Comparative study of praseodymium additives in active selenide chalcogenide optical fibers

: The choice of rare earth additive when doping chalcogenide glasses can affect their mid-infrared fiber performance. Three praseodymium additives, Pr-foil, PrCl 3 and PrI 3 are investigated in Ge-As-Ga-Se fibers. All the fibers are X-ray amorphous and the Pr(foil)-doped fiber has the lowest overall optical loss. Pumping at 1550 nm wavelength, the Pr 3+ -doped fibers exhibit photoluminescence across a 3.5 to 6 μ m span; photoluminescence lifetimes are 10 ms for 3 H 5 → 3 H 4 and 2-3 ms for ( 3 H 6, 3 F 2 ) → 3 H 5 transitions. A fast 0.21 ms decay for ( 3 F 3, 3 F 4 ) → 3 H 6 is observed only in the PrCl 3 -doped fiber due to a lower phonon energy local environment of Pr 3+ ions.


Introduction
Rare earth (RE) ion doped chalcogenide glasses exhibit discrete emissions across the mid-infrared (MIR) window, giving potential applications as active photonic devices like MIR amplifiers [1,2] and MIR molecular sensors [3][4][5] and, especially, they are the most promising materials proposed for achieving the first MIR fiber laser to work beyond 4 µm wavelength [6][7][8][9]. Although much numerical and analytical modeling work demonstrates the feasibility of MIR lasing in RE-ion-doped fibers with chalcogenide glass hosts [10][11][12][13], loss-inducing impurities and crystallization can negatively affect the quality of the fabricated RE-ion-doped chalcogenide glasses [14][15][16][17] and therefore impede realization of MIR fiber lasing. The type of RE additive used to produce the RE-ion-doping can have an important influence on chalcogenide glass-quality (e.g. crystallization, optical loss) [18].
RE additive in its elemental form, usually as a metallic foil, is a popular additive in the fabrication of RE-ion-doped chalcogenide bulk glass and fiber [19][20][21]. For example, in 2008, Park et al [20]. used a praseodymium metal additive to make Pr 3+ -doped Ge-Ga-Sb-Se bulk glasses and fibers and reported spectroscopy across the 3.5-5.5 μm wavelength range. In 2015, Zhang et al [21]. presented the thermal and optical properties of Dy 3+ -doped Ga-Sb-S glasses, made by adding Dy-foil during chalcogenide glass melting. In our previous research [18], both Dy foil and DyCl 3 additives were found to cause corrosion of the silica-glass ampoule containment during melting of Dy 3+ -doped Ge-As-Ga-Se glasses; however the Dy 3+ -doped Ge-As-Ga-Se glass made by Dy-foil addition led to better results of lowered scattering loss and crystallization and improved glass surface quality. DyCl 3 addition did not give good results at higher dopant concentrations i.e. ≥ 1000 ppmw Dy 3+ [16,22]. Nevertheless, using the DyCl 3 additive enabled an optical loss of 1.16 dB/m at 6.6 μm wavelength in a 500 ppmw (parts per million by weight) Dy 3+ (DyCl 3 )-doped Ge-As-Ga-Se fiber, without extra glass distillation and and was either re-batched as reference glass, or re-batched with Ga (5N, Testbourne) and the praseodymium additive of: Pr foil (3N, Alfa Aesar) or PrCl 3 (4N, Alfa Aesar) or PrI 3 (3N, Alfa Aesar) into a fresh silica glass ampoule of ID = 8 mm, which was sealed under vacuum (10 −3 Pa). The purity of each of the Pr additives used was the highest one available from international commercial chemical companies, including Alfa Aesar and Sigma Aldrich. The ampoule containing pre-melted Ge-As-Se glass, Ga or Pr-additive was raised to 850 °C and held isothermally for 6 hours whilst rocking ± 30 ° about a horizontal axis to homogenize the glass-melt. The chalcogenide glass melt was then quenched and annealed in situ, inside the silica-glass ampoule melt containment to form a rod of either: (i) Ge 15 As 21 Se 63 glass; (ii) 500 ppmw Pr(foil)-doped Ge 15 As 21 Ga 1 Se 63 glass; (iii) 500 ppmw Pr 3+ (PrI 3 )-doped Ge 15 As 21 Ga 1 Se 63 glass and (iv) 500 ppmw Pr 3+ (PrCl 3 )-doped Ge 15 As 21 Ga 1 Se 63 glass. The host glass was the 1 at.% Ga content glass developed as in [33]. The Pr 3+ (PrI 3 )-doped Ge-As-Ga 1 -Se glass experienced another re-melt (6 hours at 850 °C) because the glass rod preform had fractured inside the silica ampoule after the first glass-melting. Photographs of the Pr 3+ -doped Ge-As-Ga 1 -Se glass rod preforms are presented in Fig. 1; no distinct surface corrosion can be observed on any of the glass surfaces. The preforms were fiber-drawn into 230 μm diameter unstructured fibers on a customized Heathway fiber-drawing tower under N 2 ('white-spot', BOC). In this work, Ga was added to the host glass because it helped solubilize rare earth ions in chalcogenide glass matrix and is generally accepted to complex to the rare earth ions in the glass matrix [34][35][36]. However, the very low vapor pressure of Ga means it cannot be purified together with the Ge, As and Se in the distillation process [28], Therefore, the distilled Ge-As-Se glass/fiber rather than the Ge-As-Ga-Se host, was fabricated as the reference glass/fiber for optical loss comparison. Note that all the silica glass ampoules were pre-cleaned in HF aqueous and were treated for 6 hours each at 1000 °C, first in air and then under vacuum (10 −3 Pa). Key information on glass preparation and fiber fabrication is summarized in Table 1. Fig. 1. Rod preforms of 500 ppmw Pr 3+ (foil), Pr 3+ (PrI 3 ) and Pr 3+ (PrCl 3 )-doped Ge-As-Ga 1 -Se glasses, showing excellent surface quality without distinct (contamination) spots (cf [15,16].).

