OPTICAL FIBER TECHNOLOGY ARTICLE NO.
2, 387]393 Ž1996.
0044
Role of Aluminum in Ytterbium – Erbium Codoped Phosphoaluminosilicate Optical Fibers G. G. VIENNE,1 W. S. BROCKLESBY, R. S. BROWN, Z. J. CHEN, J. D. MINELLY, J. E. ROMAN, AND D. N. PAYNE Optoelectronics Research Centre, Uni¨ ersity of Southampton, Southampton SO17 1BJ, United Kingdom Received April 11, 1996; revised May 22, 1996
We report on a series of ytterbium]erbium codoped aluminophosphosilicate fibres fabricated by the modified chemical vapor deposition and solution doping technique where the concentration of aluminum is gradually increased in the region [A1] - [P]. We observe from deflection measurements that with increasing aluminum content the refractive index decreases and the phase separation increases. Laser slope efficiencies decrease when the aluminum concentration approaches the phosphorus concentration and the trend is seen to be correlated with the reduction in energy transfer rate when the aluminum concentration is increased. Raman spectra show the reduction of the highest phonon energy peak when the aluminum content is 4 4 increased. The erbium I 13rr 2 ª I 15rr 2 transition fluorescence spectrum and the lifetime of the erbium 4 S 3rr 2 level show little change from the pure phosphosilicate host for the whole range of aluminum concentrations tested. The results are explained in term of formation of AlPO4 structural units. Q 1996 Academic Press, Inc.
1. INTRODUCTION
The absorption bands of Er 3q are not only relatively sparse throughout the visible and near infrared, but are also relatively weak for wavelengths exceeding 550 nm. Sensitizing erbium-doped fibers with ytterbium allows new pumping schemes and a reduced device length. Since the first realization in 1991 of efficient silica-based ytterbium]erbium codoped fibers by the well-established modified chemical vapor deposition ŽMCVD. process w1x, unprecedented high-power amplifiers and lasers w2, 3x, soliton sources w4x, and single-frequency lasers w5, 6x have been demonstrated. There is currently a surge of interest in Yb:Er codoped cladding-pumped fibers to scale up the power of optical fiber amplifiers and lasers operating in the third optical telecommunication window w7x. These fibers are designed to take advantage of the newly available multiwatt diode 1 To whom correspondence should be addressed. Fax: q44 Ž0.1703 593142. E-mail:
[email protected].
arrays. Their structure consists of a codoped waveguide from which monomode emission at 1.5 m m is obtained, surrounded by a larger multimode waveguide in which the pump can be efficiently launched. To create this imbedded waveguide structure in a silica fiber the codoped region should have a refractive index at least 2 = 10y2 higher than silica. The emergence of wavelength division multiplexing ŽWDM. systems imposes another requirement upon our Yb:Er codoped fibers. For this application a broad fluorescence spectrum is required to obtain a flat gain over the 1530]1560 nm region w8x. All efficient Yb]Er fibers reported to date contain phosphorus. However, phosphorus is highly volatile so that numerical aperture ŽNA. higher than 0.2 Ž D n s 1.4 = 10y2 . are not easily achieved by the MCVD process. Our phosphosilicate fibers also show a particularly narrow emission spectrum around 1.5 m m. In this work we attempt to change the host glass composition to increase the NA and broaden the emission spectrum of the fiber. The range of composition is limited by an important constraint: a rapid multiphonon decay from the transferring level of erbium to the metastable level is necessary to reduce the back-transfer of energy from Er to Yb. Phosphate glasses are particularly suitable hosts because the high maximal phonon energy of these glasses results in a sufficiently rapid multiphonon decay to produce efficient Yb:Er devices w9x. In silica, aluminum is well known to broaden the emission spectrum of the rare earths and to allow a high rare-earth incorporation without clustering w10x. Alumina also has a high molar refractivity in silica Ž2.3 = 10y3 index changermol%. w11x and is not prone to evaporation w10x. Unfortunately, aluminosilicate Yb:Er codoped fibers have a smaller multiphonon decay rate than phosphosilicate fibers, which results in the presence of back-transfer of energy from Er to Yb. In phosphate glasses, addition of aluminum is well known to toughen the glass and to reduce its hygroscopy w12x. Phosphoaluminosilicate fibers have been investigated
387 1068-5200r96 $18.00 Copyright Q 1996 by Academic Press, Inc. All rights of reproduction in any form reserved.
