1. Introduction

The mid-infrared (MIR) spectral region, spanning 3-25 µm, encompasses the fundamental vibrational absorption bands of molecular solids, liquids and gases. Direct emission, bright MIR fiber lasers, of good beam-quality based on rare earth (RE-) ion doped glass fiber, promise direct molecular sensing e.g. for real-time monitoring for long-range-security (narcotics, explosives), the environment (greenhouse gases) and energy-efficiency (CO/CO₂), as well as the potential for: cutting and welding of soft materials including human tissue (e.g. new wavelengths for medical laser surgery), optical countermeasures in the 3-5 µm and 8-12 µm atmospheric windows and pumping MIR supercontinuum generation for wideband, MIR molecular sensing for in vivo early cancer diagnosis [1,2]).

Rare earth (RE-) ion doped silica-glass fiber lasers operating in the near-infrared (NIR) are of widespread utility for cutting and welding hard materials (e.g. metals) with kWatt power availability and excellent beam-quality. Silica glass solid-core fibers are opaque beyond 2 μm wavelength. So to realize MIR RE-ion fiber lasers demands lower phonon energy glass hosts. To date the longest wavelength recorded fiber-lasing is at 3.9 μm wavelength in a cooled Ho³⁺-doped ZBLAN fiber [3]. However, the relatively high phonon energy of the ZBLAN fluoride glass host precludes any longer wavelength operation.

Chalcogenide glasses are attractive hosts, having low phonon energy, down to 250 cm⁻¹ (for promoting efficient MIR RE-ion radiative processes) and exhibiting large linear refractive indices (for large RE-ion absorption and emission cross-sections) [4–9]. Heo and Shin [10] have demonstrated that, for similar electronic transitions, the multi-phonon relaxation rates in sulfide chalcogenide glasses are approximately ¹/₃ to ¹/₁₀ of those calculated for ZBLAN glasses. We have selected here a Pr³⁺-ion doped selenide chalcogenide glass matrix. The electronic energy levels of Pr³⁺ are tightly spaced (< 2500cm⁻¹), therefore the host glass must be of very low phonon energy to avoid multiphonon bridging of the levels i.e. non-radiative decay [5]. The heavier, more weakly bonded, selenide matrix reduces multi-phonon relaxation rates of excited doped-in RE-ions. Moreover, the selenides exhibit higher refractive indices than the sulfides, for greater RE-ion radiative efficiency. Finally, chalcogenide glass fibers are sufficiently mechanically and chemically robust for development [4,11].

To achieve MIR fiber lasing, fiber quality is key. Impurity absorption directly, or indirectly, resonant with the pump, or emission, wavelengths will affect fluorescence intensity and lifetime [8]. In addition, during the processing to make fiber, crystallization of the supercooled glass-melt must be avoided [8,12,13].

Most photoluminescence (PL) studies to date on Pr³⁺ doped chalcogenides have focused on NIR Pr³⁺ emission in both bulk glasses [14,15] and fiber [16], for instance for developing all-optical amplification for telecoms. at 1.3 µm. At longer wavelengths, in the MIR, numerical modeling and experimental measurement of Pr³⁺ PL in chalcogenide host bulk glasses [5,6,17,18] and optical fibers [5,6,19–23] have been previously reported, as follows.

MIR fluorescence up to 5.5 µm has been demonstrated by Shaw et al. [18] in 0.2 wt% (weight %) Pr³⁺ doped Ba-In-Ge-Ga-Se bulk glass; samples were pumped at 1.064 µm with a continuous wave (CW) Nd:YAG laser. Churbanov et al. [17] have reported MIR absorption spectra of Pr³⁺-ion doped As-S-Se-I bulk glasses, but did not measure the Pr³⁺ photoluminescence nor draw fiber. Cole et al. [20] have reported broad MIR fluorescence in the range 3 - 5 µm in core/clad., multimode, selenide fiber pumped with a 2 μm laser diode: the fiber core glasses were GeAsGaSe doped with 200 ppm (parts per million) Pr³⁺and 350 ppm Pr³⁺. Park et al. [5] also have measured broad MIR fluorescence in the range 3.5-5.5 µm, for a series of Pr³⁺ doped GeGaSbSe bulk glasses, and in 0.02 mol% (sic) Pr³⁺- doped GeGaSbSe □ber, which was ~150 mm long, excited with a 1.48 µm laser diode or 2.05 µm fiber laser [5]. Charpentier et al. [21] reported broad MIR fluorescent spectra from 3.5 – 5.5 μm wavelength in unclad Ga₅Ge₂₀Sb₁₀Se₆₅ (at% (atomic %)) fiber doped with 0.05 wt% and 1 wt% (i.e. 500ppmw and 1000 ppmw (ppm by weight), respectively) Pr³⁺ ions and pumped with a homemade Tm:YAG laser.