Characterization of bulk glass and fiber
Powder XRD patterns of each fiber were collected using a Siemens D500 system, running from 10 °2θ to 70 °2θ at step-size 0.05 °2θ per 40 seconds, totaling >12 hours for each XRD run. In fiber loss measurement, the two-groups-cleaves' cut-back method [37] was applied with a IFS 66/S, Bruker FTIR spectrometer. Then for FTIR spectroscopy of bulk glass, using the same spectrometer, bulk sample disks were sawn from the glass preforms prior to fiber drawing. All the FTIR samples were placed together on a copper sample holder for a multi-samples' polishing protocol to a 1 µm finish on the opposite parallel faces of the discs, in order to approach the same polishing quality for comparison of baseline losses. The thickness of the FTIR samples was 2.771 ± 0.005 mm. Fiber PL spectra of the Pr 3+ -doped Ge-As-Ga 1 -Se samples were collected using a pump laser at 1550 nm (FPL 1009S, Thorlabs), a monochromator (MiniMate, Spex), a lock-in amplifier (7270 DSP, Metek) and an ambient MCT detector (mercury-cadmium-telluride, PVI-6, Vigo System). As shown in Fig. 2(a), the set-up was fiber side-collection of PL intensity. In fiber PL lifetime measurements, the 1550 nm pump laser (Thorlabs) was modulated at 6.4 Hz, and used along with the monochromator (Spex), an electric-cooled MCT detector (PVI-4T-6, Vigo System) and a digital oscilloscope (Picoscope5204, Pico Technology) for the collection of fluorescent decay. As shown in Fig. 2(b), the PL lifetimes were also measured using fiber side-collection. The accuracy of a wavelength selected and measured by the monochromator system was ± 20 nm. Both fiber PL spectra and lifetimes were collected from the side of the cleaved fibers, specifically in order to minimize re-absorption [38,39] and any influence of optical loss. In this set-up, there was very short optical pathlength for the emitted light to leave the side of fiber, which had a diameter of 230 µm. Also, an effective 2 mm length of the fiber sample was used in the PL signal collection (each fiber sample was 117 mm total length but this length was mainly used for the purpose of holding the sample on a V-groove holder; the 2 mm fiber length sticking out of the sample holder was actually used for the PL collection). Also, in this side collection, a 2945 nm wavelength long-pass filter (Northumbria Optical Coatings) was applied between the fiber and monochromator to prevent any high orders of scattered light from the pump from affecting the monochromator measurements; the PrCl 3 -doped fiber and the PrI 3 -doped fiber exhibited a high scattering loss at the 1550 nm wavelength of the pump laser, whereas the Pr(foil)-doped fiber did not present distinct scattering loss at this wavelength. The system response was measured using a Globar blackbody source and all of the PL spectra in this work were corrected for this. In the measurement of PL lifetime, each presented decay plot was collected from one fiber sample with the same alignment and was measured for up to 14000 times to improve the signal-to-noise ratio. In lifetime calculation, for comparatively long transitions of >1 ms in 3 H 5 → 3 H 4 and ( 3 H 6, 3 F 2 )→ 3 H 5 , the fitting error was ~0.2 ms for the Pr 3+ (foil) and Pr 3+ (PrI 3 )-doped fibers, and the error increased to ~0.7 ms for the Pr 3+ (PrCl 3 )-doped fiber due to less signal and more noisy PL decay. For <1 ms short transition of ( 3 F 3, 3 F 4 )→ 3 H 6 , the fitting error of calculated lifetime was 0.06 ms with the Pr 3+ (PrCl 3 )-doped fiber.