388
VIENNE ET AL.
to achieve high-NA germanium-free fibers w13x. Indeed, a small addition of phosphorus ŽwPx - wA1x. has proved very useful to go beyond the 4 mol% solubility limit of alumina in silica. But, although addition of aluminum has been well studied in silica for application in active optical fibers, there is to our knowledge little information on the role of aluminum in phosphosilicate rare-earth-doped fibers. In this study we investigate whether some of the attractive features of aluminum doping of silica, also appear in a phosphosilicate host. Our approach is to start with a phosphosilicate host because it is a suitable glass for efficient Yb:Er doped fibers and to modify the fiber core by addition of aluminum in order to alter its refractive index and the fluorescence spectrum of the erbium ions. As this modification can also have consequences on the efficiency of the energy transfer, we measure the slope efficiencies of lasers with different aluminum contents. In the last part of this paper we present lifetime and Raman scattering measurements in an attempt to relate the device efficiency results to the structure of the glass. 2. EXPERIMENTAL PROCEDURE
We fabricated a series of preforms by depositing a phosphosilicate frit and doping it in an aqueous solution containing the rare earths and the aluminum precursors. We preferred to introduce the aluminum by the liquid phase but it is also possible to deposit it by the vapor phase. However, heated delivery lines are needed in this case. We gradually increased the aluminum solution strength from 0 to 24 g AlCl 3 6ŽH 2 O.r200 ml H 2 O and paid attention to keeping all the other fabrication parameters identical. Special care was also taken to dry the frits before sintering. The refractive index profiles of preforms and fibers were calculated from deflection measurements performed by York P101 and S14 profilers. The fiber diameters were chosen to correspond to a second mode cutoff wavelength of 1.3 m m. The glass compositions in the preforms were measured by energy dispersive X-ray spectrometry ŽEDS. in a JEOL-6400 scanning electron microscope. Losses in the fibers were measured by the cutback technique using an Ando AQ-6315A optical spectrum analyzer. To perform the 1.5-m m fluorescence spectra measurements we used very short lengths of fibers, typically 4 cm, to avoid distortion due to reabsorption. We also reduced progressively the excitation power at 980 nm to check that the spectra were not affected by stimulated emission. We performed two types of laser measurements. In the first one we pumped the fibers with a Nd:YAG laser, launching 1 W at 1064 nm. No reflector was used so that feedback was provided by the 3.5% Fresnel reflection of the bare fiber ends. The fiber was cutback to obtain the
maximum output power. In the second type of laser we used a 980-nm laser diode as a pump source. The input mirror reflectivities were 2% at 980 nm and 100% at 1535 nm. The output mirror reflectivities were 20% at 980 nm and 92.5% at 1535 nm. The laser length was 4.5 cm corresponding to 10-dB pump absorption. Ytterbium decays were detected with a photomultiplier. The fibers were pumped with a Q-switched Nd:YAG laser. The repetition rate was 30 Hz, allowing 95% of the excited erbium ions to relax to the ground state between each pulse. The resolution of the measurement was 100 ns. Raman spectra were acquired in a backscattering configuration with a CCD array cooled to 144 K. We used the 476.5-nm line of an argon laser to excite the samples. This excitation wavelength was chosen to avoid any absorption by the erbium and to put the 600]1500 cmy1 Raman shifted light as free of fluorescence as possible. Fluorescence from the 4S3r2 level at 550 nm was excited using an argon ion laser at 488 nm. The ion laser was modulated acousto-optically, resulting in a turnoff time of less than 100 ns. Fluorescence was detected from the side of the fiber, filtered using a Jobin]Yvon HR640 monochromator, and detected using a GaAs photocathode photomultiplier. The system time resolution was better than 100 ns. 3. RESULTS AND DISCUSSION