All of these Authors measuring the PL behavior of Pr³⁺-doped selenide glass fiber [5,20,21] added gallium to the glass formulation. This follows the work of Aitken et al. [24] who investigated clustering of rare earth ions in GeAs sulfide glasses by carrying out both Pr³⁺ rare earth (RE) ion fluorescence studies and EPR (electron paramagnetic resonance), on such glasses, with Gd³⁺ RE-ions added as a probe, and in the presence of an additive, taken from: Ga or In (or P or Sn). They found that Ga co-doping greatly enhanced rare earth ion solubility and dispersal particularly for Ga:RE-ion ratios of ≥ 10:1, as evidenced by the narrower Gd³⁺ EPR resonance and more intense photoluminescence (PL) of the Pr³⁺ doped bulk glasses. Thus 0.2 at% Ga added to 450 ppm Pr³⁺ doped GeAsS bulk glasses increased the PL ~10-fold. To explain this phenomenon, Aitken et al. [24] have suggested that Ga complexes the RE-ion in the presence of chalcogen forming ([RE³⁺] - S - [Ga^(III)]) type species in the sulfide glasses. Indium (or P or Sn) was also observed to ‘decluster’ rare earth ions, although “less efficiently” than Ga. No mention was made in this work by Aitken et al. [24] about comparison of the PL intensity of Pr³⁺ in the presence of In rather than Ga.

We have previously numerically modeled MIR laser emission in Pr³⁺-doped gallium-containing selenide chalcogenide glass fiber [19] which, assuming a fiber loss ≤ 1 dB/m at both pump and emission wavelengths and Fresnel end-reflection within the laser cavity (due to the large chalcogenide glass refractive indices), was shown potentially to give efficient MIR fiber lasing in the 4 to 5 μm region, based on ³H₅ → ground state emission. We selected the well-known GeAsSe ternary system, in which may be found many stable glass formulations and added Ga to promote dispersal, and enhance PL, of the RE-ion. Initially, we observed weak PL in Pr³⁺-GeAsGaSe bulk glasses and fiber [22]; for the results presented in this paper the collection geometry has since been improved. Both Ga and In are in Group XIII of the Periodic Table and so exhibit the same outer electronic configuration and hence similar chemical philicity for solvating RE-ions. Indium is heavier than gallium, thereby potentially promoting a lower phonon energy in the local environment of the RE-ion dopant, for improved radiative efficiency. We can find little mention in the literature of indium addition in the presence of a RE-ion dopant in chalcogenide glasses [25,26].

We have previously reported initial studies on the PL intensity in Pr³⁺-doped GeAsInSe glasses and fiber [23]. We present here the first comprehensive comparison, to our knowledge, of PL intensity and excited state lifetimes of indium-containing Pr³⁺ doped- selenide, glasses and fibers, with their gallium analogues.

In this paper: 0 to ~2000 ppmw Pr³⁺-doped GeAsGaSe bulk glasses (Ga-series) and 0 to ~2000 ppmw Pr³⁺-doped Pr³⁺-doped GeAsInSe bulk glasses (In-series) were prepared by melt-quenching. For each bulk glass, the level of Ga or In added was 1 atomic%. Also, optically-clad, multimode, step index fiber (SIF) was drawn from extruded [27] fiber-optic preforms. SIF based on a ~500 ppmw Pr³⁺- doped GeAs-(In or Ga)-Se core and a GeAs-(In or Ga, as core)-Se cladding are termed, respectively, Pr³⁺-Ga-SIF or Pr³⁺-In-SIF. Again, for each fiber, the level of Ga or In added was 1 atomic%. Both the In- and Ga-series bulk glasses were fully characterized for glass quality and optical loss was measured of the Pr³⁺-In-SIF and Pr³⁺-Ga-SIF. The absorption and emission spectra of all the Pr³⁺ bulk glasses and SIF fibers were collected, and compared. The excited state lifetimes in the Pr³⁺-doped In-series, and Pr³⁺-doped Ga-series, bulk-glasses were also compared. We report here for the first time circa doubling of the emission intensity across the 3.5 to 6 μm MIR region, and an increase in the excited state lifetime at 4.7 μm, in Pr³⁺-doped selenide-chalcogenide bulk glasses with indium addition, rather than gallium, when pumped at 1.55 μm.

2. Experimental

Bulk glass and fiber preparation

2.1.1 Bulk glass preparation

Two series: zero, 493, 994, 1476 and 2050 ppmw Pr³⁺-doped GeAsGaSe bulk glasses and zero, 527, 1027, 1473 and 2034 ppmw Pr³⁺-doped GeAsInSe bulk glasses were prepared, using commercially available elemental precursors (Ge, Ga, As, Se (all 5N to 7N5) and Pr (3N)). The host glasses GeAsInSe and GeAsGaSe were formulated to be identical in at% ratio to allow comparison of the radiative behavior of equivalently (as far as possible) Pr³⁺-doped (ppmw) samples, assuming negligible density difference for the small amounts of Ga, In and Pr³⁺ added. For each bulk glass and fiber, the 1at% Ga or 1at% In was incorporated. The accuracy of Pr doping was only known at the batching point. The praseodymium was weighed as accurately as possible to ± 0.0005 g inside an MBraun Glovebox (MBraun: < 0.1 ppm H₂O and < 0.1 ppm O₂), giving an error of a few percent maximum on the Pr amounts weighed for making bulk glass and fiber. Precursors were batched in the MBraun glovebox into purified (air-baked then vacuum-baked, each at 900°C/6 h) silica-glass ampoules. The ampoules were sealed under vacuum (10⁻³ Pa) using an oxy-propane flame. The selenide glasses were melted and homogenized at 850°C for 12-14 hours in a rocking furnace (adapted Instron, TF105/4.5/1ZF, tube ID 86mm), held vertically without rocking for refining, quenched and then annealed at the onset glass transition temperature (T_g) [28].