Fiber XRD, fiber loss and FTIR spectroscopy of bulk glass
Powder XRD patterns of the 500 ppmw Pr 3+ (foil), Pr 3+ (PrI 3 ) and Pr 3+ (PrCl 3 )-doped Ge-As-Ga 1 -Se fibers are presented in Fig. 3 and all of them indicate XRD amorphicity. In our previous work, 500 ppmw Dy 3+ (DyCl 3 )-doped Ge-As-Ga 3 -Se glass fiber exhibited a small amount of crystallization to a modified α-Ga 2 Se 3 [23]. In this work, the XRD pattern of the 500 ppmw Pr 3+ (PrCl 3 )-doped Ge-As-Ga 1 -Se glass fiber did not present any distinct crystallization peaks. Although the rare-earth ion chloride additive was different (PrCl 3 instead DyCl 3 ), it is believed that lowering the Ga content from 3 at% to 1 at% played an important role in reducing crystals in the Ge-As-Ga-Se fiber. This is supported by our previous study that lowering Ga content helped decrease Ga 2 Se 3 crystals in 3 at % Ga compared to 10 at % Ga additive in Ge-As-Ga-Se bulk glasses [15].   In Fig. 4, the Ge-As-Se fiber exhibits the lowest baseline loss of 1.2-1.3 dB/m at 2.5 to 3.4 μm wavelength and 0.9 to 1.0 dB/m at 5.6-7.0 μm wavelength, and the lowest loss is 0.87 dB/m at 6.03 μm wavelength. After including the Ga solubilizer and Pr additives to make the 500 ppmw Pr 3+ -doped Ge-As-Ga 1 -Se fibers, the background optical loss increased. Both the PrCl 3 -doped fiber and the PrI 3 -doped fiber presented a clear extra scattering loss in the NIR range. Using 1.55 μm as a wavelength reference point (which was the pump wavelength for the PL results in this work), the baseline optical loss of Ge-As-Se fiber, Pr(foil)-doped fiber, PrI 3 -doped fiber and PrCl 3 -doped fiber increased in the order: 2 dB/m, 5 dB/m, 25 dB/m and ~75 dB/m, respectively.
Among the 500 ppmw Pr 3+ -doped Ge-As-Ga 1 -Se fibers, the Pr foil-doped fiber had the lowest overall fiber baseline loss across the 2 to 9 μm range; this was 2.0-2.5 dB/m at wavelengths of 2.7-3.4 μm and 6.0-7.5 μm. Note that, although the PrCl 3 -doped fiber exhibited a large scattering loss in the NIR region, this fiber actually gave the lowest minimum loss of 1.8 dB/m at 6.93 μm wavelength amongst all of the RE-ion-doped-fibers here (see inset of Fig.  4). Finally, the PrI 3 -doped fiber presented the highest baseline loss (3 dB/m) above 6 μm wavelength. Further analysis of fiber loss was assisted by FTIR results of fiber preform glass as follows.
FTIR spectra of bulk glass samples cut and polished from the as-prepared rod-preforms before fiber-drawing of the 500 ppmw Pr 3+ (foil), Pr 3+ (PrI 3 ) and Pr 3+ (PrCl 3 )-doped Ge-As-Ga 1 -Se glasses and the Ge-As-Se glass are given in Fig. 5. In this work, the FTIR bulk glass samples underwent co-multi-polishing in an effort to reduce any variability of surface finish and optical path length and enhance the analysis of baseline loss comparison. As the background loss levels of the FTIR spectra of the bulk glass samples were not absolute, all spectra were vertical shifted to be overlapped in the 6 to 8 μm wavelength range to aid the comparison of any excess scattering loss in the NIR region. From Fig. 5, it is evident that both the PrCl 3 -doped and PrI 3 -doped bulk glasses exhibited excess NIR scattering loss compared to that of the Pr foil-doped bulk glass and the Ge-As-Se bulk glass which had almost overlapping baselines (also see inset (a) to Fig. 5). Adding the Ga and Pr additives is bound to change the properties of the Pr 3+ -doped Ge-As-Ga 1 -Se glasses: including density and/or dielectric constant fluctuations, in comparison to the base Ge-As-Se glass. This will contribute an associated variation in scattering loss [41]. However, the scattering loss induced by such compositional change is expected to be small (for As 2 Se 3 from [41] anticipated to be within the scale of dB/km) compared to the distinct extra scattering loss found in the PrCl 3 -doped and PrI 3 -doped glasses (≥20 dB/m at 1.55 µm wavelength). Note that inset (a) of Fig. 5 shows that at 1.5 μm, 1.6 μm and 2.0 μm wavelengths, the Pr(foil)-doped bulk glass exhibited unstructured Pr 3+ electronic absorption bands, but that the PrCl 3 -doped and PrI 3 -doped bulk glasses instead exhibited structured Pr 3+ electronic absorption bands. This structuring is due to Stark level absorptions because of a more ordered environment and suggests that for the as-prepared PrI 3 and PrCl 3 -doped bulk glasses, at least a part of the population of the Pr 3+ ions did not dissolve properly in the glassy matrix environment and probably remained mainly coordinated with Cl/I instead of Se, although this does not necessarily indicate clustering. Moreover, inset (b) of Fig. 5 presents the characteristics of the absorption bands around 4.5 μm wavelength. The Ge-As-Se fiber exhibited a 0.035 cm −1 absorption coefficient (equivalent to 35 dB/m) due solely to a -[Se-H] vibrational absorption band centered at 4.5 μm wavelength. However, for the Pr 3+ -doped bulk glasses, the Pr 3+ electronic absorption band and -[Se-H] vibrational absorption band overlapped in the region around 4.5 μm wavelength; the PrCl 3 -doped glass presented a lower overall band height compared to the Pr(foil)-doped and PrI 3 -doped bulk glasses. Like in [40], the absorption band at 4.5 μm wavelength was then deconvoluted to its components viz.: vibrational absorption due to extrinsic impurity in the glass: -[Se-H], and the Pr 3+ electronic absorption. The -[Se-H] loss was 38 dB/m, 32 dB/m and 28 dB/m, respectively, at 4.5 μm wavelength for the Pr foil-doped, PrCl 3 -doped and PrI 3 -doped fibers, respectively. From this result, we venture to suggest that the presence of chloride and iodide acted as a [H]-getter thereby reducing the extrinsic loss due to -[Se-H] impurity absorption. Also, the Pr 3+ absorption band at ~4.5 μm wavelength was found structured in both PrI 3 -doped and PrCl 3 -doped glasses.