The fibers contain 0.6]0.8 mol% Yb 2 O 3 and 11]13 mol% P2 O5 . The ratio of ytterbium to erbium is 20. The highest Al 2 O 3 concentration, obtained from the solution containing 24 g AlCl 3 6ŽH 2 O.r200 ml H 2 O, is 8.2 mol%. The OHy concentration is lower than 1 ppm in all the fibers so that no quenching of the erbium metastable level occurs. Figure 1 shows that the numerical aperture decreases significantly with addition of aluminum. A preform fabricated in the same way as the ones of this series but doped with erbium only ŽwErx s 0.3 mol% . and 48 g AlCl 3 6ŽH 2 O.r200 ml H 2 O exhibited an NA of 0.06, confirming the trend observed in Fig. 1. This low NA was impractical because of the weak guiding making the fiber very prone to bending losses at 1.5 m m. Figure 2 illustrates the role of the addition of aluminum on the refractive index profile ŽRIP.. Not only does the aluminum reduce the maximum refractive index but it also reduces the central dip. Indeed, the aluminum prevents the phosphorus from evaporating during the collapse of the preform just as it prevents germanium from evaporating in germanoaluminosilicate preforms w10x. There is, however, an important difference with germanoaluminosilicate preforms: although the RIP is flattened, a ripple is present all across the core. This is due to the inhomogeneity of the
ALUMINIPHOSPHOSILICATE Yb 3q]Er 3q CODOPED OPTICAL FIBERS
FIG. 1.
389
Numerical aperture vs AlCl 3 Ž6H 2 O. solution strength.
glass and can also be observed with the bare eye. The phase separation starts being noticeable at 6 g AlCl 3 6ŽH 2 O.r200 ml H 2 O and increases with increasing aluminum content. Consequently the minimum loss Žmeasured around 1.2 m m. increases with increasing aluminum content, as can be seen in Fig. 3. The preforms also showed an increasing tendency to shatter when heated or sawed with increasing aluminum content. Figure 4 shows that the fluorescence spectrum is not significantly affected by the aluminum content. All spectra had a full-width half-maximum ŽFWHM. of 20]22 nm. We note no broadening of the spectrum when the aluminum content is increased. In fact, further from the peak at 1535 nm, the fluorescence decreases with increasing aluminum content. On the other hand, we measured that an Yb]Er codoped aluminophosphosilicate fiber where
FIG. 3. Minimum loss vs AlCl 3 Ž6H 2 O. solution strength. The minimum loss was measured between 1000 and 1500 nm.
FIG. 2. Refractive index profile for two different AlCl 3 Ž6H 2 O. solution strength.
FIG. 4. Fluorescence spectra for three different AlCl 3 Ž6H 2 O. solution strength.
wPx - wA1x Ž2 mol% P2 O5 and 11 mol% Al 2 O 3 . had a FWHM of 52 nm w1x. However, this fiber showed a very poor efficiency and most of the lasing from the bare ends originated from the ytterbium rather than the erbium. We also observe in Fig. 4 that the ratio of absorption at 1480 and 1535 nm is reduced when aluminum is added. Consequently the ratio of stimulated emission is also reduced w14x. Kringlebotn et al. recently proposed to use an Yb:Er codoped fiber as a laser source for remote pumping of
390
VIENNE ET AL.