2.1.2 Fiber preparation

We selected analogous compositions of the core and cladding glasses for the ~500 ppmw Pr³⁺-doped Ga-SIF and ~500 ppmw Pr³⁺-doped In-SIF, with a 1 atomic% (at%) difference between the core/clad. glasses in each case. The refractive indices of the 493 ppmw Pr³⁺-GeAsGaSe core glass (n_core) and GeAsGaSe cladding glass (n_clad.) were measured using a Woollam UVU VASE rotating analyzer ellipsometer. The numerical aperture (NA = (n_core² – n_clad²)^1/2) of the 498 ppmw Pr³⁺: GeAsGaSe / GeAsGaSe fiber was calculated to be 0.16 at 4.7 µm wavelength.

Core and cladding glass boules were co-extruded in an in-house extruder [27] and the core/clad. fiber-optic preform was drawn to ~300 µm OD (outside diameter) on a customized Heathway fiber drawing tower in a class-10,000 cleanroom. Optical, and electron, microscopic imaging of fibers was unable to distinguish the core/clad. interface due to the close composition of the core and cladding glasses, but from our other work (including [27]) we expected the fiber core diameter to be 80-90% of the fiber OD giving a ~15-30 μm optical cladding depth.

Bulk glass characterization: amorphicity, MIR transmission and T_g

Amorphicity was determined using powder X-ray diffraction (XRD), of samples taken from the thermal-contraction-cone (located at the top of the annealed glass rod), which was likely to be the most impure part of the glass rod due to the refining steps after glass-melting, before annealing. XRD patterns were collected under ambient conditions using a Siemens Krystalloflex 810, with CuKα radiation, from 10 to 70 °2θ; diffraction angles were calculated after the subtraction of K_α2 diffraction.

NIR and MIR absorption spectra were collected for chalcogenide glass bulk samples polished to a 1 μm finish, with their two optical faces both flat and parallel and an optical pathlength of 2.5 - 3.5 mm, using a Fourier transform infrared (FT-IR) spectrometer (Bruker IFS 66/S) under ambient conditions.

The onset-T_g [28] was measured as the extrapolated onset of the T_g feature using differential scanning calorimetry (DSC, PE Pyris-1 equipment) run on 1-2 mm glass chunks inside a sealed Al pan heated at 10°C min⁻¹ under flowing Ar, against an empty Al pan reference. To ensure a comparable thermal history (which affects T_g), samples were prior heat-treated in the DSC equipment: first heating at 10°C min⁻¹ to 20°C above T_g, then cooling down to ambient at 10°C min⁻¹, before the actual DSC run.

Optical loss of fiber and PL emission of fiber and bulk glass

Loss of both the ~500 ppmw Pr³⁺-Ga-SIF and the ~500 ppmw Pr³⁺-In-SIF was measured using the cut-back method [29] by means of an IFS 66/S, Bruker FT- (Fourier-transform)-IR spectrometer. A W source, CaF₂ beamsplitter and InGaAs, and InSb, detectors were used for the 1-3 μm waveband. A Globar source, KBr beamsplitter and MCT (mercury-cadmium-telluride) detector were used for 2.75-10 μm. Around 2 m of each fiber was used for the loss measurement. Two groups of fiber cleaves were made before and after the long piece cut-back (3-4 good cleaves in each group). All fiber-cleaves were imaged using optical microscopy during the fiber loss measurement, to check measurement reliability. For the calculation of the final loss spectrum, the best cleave in each group was chosen by identifying the two cleaves with highest intensity output, based on the corresponding length of fiber [Fig. 1]. The error in measured loss spectra was estimated to be < 0.2 dB/m. This estimation of error was the result of many measurements, which lends confidence. No mode-stripping was applied to the fiber optical loss measurement which will have led to an offset error in loss.

 

Fig. 1 Optical micrographs of a “best cleave” for the cut-back measurement of the optical loss spectrum of: (a) the 498 ppmw Pr³⁺-In-SIF and (b) the 493 ppmw Pr³⁺-Ga-SIF.

Download Full Size | PDF

For PL emission, all bulk glasses and fibers were each pumped at 1.55 µm wavelength, using a 100 mW fiber-coupled single-mode, CW laser diode (Thorlabs FPL 1009S). We used a constant input power of 50.2 mW to pump the bulk glasses and 24.5 mW to pump the fibers. The laser temperature was controlled using a Peltier cooler driver (Thorlabs CAB420-15). The fluorescence signal exiting the bulk-glass, or fiber, was focused onto a slit of width 2.5 mm before being passed through a motorized monochromator (Spex MiniMate) with a diffraction grating blazed at 6 µm (51034 JobianYvon). The signal was then detected using an ambient MCT detector (Vigo System PVI-6: 2 - 6 μm) to give the relative spectral response. Data-collection was via LabView-based software: bulk glass PL data was collected over a long time period of ~13 hours, with 100 s averaging constant and 10 µm step size; however fiber PL data was collected over shorter time periods of ~3 hours due to the much stronger signal recorded. The bulk glass and fiber PL measurements were corrected for the system response, which was measured using a blackbody. The fluorescence spectra were measured under ambient atmospheric conditions; hence, CO₂ and H₂O in the optical path contributed (unwanted) extrinsic absorption bands in the 2 - 5 µm wavelength regime.