Combining optical losses shown in Fig. 4 and Fig. 5, there would have been refractive index discontinuities in the PrCl 3 -doped/PrI 3 -doped bulk glass and fiber that caused the large wavelength-dependent scattering loss. The possible sources of such 'refractive index discontinuities' were: (i) more structured but not yet crystallized Pr-Cl/Pr-I local environment sites (which would also contribute to the structured Pr 3+ absorption bands in Fig. 5); (ii) the XRD technique had a ~1 to 5% volume detection limitation (depends on type of crystals), so there was possibility of a small amount of XRD-undetectable crystals caused the extra scattering loss in the PrCl 3 /PrI 3 -doped glasses; such crystals could be either Pr 3+ -containing (e.g. PrCl 3 /PrI 3 ) or formed by the host glass elements (e.g. Ga 2 Se 3 ) -for example, Ge-modified α-Ga 2 Se 3 crystals were found in DyCl 3 -doped Ge-As-Ga-Se glasses in our previous work [16]; (iii) compositional fluctuation (including phase separation) in some sites of the host glass; (iv) fine Si-O/Pr-O-containing particles from rare earth ion corrosion during the chalcogenide glass melting in the silica-glass ampoule melt-containment. It is noted that the rare-earth-ion trichloride additives appear to exert stronger silica-glass ampoule corrosion than that of the rare-earth foil additive [18]. According to the Group of Churbanov, wavelength-dependent loss in the transparent window of chalcogenide glass fibers can be caused by excess scattering loss due to silica particles [42,43]. Even though, it should also be noted that in our previous work, a 500 ppmw Dy(DyCl 3 )-doped Ge-As-Ga 3 -Se bulk glass and fiber did not present such distinct NIR excess scattering loss [16,23]. It is possible that PrCl 3 is more aggressive than DyCl 3 in attacking the silica-glass ampoule, although the chemistry of the rare earth ions is usually similar, being dependent on inner f-electrons and shielded by outer electrons. The presence of 3 at % Ga (rather than 1 at% Ga here) in the previous case [16,23] might have helped to solubilize the DyCl 3 and reduce excess scattering loss.
In addition, we discount crystallization during reheating of the glasses to draw fiber as a main cause of scattering loss in the PrCl 3 -doped and the PrI 3 -doped fibers (Fig. 4). This is because: (i) both fibers were XRD amorphous (see Fig. 3) and (ii) the same NIR scattering occurred also in the as-prepared bulk glass fiberoptic preforms prior to the fiber drawing. Inset (a) of Fig. 5, gives the absorption coefficient (i.e. total loss coefficient) of the bulk glass of the fiberoptic preform of each of the PrCl 3 -doped and PrI 3 -doped as approximately 0.06 and 0.02 cm −1 (corresponding to 60 and 20 dB/m), respectively, higher than that of the Pr foil-doped bulk glass at 1550 nm wavelength. These values of optical loss are close to that given in Fig. 4 for a comparison of the fibers: the excess loss in the PrCl 3 -doped and PrI 3 -doped fibers, are also ~70 dB/m and 20 dB/m, respectively, compared to the loss of the Pr-foil fiber at 1550 nm wavelength. Thus, it is concluded that the fiber-drawing process did not produce distinct extra crystallization over and above the fiberoptic preform to contribute to the NIR scattering loss in the PrCl 3 -doped and PrI 3 -doped fibers. However, it cannot be ruled out that there was some crystallization in the fiberoptic preforms.