amplifiers around 1480 nm w15x. For this device the amplified stimulated emission at 1535 nm competes severely with the emission at the laser wavelength. To obtain a ratio of stimulation emission at 1480 and 1535 nm as high as possible the phosphosilicate host glass should be aluminum free. The deterioration of device efficiency with addition of aluminum already occurs in the region wPx ) wA1x, as evidenced in Fig. 5. Note that the maximum signal power versus launched pump power plotted in this figure is a good indicator of the slope efficiency of the lasers because the launched pump power was 1 W " 10%, a value far exceeding the laser threshold of typically 30 mW. The value of the threshold is expected to increase with aluminum content as for a given cutoff wavelength the pump and signal mode sizes increase with reducing NA. The increase in mode sizes is not expected to directly affect the slope efficiency of the lasers but the optimum length increases so that the background losses become more critical when the mode sizes increase. There can be several causes of the reduction in slope efficiency: hydroxyl quenching of the erbium metastable level, loss of excitation by upconversion due to interactions between erbium ions, a high background loss, or finally a deterioration of the energy transfer efficiency. The two first reasons can be ruled out, as we have checked that the concentration of hydroxyl was low enough in all fibers to avoid quenching and the rare-earth concentrations did not vary by more than 20% from fiber to fiber. However, as the pump absorption at 1064 nm is low Ž2]2.5 dBrm., the increased background loss with increasing aluminum solution strength already observed in Fig. 3 may correlate with the reduction in slope efficiency.
Indeed, we have observed in a previous study on phosphosilicate fibers with no aluminum that with 1 mol % ytterbium, the slope efficiency was reduced from 22 to 7% when the minimum loss increased from 0.2 to 0.7 dBrm with a 1064-nm pumping scheme. This study also showed that on the other hand, the slope efficiency was not significantly affected by the increase in minimum loss when pumping at 980 nm w16x. Figure 6 shows that in this case, however, even when pumping at 980 nm the slope efficiency is significantly reduced for the fiber doped with 24 g AlCl 3 6ŽH 2 O.r200 ml H 2 O. This gives us reason to believe that the reduced efficiency is related to the role of the aluminum in the energy transfer rather than the increase in minimum loss. To investigate the energy transfer between ytterbium and erbium we measured the decay of the excited ytterbium ions. Figure 7 shows the decay of low- and highaluminum-content fibers together with the double exponential fits. The two phases of the decay can be explained as follows: the closest donor]acceptor pairs rapidly undergo energy transfer initially and their population becomes depleted early in the decay because the slow erbium deexcitation prevents any more transfer from occuring within the ytterbium intrinsic excitation lifetime. The depletion continues radially outwards from the donor ions as time progresses until the only unexcited erbium ions remaining are sufficiently separated from the donor so that energy transfer does not occur. The donor decay then resumes the intrinsic exponential decay. The slow decay component shows little variation with the aluminum content. The time constant is 500 q ry 100 m s over the whole range of concentrations. On the other hand, Fig. 8
FIG. 5. Efficiency of lasers pumped at 1064 nm vs AlCl 3 Ž6H 2 O. solution strength.
FIG. 6. 980-nm pumped laser characteristics for two different AlCl 3 Ž6H 2 O. solution strengths.
ALUMINIPHOSPHOSILICATE Yb 3q]Er 3q CODOPED OPTICAL FIBERS
FIG. 7. Ytterbium decay for two different AlCl 3 Ž6H 2 O. solution strengths with biexponential fit.
shows that the fast component increases with increasing aluminum concentration and the fiber doped with 24 g AlCl 3 6ŽH 2 O.r200 ml H 2 O exhibits a much slower initial decay than all the other fibers. The loss by ytterbium fluorescence becomes significant when the transfer time is increased and the laser performances are deteriorated. We conclude that the reduction in slope efficiency is linked to the slower energy transfer when aluminum is added. So far, we have shown that addition of aluminum in the region wA1x - wPx decreases the refractive index of the
FIG. 8. strength.