To investigate the PL behavior of the ~500 ppmw and ~1000 ppmw Pr³⁺-doped Ga- and In-series of bulk glasses, the pumping face of each bulk glass sample was ground and then polished to a ¼ µm finish and the collection face of each sample was ground to be orthogonal to the pumping face and then also polished to a 1 µm finish. These two mating faces exhibited the sharpest possible (i.e. non-rounded) 90° mating-edge. To minimize self-absorption of the PL, the pump laser was carefully positioned so as to be incident at ≤ 0.5 mm from the mating-edge when fired into the polished input face of the sample. The PL was collected ≤ 2 mm from the mating-edge and out of the polished collection face, orthogonal to the input face. We endeavored to avoid collecting higher-orders of the pump laser light, originating from the grating, i.e. of wavelength nλ (where λ was 1550 nm, n is an integer). We also made every effort to avoid collecting first, or n^th-order, pump light, or PL light that had undergone random total internal reflection at surfaces inside the sample. A lock-in technique, using a mechanical chopper and lock-in amplifier, was used to process the PL signal; PL measurements took place in the range 1.5 - 6 µm and no optical filters were used. The collection optics was modified for the fiber PL: a 2.3 µm long-pass interference filter was used to eliminate the collection of the pump laser light at the exit-end of the fiber. A lock-in technique was again used to process the PL signal. No mode stripping was done.

The fluorescent decay time of the 4.7 μm emission was measured at 293K using direct modulation (8-10 Hz) of the pump laser diode at 1550 nm wavelength, via the motorized monochromator (Spex MiniMate) with a diffraction grating blazed at 6 µm (51034 JobianYvon). The lifetime decay was detected using an ambient MCT detector (Vigo System PVI-6). A digital 250 MHz PC oscilloscope (Picoscope5204 PicoTechnology) was used to analyze the signal: ~2000 points at 14 bits resolution with 10,000 averages were applied to reduce the noise. No filters were required for the lifetime decay measurements on bulk glasses. Further details about the collection geometry are given in section 3.4.

3. Results and discussion

3.1 Bulk glass characterization: amorphicity and absorption

Figure 2 shows the XRD patterns of zero to ~2000 ppmw, Pr³⁺ -doped Ga-series and Pr³⁺-doped In-series, respectively. From Fig. 2, the Ga-series glasses were amorphous up to, and including, ~1500 ppmw Pr³⁺, whereas the In-series glasses were amorphous only up to, and including, ~1000 ppmw Pr³⁺. In each case the crystallized phase was Pr ^(III) Se_1.9 [30].

 

Fig. 2 XRD patterns of: (a) zero, 493, 994, 1476 and 2050 ppmw Pr³⁺-Ga series bulk glasses, suggesting that the zero - 1476 ppmw Pr³⁺-Ga bulk glasses were amorphous and (b) zero, 527, 1027, 1473 and 2034 ppmw Pr³⁺-In series bulk glasses, suggesting that the zero - 1027 ppmw Pr³⁺ -In bulk glasses were amorphous. Patterns have been vertically shifted and sharp Bragg peaks associated with crystallization circled for convenience.

Download Full Size | PDF

FT-IR absorption spectra of the undoped GeAsGaSe and GeAsInSe bulk-glass hosts, in the 2 - 11 µm wavelength range [Fig. 3] exhibit characteristic vibrational absorption bands due to (unwanted) extrinsic impurities in the glass including: [H-Se], [H-O-H], [Ge-O] and [As-O, Si-O] at 4.5, 6.3, 7.8, 8.9 and 9.6 - 9.8 (low intensity) µm wavelength, respectively. Despite purging, there was atmospheric absorption in the optical-path: this was responsible for vibrational absorption at 4.2 µm ([O = C = O stretch of CO₂) and at least part of the vibrational absorption at 6.3µm ([H-O-H bend of H₂O) and 2.9 μm (O-H stretch of H₂O).

 

Fig. 3 FT-IR absorption coefficients in the 2 - 11 µm wavelength-range of undoped GeAsGaSe and GeAsInSe bulk, host-glass showing impurity vibrational absorption in the optical window. The sloping baseline trend mainly reflects the surface finish of the glasses; the significance of surface-defect scattering diminished with increasing wavelength. Despite purging, atmospheric vibrational absorption in the optical-path was responsible for absorption at 4.2 µm [CO₂] and at least part of the absorption at 2.9 μm [H₂O] and 6.3 µm [H₂O]. The proposed band identities are indicated; bands marked ‘?’ are still under investigation.

Download Full Size | PDF

Figure 4 presents the electronic energy level diagram of an isolated Pr³⁺ ion and shows the MIR absorptive transitions of interest here, which were identified from [31] and are in broad agreement with [18], and potential longer-wavelength radiative transitions. Figure 4 suggests that the Pr³⁺: GeAs-(Ga/In)-Se glasses and fibers may be pumped at ~1.5 µm or ~2 μm, allowing pumping to take place with commercially available sources [22].

 

Fig. 4 A simplified energy-level diagram of isolated Pr³⁺ ions with absorptive transitions of interest here, which were identified from [31] and are in broad agreement with [18], together with potential longer-wavelength radiative transitions.

Download Full Size | PDF

The absorption spectra of the 493 ppmw Pr³⁺ -Ga and 527 ppmw Pr³⁺ -In bulk glasses, in the wavelength-region 1 – 5.5 µm [Fig. 5(a)] show the main electronic absorption bands [31] due to Pr³⁺ at ~1.48, 1.57, 2.0, 2.33 and 4.5 µm, corresponding to ground-state absorption: ³H₄ to the excited-state levels ³F₄, ³F₃, ³F₂, ³H₆ and ³H₅, respectively [Fig. 4]. Again, atmospheric absorption was encountered at 2.9, 4.2 and 6.3 μm [as in Fig. 3]. [H-Se] impurity in the glass exhibited vibrational absorption at 3.5 µm and 4.6 μm [32]. Obeyance of the Beer-Lambert law was checked for the Pr³⁺-Ga and Pr³⁺-In bulk glass series [Fig. 5(b)].

