer noise figure for the amplifier. Table VI11 indicates that A l /P silica combines a very high-peak cross section with an acceptable bandwidth. Accordingly, there is little to be gained from using other glass compositions for devices pumped at 980 nm .

B. 800-nm Pump Band

Er’+ has an absorption band peaking near 800 nm which sug- gests the possibility of pumping by high-power relatively in-


244

10

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 9, NO. 2, FEBRUARY 1991

I I I I I /...

:’ \ - -.- Silicate L22 -

I f ;

1 ’ I I I f ‘*.

..... Silicate L22

20 - - Phosphate LIZ -

-..- Fluorophoshpate L11

- - - Fluorozirconate L88 - z I-

-

950 960 970 980 990 1000 I 0 1 0 1020

- E

n N

0 - z c

v)

0 Y

- -..- Phosphate L28

--. Fluorophosphate L11 - - Fluorophosphate L14 - ..... Ba-Zn-Lu-Th -

- - -

780 790 800 810 820 830 840

WAVELENGTH (nm) WAVELENGTH (nm)

Fig. I O . Absorption cross-section spectra for the 980-nm band. Fig. 11. Absorption cross-section spectra for the 800-nm band.

TABLE VI11 ‘II5l2 + ‘Il I l 2 ABSORPTION TRANSITION AT 980 nm .

Oscillator Strength Peak U Peak h FWHM

Glass ( I O - * ) cm‘) ( n m ) ( n m )

AI/P Silica 63.8 31.2 978.6 16.6 Silicate L22 20.4 9.50 980.8 22.1 Phosphate LI 2 39.5 20.1 975.2 17.7 Phosphate L28 48.7 24.7 978.8 18.1 Fluorophosphate L11 47.9 24.6 973.8 17.1 Fluorophosphate L14 45.4 21.5 973.4 17.8 Fluorozirconate F88 37.3 21.5 973.8 15.1

Sources: AI/P silica from Singh et al. [64]; all else this work

expensive AlGaAs diode laser. In most instances, however, the performance of fiber devices pumped in this band has been rather poor because the ground-state absorption transition is weak and overlaps a much stronger excited state absorption (ESA) band. These two absorption processes control the characteristics of amplifiers excited in this band and will be considered together with their dependence upon glass composition.

Fig. 11 compares 41,5/2 +

419/2 absorption cross-section spectra for several glass compositions and reveals quite significant variations for this band. One reason for the poorer performance pumping at 800 nm compared to 980 nm is immediately apparent: the cross sections for the latter band are 3-4 times larger. Relevant properties for the 800-nm band are listed in Table IX. The cross section peaks at nearly the same wavelength for all the compositions and, as with the 980-nm band, the widths fall into a relatively narrow distribution. The sole exception to these patterns is the Al /P silica which has been found to be unusual in so many other respects as well. The variation in oscillator strengths is +20% neglecting the L22 silicate, somewhat larger than for the 1500 and 980 nm transitions. The range of peak cross sections is still greater with some compositions having appreciably higher values than the AI/P silica. Although Table IX implies other glass compositions should offer significant performance improvements over the silica, the more important consideration is ESA which is not considered. Indeed, the best performance to date

1) Ground State Absorption:

TABLE IX 4115/2 + 419/2 ABSORPTION TRANSITION AT 800 nm

Oscillator Strength Peak U Peak X FWHM

Glass (IO-’) ( 10 -??cm2) ( n m ) ( n m )

AI/P Silica Silicate L22 Phosphate L12 Phosphate L28 Fluorophosphate L11 Fluorophosphate L14 Fluorozirconate F88 Ba-Zn-Lu-Th

30.3 11.5 22.5 25.1 26.8 25.4 18.6 18.9

Sources: AI/P silica from Singh’er al.

8.83 793.8 4.17 796.6 8.07 799.8

10.7 798.6 9.68 800.8 7.79 800.6 6.44 800.6 5.90 799.8

[64]; all else this work.

22.2 19.4 16.5 15.2 16.5 18.9 11.7 19.9

has been obtained pumping in the long-wavelength wing of the absorption band, well away from the peak where the ESA process dominates [ 7 8 ] . Fig. 11 reveals that this approach only works because of the fortuitous long-wavelength wing to this absorption band found for many glasses.

2) Excited-State Absorption: Fig. 12 illustrates the ESA process through an example in which the absorption of a photon in the 800-nm ground-state absorption (GSA) band excites an Er‘+ ion into the 419/2 state from which it quickly relaxes through multiphonon emission to the ‘II3l2 level. An ion in the metastable state can be promoted to a still higher level through the absorption of a second 800-nm photon. From Fig. 12 it can be seen that the GSA transition followed by the ESA transition removes two pump photons but produces only one excited ion. In most glasses the 2H,, l 2 terminal state for 800 nm ESA band quickly decays back to the 411,/2 level through multiphonon emission, the net result being the conversion of the second pump photon into heat. If one of the higher lying levels undergoes radiative decay directly to the ground state, both pump photons are lost. ESA is most severe under the conditions of high excited state population necessary for a high-gain, low-noise, traveling-wave amplifier in a three-level gain system. Because lasers operate under saturated conditions, ESA is less a problem if the inversion is kept low by limiting the cavity losses. This is consistent with demonstrations of low threshold fiber lasers [86] and [ 8 7 ] .


MINISCALCO: ERBIUM-DOPED GLASSES FOR FIBER AMPLIFIERS AT 1500 nm

.. -.

245

Fig. 12. Illustration of ESA at the 800-nm pump band. Ion excited to 419/2 level by a pump photon relaxes nonradiatively to the 41,3/2 metastable state from which it absorbs additional pump photons.