Ytterbium lifetime fast component vs AlCl 3 Ž6H 2 O. solution
391
glass, produces phase separation, does not lead to spectral broadening around 1.5 m m, and moreover reduces the device efficiency. It is clear at this stage that aluminum codoping cannot be used to improve our Yb:Er codoped amplifiers and lasers emitting at 1.5 m m. However, the measurements performed so far can provide some insight into the structure of the glass. In silica, addition of aluminum alone or phosphorus alone is well known and widely used to increase the refractive index w11x. Here, we saw that combination of the two codopants leads to much lower refractive indices than would be expected from the superposition of the two binaries Al 2 O 3 ]SiO 2 and P2 O5 ]SiO 2 . DiGiovanni also observed this behavior in the wA1x - wPx region of undoped phosphoaluminosilicate prepared by MCVD w17x. This is a clear evidence that the mixture of these two codopants in silica results in a ternary glass significantly different from the individual binary systems. A structural modification must take place when the two binary glasses are mixed together. The reduced evaporation of phosphorus when aluminum is added to the glass suggests that an unpaired aluminum ion presents a favorable site for phosphorus bonding. Finally, the fluorescence spectra provide information on the environment of the rare earths. The shape of the fluorescence spectra was almost unchanged until 12 g AlCl 3 6ŽH 2 O. and was characteristic of a phosphate host. The slight narrowing at 24 g AlCl 3 6ŽH 2 O. may be due to the reduction of available P5O bonds after pairing with aluminum. In any case no broadening is observed, suggesting that no rare earth is coordinated by wAlO4 x or wAlO6 x groups. The rare earths remain coupled to P5O sites but when 24 g AlCl 3 6ŽH 2 O. was used, not all the rare earths could be hexacoordinated by P5O sites. ŽIn phosphates, as well as in silicate and germanate glasses the ytterbium ion is sixfold coordinated w18x. The erbium ion, having almost the same size as the ytterbium ion, is also expected to be sixfold coordinated.. Finally, an interesting question still remains to be answered: What is the cause of the increase in transfer time? In particular, we want to address whether the back-transfer of energy is responsible for the deterioration in transfer efficiency. Figure 9 shows the Raman spectra for the different aluminum concentrations. Galeener and Mikkelsen showed that the high phonon energy peak around 1330 cmy1 was due to the P5O double bond vibration in pure phosphate glass w19x. Further study showed that this peak is a characteristic feature of the ultraphosphate, i.e., where phosphate PO4 tetrahedral units are linked by three bridging oxygens w20x. We normalized this peak to the 800 cmy1 peak which is associated with the silica network. In Fig. 10 we show the ratio of these two peaks for the different aluminum solution strengths. We observe that the high phonon energy peak decreases with increasing
392
VIENNE ET AL.
FIG. 11. glass.
FIG. 9.
Raman spectra for different AlCl 3 Ž6H 2 O. solution strengths.
aluminum content. The reduction of the 1330 cmy1 Raman peak with increasing aluminum doping can be attributed to the scavenging of P5O double bonds by the aluminum. Figure 11 shows the Al]O]P bonds resulting from the pairing of Al and P. Since phosphorus is electron rich when incorporated in a silica host, a PO4 unit can serve as a Lewis base by offering its double-bonded oxygen as an electron donor. Similarly, aluminum accepts an electron from the P5O double bond and may be considered a Lewis acid when coordinated tetrahedrally. In the
FIG. 10. Magnitude of the highest phonon energy peak vs AlCl 3 Ž6H 2 O. solution strength.