 

Fig. 5 (a) FT-IR spectra of 527 ppmw Pr³⁺-In and 493 ppmw Pr³⁺ -Ga bulk glass samples, showing the 1.48, 1.57 and 2.0 µm Pr³⁺ electronic absorption bands due to ground state absorption to levels ³F_4, ³F₃ and ³F_2, respectively [Fig. 4]. The spectral baselines were numerically removed. (b) Beer-Lambert plots of the 1.48, 1.57 and 2.0 µm Pr³⁺ electronic absorption bands from the ground state for the zero to 1476 ppmw Pr³⁺-Ga series glasses and zero to 1473 ppmw Pr³⁺-In series glasses. Note: (i) the 1.48 µm and 1.57 µm bands overlapped and were deconvoluted arbitrarily at their point of intersection; (ii) the 4.5µm band [Fig. 4(a)] was neglected due to a varying [H-Se] impurity vibrational absorption contribution [c.f. Fig. 3] and (iii) maximum errors were: ± 30 ppmw and ± 0.0003 cm⁻¹ μm.

Download Full Size | PDF

From Fig. 5(a), both the 493 ppmw Pr³⁺-Ga and 527 ppmw Pr³⁺-In bulk glasses exhibited similar absorption band-shapes. The 4.6 μm H-Se absorption overlapped with the Pr³⁺ electronic absorption band at 4.5 µm. The 527 ppm Pr³⁺-In bulk glass exhibited more absorption at 4.6 µm than the 493 ppmw Pr³⁺ -Ga bulk glass, probably due to differing amounts of [H-Se] impurity in the glasses.

Beer-Lambert plots of the Pr³⁺ electronic absorption bands at 1.45, 1.60 and 2.04 µm wavelength, respectively, for the zero to 1476 ppmw Pr³⁺ -Ga and zero to 1473 Pr³⁺ -In bulk glasses are shown in Fig. 5(b). The 1.45 µm and 1.60 µm bands overlapped [see Fig. 5(a)] and so were de-convoluted arbitrarily at their point of intersection. The Beer-Lambert plot of the 4.5 µm Pr³⁺ electronic absorption band was not included in Fig. 5(b), due to the varying [H-Se] extrinsic vibrational absorption contribution to this band, already noted from Fig. 3.

From Fig. 5(b), the Beer-Lambert Law was obeyed for the 1.45, 1.60 and 2.04 µm Pr³⁺ absorption bands for the zero to 994 ppmw Pr ³⁺ -Ga and zero to 1027 Pr³⁺-In bulk glasses. However, at >1000 ppmw Pr³⁺-doping in the In-series bulk glasses, the fall-off from linearity suggested that the solubility-limit of Pr³ had been exceeded in the In-series, under the current conditions (note that Fig. 2 supports this proposal showing crystallization of PrSe_1.9 at the higher dopant levels). Table 1 shows the correlation coefficients of the Beer Lambert plots presented in Fig. 5(b). Table 1 is important because it shows that the correlation coefficients of the Beer Lambert plots were close to 1 for the zero to ~1000 ppmw Pr³⁺ doping levels. This adherence to linearity lends confidence to the use of the as-batched ppmw values of Pr, throughout this work. Note also that for the Pr³⁺-Ga bulk glasses, the linearity of the Beer Lambert plot remained good to the ~1500 ppmw Pr³⁺-doping level.

Table 1. Correlation coefficients (R²) of the Lambert-Beer Law line fit shown in Fig. 5(b) for zero to 994 Pr³⁺ - GeAsGaSe and zero to 1027 Pr³⁺ -GeAsInSe bulk glasses.

View Table

3.2 Optical loss of fiber

Figure 6 shows the fiber optical loss spectra of the 493 ppmw Pr³⁺ -GeAsGaSe / GeAsGaSe SIF [22] and 498 ppmw Pr³⁺ -GeAsInSe / GeAsInSe SIF [23] across the spectral range 1 – 9 µm.

 

Fig. 6 Fiber optical loss spectrum of 493 ppmw Pr³⁺-Ga-SIF and 498 ppmw Pr³⁺-In-SIF, in the range 1 – 9 µm. Fiber OD was ~300 µm and there was a 15-30 µm thick optical cladding.

Download Full Size | PDF

From Fig. 6, the Pr³⁺-Ga-SIF exhibited a minimum loss of 2.8 ± 0.2 dB/m at 6.58 µm whereas the Pr³⁺-In-SIF exhibited a minimum loss of 4.3 ± 0.2 dB/m at 6.58 µm. Both spectra followed similar behavior in the wavelength-range 1 – 9 µm except that the [H-Se] vibrational absorption was more intense for the Pr³⁺-In-SIF, as found also when comparing the bulk glass absorption in Fig. 3.

3.3 PL emission of fiber and bulk glass

Fig. 7(a) shows the MIR PL emission spectra, from 3 – 6 μm wavelength-range of the 493 ppmw Pr³⁺ -Ga and 527 ppmw Pr³⁺ -In bulk glasses, pumped at 1550 nm / 50.2 mW. Fig. 7(b) shows the normalized PL spectra.