In principle other metastable states are possible and each will have its own ESA spectrum. As discussed in Section 11, the multiphonon emission rate for oxide glasses is high enough that only the 41,3 /2 level can be appreciably populated. Additional metastable states are found in tellurites [ 181, [ 191 and fluorides 1471, opening the possibility of additional luminescence as well as ESA transitions. Zemon and coworkers have observed ESA bands from three additional states of E?+-doped fluorozirconate glass [88], [89]. Fig. 13 shows ESA transitions originating not only from the 4F,/, level being pumped at 647 nm, but every excited state which lies below it. This rich ESA spectrum together with the bottlenecking of excitation at the higher levels [90] indicates that Er3+-doped fluoride glasses are only suitable as 1500-nm lasers and amplifiers if resonantly pumped directly into the 41,3 /2 state.

The ESA problem at the 800-nm pump band was first pointed out by Armitage et al. [91]. The process can be characterized by a parameter R,, defined as the ratio of the ESA cross section to the GSA cross section at the wavelength of interest. Subse- quent investigations have revealed that for silica R, is sensitive to the codopants [92], [93]. At 810 nm, R, = 2 for Ge silica [92] while the addition of large amounts of P reduces it to 5 1 [92], [93]. A value of = 1 was also found for A1 silica [92] and AI/P silica 1931. Codoping Ge silica with B [92] or F [93] was found to have no beneficial effect. Fluorozirconate fibers were reported to be significantly better than silica with R, < 1 for all wavelengths greater than 800 nm [93]. These results are consistent with a Judd-Ofelt analysis by Andrews et al. who predicted that glass composition could make as much as a factor of two difference in the integrated strength of the ESA transition, with fluoride and fluorophosphate glasses expected to be the best [94]. The suitability of a particular glass, however, depends upon the spectral details of the GSA and ESA bands, information which requires experiments and cannot be provided by a Judd-Ofelt analysis.

Recently Zemon and coworkers have reported ESA measurements on bulk [88] and fiber [89] Er3+-doped glasses. Fig. 14 shows that the ESA cross section uZ4 for an AI/P-silica fiber exceeds the GSA, u,~, over the complete pump band except

I 1 I I I I I I I I - -

..... Ground State Absorption - -

WAVELENGTH (nm)

Fig. 14. Comparison of ESA and GSA cross-section spectra for AI/P silica. Data are from Zemon et al. [89].

near 820 nm [89], and explains the higher gains realized for Al- silica amplifiers pumped at 820 nm [95]. In contrast, Fig. 15 reveals that for the L11 fluorophosphate GSA is larger than ESA over most of the pump band, including the peak [89]. The difference is a consequence of both the predicted reduction in ESA oscillator strength [94] and the narrower spectra for the L11 glass. The latter leads to a greater offset between the GSA and ESA bands. Other glasses have been examined and found to be intermediate in properties between these two [89]. The cross- section ratio R, can be used as a figure of merit for gain efficiency and is plotted as a function of wavelength for these two glasses in Fig. 16. To the obvious advantage of the fluorophosphate glass apparent from this figure should be added the larger peak stimulated emission cross section found in Table IV.

These spectroscopic differences have been translated into a predicted performance difference using a large-signal fiber amplifier model which includes saturation due to both forward and backward amplified spontaneous emission (ASE) [89]. The host- dependent parameters are provided in Table X for the two pump


24 6 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 9. NO. 2. FEBRUARY 1991

WAVELENGTH (nm)

Fig. 15. Comparison of ESA and GSA cross-section spectra for L 11 fluorophosphate. Data are from Zemon et al. [89].

2 2

I- o w m

P

m m

o m U

P a 1

. a m W

I I I I I I 795 800 805 810 815 820 825 830

WAVELENGTH (nm)

Fig. 16. ESA-to-GSA cross-section ratio R p as a function of wavelength for Al/P-silica and L11 fluorophosphate.

TABLE X AMPLIFIER MODELING PARAMETERS

Glass

Fluorophosphate Parameter Units AI /P Silica L11

nm IO-** cm’

-

nm cm2

nm IO-*’ cm2 lo-” cm2

ms

805.2 6.95 1.55

819.2 2.3 0.5

1531.4 65.3 55.5 10.8

801.3 9.99 0.73

814.7 2.77 0.26

1532.6 69.9 71.6

8.25

wavelengths A,, and Am, and the signal wavelengths A,, used in the calculation. For all cases the fiber had a core radius of 1.96 pm, NA of 0.229, and E?’ concentration of 2.6 X 10l8

40

35

30

25 6- s

20 5 el 15

1 0

5

0

- - - - - - - - , I . / .* - . .

I :

-

- AI/P-Silica, 805.2 nm

- , - - * AI/P-Silica, 819.2 nm - - - - Fluorophosphate, 801.3 nm - -a.- Fluorophosphate, 814.7 nm -

-

-

-

I I 0 5 1 0 15 20 25 30

FIBER LENGTH (m)

Fig. 17. Calculated gain as a function of fiber length for AI/P silica and L11 fluorophosphate amplifiers. Pump power is 100 mW at the wavelengths indicated; the effects of forward and backward ASE have been included. Data are from R. W . Davies, private communication.

cm13 to correspond to the measurements of Kimura et al. for a similar AI/Ge-silica composition [95]. The gain as a function of fiber length has been plotted in Fig. 17 for a pump power of 100 mW. The modeled AI /P-silica amplifier is close to the ex- periment [95]: 17 dB calculated versus 14 dB measured for a 5-m length pumped at 805 nm, and 31 versus 29 dB for a 26-m length pumped at 819 nm. For the L11 glass a gain of 39 dB is achieved pumping at both long and short wavelengths. This 8-dB improvement actually understates the advantages of the L11 glass because the noise properties of the amplifier have not been considered in the analysis. Since the fibers have the same Er” concentration, the shorter fluorophosphate amplifier necessarily has a higher population inversion and therefore a better noise figure in addition to a higher gain than the silica amplifier pumped at 820 nm. In an extensive survey of glasses, the L11 composition was found to be among the most favorable with respect to ESA (961, and the results presented in Fig. 17 are believed to represent close to the best improvement obtainable using a glass composition other than silica.