Structure of phosphosilicate and aluminophosphosilicate
SiO 2 ]P2 O5 ]Al 2 O 3 ternary glass the phosphorus serves to preserve the tetrahedral coordination of the aluminum. It is important to note that the almost complete disappearance of the high phonon energy peak in the fiber doped with 24 g AlCl 3 6ŽH 2 O. cannot be attributed to pairing of aluminum and phosphorus only. Indeed, the concentration of aluminum is two-thirds that of the phosphorus in this fiber and even if AlPO4 units are formed with complete efficiency, as observed in w21x, the P5O population should not be reduced by more than two-thirds. The almost complete disappearance of the high phonon energy peak may result from the partial depolymerization of the phosphate structure with the introduction of the aluminum cation}a similar effect as when Na is added to P2 O5 w16x. By looking carefully at the Raman spectra in Fig. 9 we also observe the reduction of the peak around 1230 cmy1 , arising from PO 2 antisymmetric stretching vibrations, when alumina is added. The same reduction is observed when molybdenum is added to a phosphate glass forming a MoO 3 ]P2 O5 glass w22x. The molybdenum, like the aluminum, enters the glass matrix and depolymerizes the three-dimensional phosphate network. To probe the environment of the rare earths we decided to measure the lifetime of the fluorescence of the erbium at 545 nm. It originates from the 4S3r2 level and the decay is dominated by nonradiative relaxation to the 4 F9r2 level, which is directly related to the maximum host phonon energy. Figure 12 shows that the 4S3r2 lifetime remained unchanged over the whole range of aluminum concentrations tested. The lifetime was 0.42 m s, which is in good agreement with measurements in phosphosilicate fibers from Lincoln w23x. The very small variation of the green lifetime with the aluminum content suggests that the rare earths remain coupled to P5O units over the whole range of aluminum doping tested, as was inferred from the fluorescence spectra. We can conclude that the deterioration of energy transfer with increasing aluminum content is not due to an increase in back-transfer of energy. The forward energy transfer must be affected by the disruption of the phosphate network, which may in-
ALUMINIPHOSPHOSILICATE Yb 3q]Er 3q CODOPED OPTICAL FIBERS
FIG. 12.
Green lifetimes vs AlCl 3 Ž6H 2 O. solution strength.
crease the average distance between rare-earth ions or create phase separation on a nanometer scale resulting in pockets of ytterbium ions too far apart from any erbium ion to transfer their energy. 4. CONCLUSIONS
We have shown that addition of aluminum in the region wA1x - wPx decreases the refractive index of the glass, produces phase separation, does not lead to spectral broadening around 1.5 m m, and moreover reduces the device efficiency. We have made clear that aluminum codoping cannot be used to improve our Yb:Er codoped amplifiers and lasers emitting at 1.5 m m. The reduction in slope efficiency with increasing aluminum content was seen to be caused by an increase in energy transfer time. In this series of fibers, where wA1x - wPx, the rare earths have been seen to couple preferably to P5O radicals rather than Al]O probably because all aluminum ions bond to phosphorus to create AlPO4 units. As a consequence the increase in energy transfer time was attributed to the disruption of the phosphate network rather than the presence of energy back-transfer. REFERENCES w1x J. E. Townsend, W. L. Barnes, K. P. Jedrzejewski, and S. G. Grubb, ‘‘Yb3 q sensitised Er3 q doped silica optical fiber with ultra high transfer efficiency and gain,’’ Electron. Lett., vol. 27, no. 21, 1958 Ž1991.. w2x S. G. Grubb, T. H. Windhorn, J. E. Townsend, K. P. Jedrzejewski, and W. L. Barnes, ‘‘q20 dBm erbium power amplifier pumped by a diode-pumped Nd:YAG laser,’’ in Proc. Topical Meeting on Optical Amplifiers and Their Applications, Snowmass, CO, pp. 12-1]12-4, July 1991.