 

Fig. 7 PL emission of bulk glasses: (a) 493 ppmw Pr³⁺ -Ga and 527 ppmw Pr³⁺-In and (b) as (a) but with both plots normalized at their peak to 1.

Download Full Size | PDF

From Fig. 7(a), the ~500 ppmw Pr³⁺-In bulk glass exhibited circa double the peak PL intensity at 4700nm wavelength (15.69 mV on MCT detector proportional to intensity, 2.19 x) of the ~500 ppmw Pr³⁺-Ga bulk glass (7.15 mV).

From the normalized PL spectra [Fig. 7(b)], replacing Ga with In in the ~500 ppmw Pr³⁺ -doped bulk glasses did not appear to affect the emission band-shape at the shorter wavelength edge but the band-shapes at the long wavelength edge were slightly disparate. The underlying reasons for this may be that, despite the similarity of sample size, shape and finish and optical path, controlling the MIR PL experiment is challenging, for instance in avoiding unwanted internal Fresnel reflection as mentioned above. The obvious dip at 4.2 µm wavelength on the PL spectra of both doped bulk glasses presented in Fig. 7 is due to the vibrational absorption of CO₂ which was present in the optical path and found to be present in all emission spectra presented hereon in. The smaller dip at 4.6 μm, we believe, was due to unwanted resonant vibrational absorption of impurity [H-Se] in the glass [5,6,20,21,32]. The PL signal from the 493 ppmw Pr³⁺-Ga bulk glass was noisier than the PL signal from the 527 ppmw Pr³⁺-In bulk glass due to the weaker overall collected signal intensity from the former.

Figure 8(a) shows the PL intensity of 994 ppmw Pr³⁺-Ga and 1027 ppmw Pr³⁺-In bulk glasses and Fig. 8(b) shows the normalized PL spectra.

 

Fig. 8 PL emission spectra of bulk glasses: (a) 994 ppmw Pr³⁺-Ga and 1027 ppmw Pr³⁺-In bulk glasses and (b) as (a) but peaks normalized to 1.

Download Full Size | PDF

The intensity of each of these ~1000 ppmw Pr³⁺-doped glasses was more than double the intensity of the analogous ~500 ppmw Pr³⁺-doped Ga- and In- bulk glasses [compare Figs. 8(a) and 7(a)]. The ~1000 ppmw Pr³⁺ -In bulk glass exhibited slightly less than double the PL intensity (37.00 mV on the MCT detector, 1.87x, at the maximum wavelength of 4700 nm) of the analogous ~1000 ppmw Pr³⁺ -Ga bulk glass. The normalized PL spectra of the ~1000 ppmw Pr³⁺ -Ga and ~1000 ppmw Pr³⁺-In bulk glasses [Fig. 8(b)] exhibited a similar band-shape, except for a slight difference at wavelengths > 4800 nm.

Figure 9 shows the normalized MIR PL spectrum of a 110 mm length of 498 ppmw Pr³⁺ -In-SIF and a 115 mm length of the 493 ppmw Pr³⁺ -Ga-SIF. The Pr³⁺-In-SIF exhibited a greater proportion of the PL intensity beyond 4800 nm compared to the Pr³⁺-Ga-SIF.

 

Fig. 9 Normalized PL spectra of 493 ppmw Pr³⁺-GeAsGaSe / GeAsGaSe SIF and 498 ppmw Pr³⁺-GeAsInSe / GeAsInSe SIF, each pumped at 1550 nm.

Download Full Size | PDF

Absorption and emission spectra of the ~500 ppmw Pr³⁺-Ga bulk glasses are presented in Fig. 10(a).The overlap region of the bulk glass absorption and the bulk glass emission indicates the potential for re-absorption of the emitted light. Also presented in Fig. 10(a) is the emission in the ~500 ppmw Pr³⁺-Ga-SIF. From the band-shapes it is evident that in the fiber, relative to the bulk glass, more re-absorption has taken place. Re-emission is evidenced by the greater spectral band area at longer wavelengths compared to that of the bulk glass.

 

Fig. 10 (a) Overlap of the absorption and emission bands at 4680 nm band for the 493 ppmw Pr³⁺ -Ga bulk glass and emission of 493 ppmw Pr³⁺-Ga-SIF. The absorption and emission bands have been normalized to give a peak response of 1. (b) Similar overlap of the absorption and emission bands at 4680 nm for the 527 ppmw Pr³⁺-In bulk glass together with emission from 498 ppmw Pr³⁺-In-SIF. The absorption and emission bands have been normalized to give a peak response of 1.

Download Full Size | PDF

Similarly, in Fig. 10(b), bulk glass absorption and emission of the ~500 ppmw Pr³⁺-doped indium bulk glasses along with emission of ~500 ppmw Pr³⁺-In-SIF are presented. Again, enhanced re-absorption at the shorter wavelengths and re-emission at the longer wavelengths was observed for the fiber relative to the bulk glass.

3.4 PL lifetimes of the bulk glasses

We measured the PL intensity at the 4.7 μm peak, within the 3.5 – 6.0 μm broadband emission, as a function of time for both the 493 ppmw Pr³⁺-Ga and the 527 ppmw Pr³⁺-In bulk glasses [Fig. 11(a)] and the 994 ppmw Pr³⁺-Ga and 1027 ppmw Pr³⁺-In bulk glasses [Fig. 11(b)]. The decay lifetime was deduced as the time taken for the PL intensity to fall by a fraction of 1/e. Both the ~500 ppmw and ~1000 ppmw Pr³⁺-doped In-glasses exhibited longer lifetimes (10.1 ms and 9.0 ms, respectively) than the analogously doped ~500 ppmw and ~1000 ppmw Pr³⁺- Ga-glasses (7.8 ms and 7.5 ms, respectively).