We know that 500 ppmw Pr doping in a Ge-As-Ga 1 -Se glass satisfies the recommended 10:1 ratio of the gallium: rare-earth in chalcogenide glass hosts [34,36]. However, in this work, we have found that only the 500 ppmw Pr foil appears to have been properly dissolved in the Ge-As-Ga 1 -Se host glass system. Therefore, importantly, the results of this work indicate that the ratio of gallium: rare-earth could depend on which rare earth additives are to be used. According to the order of excess scattering losses discussed above, the required Ga content to solubilize the particular form of rare earth ion additive is suggested in order of: Ga PrCl3 > Ga PrI3 > Ga Pr-foil . Figure 6 presents the PL spectra of the 500 ppmw Pr 3+ (foil), Pr 3+ (PrCl 3 ) and Pr 3+ (PrI 3 )-doped Ge-As-Ga 1 -Se fibers, on pumping at 1550 nm wavelength. From Fig. 5, the Pr 3+ absorption band of the PrCl 3 -doped chalcogenide fiber is of lower intensity than that of each of the Pr foil or PrI 3 -doped fiber. It is inferred that fewer Pr 3+ ions had successfully dissolved homogeneously in the PrCl 3 -doped chalcogenide glass; the knock-on effect is an expected lower Pr 3+ PL intensity for the PrCl 3 -doped fiber than the other two fibers, which is observed in Fig. 6. The only caveat is that this assumes a constant emission cross-section across the three types of Pr 3+ dopants. The structured shape of the Pr 3+ absorption bands and the reduction in Pr 3+ absorption band intensity observed in FTIR spectra of the PrCl 3 -and PrI 3 -doped glasses (see Fig. 5) indicate that there may be at least three population-types of the Pr 3+ ions. Firstly, there are well-behaved, ideal Pr 3+ ions which are coordinated by selenium and are bonded into the glassy network and homogeneously distributed and produce PL. Secondly there is a population of Pr 3+ ions which contribute to the PL and which exist as Pr 3+ (Cl) n and Pr 3+ (I) n halide-coordinated units in the host glass, where the first coordination sphere may be partially substituted with selenium and which may be either dissolved in the glass and distributed homogeneously or these units could be more ordered and/or clustered to explain the structuring of the Pr 3+ electronic absorption bands. The lowering of the absorption band intensity for Pr 3+ in the PrCl 3 -doped chalcogenide glass suggests that a third population of the praseodymium ions does not contribute to the absorption band at all, nor to the PL band, and may have precipitated out as oxide at the interface between the chalcogenide glass and the silica glass containment, as we found in our previously reported work, perhaps as oxide [18], although the preform surface was shiny under photo (see Fig. 1). The exact distribution of Pr-Cl/Pr-I/Pr-Se/Pr-O sites is unknown. It was found that the peak PL was repeatable at 4700 nm wavelength for all the three Pr 3+ -doped Ge-As-Ga 1 -Se fibers. Also, all the fiber PL spectra had dips at 4.5 μm wavelength due to -[Se-H] impurity underlying vibrational absorption and dips at 4.2 μm wavelength due to the external CO 2 absorption in the optical path of the PL set-up. There was a small PL peak found at 5750 nm wavelength in PrCl 3 -doped fiber; although it was noisy, the observation was repeatable and could be due the lower phonon energy local environment offered by Cl coordination of Pr 3+ which could encourage the inner, upper radiative transition: ( 3 F 3, 3 F 4 )→ 3 H 6 . The PL intensities at 4700 nm wavelength of the 500 ppmw Pr 3+ (foil), Pr 3+ (PrI 3 ) and Pr 3+ (PrCl 3 )-doped Ge-As-Ga 1 -Se fibers are shown as a function of pump power at 1550 nm wavelength in Fig. 7. A gentle sublinear behavior was observed for all fibers. This result is probably due to ground state bleaching [44].  