C. 1480-nm Pump Band

Snitzer et al. first demonstrated that an Er’+ amplifier can be excited directly into the 411,/2 metastable state by pumping at = 1480 nm [97]. In a relatively brief interval of time amplifiers pumped in the range 1470-1490 nm have come to dominate systems demonstrations and field trials because of the excellent results obtained and the commercial availability of high-power 1480-nm diode lasers. Although superficially a two-level system, resonant pumping produces gain because of the shift between the absorption and emission cross-section spectra depicted in Fig. 18 for an AI/P-silica fiber. This offset is observed for all glasses and is a consequence of nonuniformly populated Stark levels since kT at room temperature ( =200 cm- ’ ) is less than the width of both the ‘Il3/* and 4115/2 manifolds (300-400 cm-I). As discussed in Section 111, the relationship between emission and absorption is governed by (4) which describes how the spectral offset increases with decreasing temperature. Rea- sons for the excellent performance achieved by resonant pumping can also be found in Fig. 18. Despite being in the wing of



7 , I , I I ( , I , , , , , , , , , I

To

- 4 -

P z

3 - v)

2 - a

1 -

\

I I I I I I I I I I I I

. 1400 1450 1500 1550 1600 1650

WAVELENGTH (nm)

Fig. 18. Absorption and stimulated emission cross-section spectra for Al/P-silica showing offset and differences in magnitude.

the absorption band, the cross sections are comparable to those at 980 nm. Moreover, since the 4115/2 + 4113,2 absorption band is quite broad and does not vary significantly over this interval, great care need not be exercised in selecting the wavelength of the pump laser and highly multimode pumps will suffice.

There is a drawback to resonant pumping, however. Fig. 18 shows that the offset is not complete: the stimulated emission cross section is appreciable at all wavelengths for which the absorption cross section is large enough to be useful. This reduces pump absorption and prevents full inversion from being obtained even for infinite pump powers. Although high gain can still be achieved by increasing the fiber length, the restriction on inversion insures that some penalty in signal-to-noise ratio is unavoidable. McCumber’s theory provides some guidance here as well: (4) informs us that given the absorption cross section at wavelength of interest for pumping, the stimulated emission cross section at that same wavelength is a function of only the temperature, the wavelength, and the host-dependent parameter t . Because of the latter, one might reasonably expect that the performance achieved pumping at 1480 nm has some dependence on glass composition.

Resonantly pumped amplifiers have been modeled using cross sections at the pump and signal wavelengths determined for Ge/AI silica [98]. Since gain is not an issue [79], we have considered the limitations on signal-to-noise ratio by calculat- ing the degree of inversion under high pump-power conditions. Although not as quantitative as an amplifier model, this provides a rough comparison of the suitability of different glasses for 1480-nm pumping. The fraction of ions in the ‘Il3,* metastable state has been calculated as a function of wavelength for several compositions. The use of the high-power limit is con- venient since such parameters as the spontaneous and stimulated emission rates, signal power, and waveguide parameters drop out of the expression and the excited state fraction 11 is given by

where n is the total Er3+ concentration, n, is population of the are the absorption and stimulated

emission cross sections, respectively, at the pump wavelength. level, and u , ~ , ~ and

1.0 I I I I I I I , I

0.5 1 1 1 1 1 1 1 1 1 1 1 1400 1420 1440 1460 1480 1500 1520

WAVELENGTH (nm)

Fig. 19. Fraction of ions in the excited state q as a function of pump wavelength in the high-power limit.

Excited state fractions are plotted as a function of wavelength in Fig. 19. Only three compositions are shown because of the similarity between all the curves. These plots are intended to define the upper bounds on TJ and are not useful for optimizing a device under normal operating conditions because in the high- power limit most of the emphasis is placed on reducing upm. At finite pump powers U,,,, assumes more importance and the peak of 11 shifts to longer wavelengths [4 11. Table XI summarizes the most significant features for a larger number of glasses. The greatest excited state fraction that can be obtained at infinite pump power is listed together with the pump wavelength at which it is achieved. For all compositions the maximum 1 is ~ 0 . 9 . Also shown is the minimum noise figure F,,, that can be obtained for a high-gain amplifier made from that glass. The noise figure has been calculated using

where the cross-section ratio at the signal wavelength is given by R = U , ~ / ( J ~ ~ . Note that the correct expressions for the noise figure and inversion parameter include the absorption and stimulated emission cross sections at the signal wavelength. Table XI indicates there is little difference between the glasses; even in the limit of infinite pump power all have a minimum penalty of ~ 0 . 5 dB due to resonant pumping.

To obtain information pertinent to actual amplifiers, we have examined the excited state fractions at 1480 nm since this wavelength lies at the center of the range accepted as optimum for the pump powers generally available. Table XI lists the maximum excited state fractions and minimum noise figures that can be achieved at this wavelength in the limit of infinite pump power. Again, all glasses have about the same value of 7 although it is significantly lower at ~ 0 . 7 5 . The corresponding minimum degradation in signal-to-noise ratio has increased to 1.5-2 dB. These estimates are in rough agreement with measured noise figures of = 4 dB for resonant pumping [83], [99] and imply that these numbers cannot be significantly improved upon. The absorption cross sections at 1480 nm have also been included in the table and are seen to fall into the same range of values as those listed in Table VI11 for the 980-nm pump band. In general, one may conclude that glass composition should have