393
w3x J. D. Minelly, W. L. Barnes, R. I. Laming, P. R. Morkel, J. E. Townsend, S. G. Grubb, and D. N. Payne, ‘‘Diode-array pumping of ErrYb co-doped fibre lasers and amplifiers,’’ IEEE Photon. Technol. Lett., vol. 5, no. 3, 301 Ž1993.. w4x M. J. Guy, D. U. Noske, and J. R. Taylor, ‘‘Generation of femtosecond soliton pulses by passive mode locking of an ytterbium]erbium figure of eight laser,’’ Opt. Lett., vol. 18, no. 17, 1447 Ž1993.. w5x K. Hsu, C. M. Miller, J. T. Kringelbotn, and D. N. Payne, ‘‘Tunable, single-frequency Er:Yb phospho-silicate fiber Fabry]Perot lasers,’’ in ECOC’94. w6x J. T. Kringlebotn, J.-L. Archambault, L. Reekie, and D. N. Payne, ‘‘1.5 m m Er:Yb-doped fibre DFB laser,’’ in CLEO’94, cwp2, p. 261. w7x Italtel Soc Ital Telecom Spa, ‘‘Optical fiber amplifier has length of double-clad fibre with pump source coupled to length of fibre ŽEng.,’’ World wide patent application, filed 13.10.1993. w8x D. M. Spirit and M. J. O’Mahony, High Capacity Optical Transmission Explained, p. 209, Wiley, New York, 1995. w9x A. G. Murzin, D. S. Prilezhaev, and V. A. Fromzel, ‘‘Some features of laser excitation of ytterbium-erbium glasses,’’ So¨ . J. Quantum Electron., vol. 15, no. 3 Ž1985.. w10x B. J. Ainslie, ‘‘A review of the fabrication and properties of Erbium-doped fibres for optical amplifiers,’’ J. Lightwa¨ e Technol., vol. 9, no. 2 Ž1991.. w11x K. L. Walker, ‘‘Optical fibre fabrication and characteristics,’’ in OFC’87. w12x M. L. Elder, Y. T. Hayden, J. H. Campbell, S. A. Payne, and G. D. Wilke, ‘‘Thermal]mechanical and physical]chemical properties of phosphate laser glasses,’’ in American Ceramic Society Annual Meeting, Cincinnati, OH, 1991. w13x C. J. Scott, ‘‘Optimization of composition for Al 2 O 3rP2 O5-doped optical fiber,’’ in OFC’84, Paper TUM4, New Orleans. w14x D. E. McCumber, Phys. Re¨ ., vol. 134, A299 Ž1964.. w15x J. T. Kringelbotn, J.-L. Archambault, L. Reekie, R. I. Laming, and D. N. Payne, ‘‘High power Nd:YLF pumped 1490 nm grating-feedback Er 3q:Yb 3q co-doped silica fibre laser,’’ Opt. Fiber Technol., vol. 2, no. 4 Ž1996.. w16x G. G. Vienne, ‘‘Progress towards an efficient ErrYb cladding pumped fibre,’’ confidential report to ATx Telecom Systems, Inc, March 1995. w17x D. J. DiGiovanni, J. B. MacChesney, and T. Y. Kometani, ‘‘Structure and properties of silica containing aluminum and phosphorus near the AlPO4 join,’’ J. Non-Crystalline Solids, vol. 113, 58 Ž1989.. w18x C. C. Robinson and J. T. Fournier, ‘‘Co-ordination of Yb 3q in phosphate, silicate, and germanate glasses,’’ J. Phys. Chem. Solids, vol. 31, 895 Ž1970.. w19x F. L. Galeener and J. C. Mikkelsen, ‘‘The Raman spectra and structure of pure vitreous P2 O5 ,’’ Solid State Comm., vol. 30, 505 Ž1979.. w20x S. W. Martin, ‘‘Review of the structure of phosphate glasses,’’ Eur. J. Solid State Inorg. Chem. L., vol. 28, 163 Ž1991.. w21x S. G. Kosinski, D. M. Krol, T. M. Duncan, D. C. Douglass, J. B. MacChesay, and J. R. Simpson, ‘‘Raman and NMR spectroscopy of SiO 2 glasses co-doped with Al 2 O 3 and P2 O5 ,’’ J. Non-Crystalline Solids, vol. 105, 45 Ž1988.. w22x S. H. Morgan, ‘‘Raman spectra of molybdenum phosphate glasses and some crystalline analogues,’’ J. Am. Ceram. Soc., vol. 73, 753 Ž1990.. w23x J. Lincoln, ‘‘Spectroscopy of rare-earths in glass,’’ Ph.D thesis, Chap. 8, Department of Physics, University of Southampton, 1993.