The decays of the maximum intensity PL at 4700 nm wavelength in the 500 ppmw Pr 3+ (foil) and Pr 3+ (PrI 3 )-doped Ge-As-Ga 1 -Se fibers are presented in Fig. 8. It was found that two exponential decay lifetimes were required, of 10.1 ms and 3.3 ms to fit best for Pr(foil)-doped fiber and of 10.4 ms and 3.2 ms to fit best for the Pr(PrI 3 )-doped fiber. The inset of Fig. 8 of the Pr 3+ -ion simplified energy level diagram [45][46][47] shows that the shorter ~3 ms lifetime was due to the transition of ( 3 H 6, 3 F 2 )→ 3 H 5 and the ~10 ms lifetime to the transition of 3 H 5 → 3 H 4 ; a full energy level diagram of the Pr 3+ ion can be found in Weber's work [48]. From the fitting, the contributions to the steady state PL intensity at 4700 nm wavelength of 3 H 5 → 3 H 4 and ( 3 H 6, 3 F 2 )→ 3 H 5 transitions were suggested to be in a ratio of roughly 3: 1 or 3.5: 1 (judged by A n in the fit function, Fig. 8), respectively, for the Pr 3+ (foil) or Pr 3+ (PrI 3 )-doped fiber. For all the energy level diagrams presented in this work, the wavelength(s) next to the down arrow (4.0 µm, 4.8 µm and 5.2 µm) represents the nominal peak wavelength of each potential transition band [45][46][47]. Due to the thermally coupled, broad energy levels of ( 3 H 6, 3 F 2 ) and ( 3 F 3, 3 F 4 ), the transition bands of ( 3 H 6, 3 F 2 )→ 3 H 5 and ( 3 F 3, 3 F 4 )→ 3 H 6 are broad and have chance to contribute to the lifetime measurement at 4700 nm wavelength [46,49]. Fig. 8. PL decay and exponential fits of the 500 ppmw Pr 3+ (foil) and Pr 3+ (PrI 3 )-doped Ge-As-Ga 1 -Se fibers, at 4700 nm wavelength. Pump power was 65 mW, at 1550 nm wavelength. The inset presents the two lifetimes of 3 ms (( 3 H 6, 3 F 2 )→ 3 H 5 ) and 10 ms ( 3 H 5 → 3 H 4 ) in a Pr 3+ -ion simplified energy level diagram [45][46][47]. The wavelength of each down arrow is the peak wavelength of the potential PL band in each transition. Figure 9 presents the PL decay of the 500 ppmw Pr 3+ (PrCl 3 )-doped Ge-As-Ga 1 -Se fiber at the 4700 nm wavelength of maximum PL intensity. In contrast to the other two doped fibers, a rapid drop in PL intensity occurred in the first 1 ms of the Pr 3+ (PrCl 3 )-doped fiber PL decay (reproducible in the experiment), which indicates this decay region was dominated by a short lifetime. Three exponential decay lifetimes of 10.3 ms, 2.4 ms and 0.21 ms were required to best fit the PL decay of the Pr 3+ (PrCl 3 )-doped fiber, assigned to: 3 H 5 → 3 H 4 , ( 3 H 6, 3 F 2 )→ 3 H 5 and ( 3 F 3, 3 F 4 )→ 3 H 6 , respectively (see inset of Fig. 9), and it was suggested that these transitions contributed to the steady state PL intensity at 4700 nm wavelength in an approximate ratio of 4: 1: 1.3, according to the fit function (Fig. 9).

Fiber photoluminescent (PL) spectra and lifetimes, and Pr 3+ site variation
The local environments suggested for the Pr 3+ ions facilitating PL in the Pr 3+ (PrCl 3 )-doped Ge-As-Ga 1 -Se fiber are depicted in Fig. 10. Part of the Pr 3+ ions were coordinated with Se in local environment and when pumped at 1550 nm wavelength, the decay transitions of 3 H 5 → 3 H 4 and ( 3 H 6, 3 F 2 )→ 3 H 5 would be found and were observed but the ( 3 F 3, 3 F 4 )→ 3 H 6 transition would have mostly decayed non-radiatively (i.e. same as in Pr(foil)-doped fibers). Additionally, in the Pr 3+ (PrCl 3 )-doped fiber, as mentioned at least a portion of the Pr 3+ ions would have had a first coordination shell of chloride ions, which may or may not have been partially substituted with Se ions. These Pr 3+ centers may have been more ordered, may have clustered or may have formed small crystals. Nonetheless, this first coordination sphere of Cl (or Cl, Se) would have provided a local phonon energy to the Pr 3+ ions which was much lower than that of a first coordination sphere of Se alone, thus mediating the upper radiative decay in ( 3 F 3, 3 F 4 )→ 3 H 6 , which made the 0.21 ms lifetime observable. The evidence to support this hypothesis is as follows.