 

Fig. 11 PL decay at 4700 nm for pumping at 1550nm and nominally 57.8mW of: (a) 527 ppmw Pr³⁺-In and 493 ppmw Pr³⁺-Ga bulk glasses and (b) 1027 ppmw Pr³⁺-In and 994 ppmw Pr³⁺-Ga bulk glasses. For (a) and (b), single exponential decays are plotted to allow comparison.

Download Full Size | PDF

4. Discussion

4.1 Glass structure, stability and linear optical properties

Aitken et al. [24] have reported a superior EPR line narrowing of Gd³⁺ in the presence of Ga, as compared to Gd³⁺ in the presence of In, in sulfide chalcogenide glasses. These Authors concluded that Ga is a better dispersant of RE-ions in sulfide chalcogenide glasses than In. We would agree with that conclusion from our own work, as our selenide glasses doped with the Pr³⁺ rare earth ions in the presence of intended to tolerate only a lower doping level of the Pr³⁺ ions, than they did in the presence of Ga, before the selenide glass showed signs of devitrification on melt-cooling during the glass-preparation. And indeed the crystalline phase separating out in our glasses was the rare earth ion (Pr³⁺) selenide phase.

The glasses obtained after glass melting were found to be amorphous according to XRD at ≤ ~1500 ppmw Pr³⁺, for the Ga-series glasses, whereas the In-series glasses were amorphous only at ≤ ~1000 ppmw Pr³⁺. In each case the crystallized phase was Pr ^(III) Se_1.9 according to ref [30]. (but we cannot exclude other stoichiometries like: Pr ^(III) Se_~2.0). We know that the praseodymium is in the 3 + state because of the absorption and emission spectra we have presented here. It is not likely that the Pr³⁺ state could have undergone redox upon phase separation in the highly covalent selenide matrix. Moreover [33], reported that experimentally determined magnetic moments indicate Pr³⁺ for the polyselenides PrSe₂, PrSe_1.9 and PrSe_1.85.

Stoichiometric Pr₂Se₃ would exhibit a Pr³⁺: S ratio of 1.5:1. In contrast, Pr³⁺S_1.9 assigned here exhibits a Pr³⁺: S ratio of 1.9:1. The excess Se is chemically bonded as divalent -Se-Se- bridges. Doert and Graf [34] have shown that there are two crystallographic sites in PrSe₂, one of which contains an Se-Se bridge.

We recorded a tiny doublet of unassigned vibrational absorption bands just below 6 μm wavelength only in absorption spectra of the undoped, bulk, host glasses [see Fig. 3]. Therefore the doublet was not due to Pr, nor its contaminants. The absorption coefficients of the doublet ranged from 0.002 cm⁻¹, for the undoped Ga-bulk glass, to 0.005 cm⁻¹ _, for the undoped In-bulk glass; the peak absorptions of the doublet were at 5.8 and 5.95 μm wavelength for both types of glasses. The fact that the doublet appeared in the same wavelength position for both glasses means that it is unlikely to have been associated with vibrational absorption due to Ga- or In-based species. This is because the atomic mass difference of Ga and In would dictate different absorption frequencies of similar species. The doublet was not present in the Pr³⁺ doped-fiber optical loss spectra. The nearest impurity band we can find is As-OH, but this has been reported to absorb at 5.5 μm wavelength [35]

4.2 Fiber

It was not possible to resolve the core/cladding structure of the Ga-SIF and In-SIF using reflection visible microscopy [Fig. 1] as the core and cladding refractive indices were too similar. Also, fibers were too dark in transmission mode in the visible domain. Near-field imaging at 1.55 microns wavelength was not possible because the Pr³⁺- doped fiber cores were too absorbing. Near-field imaging is theoretically possible at 3-3.5 microns or around 6 microns, judging by the optical loss spectra [Fig. 6]. Previously [19], we have reported scanning electron microscopy with electron microprobe mapping analysis of a cross-section of a cleaved Pr³⁺-Ga-SIF and elemental mapping of Pr, measured against a standard Pr Foil (99.9& Alfa Aesar), but there was limited evidence of the core position.

4.3 PL behavior

In order to compare the PL of the bulk glasses we used a constant input power and ensured that the samples were the same shape and finish. However, we recognize that there will have been a small Fresnel reflection difference giving a slightly higher loss in the case of the In-bulk glasses compared to that of the Ga-bulk glasses. The bulk glass PL measurement was believed to be more controllable and more fundamentally interpretable than measurement of PL in the fiber, because injection into the fiber is a critical parameter. Here, the fiber PL results were normalized to enable only PL line-shape of the two fibers.

Shaw et al. [18] have concluded, when pumping at 1.064 µm with a CW Nd:YAG laser in Pr³⁺-doped BaInGaAsGaSe bulk glasses, that there was a rapid thermalization within each of the two pairs of close-lying upper states [³F₄ with ³F₃] and [³F₂ with ³H₆] Pr³⁺ levels in the chalcogenide glass host, with occupancy in the two lower states of at least ~90%. Such rapid thermalization of these levels could also have occurred here, for the 1.55 µm CW pump.