248 JOURNAL OF LIGHTWAVE TECHNOLOGY. VOL. 9. NO. 2, FEBRUARY 1991

TABLE XI LIMITATIONS ON EXCITED STATE FRACTIONS A N D NOISE FIGURES FOR

RESONANTLY PUMPED AMPLIFIERS ~~

At Peak of 7 ( A ) At 1480 nm

urb\ Glass A ( n m ) 7 F,,, 7 F,,, (lo-’’ cm’)

AI/P Silica 1410 0.94 3.4 0.77 4.9 2.44 Silicate L22 1429 0.89 3.5 0.74 4.4 1.37 Fluorophosphate LI1 1421 0.92 3.4 0.76 4.6 3.01 Fluorophosphate L14 1423 0.92 3.4 0.76 4.6 2.91 Fluorozirconate F88 1418 0.92 3.4 0.74 4.9 1.90

little effect upon performance for resonant pumping; all will achieve high gain but suffer a 2 1 dB noise figure penalty over what can be obtained pumping at 980 nm. AI/P silica is quite typical in this respect.

V. SUMMARY A N D CONCLUSIONS Host-dependent properties affecting the performance of Er‘+-

doped fiber amplifiers have been discussed and, to the extent to which data are available, the range of characteristics presented by the major glass compositions has been examined. If all other factors were equal, silica would be the material of choice because of its superior optical, mechanical, and chemical properties coupled with mature manufacturing techniques and its inherent compatibility with the passive transport fiber. How- ever, pure silica is a relatively poor host for rare-earth ions, dissolving only low concentrations and often providing poor spectroscopic properties. For Er” this translates into limiting concentrations to less than 100 ppm-wt and a narrow main emission peak. Codoping with Ge or G e / P does not alter these limitations. In contrast, codoping with even very low levels of A1 enables silica to be doped to rare-earth concentrations comparable to those of multicomponent silicate glasses. Surprisingly, for Er” it also results in spectroscopic properties superior to those of other silicate glasses, making AI silica the benchmark to which all other compositions must be compared.

Of the potentially useful glasses, phosphates and fluorophosphates have the largest values of both oscillator strength and peak stimulated emission cross section for the transition at 1500 nm. Silicates tend to have low oscillator strengths but high-peak cross sections by virtue of their narrow bandwidths. A1 silica, however, has the high oscillator strength, wide bandwidth, and peak emission wavelength more typical of a fluorophosphate than a silicate. Although some compositions have higher peak cross sections than A1 silica, only heavy metal fluoride glasses offer significantly wider bandwidths. At the important 980-nm pump band A1 silica has a high-peak cross section while all compositions provide roughly the same bandwidth. The absorption cross sections at = 1480 nm are all comparable, and the principal issue for this pump band is the noise figure penalty resulting from the overlap between emission and absorption. In this respect only small differences are found between glasses. A major advantage of other compositions over A1 silica occurs at the 800-nm pump band. Here fluorophosphates have lower excited-state absorption which translates into both substantially higher gain and better signal-to-noise ratio. Based upon available data, one may draw the general conclusion that in most situations A1 silica is the preferred glass for Er3+-doped fiber amplifiers at 1500 nm. The opportunity to exploit high-power

AlGaAs diode laser technology for 800-nm pumping or the need for wide-gain bandwidths, however, may make the use of other glass compositions a worthwhile endeavor.

ACKNOWLEDGMENT

The author wishes to acknowledge the invaluable assistance provided by many individuals. Stimulating discussion with L. J . Andrews, S. Zemon, R. S. Quimby, R. W. Davies, M. L. Dakss, and M . Weber were extremely valuable. Important measurements and data analysis were provided by B. A. Thompson and M . P. Singh, and R. C. Folweiler, B. T. Hall, and T. Wei were responsible for synthesizing and characterizing many of the materials discussed. I wish to particularly thank R. S. Quimby, S. Zemon, and R. W. Davies for making available their results prior to publication.

REFERENCES

[l] E . Snitzer, “Neodymium glass laser,” in Proc. 3rd Intl. Conf

[2] C . J . Koester and E. Snitzer, “Amplification in a fiber laser,”

[3] J . Stone and C. A. B u m s , “Neodymium-doped silica lasers in end-pumped fiber geometry,” Appl. Phys. Lett., vol. 23, pp. 388- 389, 1973.

[4] S. B. Poole, D. N. Payne, and M. E . Fermann, “Fabrication of low-loss optical fibres containing rare-earth ions,” Electron. Lett., vol. 21, p. 737, 1985.

[ 5 ] R. J. Mears, L. Reekie, S. B. Poole, and D . N. Payne, “Neo- dymium-doped silica single-mode fibre laser,” Electron. Lett.,

[6] R. J . Mears, L. Reekie, I. M. Jauncey, and D. N. Payne, “High- gain rare-earth-doped fiber amplifier at 1.54 pm,” in OFCIIOOC ’87, OSA 1987 Tech. Dig. Series, vol. 3, 1987, paper W12, p. 167.

[7] R. J . Mears, L. Reekie, I. M. Jauncey, a n d D . N. Payne, “Low- noise erbium-doped fibre amplifier operating at 1.54 pm,” Elec- tron. Lett., vol. 23, pp. 1026-1028, 1987.

[8] V. P. Gapontsev, S. M. Matitsin, A. A. Isineev, and V. B. Krav- chenko, “Erbium glass lasers and their applications,” Optics and Laser Techno/., pp. 189-196, Aug. 1982.

[9] N. E. Alekseev, V. P. Gapontsev, M. E . Zhabotinskii, V. B. Kravchenko, and Yu. P. Rudnitskii, Laser Phosphate Glasses. Moscow: Nauka Publishing House, trans: Berkeley Sci. Trans. Svc., 1983.