Firstly, as discussed, the PrCl 3 -fiber exhibited high NIR scattering loss, structured Pr 3+ absorption bands and a much lower PL intensity centered at 4700 nm than the Pr 3+ (foil)-doped fiber; these observations indicate that some of the PrCl 3 additive might not have dissolved properly in the Se-matrix (and the Pr 3+ bonded with Cl instead) . Secondly, in our previous work on DyCl 3 -doped Ge-As-Ga 10 -Se glasses, when the Dy 3+ concentration was <1000 ppmw, we concluded from the extended X-Ray absorption fine structure (EXAFS) [22] and also Beer-Lambert studies that the Dy 3+ fully had been incorporated into the glass network (Dy-Se coordination); yet when the Dy 3+ concentration was ≥ 1000 ppmw, Dy 3+ was shown to be in a dominantly crystalline Dy-Cl environment by the EXAFS [22]. This supports the hypothesis that a crystalline PrCl 3 environment was also possible in the Ge-As-Ga-Se glass system. Thirdly, lasing has been reported at 5.2 μm wavelength (( 3 F 3, 3 F 4 )→ 3 H 6 ) in a Pr 3+ -doped LaCl 3 crystal [50] and the host phonon energy was reported to be 210 cm −1 wavenumber [51]. In other words, the Cl-coordinated Pr 3+ local environment had very low phonon energy which would have encouraged the radiative transition of ( 3 F 3, 3 F 4 )→ 3 H 6 . As a comparison, a selenide-based chalcogenide glass is generally considered to have a phonon energy of ~350 cm -1 [52,53]. On the other hand, such Cl-coordinated sites did not happen in the Pr 3+ (foil)-doped fiber and so it is assumed that the selenide (and possibly oxide, hydroxide and -[Se-H]) coordination of Pr 3+ ions offered phonons of higher energy which could bridge the ( 3 F 3, 3 F 4 )→ 3 H 6 gap giving non-radiative decay mostly in this transition. Moreover, the Pr 3+ (Prl 3 )-doped glass exhibited a structured Pr 3+ absorption band (Fig. 5), which also indicates part of the PrI 3 did not dissolve properly in a Se coordinated environment. Residual iodide coordination of Pr 3+ in the Pr 3+ (Prl 3 )-doped glass would give an even lower local phonon energy than for chloride coordination and so one would expect to see again radiative decay due to ( 3 F 3, 3 F 4 )→ 3 H 6 and the 5750 nm wavelength PL band for the Pr 3+ (Prl 3 )-doped fiber as for the Pr 3+ (PrCl 3 )-doped fiber. However, it is known that iodine can take up the role of a chalcogen chain terminator as in: ≡Ge-S/Se-S/Se-S/Se-I or = As-S/Se-S/Se-I etc [54]. and this may be the reason why iodide was less inclined to remain coordinated to Pr 3+ . In addition, iodine is less electronegative than chlorine and so will have lowered philicity for the electropositive Pr than chlorine. Along with this, it was observed that, when compared to the PrCl 3 -doped glass/fiber, the intensities of the Pr 3+ absorption band and the PL band were distinctly higher in the PrI 3 -doped glass/fiber (see Fig. 5 and Fig. 6). This would indicate the amount of Pr 3+ coordinated with iodide in the local environment was much less and Pr-Se sites were greater in number in PrI 3 -doped fiber (compared to Pr-Cl in the PrCl 3 -doped fiber). A smaller quantity of Pr-I sites would not allow the experimental observation of the extra lifetime of ( 3 F 3, 3 F 4 )→ 3 H 6 and the 5750 nm wavelength PL band in the PrI 3 -doped fiber. Fig. 10. Schematic diagram of the proposed candidate environments of Pr 3+ ion facilitating PL in the 500 ppmw Pr 3+ (PrCl 3 )-doped Ge-As-Ga 1 -Se fiber: Se coordination of Pr 3+ ions and partial/full Cl coordination of Pr 3+ ions. In addition, simplified energy level diagrams of the Pr 3+ ions [45][46][47] reveal the PL radiative decay ( 3 F 3, 3 F 4 )→ 3 H 6 (lifetime 0.21 ms, Fig. 9) encouraged by the very low phonon energy of a chloride local environment of the Pr 3+ ions. Table 2 collates and summarizes the experimental Pr 3+ -doped Ge-As-Ga 1 -Se glass fiber emission lifetime results found here, together with the experimental results, and calculated results from the Judd-Ofelt (J-O) modelling, of comparable Pr 3+ -doped Ge-As-Ga-Se bulk glasses reported by Shaw et al. [47] and of fiber reported by Sójka et al. [46]. For the 3 H 5 → 3 H 4 transition, the experimental lifetimes found here for 500 ppmw Pr 3+ -doped Ge-As-Ga 1 -Se fibers supplied by Pr 3+ (foil), Pr 3+ (PrI 3 ) and Pr 3+ (PrCl 3 ) were of 10.1 ± 0.2 ms, 10.4 ± 0.2 ms and 10.3 ± 0.7 ms, respectively, to be compared with 10.0 ms [46] to 15.0 ms [47], estimated errors ~30% [47], from J-O modelling and a measured lifetime of 12.0 ms in an unspecified 'low concentration dopant' Pr 3+ -doped Ge-As-Ga-Se bulk glass [47] and 11.5 ms in a 115 mm long, 500 ppmw Pr 3+ (foil)-doped Ge-As-Ga-Se fiber collected at the end of the fiber [46]; note that end-collection will probably have resulted some re-absorption to cause radiation trapping and can present longer lifetime. For the ( 3 H 6, 3 F 2 )→ 