The emission here in the Pr³⁺-doped In- and Ga bulk glasses was observed to extend from 3.5 to 6 μm. Charpentier et al. [21], found a similar broad emission in Pr³⁺-doped GeSbGaSe glasses and assigned this to just two transitions: ³H₆ → ³H₅ and ³H₅ → ³H₄, stating that the ratio between these two transitions depends on the type of glass host matrix and on the re-absorption due to ³H₄→³H₅. Our previous modeling of Pr³⁺-doped GeAsGaSe supports this conclusion [19,22]. On the other hand, Cole et al. [20] have reported MIR fluorescence in the range 3 - 5 µm from core/clad. multimode selenide fiber pumped with a 2 μm laser diode: the fiber core glass, GeAsGaSe (atomic% (at%)), was doped with 200 ppm and 350 ppm Pr³⁺. The observed MIR emission in the “3-5 µm band” was attributed to all possible overlapping bands and due to: ¹G₄ → (³F₄, ³F₃) [3.3 μm and 2.9 μm]; ³F₄ → ³F₂ [5.4 μm, sic]; (³F₄, ³F₃) → ³H₆ [3.9 μm, 4.8 μm] ; (³H₆, ³F₂) → ³H₅ [4.5 μm, 3.4 μm] and ³H₅ → ³H₄ (N.B. ³H₄ was misprinted in the original as ³H₆) [4.9 μm] transitions. In the absence of any evidence to dismiss any of these, then we agree that all of these transitions should be considered.

For optimal emission, the [H-Se] impurity absorption at 4.6 μm [32] in Pr³⁺-ion doped selenide glasses should be minimized but it is usually observed [5,6,20,21], as found here. We measured the excited state lifetime at 4.7 μm wavelength because this was the peak of the broadband emission in the 3.5 μm to 6 μm range. However, this 4.7 μm peak lay close to the 4.6 μm [H-Se] vibrational absorption and therefore the true peak emission may lie at shorter wavelengths than 4.7 μm, and maybe as low as 4.5 μm.

The Pr³⁺-doped bulk glass absorption and emission [Fig. 10] overlapped in the 3.5 - 6 μm band. Hence for fibers, the shorter wavelength emission was observed to be self-absorbed. Interestingly, re-emission was observed in the fiber at longer wavelengths. This could indicate that the unwanted resonant vibrational absorption at 4.6 μm wavelength, due to the presence of [H-Se] impurity in the glass, may not be too much of a dominating factor. The enhanced emission at longer wavelengths may be a combination of re-absorption along the fiber and the normalization process. There is likely to be low, or no, population inversion under the conditions used here, as the pump power was low. Emission power out from the fibers has not been reported here, only normalized emission to compare the PL line-shapes. It was observed that the fiber emission intensity was linear with pump power, over the power range used here.

The experimental lifetimes measured here (from 7.5 to 10.1 ms) were longer than those reported [18] of 2.5 ± 0.9 ms for 1000 ppmw Pr³⁺-doped BaInGaGeSe bulk glass, at 4.5-5 μm using a 1.55 μm pump, for which the calculated radiative lifetime was 12.9 ms [18].

PL intensity of the Pr³⁺ -In glasses measured here exhibited about double the emission intensity of the Pr³⁺-Ga glasses, for both of the ~500 ppmw, and 1000 ppmw, RE-ion doping levels. Moreover, the excited state lifetimes of the ~500 ppmw and 1000 ppmw Pr³⁺-In glasses were found to be longer than those of the ~500 ppmw and 1000 ppmw Pr³⁺-Ga glasses. These superior PL results found for the Pr³⁺ -In glasses support the hypothesis that the Ga and In each can complex the RE-ions in these glasses to form: ([RE³⁺] - Se - [Ga^(III) or In^(III)]) units as part of the glass matrix and that indium provides a lower phonon energy local environment in the glass structure than gallium.

5. Conclusions

Pr³⁺ was successfully dissolved in GeAsGaSe bulk glasses up to ~1500 ppmw and in GeAsInSe bulk glasses up to ~1000 ppmw. ~500 ppmw Pr³-doped optically clad fiber was successfully fabricated with minimum optical losses of 2.8 dB/m at 6.58 μm for Pr³⁺-Ga-SIF and 4.3 dB/m at 6.58 μm for Pr³⁺-In-SIF. Bulk glasses doped with up to ~1000 ppmw Pr³⁺, and fiber doped with ~500 ppmw Pr³⁺, gave broadband emission from 3.5 - 6 μm, when pumped at 1.55 μm wavelength. The fluorescence intensity was found to be approximately double for the ~500 ppmw and ~1000 ppmw Pr³⁺-In bulk glasses compared to the analogously doped Ga bulk glasses. Moreover, the 500 ppmw and ~1000 ppmw Pr³⁺- In bulk glasses exhibited longer decay-lifetimes, at the 4.7 μm peak emission intensity when pumped at 1.55 μm compared to the analogously doped Ga bulk glasses. In all cases the emission behavior of the Pr³⁺-ion suggests that the addition of In in the glass formulation, as opposed to Ga, is favorable in providing a lower phonon energy local environment for the Pr³⁺ion. This supports the existence of [Pr³⁺-Se-In^(III)] complexation in the host glass matrix with the result of not only aiding RE-ion solubility but also enhancing the emission behavior of the active RE-ion.