[IO] S . E. Stokowski, R. A. Saroyan, and M. J . Weber, “Nd-doped laser glass spectroscopic and physical properties, ” Lawrence Livermore National Laboratory Report M-095, Rev. 2, Nov. 1981.

[ I l l M. J. Weber, “Science and technology of laser glass,” J . Non- Cryst. Solids, vol. 123, pp. 208-222, 1990.

[12] R. Reisfeld and Y. Eckstein, “Dependence of spontaneous emission and nonradiative relaxations of Tm3+ and Er” on glass host and temperature,” J. Chem. Phys., vol. 63, pp. 4001-4012, 1975.

[I31 C. B. Layne, W. H. Lowdermilk, and M. J . Weber, “Multi- phonon relaxation of rare-earth ions in oxide glasses,” Phys. Rev.

[14] C . B. Layne and M. J. Weber, “Multiphonon relaxation of rare- earth ions in beryllium-fluoride glass,” Phys. Rev. B , vol. 16,

151 M. D. Shinn, W. A. Sibley, M. G. Drexhage, a n d R . N. Brown, “Optical transitions of Er?’ ions in fluorozirconate glass,” Phys. Rev. B , vol. 27, pp. 6635-6648, 1983.

161 E. R. Taylor et al., “Application-specific optical fibres manu- factured from multicomponent glasses,” Mater. Res. Soc. Proc.,

171 R. Reisfeld and Y . Eckstein, “Optical spectra of erbium and thu- lium in germanate glass,” J. Non-Cryst. Solids, vol. 12, pp. 357- 376, 1973.

181 R. Reisfeld and Y . Eckstein, “Radiative and nonradiative transition probabilities and quantum yields for excited states of E?+

Quantum Electron (Paris, France), 1963, pp. 999-1019.

Appl. Opt., vol. 3, pp. 1182-1186, 1964.

vol. 21, pp. 738-740. 1985.

B, vol. 16, pp. 10-20, 1977.

pp. 3259-3261, 1977.

vol. 172, pp. 321-327, 1990.



in germanate and tellurite glasses,” J . Non-Crysr. Solids, vol.

[19] R. Reisfeld and Y. Eckstein, “Intensities of electronic transitions and quantum efficiencies of the green luminescence of Er3+ in tellurite glasses,” Solid State Commun., vol. 13, pp. 741-744, 1973.

[20] R. Wyatt, “Spectroscopy of rare earth doped fibres,” Fibre Laser Sources and Amplifiers, M. J . F. Digonnet, Ed., Proc. SPIE, vol. 1171, 1990,pp. 54-64.

[21] K. Arai et al., “Aluminum or phosphorus co-doping effects on the fluorescence and structural properties of neodymium-doped silica glass,” J . Appl. Phys., vol. 59, pp. 3430-3436, 1986.

[22] F. E. Auzel, “Materials and devices using double-pumped phos- phors with energy transfer,” Proc. IEEE, vol. 61, pp. 758-786, 1973.

[23] J . C. Wright, “Up-conversion and excited state energy transfer in rare earth doped materials,” in Radiationless Processes in Molecules and Condensed Phases, F. K. Fong, Ed. Berlin: Sprin- ger-Verlag, 1976, pp. 239-295.

[24] B. J. Ainslie et al., “Erbium doped fibres for efficient optical amplifiers,” IEE Proc., vol. 137, Pt. J , pp. 205-208, 1990.

[25] N. Kagi, A. Oyobe, and K. Nakamura, “Efficient optical amplifier using a low-concentration erbium-doped fiber, ” IEEE Pho- ton. Technol. Lett., vol. 2, pp. 559-561, 1990.

1261 R. I. Laming, D. N. Payne, F. Meli, G. Grasso, and E. J . Tar- box, “Saturated erbium-doped fibre amplifiers,” in Tech. Dig. Opt. Ampl$ers Applications, 1990 (Optical Society of America, Washington, DC), 1990, vol. 13, pp. 16-19.

1271 S . P. Craig-Ryan, J . F. Massicott, M. Wilson, B. J. Ainslie, and R. Wyatt, “Optical study of low concentration Er” fibres for efficient power amplifiers,” in ECOCPO, Proc., vol. I , pp. 571- 574, 1990.

1281 J . F. Massicott, R. Wyatt, B. J. Ainslie, and S . P. Craig-Ryan, “Efficient, high power, high gain Er” doped silica fibre amplifier,” Fiber Laser Sources and Amplifiers 11, M. J . F. Digonnet, Ed., Proc. SPIE, vol. 1373, in press.

1291 B. J . Ainslie “Optical and structural investigation of Nd” in silica-based fibers,” J . Materials Sci. Lett., vol. 6, pp. 1361-1363, 1987.

[301 H. Namikawa, A. Kazuo, K. Kumata, Y. Ishii, and H. Tanaka, “Preparation of Nd-doped SiO, glass by plasma torch CVD,” Jpn. J . Appl. Phys., vol. 21, pp. L360-L362, 1982.

[31] M. Shimizu, M. Yamada, M. Horiguchi, and E. Sugita “Con- centration effects on optical amplification characteristics of Er- doped silica single-mode fibers,” IEEE Photon. Technol. Lett.,

[32] B. J . Ainslie, S. P. Craig, and S. T. Davey, “Fabrication and optical properties of Nd’ ’ in silica-based optical fibers,” Mater. Lett., vol. 5 , pp. 143-146, 1987.

[33] J . R. Armitage, “Three-level fiber laser amplifier: A theoretical model,” Appl. Opt. , vol. 27, pp. 4831-4836, 1988.

[34] B. J. Ainslie, S. P. Craig, S. T. Davey. and B. Wakefield, “The fabrication, assessment and optical properties of high-concentration Nd3+- and Er?+-doped silica-based fibres,” Mater. Lett., vol.