3 H 5 transition, the 500 ppmw Pr 3+ (foil) and Pr 3+ (PrI 3 )-doped Ge-As-Ga 1 -Se fibers here exhibited experimental lifetimes of 3.3 ± 0.2 ms and 3.2 ± 0.2 ms, respectively, and the Pr 3+ (PrCl 3 )-doped Ge-As-Ga 1 -Se fiber gave a slightly shorter lifetime of 2.4 ± 0.7 ms, to be comparable with a calculated lifetime of 3.4 ms and a measured lifetime of 2.7 ms in the literature [1,47]. The extra radiative lifetime of 0.21 ± 0.06 ms observed only for the Pr 3+ (PrCl 3 )-doped Ge-As-Ga 1 -Se fiber here -assigned as ( 3 F 3, 3 F 4 )→ 3 H 6, had theoretical and experimental lifetimes reported as 0.29 ms and 0.25 ms, respectively, by Shaw et al [47]. However, unlike in the current work, Shaw  F 4 )→ 3 H 6 (0.21 ms compared to 0.25 and 0.29 ms) (although the bigger error in lifetimes in the PrCl 3 -doped fiber should be noted due to a lower signal-to-noise ratio), which is a hint of lifetime-quenching. Any lifetime-quenching, if it occurred, may have been due to Pr-Cl clustering in the PrCl 3 -doped fiber. On the other hand, because lifetime quenching was not really significant in the PrCl 3 -doped fiber, it suggests that the majority of the Pr-Cl sites were distributed more evenly instead of any clustering of Pr 3+ . In Table 2, the lifetimes of PrI 3 -doped fiber are found to be comparable with the values of the Pr-foil doped fiber, which indicates no distinct Pr 3+ ion clustering had occurred in the PrI 3 -doped fiber. This chimes with the discussion above that the sites of Pr-I coordination in the PrI 3 -doped fiber were lower in number density than the Pr-Cl sites in the PrCl 3 -doped fiber. Thus, the amount of Pr-I clustering would also have tended to be less.  [2], respectively, used in published modeling work. This indicates that the Pr 3+ -doped glass fiber system of this work has the potential to be developed into fiber amplifiers and/or lasers. However, the optical fiber loss of this work (~30 dB/m at 4.5 µm wavelength, due to the Se-H contamination band) is much higher than the losses used in the modelling work, where the optical loss assumed is generally 1 dB/m [1,29,31,32], or 3-7 dB/m [2,31,32] for signal wavelengths of 4.3-5.0 µm. The optical loss is a key issue that presently prevents the realization of MIR fiber amplification and lasing in chalcogenide glasses, and is an important research challenge to be solved.

Conclusions
The work presents a comparative study of 500 ppmw Pr 3+ (introduced by Pr foil, PrI 3 or PrCl 3 additive) doped Ge-As-Ga 1 -Se glass fibers. No crystallization peak was found in powder XRD patterns of any of the Pr 3+ -doped fibers. Fiber loss spectra showed that the Pr(foil)-doped fiber had the lowest overall background optical loss of ~2.0-2.5 dB/m across the 2-9 μm window. Although the PrCl 3 -doped fiber presented excess NIR scattering loss below 4 μm wavelength, it gave the lowest loss of all doped fibers here of 1.8 dB/m at 6.9 μm wavelength. Excess NIR scattering loss was also observed in the PrI 3 -doped fiber below 4 μm wavelength but lower than that of the PrCl 3 -doped fiber. FTIR spectra of the preform bulk glass (i.e. the glass before fiber drawing) showed that most of the extra NIR scattering loss in the PrI 3 -doped and PrCl 3 -doped fibers was already in the preforms due to RE additive solubility and not caused by crystallization from fiber-drawing; in addition, structured Pr 3+ absorption bands were present in FTIR spectra of the PrI 3 -and PrCl 3 -doped Ge-As-Ga 1 -Se preforms.
PL spectra and lifetimes under 1550 nm pumping of the Pr 3+ -doped Ge-As-Ga 1 -Se fibers showed reproducibly that PL intensity at 4700 nm wavelength was in the order: Pr(foil)-doped fiber > PrI 3 -doped fiber > PrCl 3 -doped fiber. The PL intensity at 4700 nm wavelength of all doped fibers, increased with a gentle sublinear behavior (proposed due to ground state bleaching), as incident pump power increased. The Pr 3+ -doped fibers had radiative lifetimes of 10.1-10.4 ms for the transition of 3 H 5 → 3 H 4 , and 2.4-3.3 ms for the transition of ( 3 H 6, 3 F 2 )→ 3 H 5 . The local Pr 3+ ions environment in the PrCl 3 -doped fiber was suggested to have at least a portion of Pr-Cl to account for the extra 0.21 ms radiative lifetime found for the transition of ( 3 F 3, 3 F 4 )→ 3 H 6 , but not observed in the Pr(foil) and PrI 3 -doped fibers. The content of Ga required to solubilize the same amount of Pr 3+ ions is suggested to be in order of: Ga PrCl3 > Ga PrI3 > Ga Pr-foil , and future work can be carried out to study the optimum Ga ratio for REI 3 and RECl 3 dopants, along with future investigation by glass structural experiments (e.g. high intensity X-ray diffraction) and a thermal analysis with microscopy study. Also, the preparation of high purity rare earth ion doped chalcogenide glass fiber continues to be a vital research direction.