[35] J . R. Simpson and J . B. MacChesney, “Optical fibres with an Al,03-doped silicate core composition,” Electron, Lett., vol. 19,

[36] H. Imai, K . Arai, Y . Fujino, Y . Ishii, and H. Namikawa, “The role of fluorine as a codopant in cerium activated silica glass,’’ Phys. and Chem. Glasses, vol. 29, pp. 54-58, 1988.

[37] N. Sandoe, P. H. Sarkies, and S. Parke. “Variation of Er3+ cross section for stimulated emission with glass composition,” J . Phys.

[38] E. Desurvire and J . R. Simpson, “Amplification of spontaneous emission in erbium-doped single-mode fibers,” J . Lightwave Technol., vol. 7, pp. 835-845, 1989.

[391 C. G. Atkins et al. , “High-gain, broad spectral bandwidth erbium-doped fibre amplifier pumped near 1 .5 pm,” Electron. Lett.,

1401 W. J . Miniscalco, L. J . Andrews, B. A. Thompson, T. Wei, and B. T. Hall, “41,3/2 ++ 41 , s / , emission and absorption cross sections for Er3+ doped glasses,” in Tunable Solid State Laser, M. L. Shand and H. P. Jenssen, Eds., OSA Proc. Series, vol. 5 , 1989, pp. 71-83.

[41] W. J . Miniscalco, L. J . Andrews, B. A. Thompson, T. Wei, and B. T. Hall, “The effect of glass composition on the performance

15, pp. 125-140, 1974.

vol. 2 , pp. 43-45, 1990.

6, pp. 139-144, 1988.

pp. 261-262, 1983.

D., vol. 5 , pp. 1788-1799, 1972.

vol. 2.5, pp. 910-91 1, 1989.

of Er” fiber amplifiers,” Fiber Laser Sources and Ampl$ers, M. J . F. Digonnet, Ed., Proc. SPIE, vol. 1171, 1990, pp. 93- 102.

1421 K. Dybdal, N. Bjerre, J . E. Pedersen, and C. C. Larsen, “Spec- troscopic properties of Er-doped silica fibers and preforms,” Fi- ber Laser Sources and Amplifiers, M. J . F. Digonnet, Ed., SPIE,

[43] W. J . Miniscalco er al., “Glasses for Er3+-doped fiber lasers and optical amplifiers,” Mar. Res. Soc. Proc., vol. 172, pp. 329- 334, 1990.

[44] M. Yamada, M. Shimizu, M. Horiguchi, M. Okayasu, and E. Sugita, “Gain characteristics of an Er3+-doped multicomponent glass single-mode optical fiber,” IEEE Photon. Technol. Lett.,

[45] T. Yamashita, “Nd- and Er-doped phosphate glass for fiber laser,” Fiber Laser Sources and Amplifiers, M. J . F. Digonnet, Ed., Proc. SPIE, vol. 1171, pp. 291-297, 1990.

[46] S. G . Grubb, W. L. Barnes, E. R. Taylor, and D. N. Payne, “Diode-pumped 1.36 pm Nd-doped fibre laser,” Electron. Lett.,

[47] P. W. France and M. C. Brierley, “Progress in fluoride fibre lasers and amplifiers,” Fiber Laser Sources and Amplifiers 11, M. J . F. Digonnet, Ed., SPIE vol., 1373, in press.

1481 S. Zemon, G . Lambert, W. J . Miniscalco, L. J . Andrews, and B. T. Hall, “Characterization of Er3+-doped glasses by fluorescence line narrowing,” Fiber Laser Sources and Ampl$ers, M. J. F. Digonnet, Ed., Proc. SPIE, vol. 1171, pp. 219-236, 1990.

1491 E. Desurvire and J. R. Simpson, “Evaluation of 4115/2 and 41,3/2 Stark-level energies in erbium-doped aluminosilicate glass fiber,” Opt. Lett., vol. 15, pp. 547-549, 1990.

[SO] E. Desurvire, J. L. Zyskind, and J . R. Simpson, “Spectral gain hole-burning at 1.53 pm in erbium-doped fiber amplifiers,” IEEE Photon. Technol., Lett., vol. 2, pp. 246-248, 1990.

1511 D. W. Hall, R. A. Haas, W. F. Krupke, and M. J . Weber, “Spectral and polarization hole burning in neodymium glass lasers,” IEEE J . Quantum Electron., vol. QE-19, pp. 1704-1717, 1983.

[52] D. W. Hall, W. F. Hagen, and M. J. Weber, “Modeling broad- band and multiple-frequency energy extraction from glass laser amplifiers,” IEEE J . Quantum Electron., vol. QE-22, pp. 793- 796, 1986.

1531 E. Desurvire, J . W. SulhofF, J. L. Zyskind, and J . R. Simpson, “Study of spectral dependence of gain saturation and etfect of inhomogeneous broadening in erbium-doped aluminosilicate fiber amplifiers,” IEEE Photon. Technol. Lett., vol. 2, pp. 653- 655, 1990.

[54] R. I. Laming, L. Reekie, P. R. Morkel, and D. N. Payne, “Mul- tichannel crosstalk and pump noise characterisation of E?+-doped fibre amplifier pumped at 980 nm.” Electron. Lett., vol. 25, pp.

[SS] J . L. Zyskind, E. Desurvire, J. W. Sulhoff, and D. J . Di- Giovanni, ‘‘Spectral gain hole-burning in an erbium-doped fiber amplifier with Ge02 : Si02 core,” Technical Dig. on Optical Am- plijiers and their Applications, 1990 (Optical Society of America, Washington, DC), 1990, vol. 13, pp. 56-59.

[56] C. C. Robinson, “Multiple sites for Er3+ in alkali silicate glasses 1,”J. Non-Cryst. Solids, vol. 15, pp. 1-10, 1974.

[57] C. C. Robinson, “Multiple sites for E?’ in alkali silicate glasses 11,” J . Non-Cryst. Solids, vol. 15, pp. 11-29, 1974.

[58] C. Brecher and L. A. Riseberg, “Laser-induced fluorescence line narrowing in Eu glass: A spectroscopic analysis of coordination structure,” Phys. Rev. B , vol. 13, pp. 81-93, 1976.

[59] C. Brecher and L. A. Riseberg, “Laser-induced line narrowing of Eu3+ fluorescence in fluoroberyllate glass: Site-dependent spectroscopic properties and their structural implications, ” Phys. Rev. B , vol. 21, pp. 2607-2618, 1980.

[60] M. J . Weber and S. A. Brawer, “Investigations of glass structure using fluorescence line narrowing and molecular dynamics simulations,” J . Phys. (Paris) vol. 43, Colloque C9, pp. 291-301, 1982.

[61] B. R. Judd, “Optical absorption intensities of rare-earth ions,” Phys. Rev., vol. 127, pp. 750-761, 1962.

[62] G. S. Ofelt, “Intensities of crystal spectra of rare-earth ions,” J . Chem. Phys., vol. 37, pp. 511-520, 1962.

[63] W. T. Carnall, H. Crosswhite, and H. M. Crosswhite, “Energy level structure and transition probabilities of the trivalent lan- thanides in LaF,,” Argonne National Lab. Rep., 1978.

vol. 1171, pp. 209-218, 1990.

vol. 2, pp. 656-658, 1990.

vol. 26, pp. 121-122, 1990.

455-456, 1989.


250 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 9. NO. 2, FEBRUARY 1991

[64] M. P. Singh, D. W. Oblas, J . 0. Reese, W. J . Miniscalco, and T. Wei, “Measurement of the spectral dependence of absorption cross section for erbium-doped single-mode fiber,’’ Symp. on Op- tical Fiber Measurements, 1990, NIST Special Publication 792,

1651 D. E. McCumber, “Theory of phonon-terminated optical ma- sers,” Phys. Rev., vol. 134, pp. A299-A306, 1964.

[66] M. J . Pettitt, R. A. Baker, and A. Hadjifotiou, “Crosstalk in erbium doped fibre amplifiers,” Electron. Lett., vol. 25, pp. 416- 417, 1989.

L671 E. Desurvire, C. R. Giles, and J . R. Simpson, “Gain saturation effects in high-speed, multichannel erbium-doped fiber amplifiers at 1.53 pm,” J. Lightwave Technol., vol. 7, pp. 2095-2104, 1989.

[68] P. R. Morkel and R. I. Laming, “Theoretical modeling of erbium-doped fiber amplifiers with excited state absorption,” Opt.

[69] W. L. Barnes, R . I. Laming, P. R . Morkel, and E. J . Tarbox, “Absorption-emission cross-section ratio for Er” doped fibers at 1.5 pm,” CLEO’90, OSA 1990 Tech. Dig. Series, vol. 7, p. 50, 1990.

(701 R. S. Quimby and W. J . Miniscalco, “Determination of stimulated emission cross sections for Er-doped glass,” OSA 1990 An- nual Meeting, 1990, Paper MCC2.

[71] W. J . Miniscalco and R. S. Quimby, “General procedure for the analysis of Er3+ cross sections,” to be published in Opt. Lett.

[72] J . R. Armitage, “Spectral dependence of the small-signal gain around 1.5 p n in erbium doped silica fiber amplifiers,” IEEE J . Quantum Electron., vol. 26, pp. 423-425, 1990.

[73] P. C. Becker, A. Lidgard, J . R. Simpson, and N. A. Olsson, “Erbium-doped fiber amplifier pumped in the 950-1000 nm region,” IEEE Photon. Technol. Lett., vol. 2, pp. 35-37, 1990.

1741 J . Lucas, “Rare earths in fluoride glasses,” J . Less-Common Metals, vol. 112, pp. 27-47, 1985.

[75] C. A. Millar and P. W. France, “Diode-laser pumped erbium- doped fluorozirconate fibre amplifier for the 1530 nm communications window,” Electron. Lett., vol. 26, pp. 634-635, 1990.

(761 Y. Miyajima, T. Sugawa, and T. Komukai, “20 dB gain at 1.55 pm wavelength in 50 cm long Er’+-doped fluoride fibre amplifier,” Electron. Lett., vol. 26, pp. 1527-1528, 1990.

(771 B. Pedersen et al., “Experimental verification of an accurate model for erbium doped fiber amplifier in a gain range from 0 to 40 dB,” in Tech. Dig. Opt. Amplifers Applications, 1990 (Op- tical Society of America, Washington, DC), vol. 13, pp. 60-63.

[78] M. Nakazawa, Y. Kimura, and K. Suzuki, “High gain erbium fibre amplifier pumped by 800 nm band,” Electron. Lett., vol.

[79] Y. Kimura, K. Suzuki, and M. Nakazawa, “46.5 dB gain in E8+- doped fibre amplifier pumped by 1.48 p n GaInAsP laser diodes,” Electron. Lett., vol. 25, 1656-1657, 1989.

[SO] K. Suzuki, Y. Kimura, and M. Nakazawa, “High gain Er‘+- doped fibre amplifier pumped by 820 nm GaAlAs laser diodes,” Electron. Lett., vol. 26, pp. 948-949, 1990.

[81] M. Nakazawa, Y. Kimura, and K. Suzuki, “An ultra-efficient erbium-doped fiber amplifier of 10.2 dB/mW at 0.98 pm pumping and 5.1 dB/mW at 1.48 pm pumping,” Topical Meeting Op- tical Ampl$ers Applications, OSA 1990 Tech. Dig. Series, vol. 13, Postdeadline Paper PDPl.

[82] J . L. Zyskind, D. J . DiGiovanni, J . W. Sulhoff, P. C. Becker, and C. H. Brito Cruz, “High performance erbium-doped fiber amplifier pumped at 1.48 pm and 0.97 pm,” in Tech. Dig. Op- tical Amplifiers Applications, 1990 (Optical Society of America, Washington, DC), vol. 13, Postdeadline Paper PDP6.

1831 M. Yamada et al . , “Noise characteristics of Er’+-doped fiber amplifiers pumped by 0.98 and 1.48 pm laser diodes,” Photon. Technol. Lett., vol. 2 , pp. 205-207, 1990.

[84] R . I. Laming and D. N. Payne, “Noise characteristics of erbium- doped fiber amplifier pumped at 980 nm,” IEEE Photon. Tech- no/. Lett., vol. 2, pp. 418-421, 1990.

(851 W. J . Miniscalco, B. A. Thompson, E. Eichen, and T. Wei, “Very high gain Er” fiber amplifier pumped at 980 nm,” OFC’90, OSA Tech. Dig. Series, 1990, vol. I , p. 192, 1990.

[86] C . A. Millar, I. D. Miller, B. J. Ainslie, S. P. Craig, and J . R.

pp. 93-96, 1990.

Lett., vol. 14, pp. 1062-1064, 1989.

26, pp. 548-550, 1990.

Armitage, “Low-threshold cw operation of an erbium-doped fibre laser pumped at 807 nm wavelength,” Electron. Lett., vol. 23,

[87] L. Reekie, I. M. Jauncey, S. B . Poole, and D. N. Payne, “Diode- laser pumped operation of an Er‘+-doped single-mode fibre laser,” Electron. Lett., vol. 23, pp. 1076-1078, 1987.

I881 S. Zemon, G. Lambert, W. J . Miniscalco, L. J . Andrews, and B. T. Hall, “Excited state cross sections for Er-doped glasses,” Mat. Res. Soc. Proc., vol. 172, pp. 335-340, 1990.

[89] S. Zemon et al., “Excited state cross sections for Er-doped glasses,” Fiber Laser Sources and Ampl@ers 11, M. J . F. Di- gonnet, Ed., SPIE 1373, in press.

1901 R. S. Quimby, “Output saturation in fiber lasers,” Appl. Opt.,

[Yl] J . R. Armitage, C. G . Atkins, R. Wyatt, B. J . Ainslie, and S. P. Craig, “Spectroscopic studies of Er3+ doped single-mode silica fibre,” Topical Meeting on Tunable Solid State Lasers, Tech. Dig. Series, 1987, vol. 20 (Optical Society of America, Washington,

(921 R . I . Laming, S. B. Poole, and E. J . Tarbox, “Pump excited- state absorption in erbium-doped fibers,” Opt. Lett., vol. 13, pp.

[93) C. G . Atkins, J . R . Armitage, R . Wyatt, B. J . Ainslie, and S. P. Craig-Ryan, “Pump excited state absorption in Er“ doped optical fibers,” Opt. Commun., vol. 73, pp. 217-222, 1989.

[94] L. J . Andrews, W. J . Miniscalco, and T. Wei, “Excited state absorption of rare earths in fiber amplifiers,” in Proc. ofthe First Inti. School on Excited States of Transition Elements (World Sci- entific, Singapore, 1989), pp. 9-30.

[95] Y. Kimura, K. Suzuki, and M. Nakazawa, “High gain erbium- doped fiber amplifier pumped in the 0.8 p n pump band,” ECOC’90, Proc., vol. I , pp. 103-106, 1990.

pp. 865-866, 1987.

vol. 29, pp. 1268-1276, 1990.

DC), pp. 193-196.

1084-1086, 1988.

1961 S. Zemon, private communication. [97] E. Snitzer, H. Po, F. Hakimi, R. Tumminelli, and B. C. Mc-

Collum, “Erbium fiber laser amplifier at 1.55 pm with pump at 1.49 p n and Yb sensitized Er oscillator,” OFCIOFS’88, 1988, OFC Postdeadline Paper PD2.

[98] E. Desurvire, “Analysis of erbium-doped fiber amplifiers pumped in the 41,s/2-41,3/2 band,” IEEE Photon. Technol. Lett.. vol. 1,

[99] C. R. Giles, E. Desurvire, J . L. Zyskind, and J . R. Simpson, “Noise performance of erbium-doped fiber amplifier pumped at 1.49 pm, and application to signal preamplification at 1.8 Gbits/s,” IEEE Photon. Technol. Lett., vol. I , pp. 367-369, 1989.

pp. 293-296, 1989.

*

William J. Miniscalco received the Ph.D. degree in physics from the University of Illinois, Urbana, in 1977 for research on phase transitions of electrons and holes in semiconductors.

Subsequently he joined the Physics Depart- ment of the University of Wisconsin where he undertook optical investigations of transition metal and rare-earth doped insulators. Since 1979 he has been at GTE Laboratories Incor- porated, Waltham, MA, where he is now a Principal Member of Technical Staff in the area

of optical communications. He is currently the principal investigator of projects on doped active fibers and ultratrace impurity analysis of glasses using selectively excited photoluminescence. Previously he was principal investigator of programs to characterize radiation-induced defects in silica fiber and develop new materials for tunable solid state lasers. His research has been in the areas of the electronic structure of magnetic semiconductors and doped insulators, phase transitions and critical phenomena, excited state dynamics of optical ions, electronic defects in solids, and nonlinear magneto-optics.

Dr. Miniscalco is a member of the American Physical Society, the Optical Society of America, and Sigma Xi.


Erbium-doped glasses for fiber amplifiers at 1500 nm...

Documents

Transcript of Erbium-doped glasses for fiber amplifiers at 1500 nm...